Vola

Marc Volume A: Theory and User Information
Marc® 2021.1
Volume A: Theory and User Information
Corporate Europe, Middle East, Africa
MSC Software Corporation MSC Software GmbH
5161 California Ave., Suite 200 Am Moosfeld 13
University Research Park 81829 Munich, Germany
Irvine, CA 92617 Telephone: (49) 89 431 98 70
Telephone: (714) 540-8900 Email: europe@mscsoftware.com
Email: americas.contact@mscsoftware.com
Japan Asia-Pacific
MSC Software Japan Ltd. MSC Software (S) Pte. Ltd.
KANDA SQUARE 16F 100 Beach Road
2-2-1 Kanda Nishikicho #16-05 Shaw Tower
Chiyoda-ku Singapore 189702
Tokyo 101-0054, JAPAN Telephone: 65-6272-0082
Telephone: (81) (3)-6275-0870 Email: APAC.Contact@mscsoftware.com
Email: MSCJ.Market@mscsoftware.com
Worldwide Web
www.mscsoftware.com
Support
http://www.mscsoftware.com/Contents/Services/Technical-Support/Contact-Technical-Support.aspx
Disclaimer

User Documentation: Copyright  2021 MSC Software Corporation. All Rights Reserved.
This document, and the software described in it, are furnished under license and may be used or copied only in accordance with the
terms of such license. Any reproduction or distribution of this document, in whole or in part, without the prior written authorization of
MSC Software Corporation is strictly prohibited.
MSC Software Corporation reserves the right to make changes in specifications and other information contained in this document
without prior notice. The concepts, methods, and examples presented in this document are for illustrative and educational purposes
only and are not intended to be exhaustive or to apply to any particular engineering problem or design. THIS DOCUMENT IS
PROVIDED ON AN “AS-IS” BASIS AND ALL EXPRESS AND IMPLIED CONDITIONS, REPRESENTATIONS AND WARRANTIES,
INCLUDING ANY IMPLIED WARRANTY OF MERCHANTABILITY OR FITNESS FOR A PARTICULAR PURPOSE, ARE
DISCLAIMED, EXCEPT TO THE EXTENT THAT SUCH DISCLAIMERS ARE HELD TO BE LEGALLY INVALID.
MSC Software logo, MSC, MSC Nastran, Adams, Dytran, Marc, Mentat, and Patran are trademarks or registered trademarks of
MSC Software Corporation or its subsidiaries in the United States and/or other countries.
NASTRAN is a registered trademark of NASA. Python is a trademark of the Python Software Foundation. Parasolid is a registered
trademark of Siemens Product Lifecycle Management Software, Inc. All other trademarks are the property of their respective owners.
This software may contain certain third-party software that is protected by copyright and licensed from MSC Software suppliers.
Additional terms and conditions and/or notices may apply for certain third party software. Such additional third party software terms
and conditions and/or notices may be set forth in documentation and/or at http://web.mscsoftware.com/thirdpartysoftware (or successor
website designated by MSC from time to time).
The MSC Software Logo, MSC, MSC Nastran, Marc, Patran, Dytran, and Laminate Modeler are trademarks or registered trademarks
of MSC Software Corporation in the United States and/or other countries. Hexagon and the Hexagon logo are trademarks or registered
trademarks of Hexagon AB and/or its subsidiaries.
METIS is copyrighted by the regents of the University of Minnesota. MS MPI is developed by Microsoft Corporation. PCGLSS 8.0,
Copyright  1992-2018 Computational Applications and System Integration Inc. All rights reserved. Portions of this software are
owned by Siemens Product Lifecycle Management Software, Inc. Copyright 2020.
Use, duplication, or disclosure by the U.S. Government is subject to restrictions as set forth in FAR 12.212 (Commercial Computer
Software) and DFARS 227.7202 (Commercial Computer Software and Commercial Computer Software Documentation),
as applicable.
U.S. Patent 9,361,413
MA:V2021.1:Z:Z:Z:DC-VOL-A-PDF
Documentation Feedback
At MSC Software, we strive to produce the highest quality documentation and welcome your
feedback. If you have comments or suggestions about our documentation, write to us at:
documentation-feedback@mscsoftware.com
Please include the following information with your feedback:
• Document name
• Release/Version number
• Chapter/Section name
• Topic title (for Online Help)
• Brief description of the content (for example, incomplete/incorrect information,
grammatical errors, information that requires clarification or more details and so on.)
• Your suggestions for correcting/improving documentation
You may also provide your feedback about MSC Software documentation by taking a short 5-minute
survey at: http://msc-documentation.questionpro.com
Note: The above mentioned e-mail address is only for providing documentation specific
feedback. If you have any technical problems, issues, or queries, please visit the
Technical Support Contact Information.
C O N T E N T S
Contents
Contents
Table of Contents
Preface  About This Guide, 20

 Purpose of This Guide, 20
 Contents of This Guide, 21
 Typographical Conventions, 22
 Accessing Marc Manuals, 23
 Downloading the PDF Documentation Files, 24
 Navigating the PDF Files, 24
 Printing the PDF Files, 24
 Training and Internet Resources, 25
 Technical Support, 27
 Visit SimCompanion, 27
 Help Us Help You, 28
1
The Marc System  Overview, 30
 Marc Programs, 30
 Marc for Analysis, 31
 Mentat or Patran for GUI, 31
 Structure of Marc, 32
 Procedure Library, 32
 Material Library, 32
 Element Library, 32
 Program Function Library, 32
 Features and Benefits of Marc, 32
2
Program Initiation  Overview, 35
 Workspace Requirements, 35
 Marc Workspace Requirements, 35
 File Units, 37
 Program Initiation, 40
 Examples of Running Marc Jobs, 42
3
Data Entry  Overview, 46
 Input Conventions, 46
 Input of List of Items, 47
 Examples of Lists, 49
 Input File Version 13, 49
 Table Driven Input, 52
 Table Input, 52
 Parameters, 57
 Model Definition Options, 57
 History Definition Options, 57
 REZONE Option, 58
4
Structural Procedure Library  Overview, 60
 Linear Analysis, 60
 Accuracy, 62
 Error Estimates, 62
 Adaptive Meshing, 62
 Fourier Analysis, 62
 Nonlinear Analysis, 65
 Geometric Nonlinearities, 69
 Eulerian Formulation, 77
 Arbitrary Eulerian-Lagrangian (AEL) Formulation, 79
 Shell Thickness Update, 79
 Nonlinear Boundary Conditions, 80
 Buckling Analysis, 82
 Perturbation Analysis, 83
 Computational Procedures for Elastic-Plastic Analysis, 89
 Creep, 101
 Viscoelasticity, 105
 Viscoplasticity, 106
 Fracture Mechanics, 107
 Linear Fracture Mechanics, 107
 Nonlinear Fracture Mechanics, 110
 Numerical Evaluation of the J-integral, 111
 Numerical Evaluation of the Energy Release Rate with the VCCT Method, 113
 Automatic Crack Propagation, 119
 Dynamic Fracture Methodology, 129
 Dynamic Crack Propagation, 130
 Crack Initiation, 130
 Mesh Splitting, 132
 Mesh Splitting Along Edges or Faces, 132
 Mesh Cutting, 134
 Dynamics, 135
 Modal (Eigenvalue) Analysis, 135
 Harmonic Response, 139
 Spectrum Response, 142
 Transient Analysis, 146
 Inertia Relief, 158
 Rigid Body Mode Evaluation, 158
 Rigid-Plastic Flow, 161
 Steady State Analysis, 162
 Technical Background, 162
 Superplasticity, 163
 Soil Analysis, 165
 Technical Formulation, 166
 Mechanical Wear, 170
 Design Sensitivity Analysis, 172
 Theoretical Considerations, 174
 Design Optimization, 175
 Approximation of Response Functions Over the Design Space, 176
 Improvement of the Approximation, 178
 The Optimization Algorithm, 178
 Marc User Interface for Sensitivity Analysis and Optimization, 179
 Define Initial State with Results from a Previous Analysis, 181
 Pre State, 181
 Model Sections, 183
 Steady State Rolling Analysis, 185
 Kinematics, 186
 Inertia Effect, 187
 Rolling Contact, 188
 Steady State Rolling with Marc, 188
 Structural Zooming Analysis, 188
 Element Types Supported, 189
 Uncoupled Thermal Stress Analysis, 190
 Cure-Thermal-Mechanically Coupled Analysis, 191
 Cure Kinetics, 192
 Cure Shrinkage Strain, 195
 References, 197
5
Material Library  Overview, 201
 Linear Elastic Material, 201
 Composite Material, 203
 Layered Materials, 204
 Classical Lamination Theory for Multi-Layered Shells, 207
 Material Preferred Direction, 208
 Material Dependent Failure Criteria, 214
 Interlaminar Shear for Thick Shell, Beam, Solid Shell, and 3-D Composite Brick Elements, 233
 Interlaminar Stresses for Continuum Composite Elements, 235
 Progressive Composite Failure, 235
 Mixture Model, 238
 Gasket, 241
 Constitutive Model, 242
 Nonlinear Hypoelastic Material, 247
 Thermo-Mechanical Shape Memory Model - Type 1, 261
 Transformation Induced Deformation, 263
 Constitutive Theory, 265
 Phase Transformation Strains, 265
 Experimental Data Fitting for Thermo-mechanical Shape Memory Alloy, 267
 Thermo-Mechanical Shape Memory Model - Type 2, 271
 Calibration of the Parameters Related to Tension/Compression Sensitivity, 274
 Mechanical Shape Memory Model, 280
 Experimental Data Fitting for Mechanical Shape Memory Alloy, 284
 Conversion from Thermo-Mechanical to Mechanical SMA, 285
 Elastomer, 286
 Updated Lagrange Formulation for Nonlinear Elasticity, 305
 Time-independent Inelastic Behavior, 306
 Yield Conditions, 308
 Barlat 2005 Yield Functions, 314
 Vegter Yield Criterion, 315
 Mohr-Coulomb Material (Hydrostatic Stress Dependence), 319
 Buyukozturk Criterion (Hydrostatic Stress Dependence), 320
 Powder Material, 321
 Obtaining Crush Curve and Shear Failure Parameters by Curve Fitting in Marc, 331
 Work or Strain Hardening, 334
 Flow Rule, 341
 Constitutive Relations, 342
 Time-independent Cyclic Plasticity (Chaboche Model), 344
 Annealing, 353
 Time-dependent Inelastic Behavior, 355
 Creep (Maxwell Model), 359
 Oak Ridge National Laboratory Laws, 363
 Swelling, 365
 Viscoplasticity, 366
 Time-dependent Cyclic Plasticity (Chaboche Model), 366
 Anand Solder Model, 367
 Viscoelastic Material, 369
 Bergström-Boyce Model, 381
 Parallel Rheological Framework (PRF), 383
 Narayanaswamy Model, 384
 Frequency-dependent Material Behavior, 388
 Viscoelastic Material Behavior in the Frequency Domain, 389
 Amplitude Dependent Relaxation in the Frequency Domain, 405
 Thermo-Rheologically Simple Material Behavior in the Frequency Domain, 429
 Deformation Dependent Relaxation in the Frequency Domain, 430
 Using Tables to Define Storage and Loss Moduli, 432
 Harmonic Equations of Motion, 435
 Performing Viscoelastic Analysis in the Frequency Domain, 439
 Temperature Effects and Coefficient of Thermal Expansion, 445
 Piecewise Linear Representation, 446
 Temperature-Dependent Creep, 447
 Coefficient of Thermal Expansion, 447
 Phase Transformations, 448
 Phases of Steel, 452
 Transformation Temperatures, 454
 Heat Treatment, 456
 Material Input, 457
 Time-Temperature-Transformation, 459
 Low Tension Material, 461
 Uniaxial Cracking Data, 462
 Low Tension Cracking, 462
 Tension Softening, 462
 Crack Closure, 463
 Crushing, 463
 Analysis, 463
 Soil Model, 463
 Elastic Models, 464
 Cam-Clay Model, 464
 Evaluation of Soil Parameters for the Critical State Soil Model, 466
 Damage Models, 475
 Ductile Metals, 475
 Elastomers, 478
 Fatigue Analysis, 482
 Cohesive Zone Modeling, 485
 Nonstructural Materials, 494
 Heat Transfer Analysis, 494
 Piezoelectric Analysis, 494
 Thermo-Electrical Analysis, 495
 Coupled Electrical-Thermal-Mechanical Analysis, 495
 Hydrodynamic Bearing Analysis, 495
 Fluid/Solid Interaction Analysis – Added Mass Approach, 495
 Electrostatic Analysis, 495
 Magnetostatic Analysis, 495
 Magnetodynamic Analysis, 495
 Coupled Electrostatic-Structural, 496
 Acoustic Analysis, 496
 Fluid Analysis, 496
 References, 496
6
Contact  Introduction, 502
 Definition of Contact Bodies, 502
 Contact Tables, 506
 Motion of Bodies, 506
 Initial Conditions, 508
 Initial Gap, 509
 Node-to-Segment Contact, 510
 Numbering of Contact Bodies, 510
 Detection of Contact, 513
 Shell Contact, 515
 Neighbor Relations, 516
 Implementation of Constraints, 517
 Friction Modeling, 519
 Anisotropic Friction Model, 530
 Separation, 533
 Release, 534
 Glue Contact Conditions, 534
 Coupled Analysis, 540
 Joule Heating, 542
 Contact in Electrostatic or Piezoelectric Analysis, 543
 Contact in Magnetostatic Analysis, 544
 Contact in Magnetodynamic Analysis, 544
 Contact in Soil Analysis, 544
 Contact in Acoustic Analysis, 544
 Element Considerations, 544
 Shell Elements, 547
 Dynamic Impact, 547
 Global Remeshing and Rezoning, 548
 Local Adaptivity, 548
 Results Evaluation, 549
 Tolerance Values, 551
 Numerical Aspects of the Direct Constraint Procedure, 553
 Solution Strategy for Deformable Contact, 560
 Iterative Penetration Checking, 561
 Node to Segment Linear Contact, 562
 Instabilities, 562
 Segment-to-Segment Contact, 563
 Basic Theory for Normal and Frictional Contact, 563
 Contact Body Definition, 567
 Beam Contact, 568
 Contact Detection, 572
 Force Vector and Stiffness Matrix Contribution, 574
 Iterative Penetration Checking and Contact Stress Update, 576
 Separation, 577
 Finite Sliding for Deformable Contact, 578
 Results Evaluation, 578
 Multi-physics, 581
 Trouble Shooting, 581
 Segment to Segment Linear Contact, 582
 Interference Fit, 583
 Limitations of Interference Fit, 586
 Symmetry Conditions, 586
 References, 587
7
Introduction to Mesh Definition  Overview, 590
 Direct Input, 590
 Element Connectivity Data, 590
 Nodal Coordinate Data, 596
 Activate/Deactivate, 596
 User Subroutine Input, 597
 MESH2D, 597
 Block Definition, 597
 Merging of Nodes, 597
 Block Types, 598
 Symmetry, Weighting, and Constraints, 600
 Additional Options, 601
 Mentat, 602
 FXORD Option, 602
 Major Classes of the FXORD Option, 603
 Recommendations on Use of the FXORD Option, 607
 Incremental Mesh Generators, 607
 Bandwidth Optimization, 608
 Rezoning, 608
 Substructure, 609
 Scaling Element Stiffness, 612
 BEAM SECT Parameter, 612
 Orientation of the Section in Space, 613
 Definition of the Section, 613
 Error Analysis, 619
 Local Adaptivity, 619
 Number of Elements Created, 619
 Boundary Conditions, 620
 Location of New Nodes, 621
 Adaptive Criteria, 622
 Automatic Global Remeshing, 625
 Remeshing Region – Partial Remeshing, 629
 Remeshing Criteria, 630
 Meshing Techniques, 632
 Mesh Density Control, 634
 Geometry Preservation, 639
 Remeshing with Self Contact, 641
 Remeshing with Boundary Conditions, 642
 History Data Mapping Technique, 644
 Trimming, 644
8
Boundary Conditions  Loading, 649
 Loading Types, 649
 Face ID for Distributed Loads, Fluxes, Charge, Current, Source, Films, and Foundations, 658
 Mechanical Loads, 663
 Fluid Drag and Wave Loads, 666
 Cavity Pressure Loading, 667
 Cyclic Loading, 672
 Thermal Loads, 673
 Initial Stress and Initial Plastic Strain, 673
 Inherent strains, 674
 Heat Fluxes, 676
 Mass Fluxes and Restrictors, 677
 Electrical Currents, 677
 Electrostatic Charges, 678
 Acoustic Sources, 678
 Piezoelectric Loads, 678
 Electrostatic-Structural Loads, 679
 Magnetostatic Currents, 679
 Magnetodynamic Currents and Charges, 679
 Kinematic Constraints, 679
 Transformation of Degree of Freedom, 681
 Shell Transformation, 682
 Tying Constraint, 684
 Rigid Link Constraint, 693
 Shell-to-Solid Tying, 694
 Rigid Tying to a Surface Patch, 694
 Overclosure Tying, 696
 Cross Section, 700
 Insert, 702
 AUTOMSET, 702
 Support Conditions, 703
 Bushings, 703
 Cyclic Symmetry, 706
 MSC Nastran RBE2 and RBE3, 709
 Beam - Shell Offsets, 714
 Pin Code for Beam Elements, 717
 Mesh Independent Connection Methods, 717
 CWELD patch-to-patch connections, 720
 Parameters for the Projection Process, 724
 Internal Constraints to Connect Two Surfaces, 726
 Multiple Surface Connections, 727
 Coupled Thermo-mechanical Analysis, 727
 Rules for usage of the GS-node or the GA-, GB-node Pair in the Projection Process, 727
 CWELD Properties, 729
 CWELD Input Styles, 732
 CWELD Error and Warning Messages, 733
 CWELD Output, 735
 Two Dimensional Models, 735
 CFAST Connections, 736
 CFAST Properties, 736
 SWLDPRM, Special CWELD/CFAST Parameters, 737
9
Element Library  Truss Elements, 745
 Membrane Elements, 746
 Continuum Elements, 746
 Beam Elements, 747
 Plate Elements, 748
 Heat Transfer Elements, 748
 Coupled Electrostatic-Structural, 749
 Fluid/Solid Interaction, 749
 Soil Analysis, 750
 Fluid Analysis, 750
 Special Elements, 750
 Gap-and-Friction Elements, 750
 Pipe-bend Element, 751
 Curved-pipe Element, 751
 Shear Panel Element, 751
 Cable Element, 751
 Rebar Elements, 751
 Interface Elements, 751
 Incompressible Elements, 752
 Large Strain Elasticity, 752
 Large Strain Plasticity, 752
 Rigid-Plastic Flow, 752
 Constant Dilatation Elements, 753
 Elements with Strain Smoothing, 753
 Reduced Integration Elements, 754
 Continuum Composite Elements, 754
 Fourier Elements, 754
 Semi-infinite Elements, 754
 Cavity Surface Elements, 755
 Assumed Strain Formulation, 755
 Follow Force Stiffness Contribution, 755
 Explicit Dynamics, 755
 Adaptive Mesh Refinement, 756
10
Nonstructural Procedure  Overview, 758
Library  Heat Transfer, 758
 Thermal Contact, 760
 Convergence Controls, 760
 Temperature Effects, 762
 Surface Energy, 768
 Mass Conservation Equation, 776
 Surface Recession Rate, 776
 Surface Recession and Remeshing, 780
 Pyrolysis, 791
 Coking, 794
 Monitoring Thermal Degradation, 798
 Presentation of the Energy Equation, 799
 Welding, 801
 Radiation, 811
 Conrad Gap, 825
 Channel, 826
 Output, 827
 Molecular Diffusion, 829
 Technical background, 829
 Diffusion, 831
 Hydrodynamic Bearing, 835
 Coils, 861
 Circuit Approach for Magnetostatics, Harmonic, and Transient Magnetodynamic Analysis, 864
 Strain Based Piezoelectric Coupling, 876
 Rigid Cavity Acoustic Analysis, 877
 Fluid Mechanics, 879
 Finite Element Formulation, 882
 Penalty Method, 884
 Solid Analysis, 885
 Solution of Coupled Problems in Fluids, 885
 Degrees of Freedom, 886
 Element Types, 886
 Nonlocal Analyses, 887
 Coupled Analyses, 888
 Thermal Mechanically Coupled Analysis, 891
 Coupled Acoustic-Structural Analysis, 893
 Fluid/Solid Interaction – Added Mass Approach, 897
 Coupled Electrostatic-Structural Analysis, 898
 Coupled Magnetostatic-Structural Analysis, 900
 Coupled Thermal-Electrical Analysis (Joule Heating), 902
 Coupled Electrical-Thermal-Mechanical Analysis, 906
 Coupled Magnetostatic-Thermal Analysis, 907
 Coupled Magnetodynamic-Thermal Analysis, 908
 Coupled Magnetodynamic-Thermal-Structural Analysis, 910
 Coupled Diffusion/Thermal Analysis, 913
 Molecular Diffusion Coupled Analyses, 923
 References, 925
11
Solution Procedures for  Overview, 928
Nonlinear Systems  Considerations for Nonlinear Analysis, 928
 Behavior of Nonlinear Materials, 929
 Scaling the Elastic Solution, 929
 Load Incrementation, 929
 Selecting Load Increment Size, 931
 Fixed Load Incrementation, 931
 User Defined Time Incrementation, 931
 Arc Length Method, 932
 Residual Load Correction, 940
 Restarting the Analysis, 941
 Newton-Raphson Method, 942
 Modified Newton-Raphson Method, 943
 Strain Correction Method, 944
 Direct Substitution, 945
 Arc-length Methods, 945
 Convergence Controls, 952
 Equilibrium per node, 955
 Singularity Ratio, 956
 AutoSPC, 957
 Solution of Linear Equations, 957
 Direct Methods, 958
 Iterative Methods, 958
 Mixed Direct-Iterative Solver, 959
 Preconditioners, 959
 Storage Methods, 960
 Nonsymmetric Systems, 960
 Complex Systems, 961
 Iterative Solvers, 961
 Parallel Solvers, 962
 Basic Theory, 962
 Flow Diagram, 963
 Remarks, 964
 References, 965
12
Parallel Processing  Overview, 967
 Domain Decomposition, 967
 Element Loop Parallelism, 967
 Matrix Solution, 967
 Different Types of Machines, 967
 Supported and Unsupported Features, 968
 Matrix Solvers, 970
 Contact, 971
 GPU Support, 971
 Technical Discussion, 972
 Input, 972
 Examples, 972
 Domain Decomposition, 973
 Running a Parallel Job, 973
 Domain Decomposition Methods, 975
13
Output Results  Overview, 981
 Workspace Information, 981
 Increment Information, 984
 Summary of Loads, 985
 Timing Information, 985
 Singularity Ratio, 986
 Convergence, 986
 Selective Printout, 986
 Options, 986
 Grid Force Balance, 988
 User Subroutines, 988
 Restart, 988
 Element Information, 988
 Mentat Computed Element Results, 990
 Solid (Continuum) Elements, 990
 Solid-shell Elements, 990
 Beam Elements, 991
 Gap Elements, 992
 Linear and Nonlinear Springs, 992
 Joule Heating Analysis, 993
 Piezoelectric or Electrostatic-Structural Analysis, 993
 Magnetodynamics Analysis, 993
 Nodal Information, 994
 Stress Analysis, 994
 Reaction Forces, 995
 Residual Loads, 995
 Dynamic Analysis, 995
 Joule Heating Analysis, 995
 Rigid-Plastic Analysis, 995
 Piezoelectric or Electrostatic-Structural Analysis, 996
 Supplementary Information, 996
 Contact Analysis, 996
 Post File, 996
 Forming Limit Parameter (FLP), 997
 Program Messages, 1000
 Marc HyperMesh Results Interface, 1001
 Marc SDRC I-DEAS Results Interface, 1001
 Marc HDF Results Interface, 1002
 HDF File Creation, 1002
 HDF File Post-Processing, 1003
 HDF File Layout, 1004
 Marc - ADAMS Results Interface, 1005
 Status File, 1007
 Element Group Information, 1007
 References, 1008
14
Code Coupling Interfaces  Overview, 1010
 General Code Coupling, 1010
 Coupling Regions, 1012
 Time Step Control, 1014
 Parallel Processing, 1014
 Marc Co-simulation Adapter, 1015
 ACSI Adapter, 1015
 Communication Architecture, 1017
 Co-simulation Setup, 1017
 Interpolation/Extrapolation Algorithms, 1017
 Co-simulation Algorithm, 1018
 CoSim Adapter, 1020
A
Finite Element Technology in  Governing Equations of Various Structural Procedures, 1023
Marc  System and Element Stiffness Matrices, 1025
 Load Vectors, 1026
B
Finite Element Analysis of NC  General Description, 1029
Machining Processes  NC Files (Cutter Shape and Cutter Path Definition), 1029
 Intersection Between Finite Element Mesh and Cutter, 1031
 Deactivation of Elements, 1031
 Adaptive Remeshing, 1031
 Typical Features for Machining, 1032
 Input of Initial Stresses, 1032
C
Material Data Files and Data  External Material Data Files, 1034
Encryption  Data Encryption, 1034
INDEX
Preface
Preface
 About This Guide 20


Purpose of This Guide 20
 Contents of This Guide 21
 Typographical Conventions 22
 Accessing Marc Manuals 23
 Training and Internet Resources 25
 Technical Support 27
20 Marc Volume A: Theory and User Information
About This Guide

The Marc Volume A: Theory and User Information, the first in a series of five volumes documenting the Marc Finite Element
program, organizes the features and operations of the Marc program sequentially. It gives information about Marc’s scope
and capabilities to model complex physical problems and the theoretical basis of the computational techniques used to solve
them. You will find references to the following documents throughout this guide:
Marc Documentation
TITLE VOLUME
Element Library Volume B
Program Input Volume C
User Subroutines and Special Routines Volume D
Demonstration Problems Volume E
Release Guide
Installation and Operations Guide
Python: Tutorial and Reference Manual
Purpose of This Guide

The purpose of this volume is:
1. To help you define your finite element problem by describing Marc’s capabilities to model physical problems.
2. To identify and describe complex engineering problems and introduce Marc’s scope and capabilities for solving
these problems.
3. To assist you in accessing Marc features that are applicable to your particular problems and to provide you with
references to the rest of the Marc literature.
4. To provide you with the theoretical basis of the computational techniques used to solve the problem.
CHAPTER 21
Preface
Contents of This Guide

The principal categories of information are found under the following titles:
Chapter Title Description

Chapter 1 The Marc System Demonstrates Marc programs to perform stress analysis, structure
of Marc and feature and benefits of Marc.
Chapter 2 Program Initiation Describes commands/procedures along with examples that
facilitate the execution of Marc. It also encompasses workspace
requirements and FORTRAN File Units Used by Marc.
Chapter 3 Data Entry Shows different approaches to defining the data inputs such as list
of items, table, parameter, model definition options, history
definition options, rezone option.
Chapter 4 Structural Procedure Library Expands the analysis procedures in Marc applicable to structural
problems. These procedures range from simple linear elastic
analysis to complex nonlinear analysis. It also provides technical
background information as well as usage information about
these capabilities.
Chapter 5 Material Library Describes the basic information about material models available
in Marc and the behavior of various types of engineering materials
specifying the data required by the program for each material.
Chapter 6 Contact Demonstrates modeling the contact phenomena for the
simulation of most physical problems. Phenomena mainly
associate with detect the motion of the bodies, apply a constraint
to avoid penetration, and apply appropriate boundary conditions
to simulate the frictional behavior and heat transfer.
Chapter 7 Introduction to Mesh Definition Introduces the internally available techniques for mesh definition
process of converting a physical problem into discrete geometric
entities for the purpose of analysis.
Chapter 8 Boundary Conditions Depicts different kinds of loading, kinematic constraints and
mesh independent connection methods required for analyses.
Chapter 9 Element Library Describes Marc’s extensive element library which allows
you to model various types of one-, two-, and three-dimensional
structures, such as plane stress and plane strain structures,
axisymmetric structures, full three-dimensional solid structures,
and shell-type structures.
Chapter 10 Nonstructural Procedure Library Demonstrates the nonstructural analysis procedures available in
Marc assisting several areas including heat transfer, hydrodynamic
bearing, electrostatic, magnetostatic, electromagnetic,
piezoelectric, fluid mechanics, and coupled analysis.
Chapter Title Description

Chapter 11 Solution Procedures for Nonlinear Discusses the solution schemes in Marc for nonlinear problems.
Systems Issues of convergence controls, singularity ratio, and available
solvers for linearized system of equations are also discussed
Chapter 12 Parallel Processing Describes Marc’s capability of using multiple processors at a time
when performing an analysis. A detailed list of unsupported
features is given later on in this chapter.
Chapter 13 Output Results Summarizes the information that Marc provides in the output. It
includes workspace information, increment information, selective
printout, restart, element information, nodal information, post
file, forming limit parameter, program messages and many more.
Chapter 14 Code Coupling Interfaces Demonstrates the co-simulation interfaces to couple Marc with
external solvers. The interface allows code coupling libraries such
as MpCCI to couple Marc with commercial CFD codes, such as
STAR-CD and Fluent. The interface may also be used for
developing dedicated interfaces to external solvers, for applying
complex boundaries to certain regions of the model, or for
dedicated postprocessing.
Appendix A Finite Element Technology in Marc Discusses finite element technology employed in Marc. It
comprises governing equations of various structural procedures,
system and element stiffness matrices, load vectors.
Appendix B Finite Element Analysis of NC Machining Depicts finite element technology with reference to NC
Processes machining processes employed in Marc to predict the distortion
during the 3-D bulk machining so that engineers can optimize
their structure design and machining processes.
Appendix C Material Data Files and Data Encryption Describes how material data properties can be used by referring
external files. It also explains data encryption mechanism for
material files to protect the sensitive information against
unauthorized use.
Typographical Conventions
The section provides a brief overview of the typographical conventions used in the document to help you better follow the
Marc and Mentat documentation.
CHAPTER 23
Preface
This section describes some syntax that will help you in understanding text in the various chapters and thus in facilitating
your learning process. It contains stylistic conventions to denote user action, to emphasize particular aspects of Marc and
Mentat to signal other differences within the text.
Adobe Garamond Body and general text
Pro
Courier New  Represents command-line options of Marc and Mentat.
 Directory names and paths
 File names and Paths
 Linux terminal script
Example: lmreread -c <parent>/msc/MSC.Licensing/licenses/license.dat
Bold Text  Highlights
 Dialog box
 Buttons
 Menus
 The commands/user inputs for all descriptions related to terminal commands.
 Default values
Example: [root@vm-tmrhel73 MSC]# ./msc_licensing_helium_linux64.bin
HelveticaNeueLT  Hyperlinks
Pro Cn 57  Weblinks
Example: Appendix. A: Microsoft Windows: Marc Subdirectories and Installation
Italic Text Represents references to books.
Example: Volume A: Theory and User Information

20XX Represents the latest version number. If the release is 2019, XX stands for 19.
Accessing Marc Manuals

This section describes how to access the Marc documentation outside of MSC Software. Marc documentation is available
through PDF files. The PDF files can be obtained from the following sources:
 Marc documentation installer
 SimCompanion
 Combined documentation
The PDF documentation files are appropriate for viewing and printing with Adobe Acrobat Reader (version 5.0 or higher),
which is available for most Windows and Linux systems. These files are identified by a .pdf suffix in their file names.
Downloading the PDF Documentation Files

You can download the PDF documentation from SimCompanion (http://simcompanion.mscsoftware.com).
Navigating the PDF Files

For the purpose of easier online document navigation, the PDF files contain hyperlinks in the table of contents and index.
In addition, links to other guides, hyperlinks to all cross-references to chapters, sections, figures, tables, bibliography, and
index entries have been applied.
Printing the PDF Files

Adobe Acrobat PDF files are provided for printing all or part of the manuals. You can select the paper size to which you
are printing in Adobe Acrobat Reader by doing the following:
1. Click File.
2. Select the Print.... option. The Print dialog box is displayed.
3. Select Page Setup....
4. Choose the required paper size in the Page Setup menu.
The PDF files are recommended when printing long sections since the printout will have a higher quality.
If the page is too large to fit on your paper size, you can reduce it by doing the following:
1. Select the File -> Print.
2. Under Page Scaling, choose the Shrink to Printable Area option.
CHAPTER 25
Preface
Training and Internet Resources

Information about MSC's products, services and latest events is available on our website www.mscsoftware.com.
The information about MSC Seminars is available on the training link http://www.mscsoftware.com/msc-training. You can also
use this link to schedule the seminars.
If you are a new Marc user, we recommend the following courses:
MAR101 - Basic Nonlinear Analysis using Marc and Mentat

The purpose of this course is to introduce the new Marc user to both Marc and Mentat by lectures and hands on modeling
of nonlinear problems.
Pre-requisites:
A basic knowledge of statics and strength of materials is highly recommended. Previous finite element analysis experience
is recommended.
Topics:
 Introduction to Mentat
 Nonlinear Finite Element Analysis
 Resolving Convergence Problems
 Numerical Analysis of Nonlinear Problems
MAR102 - Advanced Nonlinear Analysis using Marc and Mentat

The purpose of this course is to enhance the current Marc user's understanding of modeling nonlinear problems. Lectures
are supported by hands-on modeling of nonlinear problems.
Pre-requisites:
A basic knowledge of nonlinear simulations - Familiarity with Mentat 2011 - Completion of MAR101 (Basic Nonlinear
Analysis using Marc and Mentat) or equivalent experience.
Topics:
 Material Nonlinearity
 Contact
 Adaptive Meshing
 User Subroutines in Marc
 Heat Transfer and Thermal Stresses
 Global - Local (Structural Zooming) Analysis in Marc
 Restarts
 Performance
 Workshop Problems
MAR103 - Experimental Elastomer Analysis

The purpose of this course is to provide a fundamental understanding of how material testing and finite element analysis
are combined to improve the design of rubber and elastomeric products.
Pre-requisites:
A basic knowledge of statics and strength of materials is highly recommended. Previous finite element analysis experience
is recommended. And the knowledge of elastomeric materials.
Topics:
 Introduction
 Overview of Elastomer Testing and Analysis
 Uniaxial Tension/Compression Testing and Analysis
 Biaxial Tension/Compression Testing
 Pure Shear Testing
 Product Simulations with Specimen Data
MAR120 - Basic Nonlinear Analysis using Marc and Patran

MAR120 covers the use of Marc and Patran or AFEA (the interlocked combination of Patran and Marc) for the solution
of complex engineering problems. Students who successfully complete this course will be able to: create finite element
models representing nonlinear physical phenomena; select appropriate element types and mesh densities; understand the
limitations of solving nonlinear FEA problems; select solution types for various nonlinear phenomena such as nonlinear
dynamics, metal forming, elastomers, and contact problems; select error tolerance parameters and properly use automatic
time-stepping techniques; and understand the basis of large deformation, rotation, and strain finite element analysis. Patran
provides a Marc Preference which directly supports most Marc features and indirectly supports all Marc features. MSC
customers that have been using Advanced FEA (which is replaced by AFEA) for meeting their analysis needs will find this
new Marc Preference to be the ideal environment to continue their work. They are especially encouraged to attend this
course. All the class practice (16 exercises) is made using Patran and Marc rather than Marc and Mentat. Engineers who
have attended the MAR101 and MAR102 will also benefit from attending this class if they intend to use the Patran Marc
Preference.
Pre-requisites:
A basic knowledge of statics and strength of materials is highly recommended and previous finite element analysis
experience is recommended.
MAR121 - Advanced Nonlinear Analysis using Marc and Patran

The purpose of this course is to enhance the current Marc user's understanding of modeling nonlinear problems. Lectures
are supported by hands-on modeling of nonlinear problems.
Pre-requisites:
MAR120 - Basic Nonlinear Analysis using Marc and Patran
CHAPTER 27
Preface
Topics:
 Expand knowledge from MAR120 (Basic Nonlinear Analysis using Marc and Patran) course
 Practical aspects of rubber simulation
 Creep
 Superplastic forming
 Composite failure techniques
 Advanced contact techniques
 Adaptive meshing
 User subroutines
 Global/Local modeling
 Heat transfer and thermal stress
 Coupled Thermal/Structural analysis
 Restarts
 Performance
 Workshop Problems
Technical Support
If you encounter difficulties while using Marc, first please refer to the section(s) of the manual containing information on
the commands you are trying to use or the type of problem you are trying to solve.
Visit SimCompanion
The product documentation is available in SimCompanion (http://simcompanion.mscsoftware.com). The SimCompanion gives
you access to a wealth of resources for MSC Software products. You will find various information such as:
 Product documentations
 Knowledge base articles
 Product error lists (fixed and known issues for each release)
 SimAcademy webinars
 Product and support contact information
SimCompanion is a searchable database which allows you to find articles relevant to your inquiry. Valid MSC customer
entitlement and login is required to access the database and documents. It is a single sign-on that gives you access to product
documentation for complete list of products from MSC Software, allows you to manage your support cases, and participate
in our discussion forums.
Help Us Help You

Clients frequently call up the support engineers at MSC Software with enquiry regarding models that do not run correctly.
Our technical support staff can help you much more efficiently and effectively if you are working with a small model, since
debugging a small model is much easier, and the turnaround time to rerun a (hopefully) corrected test model is minutes
rather than hours.
 For information on the latest events, products and services for all products, refer to the MSC Software corporate
site (www.mscsoftware.com).
 For technical support phone numbers and contact information, please visit:
http://www.mscsoftware.com/Contents/Services/Technical-Support/Contact-Technical-Support.aspx.
Chapter 1: The Marc System
1 The Marc System
 Overview 30

Marc Programs 30
 Structure of Marc 32
 Features and Benefits of Marc 32
Overview
The Marc system contains a series of integrated programs that facilitate analysis of engineering problems in the fields of
structural mechanics, heat transfer, and electromagnetics. The Marc system consists of the following programs:
 Marc for Analysis
 Mentat or Patran for GUI (For a detailed description of the supported functionalities by Patran, refer to the Patran
Marc Preference Guide.)
These programs work together to:
 Generate geometric information that defines your structure (Marc and Mentat or Patran)
 Analyze your structure (Marc)
 Graphically depict the results (Marc and Mentat or Patran)
Figure 1-1 shows the interrelationships among these programs. Marc Programs discusses the Marc component programs.
MARC AFEA
Preprocessing Mentat or Patran
Marc
Analysis
Postprocessing Mentat or Patran
Figure 1-1 The Marc System
Marc Programs
The Marc Programs will be discussed in the following sections:
 Marc for Analysis
 Mentat or Patran for GUI
CHAPTER 1 31
The Marc System
Marc for Analysis

You can use Marc to perform linear or nonlinear stress analysis in the static and dynamic regimes, to perform heat transfer
analysis and electromagnetic analysis. The nonlinearities may be due to either material behavior, large deformation, or
boundary conditions. An accurate representation accounts for these nonlinearities.
Physical problems in one, two, or three dimensions can be modeled using a variety of elements. These elements include
trusses, beams, membranes, shells, and solids. Mesh generators, graphics, and postprocessing capabilities, which assist you
in the preparation of input and the interpretation of results, are all available in Marc. The equations governing mechanics
and implementation of these equations in the finite element method are discussed in Chapters 4, 10, 5, 6, and 11.
Mentat or Patran for GUI

Mentat is an interactive computer program that prepares and processes data for use with the finite element method.
Interactive computing can significantly reduce the human effort needed for analysis by the finite element method.
Graphical presentation of data further reduces this effort by providing an effective way to review the large quantity of data
typically associated with finite element analysis.
An important aspect of Mentat is that you can interact directly with the program. Mentat verifies keyboard input and
returns recommendations or warnings when it detects questionable input. Mentat checks the contents of input files and
generates warnings about its interpretation of the data if the program suspects that it may not be processing the data in the
manner in which you, the user, have assumed. Mentat allows you to graphically verify any changes the input generates.
Mentat can process both two- and three-dimensional meshes to do the following:
 Import CAD geometry
 Modify solid geometry
 Generate and display a mesh
 Generate and display boundary conditions and loadings
 Perform postprocessing to generate contour, deformed shape, time history plots, path plots and particle tracking
The data that is processed includes:
 Nodal coordinates
 Element connectivity
 Nodal boundary conditions
 Nodal coordinate systems
 Element material properties
 Element geometric properties
 Element loads
 Initial conditions
 Contact conditions
 Constraints, RBE2, RBE3
 Crack definition
 Tables
 Weld definition
 Adaptive mesh criteria
 Element and nodal sets
 Loadcase definition
 Job definition
Structure of Marc
Marc has four comprehensive libraries, making the program applicable to a wide range of uses. These libraries contain
structural procedures, materials, elements, and program functions. The contents of each library are described below.
Procedure Library
The structural procedure library contains procedures such as static, dynamic, creep, buckling, heat transfer, fluid
mechanics, and electromagnetic analysis.
The procedure library conveniently relates these various structural procedures to physical phenomena while guiding you
through modules that allow, for example, nonlinear dynamic and heat-transfer analyses.
Material Library
The material library includes many material models that represent most engineering materials. Examples are the inelastic
behavior of metals, soils, and rubber material. Many models exhibit nonlinear properties such as plasticity, viscoelasticity,
and hypoelasticity. Linear elasticity is also included. All properties may depend on temperature.
Element Library
The element library contains over 225 elements. This library lets you describe any geometry under any linear or nonlinear
loading conditions.
Program Function Library

The program functions such as selective assembly, user-supplied subroutines, and restart, are tailored for user-friendliness
and are designed to speed up and simplify analysis work. Marc allows you to combine any number of components from
each of the four libraries and, in doing so, puts at your disposal the tools to solve almost any structural mechanics problem.
Features and Benefits of Marc

Since the mid-1970s, Marc has been recognized as the premier general purpose program for nonlinear finite element
analysis. The program’s modularity leads to its broad applicability. All components of the structural procedure, material,
and element libraries are available for use, allowing virtually unlimited flexibility and adaptability.
Marc has helped analyze and influence final design decisions on:
CHAPTER 1 33
The Marc System
 Aerospace panels
 Automotive Metallic Parts (Transmissions, Differential, Doors)
 Automotive Rubber Parts (Seals, Tires, Belts, Constant Velocity Boots, Mounts)
 Bearings
 Biomedical Implants
 Biomedical Devices
 Bulk Forming (Forging, Extrusion)
 Civil Engineering Structures, Roofing systems
 Cutting
 Earthquakes
 Electronic Chips
 Engines (Pistons, Gaskets, Bolts)
 Heat Treatment
 Jet Engines (Blades, Vanes, Combustion Liners)
 Nuclear Reactors, Housing, Piping systems
 Offshore Piping, Risers, Seals
 Offshore Platforms
 Rail Systems
 Rocket Motor Engines
 Sheet Forming (Drawing, Stamping)
 Ship hulls
 Space Vehicles
 Transformers
 Welding
 Windmills
Marc’s clients gained the following benefits not attainable through other numerical or experimental techniques. These
benefits include:
 Accurate results for both linear and nonlinear analysis
 Better designs, which result in improved performance and reliability
 The ability to model complex structures and to incorporate geometric and material nonlinear behavior
 Documentation, technical support, consulting, and education provided by MSC Software Corporation
 Availability of Marc on most computers
 Efficient operation
Chapter 2: Program Initiation
2 Program Initiation
 Overview 35

Workspace Requirements 35
 File Units 37
 Program Initiation 40
 Examples of Running Marc Jobs 42
CHAPTER 2 35
Program Initiation
Overview
This chapter explains how to execute Marc on your computer. Marc runs on many types of machines. All Marc capabilities
are available on each type of machine; however, program execution can vary among machine types. The allocation of
computer memory depends on the hardware restrictions of the machine you are using. For details on the types of machines
and operating systems, refer to the Installation and Operations Guide.
Workspace Requirements
Computing the amount of workspace required by Marc is a complex function of many variables. The most efficient method
is to use the default values for the allocation. The program dynamically acquires memory if necessary and if available. In
some situations, it is advantageous to initially allocate an amount of memory as described below. The following sections
discuss workspace requirements for Marc.
Marc Workspace Requirements

The workspace used by Marc is allocated in separate parts. One part is referred to as general memory and contains items as
follows:
 Element stiffness matrices
 Assembled global stiffness
 Mass matrices and decomposed operator matrix (for certain matrix solvers)
The initial amount of memory for this part can be entered by the user and the program dynamically allocates more memory
if necessary. If no initial memory is specified, it is automatically allocated as needed. Other parts of the dynamically
allocated workspace can not be influenced by the user. This includes data for elements, vectors, tables, sets, contact bodies,
kinematic boundary conditions, transformations, tying, the so-called incremental backup, solver workspace for certain
solvers among other things. These are all allocated separately. The incremental backup is an extra copy of stress tensors and
similar quantities and is used for the Newton-Raphson iterations in a non-linear analysis. If the cut-back feature is activated,
more data is stored in this part to allow for redoing the increment with a modified time step if a failure occurs.
General Memory stored in common/space/
Basic Data Assembled Stiffness Matrix (2) Decomposed Stiffness Matrix (3)
Second Group of Memory stored in individual dynamically allocated vectors
Element Data (1) Vectors Sets Tables Contact Data Boundary Conditions Transformation Tying Attach
Incremental Backup (4) Decomposed Stiffness Matrix (5)
1. Element Data memory allocation is reduced if ELSTO is used. This also reduces Incremental Backup memory.
2. Assembled stiffness matrix memory allocation is reduced if out-of-core solver is used.
3. Decomposed Stiffness matrix is for solver type 0, 4, and if Cholesky preconditioner is used, solver type 2.
4. Incremental Backup memory allocation is reduced if IBOOC is used.
5. Decomposed Stiffness matrix is for solver type 6 and 8.

During the analysis, some of the blocks may grow because of changes in the model. In particular because of local or global
adaptive meshing, the Element Data, Vector Data, Assembled Stiffness Matrix, and Decomposed Stiffness Matrix may
grow in size because of the addition of elements and nodes. The Assembled Stiffness and Decomposed Stiffness matrices
also expand due to changes in the bandwidth of the system. Changes in the bandwidth occur when deformable-deformable
or self-contact occur. These changes can have a dramatic effect of the amount of memory required. The program
dynamically requests additional memory. If this memory is not available, it activates one or more of the out-of-core
options.
The amount of memory that can be used by an analysis is limited by the hardware employed in the operating system and
by internal restrictions within Marc. A failure to allocate memory can occur if other processes are using large amounts of
memory. A typical output when a memory request fails is:
memory request of 250000000 words failed
This occurs when Marc sends a memory allocation request (using the C function malloc) and the system refuses the
request.
The memory allocation for the general memory and matrix solver can be affected by the user. The ALLOCATE parameter
specifies the amount of memory that is initially allocated for this part. This amount is allocated regardless of whether it is
used or not. The default is set to a small value and the memory grows as needed.
For large problems, you may want to get an estimate of the workspace requirements for running a job without actually
executing the analysis. To do this, insert the STOP parameter to exit the program normally after the workspace is allocated.
Marc prints out a summary of the memory needed. It should be noted that the reported workspace with this option is only
an estimate. Some parts of the memory will grow during the analysis and there is no way to predict this without actually
running the analysis. For setting the appropriate initial allocation to avoid memory growth of the general memory (which
can be inefficient), one should look at the entry on general memory of the summary printout. Please note that the ALLOCATE
parameter value in a parallel run refer to the complete model; the specified amount is divided equally among the domains.
Chapter 12 Output Results describes the Marc output related to memory in more detail.
By default, the data is stored in-core. There are three out-of-core storage options in Marc as follows:
1. Out-of-core element data storage (the ELSTO parameter also implies IBOOC):
The out-of-core element data option stores element arrays (strains, stresses, temperatures, etc.) on a file (Fortran
unit 3). Data connected with storage of all element quantities occupy a large amount of space for the more complex
shell or three-dimensional elements. Putting this data out-of-core leads to a slowdown of the execution (disk access
is, in general, slower than memory access) but the effect of this is usually not severe. The ELSTO parameter is used
to set this option.
2. Out-of-core storage of incremental backup (the IBOOC parameter):
The out-of-core option for incremental backup stores this data in a file (Fortran unit 29). If the element data is out-
of-core, the incremental backup is automatically out-of-core as well. The slow-down related to incremental backup
out-of-core is much less than the one related to element data out-of-core.
3. Out-of-core matrix solution:
The out-of-core matrix solution has a more severe effect on the execution speed. Not all matrix solvers support this
option. It is not supported by the iterative sparse solver (Marc solver number 2). The multifrontal direct solver
(solver 8), the Pardiso solver (solver 11) and MUMPS (solver type 12) supports an out-of-core matrix solution.
CHAPTER 2 37
Program Initiation
In the case that a memory allocation request fails, Marc automatically switches to out-of-core storage unless it is too late to
do this at the point where the allocation failure occurs. If the memory allocation for element data or incremental backup
fails, this part is put out-of-core. The incremental backup part can also automatically be put out-of-core to free up memory
for the matrix solver. If the default solver is used (multifrontal, solver 8), there is a special way for checking if out-of-core
is needed. The solver workspace is allocated in two parts. One in general memory for the assembly and one in separate
memory. If the in-core assembly part takes more than 25% of the available memory, the matrix assembly is done out-of-
core. This ensures that this decision is done early enough so there is memory left for other parts of the job. The actual LDU
decomposition of the matrix is done out-of-core if the in-core decomposition requires more than the available memory to
avoid the usage of swap memory during the decomposition.
It is possible to limit the amount of memory that Marc will use. The analysis will stop if any portion of the code would
request more total memory than the maximum specified. The default for this limit is the physical amount of memory on
the machine. There is no attempt to figure out the amount of available memory. The memory limit can be set in the
installation by adding a variable called MEMLIMIT to the file tools/include (for Windows, tools\include.bat)
in the installation directory. This variable specifies the upper memory bound in Mbytes. It can also be specified with the
start-up script run_marc with the option -ml new_memlimit (not available through Mentat).
File Units
Marc uses auxiliary files for data storage in various ways. Particular Fortran unit numbers are used for certain program
functions (for example, ELSTO, RESTART). Table 2-1 lists these file unit numbers.
Notes: 1. On most systems, the files are referenced by file names as well as by the file.
2. Files with spaces in their names are only supported on Windows.
Several of these files are necessary for solving most problems. The program input file and program output file are always
required.
Table 2-1 FORTRAN File Units Used by Marc

File name Unit Description File Type
jidname.log 0 Analysis sequence log file sequential access, formatted
jidname.t01 1 Usually contains mesh data random access, formatted
jidname.t02 2 OOC* solver scratch file random access, binary
jidname.t03 3 Element data storage (see ELSTO parameter) random access, binary
jidname.dat 5 Data input file sequential access, formatted
jidname.out 6 Output file sequential access, formatted
jidname.t08 8 Restart file, written out sequential access, binary
ridname.t08 9 Restart file to be read in from a previous job sequential access, binary
jidname.t11 11 OOC* solver scratch file sequential access, binary
*OOC denotes Out-Of-Core solution.
Table 2-1 FORTRAN File Units Used by Marc (continued)

jidname.t14 14 OOC* solver scratch file random access, binary
jidname.t16 16 Post file, written out sequential access, binary
ridname.t16 17 Post file to be read in from a previous job sequential access, binary
jidname.t18 18 Mesh optimization correspondence table sequential access, formatted
jidname.fem 18 From Marc to external mesher sequential access, formatted
jidname.t19 19 Post file, written out sequential access, formatted
ridname.t19 20 Post file to be read in from a previous job sequential access, formatted
jidname_j_.dat 21 Temporary input file when cut-back is used. sequential access, formatted
jidname.t22 22 Subspace iteration scratch file random access, binary
jidname.t23 23 Fluid-solid interaction file sequential access, binary
pidname.t19 24 Temperature post file for CHANGE STATE or post file sequential access, formatted
from previous analysis for PRE STATE or GLOBALLOCAL
or MAP TEMP.
pidname.t16 25 Temperature post file for CHANGE STATE or post file sequential access, binary
from previous analysis for PRE STATE or GLOBALLOCAL
or MAP TEMP.
jidname.t29 29 Incremental backup file when ELSTO, IBOOC is used, or sequential access, binary
insufficient memory exists.
jidname.t31 31 Substructure results file for writing sequential access, binary
jidname.t33 33 Lanczos scratch file sequential access, binary
sidname.t31 35 Substructure results file for reading sequential access, binary
material.mat 38 Material data base file sequential access, formatted
jidname.g 39 Intergraph post file sequential access, formatted
jidname.unv 40 I-DEAS Universal post file sequential access, formatted
jidname.t41 41 Post file – Domain Decomposition sequential access, binary
ridname.t42 42 Post file – Domain Decomposition sequential access, formatted
jidname.opt 45 Duplicate loadcase data file during design optimization sequential access, formatted
run
jidname.t46 46 Design optimization scratch file sequential access, binary
jidname.trk 47 New particle tracking file sequential access, formatted
ridname.trk 48 Old particle tracking file sequential access, formatted
CHAPTER 2 39
Program Initiation

userspecified 49 User default file (see Marc Volume C: Program Input, sequential access, formatted
Appendix C: Default File)
jidname.vfs or 50 Viewfactors for cavity number xx sequential access, formatted
jidname_CXX.vfs
jidname.lck 51 Locking of post file sequential access, formatted
jidname.cnt 52 Dynamic control file sequential access, formatted
jidname.mfd 52 rebar - Mentat interface sequential access, formatted
jidname-bbc.mfd 52 beam-beam contact - Mentat interface sequential access, formatted
jidname.seq 53 Sequence option sequential access, formatted
jidname.rst 54 loadcase data sequential access, formatted
jidname.mesh 55 User supplied mesh sequential access, formatted
jidname.feb 55 From 3-D mesher to Marc sequential access, formatted
jidname.pass 56 Auto restart command line sequential access, formatted
jidname.rms 57 2-D outline file for remeshing sequential access, formatted
jidname.domesh 59 Lock files indicating meshing status sequential access, formatted
jidname.donemesh
“do mesh” and “done mesh”
jidname.doview 59 Lock file indicating viewfactor calculation status sequential access, formatted
jidname.doneview
jidname.sltrk 60 New streamline tracking file sequential access, formatted
ridname.sltrk 61 Old streamline tracking file sequential access, formatted
jidname.sts 67 Analysis progress reporting file sequential access, formatted
bbctch.noconv 80 beam-beam contact information sequential access, formatted
jidname.t81 81 Multifrontal OOC scratch file random access, binary
jidname.t82 82 Multifrontal OOC scratch file random access, binary
jidname.t83 83 Multifrontal OOC scratch file sequential access, binary
jidname.t84 84 Multifrontal OOC scratch file sequential access, binary
jidname.t85 85 Multifrontal DDM scratch file sequential access, binary
jidname.fld 91 Forming Limit input file sequential access, formatted

jidnamd.stm 91 Output at streamline integration points - PYROLYSIS sequential access, formatted
parameter.
jidname.rec 96 output file of recession surface - ABLATION parameter. sequential access, formatted
filename.apt 94 APT file - machining option sequential access, file
filename.ccl 95 CL file - machining option sequential access, file
EXITMSG 97 Exit messages sequential access, formatted
USRDEF 98 User global default file (see Marc Volume C: Program sequential access, formatted
Input, Appendix C: Default File)
jidname.t08 99 Base restart file for DDM sequential access, formatted
jidname.grd 103 Grid Force Balance Output sequential access, formatted
user specified 110 - 119 Include files for input sequential access, formatted
jidnam-dmig* 120 - 130 DMIG output files. sequential access, formatted
jidname.hmr N/A Hypermesh results file sequential access, binary C file
jidname.dump N/A Scratch file used during memory reallocation on sequential access, binary C file
Windows if in-core reallocation fails.
Program Initiation
Procedures (shell script) are set up that facilitate the execution of Marc on most computers. These procedures invoke
machine-dependent control or command statements. These statements control files associated with a job.
This shell script submits a job and automatically takes care of all file assignments. This command must be executed at the
directory where all input and output files concerning this Marc job are available. To use this shell script, every Marc job
should have a unique name qualifier and all Marc output files connected to that job uses this same qualifier. For restart,
post, change state, all default Marc Fortran units should be used.
To actually submit a Marc job, the following command should be used:
run_marc -prog prog_name -jid job_name -rid rid_name -pid pid_name \
-sid sid_name -queue queue_name -user user_name -back back_value \
-ver verify_value -save save_value -vf view_name -def def_name \
-nprocd number_of_processors (for Single Input file runs, use -nps
number_of_processors) -nts number_of_threads -nte number_of_element_threads \
-dir directory_where_job_is_processed -itree message_passing_type \
-host hostfile (for running over the network) -pq queue_priority \
-at date_time -comp compatible_machines_on_network -cpu time_limit \
-nsolver number of processors (MUMPS) -ml memory_limit -mpi intelmpi \
-gpuid GPU_card_number
where the \ provides for continuation of the command line.
CHAPTER 2 41
Program Initiation
Table 2-2 Keyword Descriptions*

Keyword Options Description
-jid (-j) job_name Input file (job) name identification.
-prog (-pr) progname Run saved executable progname.marc from a previous job
(see -user and -save).
-user (-u) user_name User subroutine user_name.f is used to generate a new executable
program called user_name.marc (see -save and -prog).
-save (-sa) no Do not save the new executable program user_name.marc.
yes Save the executable program user_name.marc for a future time (see -
prog and -user).
-rid -(r) restart_name Identification of previous job that created RESTART file.
-pid (-pi) post_name Identification of previous job that created the post file.
-sid (-si) substructure Identify the job that contains the solution to the external nodes of the
superelement.
-back (-b) yes Run Marc in the background.
no Run Marc in the foreground.
-ver (-v) yes Ask for confirmation of these input options before starting the job.
Start the job immediately.
no
-def (-de) default_file File name containing user defined default data.
-nprocd (-np) number Number of domains for parallel processing.
-nprocds (-nps) number Number of domains for parallel processing using a Single Input file.
-nts (-nthread) number Number of threads per parallel matrix solver (solver 8, 9, and 11).
-nte number Number of threads used for stiffness matrix assembly, mass matrix assembly
and stress recovery phase parallelization. For DDM, this is the total number
of threads across all domains.
-nsolver number Number of processes to use for the MUMPS parallel matrix solver (solver
12).
-dir directory_name Pathname to directory where the job I/O should take place.
Defaults to current directory.
-sdir directory_name Directory where scratch files are placed. Defaults to -dir.
-host (-ho) hostfile Specify the name of the host file for running over a network (default is
execution on one machine only in which case this option is not needed).
-ci yes Copy input files automatically to remote hosts for a network run,
if necessary.
no
*Default options are shown in bold.
Table 2-2 Keyword Descriptions* (continued)

Keyword Options Description
-cr yes Copy post files automatically from remote hosts used for a network run,
if necessary.
no
-vf viewfactor_name Name of file containing viewfactors for radiation from previous analysis or
from Mentat using either the Monte Carlo or Hemicube method.
-mpi intelmpi Select which version of mpi will be used for Domain Decomposition or the
msmpi MUMPS solver. If not specified, the appropriate default shown below will
be used:
Platform Default MPI Alternative MPI
Windows 64 intelmpi msmpi
Linux 64 intelmpi
-ml memlimit Memory limit for deciding if the solver should go out-of-core. Specified in
Mbyte. On Windows machines with minimal paging space, it is
recommended to set the memory limit to 80-90% of the physical memory.
Defaults to the physical amount of memory on the machine.
-gpuid <id:id:id> auto GPU card number or auto for automatic selection by program. For parallel
(DDM) jobs, multiple card IDs can be specified if available.
*Default options are shown in bold.
Examples of Running Marc Jobs

Example 1:
run_marc -jid e2x1
This runs the job e2x1 in the background using a single processor.
The input file is e2x1.dat in the current working directory.
Example 2:
run_marc -jid e2x14 -user u2x14 -sav y -nprocd 4
This runs the job e2x14 in the background with four processors. The user subroutine is linked with the Marc library and
a new execu module is created as u2x14.marc and saved in the current working directory after completion of the job.
Example 3:
run_marc -jid e2x14a -prog u2x14 -nprocd 4
Use the above saved module u2x14.marc to run the job e2x14a in the background with four processors.
CHAPTER 2 43
Program Initiation
Example 4:
run_marc -jid e3x2a -v no -b no -nprocd 2
Run the job e3x2a in the foreground with two processors. The job runs immediately without verifying any arguments
interactively. If there are any input errors in the arguments, the job does not run and the error message is sent to the screen.
Example 5:
run_marc -jid e3x2b -rid e3x2a
Run the job e3x2b in the background using a single processor. The job uses e3x2a.t08, which is created from Example
4, as restart file.
Example 6:
run_marc -jid e2x1 -nprocd 2
Runs a two processor job on a single parallel machine.
Example 7:
run_marc -jid e2x1 -nprocd 2 -host hostfile
Runs a two-processor job over a network. The hosts are specified in the file hostfile.
Example 8:
run_marc -jid e2x1 -nps 2
Runs a two-processor job on a single parallel machine using a single input file.
Example 9:
run_marc -jid e2x1 -nps 2 -host hostfile
Runs a two-processor job over a network using a single input file. The hosts are specified in the file hostfile.
Example 10:
run_marc -jid example -nts 4
Runs a job using four threads on a machine that supports parallel processing. The Pardiso solver should also be activated.
On a host machine with two GPGPU cards and four CPUs:
Example 11:
run_marc -jid example -gpuid 0
Runs a job with GPU ID number 0.
Example 12:
Runs a job with GPU ID number 1.
Example 13:
run_marc -jid example -gpuid 0:1 -nprocd 2
Runs a two processor job with GPU ID number 0 and 1.
Example 14:
run_marc -jid example -gpuid 0:0:1:1 -nprocd 4
Runs a four processor job with the first two domains on GPU ID 0 and last two domains on GPU ID 1.
Example 15:
This will be translated as follows:
Example 16:
run_marc -jid example -gpuid auto
Runs a job with program chosen GPU ID.
Chapter 3: Data Entry
3 Data Entry
 Overview 46

Input Conventions 46
 Input File Version 13 49
 Table Driven Input 52
 Parameters 57
 Model Definition Options 57
 History Definition Options 57

REZONE Option 58
Overview
The input data structure is made up of three logically distinct sections:
1. Parameters describe the problem type and size.
2. Model definition options give a detailed problem description.
3. History definition options define the load history.
Input data is organized in (optional) blocks. Key words identify the data for each optional block. This form of input enables
you to specify only the data for the optional blocks that you need to define your problem. The various blocks of input are
optional in the sense that many have built-in default values which can be used by Marc in the absence of any explicit input
from you.
Input Conventions
Marc performs all data conversion internally so that the system does not abort because of data errors made by you. The
program reads all input data options alphanumerically and converts them to integer, floating point, or keywords, as
necessary. Marc issues error messages and displays the illegal option image if it cannot interpret the option data field
according to the specifications given in the manual. When such errors occur, the program attempts to scan the remainder
of the data file and ends the run with an exit error message at the END OPTION option or at the end of the input file.
Two input format conventions can be used:
1. Fixed format.
2. Free format.
You can mix fixed and free format options within a file, but you can only enter one type of format on a single option.
The syntax rules for fixed fields are as follows:
 You must right-justify integers in their fields. (The right blanks are filled with zeroes).
 Give floating point numbers with or without an exponent. If you give an exponent, it must be preceded by the
character E or D and must be right-justified.
The syntax rules for free fields are as follows:
 Check that each option contains the same number of data items that it would contain under standard fixed-format
control. This syntax rule allows you to mix fixed-field and free-field options in the data file because the number of
options you need to input any data list are the same in both cases.
 Separate data items on a option with a comma. The comma can be surrounded by any number of blanks. Within
the data item itself, no embedded blanks can appear.
 Give floating point numbers with or without an exponent. If you use an exponent, it must be preceded by the
character E or D and must immediately follow the mantissa (no embedded blanks).
 Give keywords exactly as they are written in the manual. Embedded blanks do not count as separators here (for
example, BEAM SECT is one word only).
 If a option contains only one free-field data item, follow that item with a comma. For example, the number “1”
must be entered as 1, if it is the only data item on a option. If the comma is omitted, the entry is treated as fixed
format and may not be properly right-justified.
CHAPTER 3 47
Data Entry
 If the EXTENDED parameter is used, integer data is given using 10 fields as opposed to 5 fields. This allows very large
models to be included in Marc. Additionally, real numbers are entered using 20 or 30 fields as opposed to 10 or
15. This allows increased accuracy when reading in data.
 All data can be entered as uppercase or lowercase text.
Input of List of Items

Marc often requests that you enter a list of items in association with certain program functions. As an example, these items
can be a set of elements as in the ISOTROPIC option, or a set of nodes as in the POINT LOAD option. Eighteen types of items
can be requested:
1. Element numbers 2. Body numbers

3. Node numbers 4. Edge pairs
5. Degree of freedom numbers 6. Face pairs
7. Integration point numbers 8. Oriented curves
9. Layer numbers 10. Oriented surfaces
11. Increment numbers 12. CAD body vertex pairs
13. Points IDs 14. CAD body edge pairs
15. Curves IDs 16. CAD body face pairs
17. Surfaces IDs 18. CAD body numbers
This list can be entered using either the OLD format (compatible with the G, H, and J versions of Marc) or the NEW
format (the K version).
Using the OLD format, you can specify the list of items in three different forms. You can specify:
1. A range of items as:
m n p
which implies items m through n by p. If p is not specified, the program assumes it is 1. Note that the range can
either increase or decrease.
2. A list of items as:
-n a1 a2 a3 ... an
which implies that you should give n items, and they are a1 ,...an.
3. A set name as:
MYSET
which implies that all items previously specified in the set MYSET are used. Specify the items in a set using the DEFINE
model definition option.
Using the NEW format, you can express the list of items as a combination of one or more sublists. These sublists can be
specified in three different forms. The following operations can be performed between sublists:
AND
INTERSECT
EXCEPT
When you form a list, subsets are combined in binary operations (from left to right). The following lists are examples.
1. SUBLIST1 and SUBLIST2
This list implies all items in subsets SUBLIST1 and SUBLIST2. Duplicate items are eliminated and the resulting
list is sorted.
2. SUBLIST1 INTERSECT SUBLIST2
This list implies only those items occurring both in subsets SUBLIST1 and SUBLIST2; the resulting list is sorted.
3. SUBLIST1 EXCEPT SUBLIST2
This list implies all items in subset SUBLIST1 except those which occur in subset SUBLIST2; the resulting list is
sorted.
4. SUBLIST1 AND SUBLIST2 EXCEPT SUBLIST3 INTERSECT SUBLIST4
This list implies the items in subsets SUBLIST1 and SUBLIST2 minus those items that occur in subset SUBLIST3.
Then, if the remaining items also occur in subset SUBLIST4, they are included in the list.
Sublists can have several forms. You can specify:
1. A range of items as:
m TO n BY p
or
m THROUGH n BY p
which implies items m through n BY p. If “BY p” is not included, the program assumes “BY 1”. Note that the
range can either increase or decrease.
2. A string of items as:
a1 a2 a3 ... an
which implies that n items are to be included. If continuation options are necessary, then either a C or
CONTINUE should be the last item on the option.
3. A set name as:
MYSET
which implies that all items you previously specified to be in the set MYSET are used. You specify the items in a set
using the DEFINE option.
Note: INTERSECT or EXCEPT cannot be used when defining lists of degrees of freedom.
CHAPTER 3 49
Data Entry
In a list, edges, faces, oriented curves, and oriented surfaces are entered as pairs (i:j) where i is the user element ID and
j is the edge id or face id. The edge id/face id for the different element classes is given in Marc Volume C: Program Input,
Chapter 1.
An oriented curve or an oriented surface is used in conjunction with shell elements to indicate if the top surface or the
bottom surface of the shell is to be used.
There are two types of edge and face sets; those expressed in Marc or Mentat convention. The edge/face ID in the Marc
convention is one greater than the Mentat convention.
For example, to specify edge 1 on elements 1 to 20, one would use:
1:1 TO 20:1
Examples of Lists
This section presents some examples of lists and entry formats.
Use the DEFINE model definition option to associate a list of items with a set name with items. Three sets are defined below:
FLOOR, NWALL, and WWALL.
 DEFINE NODE SET FLOOR contains:
1 TO 15 (1,2,3,4,5,6,7,8,9,10,11,12,13,14,15)
 DEFINE NODE SET NWALL contains:
5 TO 15 BY 5 AND 20 TO 22 (5,10,15,20,21,22)
 DEFINE NODE SET WWALL contains:
11 TO 20 (11,12,13,14,15,16,17,18,19,20)
Some possible lists are:
 NWALL AND WWALL, which would contain nodes:
5 10 11 12 13 14 15 16 17 18 19 20 21 22
 NWALL INTERSECT WWALL, which would contain nodes:
15 20
 NWALL AND WWALL EXCEPT FLOOR, which would contain nodes:
16 17 19 20 21 22
Input File Version 13

Material properties are assigned to elements by specifying the list of elements in, for example, the ISOTROPIC option, as
explained in the preceding section. Input file version 13 provides an alternative way to assign a material to a group of
elements. If the input file version is set to 13 via the VERSION parameter option, then the material must be assigned to a
group of elements, by specifying the material number on the CONNECTIVITY option that defines the elements. The element
list in the ISOTROPIC option must be empty in this case (i.e. it must consist of a single blank line). Similarly, element
geometry (defined by the GEOMETRY, GAP DATA or PBUSH options), material orientation (defined by the ORIENTATION option),
and contact body (defined by the CONTACT or THERMAL CONTACT options), must be assigned in the same way by specifying
the element geometry number, material orientation number and contact body number on the CONNECTIVITY option. The
latter can be repeated as often as necessary for each unique combination of element geometry, material, material orientation
and contact body in the model.
Figure 3-1 Solid, Sheet and Wire CAD Bodies
In addition, input file version 13 supports CAD bodies. CAD bodies are geometrical entities that can be used in two
different ways in the finite element analysis:
1. CAD bodies can be used to apply physical properties (element geometry, material, material orientation, contact
body), local coordinate systems, initial conditions and boundary conditions to the model, independent of the finite
element mesh;
2. CAD bodies can be associated with rigid contact bodies in a contact analysis for visualization purposes while post-
processing in Mentat.
Three types of CAD bodies are recognized (see Figure 3-1):
a. Solid bodies (volumes):
A solid body is a 3-D geometrical volume. The boundary is formed by a number geometrical surfaces, called
faces, that enclose the volume. Each face has a closed loop of geometrical curves, called edges, that define the
boundary of the face. The end points of the edges are called vertices. A solid body represents a 3D volumetric
body. If used to apply physical properties, etc., to the model, the body must be meshed with volume elements,
that is, the body must have a finite element mesh consisting of volume elements attached to it. In a 3D contact
analysis, the body can be associated with a rigid contact body for visualization purposes. In this case, the body
must not be meshed.
CHAPTER 3 51
Data Entry
b. Sheet bodies (surfaces):

A sheet body is a geometrical surface that consist of one or more faces. The faces usually do not enclose a
volume. A sheet body can be used in both 2-D and in 3-D analyses to apply physical properties, etc. to the
model. In a 2-D planar or axisymmetric analysis, the sheet body represents a 2-D region and must lie in the
global xy-plane or zr-plane. The body must be meshed with planar or axisymmetric elements. In a 3-D analysis,
the sheet body represents a shell structure in 3-D space and must be meshed with shell or membrane elements.
In a 3-D contact analysis, the body can be associated with a rigid contact body for visualization purposes. In
that case, the body represents the outer surface of the rigid contact body and must not be meshed.
c. Wire bodies (curves):
A wire body is a geometrical curve that consists of one or more edges and has no faces. Like the sheet body, a
wire body can also be used in both 2-D and in 3-D analyses to apply physical properties, etc. to the model. In
a 2-D planar analysis, the body represents a 2-D beam body and must lie in the global xy-plane. It must be
meshed with 2-D beam or truss elements. In an axisymmetric analysis, the body represents an axisymmetric
shell structure. The body must lie in the global zr-plane and must be meshed with axisymmetric shell elements.
In a 3-D analysis, the body represents a 3-D beam body and must be meshed with beam or truss elements. In
a 2-D contact analysis, the body can be associated with a rigid contact body for visualization purposes. In that
case, the body represents the outer surface of the rigid contact body and must not be meshed.
A CAD body is defined by the CAD BODIES option, in which the number of vertices, edges and faces of the body must be
specified. In addition, the name of a Parasolid file that contains the actual geometry of the body can be supplied. This file
(and the actual geometry) is not used in the analysis, but is primarily intended for visualization of the body during post-
processing in Mentat. Element geometry, material, material orientation and contact body are assigned to the CAD body
via the CAD BODIES option.
Elements are associated with a CAD body by specifying the CAD body number on the CONNECTIVITY option that defines
the elements. These elements inherit the element geometry, material, material orientation and contact body from the CAD
body. Nodes can be associated with vertices of a CAD body via the ATTACH NODE option, edges of finite elements can be
associated with edges of a CAD body via the ATTACH EDGE option, and faces of finite elements can be associated with faces
of a CAD body via the ATTACH FACE option.
If Table Driven Input is used (see the next Section), then initial conditions, such as the initial temperature, can be applied
to CAD bodies. The initial condition is automatically applied to the elements, or the nodes of these elements, associated
with the CAD body. Similarly, boundary conditions, such as fixed displacements, pressure loads or volume loads can be
applied to the vertices, the edges or the faces of the body or (in case of a volume load) to the body itself and are automatically
applied to the nodes, element edges, element faces or elements associated with these CAD body entities.
Local coordinate systems can be assigned to the vertices, edges or faces of a CAD body or to the CAD body itself using the
TRANSFORMATION, COORD SYSTEM and UTRANFORM options. If a local coordinate system is assigned to a CAD body entity, then
kinematic boundary conditions and concentrated nodal loads on this entity must be input in the local system. The
coordinate systems are automatically assigned to the nodes associated with these entities. For example, a local coordinate
system assigned to a face of a CAD body is automatically assigned to the nodes of the element faces attached to that face.
Similarly, a local coordinate system assigned to a CAD body is automatically assigned to the nodes of the elements
associated with that CAD body.
Table Driven Input

An alternative approach to defining the data which is based upon defining nonlinear material properties and variations in
the boundary conditions using tables. This input procedure is activated by use of the TABLE parameter. There are six aspects
to this input procedure:
1. The data associated with material properties, boundary conditions, and contact may reference a function of up to
four variables defined in the TABLE model definition option.
2. Boundary conditions and initial conditions are input in a consistent manner and are associated with a name. They
are not applied unless they have been activated by the LOADCASE model and history definition option.
3. Based upon the use of tables and loadcases, all boundary conditions may be defined in the model definition section.
4. Mechanical boundary conditions are, in general, entered as total values at the end of the loadcase.
5. Boundary conditions can be applied to either finite element entities (such as nodes, elements, edges or faces) or they
may be applied to geometric entities (such as points, curves, surfaces, CAD body vertices, CAD body edges, CAD
body faces, or CAD bodies). The boundary conditions are applied to the finite element based upon the current
attach information. This is updated for local adaptive meshing (2-D or 3-D) or global adaptive meshing in 2-D.
6. Each history definition section provides information on the type of analysis to be performed, the duration, and the
active boundary conditions. The LOADCASE history definition option is used to perform the latter.
Table Input
Many physical quantities cannot be represented by constants. This includes boundary conditions that are dependent upon
spatial location and/or time, material properties dependent upon temperature or material damage, contact parameters such
as the coefficient of friction and many more. In previous versions, this data could be input through a variety of options
including TEMPERATURE EFFECTS, WORK HARD, STRAIN RATE, by repetitive input of boundary conditions, or, in the most
general case, by user subroutines.
1 time 32 current 63 3rd state variable

current radius  2
 -----------------------------------
2 normalized time 33  radius of throat
(see THROAT) 64 4th state variable
3 increment number 34  p (pyrolysis damage). 65 5th state variable
4 normalized increment 35  w (water vapor fraction). 66 loadcase number
5 x coordinate 36  c (coking damage). 67 degree of cure
6 y coordinate 37 gasket closure distance 68 magnetic field intensity
7 z coordinate 38 displacement magnitude or, when 69 equivalent mechanical strain
used for springs/foundations,
displacement in the direction of the
spring/foundation
8 2 2 2 39 stress rate 70 1st strain invariant
s = x +y +z
9  angle 40 experimental data 71 2nd strain invariant
10 mode number 41 porosity 72 3rd strain invariant
11 frequency 42 void ratio 73 any strain component
CHAPTER 3 53
Data Entry
·c
12 temperature 43  (equivalent creep strain rate) 74 damage
13 function 44 minor principal total strain 75 accumulated crack growth
14 fourier 45 distance from neutral axis 76 relative sliding velocity
 – t  2 +t  2 
p
15  (equivalent plastic strain) 46 normalized distance from neutral axis 77 damping ratio
 – 1 +1 
·
16  (equivalent strain rate) 47 local x-coordinate of layer point for 78 log frequency (base 10)
open or closed section beam
17 ·  m (normalized
Bg = m 48 local y-coordinate of layer point for 79 cooling rate
mass flow rate). open or closed section beam
18 arc length 49 1st isoparametric coordinate 83 equivalent plastic strain at
(not available in this release) network 1 (PRF) model
19 (relative)* density 50 2nd isoparametric coordinate 84 equivalent plastic strain at
(not available for shells) (not available in this release) network 2 (PRF) model
20  (equivalent stress) 51 wavelength (used in spectral 85 equivalent plastic strain at
radiation) network 3 (PRF) model
21 magnetic induction creep strain  er 86 equivalent plastic strain at
52 network 4 (PRF) model
22 velocity 53 pressure or primary quantity in 87 equivalent plastic strain at
diffusion network 5 (PRF) model
23 particle diameter 54 equivalent strain rate for 88 equivalent plastic strain at
nonNewtonian viscosity network 6 (PRF) model
24 x 0 initial coordinate 55 normalized arc distance 89 equivalent plastic strain at
network 7 (PRF) model
25 y 0 initial coordinate 56 distance to other contact surface (near 90 equivalent plastic strain at
contact only) network 8 (PRF) model
26 z 0 initial coordinate 57 term of series 91 vibration amplitude
27 2 2 2 58 hydrostatic stress 92 log vibration amplitude
s0 = x0 + y0 + z0
(base 10)
28 contact force F 59 hydrostatic strain 93 octahedral shear strain
(deviatoric or coupled
response) or volumetric
deformation
contact body M ·
B g p = m
29 60 g p  m . 94 strain energy function value
30  n (normal stress) ·
B g w = m g w  m . 95 user-defined invariant
61 quantity (UINVAR)
31 voltage 62 2nd state variable -1 to parametric variable 1 to 100
-100
* Independent variable 19 is the user defined density; it is updated for Pyrolysis. In the case of Shima powder model, it
is the relative density.
Independent variable types (5, 6, 7) will have the following characteristics.

1. When used to evaluate material behavior, they will be original integration point coordinates unless the updated
Lagrange procedure is used.
2. When used to evaluate DIST LOADS, they will be the original integration point coordinates unless the FOLLOW FOR
parameter is used.
3. When used to evaluate boundary conditions FIXED DISP, etc. or POINT LOAD, etc., they will be the original nodal
coordinates unless the FOLLOW FOR parameter is used.
A mathematical formula may be either 80 characters or 160 characters long if extended input format is used. The formula
is defined in terms of independent variables v 1 , v 2 , v 3 , and/or v 4 , where the meaning of those variables is based on the
variable type defined in the 3rd data block.
The evaluation is based upon usual mathematical standards moving from left to right with the conventional rules of the
use of parentheses. The following mathematical symbols/operations are available.
+ addition ^ exponential
- subtraction ! factor
* multiplication % mod
/ division
CHAPTER 3 55
Data Entry
Function
F1
v1 x1 x2 x3 x4 v1
Allowing Extrapolation No Extrapolation

Data Entered
x 1 f 1
x 2 f 2
x 3 f 3
x 4 f 4
Given x 0
with extrapolation
f1 – f2
f 0 = f 1 +  ------------------   x 0 – x 1 
 x 1 – x 2
without extrapolation
f0 = f1
In addition to v 1 , v 2 , v 3 , and v 4 , the following constants may be used in the equation:
pi 
e exponent
tz offset temperature entered via the PARAMETERS model definition option
q Activation energy entered via MATERIAL DATA model definition option
r Universal gas constant entered via the PARAMETERS model definition option
sb Stefan Boltzman constant entered via the PARAMETERS model definition option
The following mathematical functions may be used in an equation:
cos cosine (x) x in radians

sin sine (x) x in radians
tan tangent (x) x in radians
dcos cosine (x) x in degrees
dsin sine (x) x in degrees

dtan tangent (x) x in degrees
acos inverse cosine (x) f in radians
asin inverse sine (x) f in radians
atan inverse tangent (x) f in radians
atan2 inverse tangent (x,y) f in radians
dacos inverse cosine (x) f in degrees
dasin inverse sine (x) f in degrees
datan inverse tangent (x) f in degrees
datan2 inverse tangent (x,y) f in degrees
log log based 10
ln natural log
exp exponent
cosh hyperbolic cosine
sinh hyperbolic sine
tanh hyperbolic tangent
acosh inverse hyperbolic cosine
asinh inverse hyperbolic sine
atanh inverse hyperbolic tangent
sqrt square root
rad convert degrees to radians
deg convert radians to degrees
abs obtain absolute value
int truncates the value to whole
frac take the fractional value
max takes the maximal value
min takes the minimal value
mod return the remainder of x, based on y
mod(x,y) = x - y * int (x/y)
The equation is evaluated based upon the current value of the independent variable. It is your responsibility to make sure
that the equation may be evaluated for the potential values of the independent values. For example, if the function is 1  v 1 ,
then if v 1 = 0.0 , an error will occur.
CHAPTER 3 57
Data Entry
Parameters
This group of parameters allocates the necessary working space for the problem and sets up initial switches to control the
flow of the program through the desired analysis, This set of input must be terminated with an END parameter. The input
format for these parameters is described in Marc Volume C: Program Input.
Model Definition Options

This set of data options enters the initial loading, geometry, and material data of the model and provides nodal point data,
such as boundary conditions. Model definition options are also used to govern the error control and restart capability.
Model definition options can also specify print-out and postprocessing options. The data you enter on model definition
options provides the program with the necessary information for determining an initial elastic solution (zero increment
solution). When boundary conditions reference time dependent tables, the transient nonlinear behavior can be defined.
The transient period is defined in the history definition section. This group of options must be terminated with the END
OPTION option. The input format for these options is described in Marc Volume C: Program Input.
History Definition Options

This group of options provides the load incrementation and controls the program after the initial elastic analysis. History
definition options also include blocks which allow changes in the initial model specifications. Each set of load sets must be
terminated with a CONTINUE option. This option requests that the program perform another increment or series of
increments if you request the auto-incrementation features. The input format for these options is described in Marc Volume
C: Program Input.
A typical input file setup for the Marc program is as follows:
 Marc Parameter
Terminated by an END parameter
 Marc Model Definition Options
(Zero Increment)
Terminated by an END OPTION option
 Marc History Definition
Data for the First Increment
Terminated by a CONTINUE option
 (Additional History Definition
Option for the second, third, ..., Increments)
Figure 3-2 is a dimensional representation of the Marc input data file.
Proportional
Increment
Load Auto Load
Incrementation Etc.
Linear and Nonlinear Analysis
Connectivity
Requiring Incrementation
Coordinates
Model Fixed Displacements
Definition Etc.
Linear Analysis
Title
Parameter Sizing
Etc.
Figure 3-2 The Marc Input Data File
REZONE Option
When the REZONE option is inserted into the input file and the manual procedure is used, the program reads additional data
options to control the rezoning steps. These options must immediately follow the END OPTION option or a CONTINUE history
definition option.
You can select as many rezoning steps in one increment as you need. Every rezoning step is defined by the data, starting
with the REZONE option and ending with the CONTINUE option. The END REZONE option terminates the complete set of
rezoning steps that form a complete rezoning increment. Follow the rezoning input with normal history definition data,
or again by rezoning data. The input format for these options is described in Marc Volume C: Program Input.
Chapter 4: Structural Procedure Library
4 Structural Procedure
Library
 Overview 60

Linear Analysis 60

Nonlinear Analysis 65
 Fracture Mechanics 107
 Dynamics 135

Inertia Relief 158

Rigid-Plastic Flow 161
 Superplasticity 163

Soil Analysis 165
 Mechanical Wear 170
 Design Sensitivity Analysis 172

Design Optimization 175

Define Initial State with Results from a Previous Analysis 181
 Structural Zooming Analysis 188

Cure-Thermal-Mechanically Coupled Analysis 191
 Fatigue Analysis 196

References 197
Overview
This chapter describes the analysis procedures in Marc applicable to structural problems. These procedures range from
simple linear elastic analysis to complex nonlinear analysis. A large number of options are available, but you need to
consider only those capabilities that are applicable to your physical problem. This chapter provides technical background
information as well as usage information about these capabilities.
Linear Analysis
Linear analysis is the type of stress analysis performed on linear elastic structures. Because linear analysis is simple and
inexpensive to perform and generally gives satisfactory results, it is the most commonly used structural analysis.
Nonlinearities due to material, geometry, or boundary conditions are not included in this type of analysis. The behavior of
an isotropic, linear, elastic material can be defined by two material constants: Young’s modulus E , and Poisson’s ratio v .
Marc allows you to perform linear elastic analysis using any element type in the program. Various kinematic constraints
and loadings can be prescribed to the structure being analyzed; the problem can include both isotropic and anisotropic
elastic materials.
The principle of superposition holds under conditions of linearity. Therefore, several individual solutions can be
superimposed (summed) to obtain a total solution to a problem.
Linear analysis does not require storing as many quantities as does nonlinear analysis; therefore, it uses the core memory
more sparingly. The ELASTIC parameter uses the assembled and decomposed stiffness matrices to arrive at repeated solutions
for different loads.
Note: Linear analysis is always the default analysis type in the Marc program.
Linear analysis in Marc requires only the basic input. Table 4-1 shows a subset of the Marc options and parameters which
are often used for linear analysis.
CHAPTER 4 61
Structural Procedure Library
Table 4-1 Basic Input

Type Name
Parameter TITLE
SIZING
ELEMENTS
ELASTIC
ALL POINTS
CENTROID
ADAPTIVE
FOURIER
END
Model Definition CONNECTIVITY
COORDINATES
GEOMETRY
ISOTROPIC
FIXED DISP
DIST LOADS
POINT LOAD
CASE COMBIN
END OPTION
More complex linear analyses require additional data blocks.

1. The ELASTIC parameter allows solutions for the same structural system with different loadings (multiple loading
analysis). When using the ELASTIC parameter, you must apply total loads, rather than incremental quantities (for
example., total force, total moment, total temperature) in subsequent increments.
2. The RESTART option, used with the ELASTIC parameter and/or CASE COMBIN option, stores individual loadcases in a
restart file.
3. The CASE COMBIN model definition option combines the results obtained from different loading cases previously
stored in a restart file.
4. The ADAPTIVE option can be used to improve the accuracy of the analysis.
5. The LORENZI option allows the study of problems of fracture mechanics.
6. The FOURIER option allows the analysis of axisymmetric structures subjected to arbitrary loadings.
7. The ORTHOTROPIC or ANISOTROPIC model definition option activates the anisotropic behavior option. In addition, the
ANELAS, HOOKLW, ANEXP, and ORIENT user subroutines define the mechanical and thermal anisotropy and the
preferred orientations.
8. You can use both the linear SPRINGS and FOUNDATION options in a linear stress analysis.
Accuracy
It is difficult to predict the accuracy of linear elastic analysis without employing special error estimation techniques. An
inaccurate solution usually exhibits itself through one or more of the following phenomena:
 Strong discontinuities in stresses between elements
 Strong variation in stresses within an element
 Stresses that oscillate from element to element
Error Estimates
The ERROR ESTIMATE option can also be used to obtain an indication of the quality of the results. You can have the program
evaluate the geometric quality of the mesh by reporting the aspect ratios and skewness of the elements. In a large
deformation updated Lagrange analysis, you can also observe how these change during the analysis, which indicates mesh
distortion. When the mesh distortion is large, it is a good idea to do a rezoning step.
The ERROR ESTIMATE option can also be used to evaluate the stress discontinuity between elements. Marc first calculates a
nodal stress based upon the extrapolated integration point values. These nodal values are compared between adjacent
elements and reported. Large discrepancies indicate an inability of the mesh to capture high stress gradients, in which case
you should refine the mesh and rerun the analysis or use local adaptive meshing.
The ERROR ESTIMATE option can be used for either linear or nonlinear analysis.
Adaptive Meshing
The ADAPTIVE option can be used to insure that a certain level of accuracy is achieved. The elastic analysis is repeated with
a new mesh until the level of accuracy requested is achieved.
Fourier Analysis
Through Fourier expansion, Marc analyzes axisymmetric structures that are subjected to arbitrary loading. The FOURIER
option is available only for linear analysis.
During Fourier analysis, a three-dimensional analysis decouples into a series of independent two-dimensional analyses,
where the circumferential distribution of displacements and forces are expressed in terms of the Fourier series. Both
mechanical and thermal loads can vary arbitrarily in the circumferential direction. You can determine the structure’s total
response from the sum of the Fourier components.
The Fourier formulation is restricted to axisymmetric structures with linear elastic material behavior and small strains and
displacements. Therefore, conditions of linearity are essential and material properties must remain constant in the
circumferential direction.
To use Fourier expansion analysis in Marc, the input must include the following information:
 The FOURIER parameter allocates storage for the series expansion.
CHAPTER 4 63
 Fourier model definition blocks for as many series as are needed to describe tractions, thermal loading, and
boundary conditions. Number the series sequentially in the order they occur during the FOURIER model definition
input. Three ways to describe the series are listed below:
• Specify coefficients a 0 a 1 b 1  on the Fourier model definition blocks.
• Describe F    (where  is the angle in degrees about the circumference) in point-wise fashion with an
arbitrary number of pairs   F     given on the blocks. Marc forms the corresponding series coefficients.
• Generate an arbitrary number of   F     pairs using the UFOUR user subroutine and let the program calculate
the series coefficients.
You can obtain the total solution at any position around the circumference by superposing the components already
calculated after completion of all increments required by the analysis. The CASE COMBIN option calculates this total solution
by summing the individual harmonics which are stored in the restart file.
The number of steps or increments needed for analysis depends on the number of harmonics that are chosen. For a full
analysis with symmetric and antisymmetric loadcases, the total number of increments equals twice the number of
harmonics. Table 4-2 shows which Fourier coefficients are used for a given increment.
Table 4-2 Fourier Coefficients – Increment Number

LOAD TERMS
INC. 1st DOF,Z 2nd DOF,R 3rd DOF,
0 a0 a0 0
1 0 0 a0
2 a1 a1 b1
3 b1 b1 a1
. . . .
. . . .
. . . .
2n an an bn
2n + 1 bn bn an
The magnitude of concentrated forces should correspond to the value of the ring load integrated around the circumference.
Therefore, if the Fourier coefficients for a varying ring load p    are found from the   p     distribution, where p   
has the units of force per unit length, the force magnitude given in the point load block should equal the circumference of
the loaded ring. If p    is in units of force per radian, the point load magnitude should be 2 .
The Fourier series can be found for varying pressure loading from   p     input with p expressed in force per unit area.
Marc calculates the equivalent nodal forces and integrates them around the circumference. The distributed load magnitude
in the distributed loads block should be 1.0.
Table 4-3 shows the elements in the program that can be used for Fourier analysis.
Table 4-3 Elements Used for Fourier Analysis

Element Type Description
62 8-node
73 8-node with reduced integration
63 8-node for incompressible behavior
74 8-node for incompressible behavior with reduced integration
90 3-node shell
Technical Background
The general form of the Fourier series expansion of the function F    is shown in the equation below:

F    = a0 +   a n cos n + b n sin n  (4-1)
n = 1
This expression expands the displacement function in terms of sine and cosine terms. A symmetric and an antisymmetric
problem are formulated for each value of n .
The displacements for the symmetric case, expressed in terms of their nodal values, are:
u n =  N 1 N 2   cos n  u n  e
v n =  N 1 N 2   cos n  v n  e (4-2)
wn =  N 1 N 2   sin n   w n  e
Nodal forces are:

n
Z = Z0 +  Z n cos n
1
n
(4-3)
R = R0 +  R n cos n
1
n
T = T0 +  T n sin n
1
The value n = 0 is a special case in Fourier analysis. If only the symmetric expansion terms are used, the formulation
defaults to the fully axisymmetric two-dimensional analysis. The antisymmetric case for n = 0 yields a solution for the
variable  that corresponds to loading in the tangential direction. Analyze axisymmetric solids under pure torsion in this
way.
CHAPTER 4 65
Modal Shapes and Buckling Load Estimations During a Fourier Analysis

During a Fourier analysis, Marc can be asked to estimate both the modal shapes and buckling loads for each harmonic in
the analysis. In either case, the program performs a Fourier analysis first and then estimates the modal shapes/buckling load
at prescribed harmonic numbers. In addition to the input data required for a Fourier analysis (FOURIER parameter and
FOURIER model definition option), the following must also be added: DYNAMIC parameter and MODAL INCREMENT model
definition option; BUCKLE parameter and BUCKLE INCREMENT model definition option, for Fourier modal shape and Fourier
buckling load estimations, respectively.
In the Fourier modal analysis, the mass matrix in the eigenvalue equation is a constant matrix. The stiffness matrix in the
eigenvalue equation is the one associated with a prescribed harmonic of the Fourier analysis. The expression of the
eigenvalue equation is:
 K m  –  2  M 0  = 0 (4-4)
m 0
where  K  is the stiffness matrix associated with the mth harmonic of the Fourier analysis and  M  is a constant matrix.
Multiple modes for each harmonic can be extracted.
Similarly, in a Fourier buckling analysis, the stiffness matrices in the eigenvalue equation are (respectively); the linear elastic
stiffness matrix and the geometric stiffness matrix associated with the prescribed harmonic of the Fourier analysis.
The eigenvalue equation is expressed as:
m m
 K  –   K g  = 0 (4-5)
m m
where  K  is the linear elastic stiffness matrix and  K g  is the geometric stiffness matrix, associated with the mth
harmonic of the Fourier analysis. The stresses used in the calculation of the geometric stiffness matrix are those associated
with the symmetric loadcase, m = 0 . Multiple buckling load estimations for each harmonic are also available.
Nonlinear Analysis
The finite element method can be used for nonlinear, as well as linear, problems. Early development of nonlinear finite
element technology was mostly influenced by the nuclear and aerospace industries. In the nuclear industry, nonlinearities
are mainly due to the nonlinear, high-temperature behavior of materials. Nonlinearities in the aerospace industry are
mainly geometric in nature and range from simple linear buckling to complicated post-bifurcation behavior. Nonlinear
finite element techniques have become popular in metal forming manufacturing processes, fluid-solid interaction, and fluid
flow. In recent years, the areas of biomechanics and electromagnetics have seen an increasing use of finite elements.
A problem is nonlinear if the force-displacement relationship depends on the current state (that is, current displacement,
force, and stress-strain relations). Let u be a generalized displacement vector, P a generalized force vector, and K the
stiffness matrix. The expression of the force-displacement relation for a nonlinear problem is:
P = K  P u u (4-6)
Linear problems form a subset of nonlinear problems. For example, in classical linear elastostatics, this relation can be
written in the form:
P = Ku (4-7)
where the stiffness matrix K is independent of both u and P . If the matrix K depends on other state variables that do
not depend on displacement or loads (such as temperature, radiation, moisture content, etc.), the problem is still linear.
Similarly, if the mass matrix is a constant matrix, the following undamped dynamic problem is also linear:
P = Mu·· + Ku (4-8)
There are three sources of nonlinearity:
1. Material
2. Geometric
3. Nonlinear boundary conditions
Material:
Material nonlinearity results from the nonlinear relationship between stresses and strains. Considerable progress has been
made in attempts to derive the continuum or macroscopic behavior of materials from microscopic backgrounds, but, up
to now, commonly accepted constitutive laws are phenomenological. Difficulty in obtaining experimental data is usually a
stumbling block in mathematical modeling of material behavior. A plethora of models exist for more commonly available
materials like elastomers and metals. Other material model of considerable practical importance are: composites,
viscoplastics, creep, soils, concrete, powder, and foams. Figure 4-1 shows the elastoplastic, elasto-viscoplasticity, and creep.
Although the situation of strain hardening is more commonly encountered, strain softening and localization has gained
considerable importance in recent times.
 
 
Elasto-Plastic Behavior Elasto-Viscoplastic Behavior
c
t
Creep Behavior
Figure 4-1 Material Nonlinearity
Geometric:
Geometric nonlinearity results from the nonlinear relationship between strains and displacements on the one hand and the
nonlinear relation between stresses and forces on the other hand. If the stress measure is conjugate to the strain measure,
CHAPTER 4 67
both sources of nonlinearity have the same form. This type of nonlinearity is mathematically well defined, but often
difficult to treat numerically. Two important types of geometric nonlinearity occur:
a. The analysis of buckling and snap-through problems (see Figure 4-2 and Figure 4-3).
P Linear
P S
u Pc Neutral
Uns
u
Figure 4-2 Buckling
Figure 4-3 Snap-Through
b. Large strain problems such as manufacturing, crash, and impact problems. In such problems, due to large strain
kinematics, the mathematical separation into geometric and material nonlinearity is nonunique.
Boundary Conditions:
Boundary conditions and/or loads can also cause nonlinearity. Contact and friction problems lead to nonlinear boundary
conditions. This type of nonlinearity manifests itself in several real life situations; for example, metal forming, gears,
interference of mechanical components, pneumatic tire contact, and crash (see Figure 4-4). Loads on a structure cause
nonlinearity if they vary with the displacements of the structure. These loads can be conservative, as in the case of a
centrifugal force field (see Figure 4-5); they can also be nonconservative, as in the case of a follower force on a cantilever
beam (see Figure 4-6). Also, such a follower force can be locally nonconservative, but represent a conservative loading system
when integrated over the structure. A pressurized cylinder (see Figure 4-7) is an example of this.
Figure 4-4 Contact and Friction Problem
Figure 4-5 Centrifugal Load Problem (Conservative)
Figure 4-6 Follower Force Problem (Nonconservative)

CHAPTER 4 69
Figure 4-7 Pressurized Cylinder (Globally Conservative)
The three types of nonlinearities are described in detail in the following sections.
Geometric Nonlinearities
Geometric nonlinearity leads to two types of phenomena: change in structural behavior and loss of structural stability.
There are two natural classes of large deformation problems: the large displacement, small strain problem and the large
displacement, large strain problem. For the large displacement, small strain problem, changes in the stress-strain law can
be neglected, but the contributions from the nonlinear terms in the strain displacement relations cannot be neglected. For
the large displacement, large strain problem, the constitutive relation must be defined in the correct frame of reference and
is transformed from this frame of reference to the one in which the equilibrium equations are written.
The collapse load of a structure can be predicted by performing an eigenvalue analysis. If performed after the linear solution
(increment zero), the Euler buckling estimate is obtained. An eigenvalue problem can be formulated after each increment
of load; this procedure can be considered a nonlinear buckling analysis even though a linearized eigenvalue analysis is used
at each stage.
The kinematics of deformation can be described by the following approaches:
A. Lagrangian Formulation
B. Eulerian Formulation
C. Arbitrary Eulerian-Lagrangian (AEL) Formulation
The choice of one over another can be dictated by the convenience of modeling physics of the problem, rezoning
requirements, and integration of constitutive equations.
Lagrangian Formulation
In the Lagrangian method, the finite element mesh is attached to the material and moves through space along with the
material. In this case, there is no difficulty in establishing stress or strain histories at a particular material point and the
treatment of free surfaces is natural and straightforward.
The Lagrangian approach also naturally describes the deformation of structural elements; that is, shells and beams, and
transient problems, such as the indentation problem shown in Figure 4-8.
sz
u
Figure 4-8 Indentation Problem with Pressure Distribution on Tool
This method can also analyze steady-state processes such as extrusion and rolling. Shortcomings of the Lagrangian method
are that flow problems are difficult to model and that the mesh distortion is as severe as the deformation of the object.
Severe mesh degeneration is shown in Figure 4-9b. However, recent advances in adaptive meshing and rezoning have
alleviated the problems of premature termination of the analysis due to mesh distortions as shown in Figure 4-9c.
The Lagrangian approach can be classified in two categories: the total Lagrangian method and the updated Lagrangian
method. In the total Lagrangian approach, the equilibrium is expressed with the original undeformed state as the reference;
in the updated Lagrangian approach, the current configuration acts as the reference state. The kinematics of deformation
and the description of motion is given in Figure 4-10 and Table 4-4.
(a) Original (b) Deformed Mesh

(Undeformed Mesh) Before Rezoning
(c) Deformed Mesh

After Rezoning
Figure 4-9 Rezoning Example

CHAPTER 4 71
Previous f
t=n u
Current
t=n+1
Fn
un + 1
un
F
Reference
t=0
Fn+1 = fFn
Figure 4-10 Description of Motion
Depending on which option you use, the stress and strain results are given in different form as discussed below. If the LARGE
DISP or LARGE STRAIN parameters are not used, the program uses and prints engineering stress and strain measures. These
measures are suitable only for analyses without large incremental or total rotation or large incremental or total strains.
LARGE DISP
Using the LARGE DISP parameter, Marc uses the total Lagrangian method. The program uses and prints second Piola-
Kirchhoff stress and Green-Lagrange strain. These measures are suitable for analysis with large incremental rotations and
large incremental strains.
Table 4-4 Kinematics and Stress-Strain Measures in Large Deformation

Configuration Measures Reference (t = 0 or n) Current (t = n + 1)
Coordinates X x
Deformation Tensor C (Right Cauchy-Green) b (Left Cauchy-Green)
Strain Measure E (Green-Lagrange) e (Logarithmic)
F (Deformation Gradient)
Stress Measure S (second Piola-Kirchhoff )  (Cauchy)
P (first Piola-Kirchhoff )
LARGE STRAIN
With the LARGE STRAIN parameter, Marc offers a convenient option to flag the appropriate large displacement / large strain
formulation for a particular element / material combination. The default formulation preference is Updated Lagrange
which uses Cauchy stresses and true strains. This is suitable for analyses with large elastic and plastic strains. Stress and
strain components are printed with respect to the current state. The LARGE STRAIN parameter is normally specified as
LARGE STRA, N where N can take three different values 1, 2 or 4 (default = 1):
1. LARGE STRA,1: this favors the Additive Decomposition Updated Lagrange procedure at the global level. For
specific element and material combinations, it may override the global parameters and replace them with more
suitable local parameters. Also, for certain cases (for example, analysis with mechanical shape memory alloy, analysis
with hyperelastic materials), it may change the global parameters from LARGE STRA,1 to LARGE STRA,2. The
user has no flexibility in controlling the global parameters and limited flexibility in controlling the local parameters.
More information on this is provided in the next section.
2. LARGE STRA,2: this favors the Multiplicative Decomposition Updated Lagrange procedure at the global level.
For specific element and material combinations, it may override the global parameters and replace them with more
suitable local parameters. The user has no flexibility in changing the global or local parameters.
3. LARGE STRA,4: this does not favor any particular procedure at the global level. For each element and material
combination, the optimal choice of formulation flags is automatically determined by the program. The user has
complete flexibility in changing these local parameters. In general, the preference at the local level is for the updated
Lagrange procedure.
For more details on the criteria followed in the selection of a procedure, the user is referred to the LARGE STRAIN parameter
in Marc Volume C: Program Input.
Theoretically and numerically, if formulated mathematically correct, the Total Lagrange and Updated Lagrange
formulations yield exactly the same results. However, integration of constitutive equations for certain types of material
behavior (for example, plasticity) make the implementation of the total Lagrange formulation inconvenient. If the
constitutive equations are convected back to the original configuration and proper transformations are applied, then both
formulations are equivalent. However, for deformations involving excess distortions, ease of rezoning favors the updated
Lagrangian formulation. This is reflected in the fact that a rezoned mesh in the current state is mapped back to excessively
distorted mesh leading to negative Jacobian in the total Lagrangian formulation.
The terminology total and updated Lagrangian has been used with some vagueness [1, 2]. In this document, for a sequence
of incremental motions at t = 0 1 2 n and n + 1 , the total Lagrangian formulation entails the use of t = 0
configuration as reference; while in the updated Lagrangian configuration, the t = n + 1 (unequilibriated) configuration
is the reference.
Total Lagrangian Procedure

The total Lagrangian procedure can be used for linear or nonlinear materials, in conjunction with static or dynamic
analysis. Although this formulation is based on the initial element geometry, the incremental stiffness matrices are formed
to account for previously developed stress and changes in geometry.
This method is suitable for the analysis of nonlinear elastic problems (for instance, with the MOONEY, OGDEN, or NLELAST
material behavior or the HYPELA2 user subroutine). The total Lagrangian approach is also useful for problems in plasticity
and creep, where moderately large rotations but small strains occur. A case typical in problems of beam or shell bending.
However, this is only due to the approximations involved.
To activate the large displacement (total Lagrangian approach) option in Marc, use the LARGE DISP parameter. Include the
FOLLOW FOR parameter for follower force (for example, centrifugal or pressure load) problems. This parameter forms all
distributed loads on the basis of the current geometry. Do not use the CENTROID parameter with this parameter. Always use
CHAPTER 4 73
residual load corrections with this parameter. To input control tolerances for large displacement analysis, use CONTROL
In the total Lagrangian approach, the equilibrium can be expressed by the principle of virtual work as:
0 0
 S ij E ij dV =  b i  i dV +  t i  i dA (4-9)
V0 V0 A0
0
Here S ij is the symmetric second Piola-Kirchhoff stress tensor, E i j , is the Green-Lagrange strain, b i is the body force in
0
the reference configuration, t i is the traction vector in the reference configuration, and  i is the virtual displacements.
Integrations are carried out in the original configuration at t = 0 . The strains are decomposed in total strains for
equilibrated configurations and the incremental strains between t = n and t = n + 1 as:
n+1 n
Ei j = E ij + E ij (4-10)
n
while the incremental strains are further decomposed into linear, E i j and nonlinear, E i j parts as:
n
E i j = Eij + E ij
where E is the linear part of the incremental strain expressed as:

n
1 u u 1  u k  u k  u kn  u k
E = --- ------------i + ------------j + ---  ---------  ------------- +  ---------  ------------- (4-11)
2 X j X i 2  X i  X j   X j  X i 
n
The second term in the bracket in Equation (4-11) is the initial displacement effect. E is the nonlinear part of the
incremental strain expressed as:
n 1 u k u k
E = ---  -------------  ------------- (4-12)
2  X i   X j 
Linearization of equilibrium of Equation (4-9) yields:

 K 0 + K 1 + K 2 u = F – R (4-13)
where K 0 is the small displacement stiffness matrix defined as:
0 0
 K 0  ij =   imn D m np q  p qj dV
V0
K 1 is the initial displacement stiffness matrix defined as:
u u u 0 u u
 K1 i j =    imn D m n pq  p qj +  i m n D m n pq  p qj +  i m n D m n p q  p q j  dv
V0
0 u
in the above equations,  i m n and  i m n are the constant and displacement dependent symmetric shape function gradient
matrices, respectively, and D m n pq is the material tangent,
and K 2 is the initial stress stiffness matrix
 K2 i j =  N i k N j l S k l dV
V0
in which S kl is the second Piola-Kirchhoff stresses and N i k is the shape function gradient matrix.
Also, u is the correction displacement vector. Refer to CHAPTER 11 in this manual for more details on the solution
procedures. F and R are the external and internal forces, respectively.
This Lagrangian formulation can be applied to problems if the undeformed configuration is known so that integrals can
be evaluated, and if the second Piola-Kirchhoff stress is a known function of the strain. The first condition is not usually
met for fluids, because the deformation history is usually unknown. For solids, however, each analysis usually starts in the
stress-free undeformed state, and the integrations can be carried out without any difficulty.
For viscoelastic fluids and elastic-plastic and viscoplastic solids, the constitutive equations usually supply an expression for
the rate of stress in terms of deformation rate, stress, deformation, and sometimes other (internal) material parameters. The
relevant quantity for the constitutive equations is the rate of stress at a given material point.
It, therefore, seems most obvious to differentiate the Lagrangian virtual work equation with respect to time. The rate of
virtual work is readily found as:
· v k  k · ·
 S i j E ij + S ij --------- ------------- dV =  b i  i dV +  t i  i dA (4-14)
X i X j
V0 V0 A0
This formulation is adequate for most materials, because the rate of the second Piola-Kirchhoff stress can be written as:
· · ·
S i j = S ij  E kl S mn E p q  (4-15)
For many materials, the stress rate is even a linear function of the strain rate
· ·
S i j = D ijkl  S mn E pq E k l (4-16)
Equation (4-14) supplies a set of linear relations in terms of the velocity field. The velocity field can be solved non iteratively
and the displacement can be obtained by time integration of the velocities.
The second Piola-Kirchhoff stress for elastic and hyperelastic materials is a function of the Green-Lagrange strain defined
below:
CHAPTER 4 75
S ij = S ij  E kl  (4-17)
If the stress is a linear function of the strain (linear elasticity)
S i j = D ijkl E kl (4-18)
the resulting set of equations is still nonlinear because the strain is a nonlinear function of displacement.
Updated Lagrangian Procedure

The Updated Lagrange formulation takes the reference configuration at t = n + 1 . True or Cauchy stress and an
energetically conjugate strain measure, namely the true strain, are used in the constitutive relationship.
The updated Lagrangian approach is useful in:
a. Analysis of shell and beam structures in which rotations are large so that the nonlinear terms in the curvature
expressions may no longer be neglected, and
b. Large strain elasticity and plasticity analysis.
In general, this approach can be used to analyze structures where inelastic behavior (for example, plasticity, viscoplasticity,
or creep) causes the large deformations. The (initial) Lagrangian coordinate frame has little physical significance in these
analyses since the inelastic deformations are, by definition, permanent.
The LARGE STRAIN parameter generally specifies large strain analysis within the framework of Updated Lagrange
formulation. It signals Marc to calculate a geometric stiffness matrix and the initial stress stiffness matrix based on the
current deformed configuration.
Note: Because large strain analysis involves nonlinearity, the CENTROID parameter must not be used with this option.
For large strain analysis for rubber-like materials with incompressibility (such as materials defined with MOONEY, OGDEN,
GENT, ARRUDBOYCE, and ANISO HYPE model definition options), Marc uses a mixed formulation, in which both the
displacement and the hydrostatic pressure are independent variables, to overcome the numerical difficulties resulting from
the volumetric constraints. For compressible hyperelastic materials defined with FOAM model definition option, Marc uses
conventional displacement formulation.
For large strain elastic-plastic analysis, the default procedure in Marc uses a procedure based on an additive decomposition
of incremental strain into an elastic part and a plastic part, together with a mean normal return-mapping algorithm. In this
case, volumetric strain in a lower-order plane strain, axisymmetric or 3-D brick element is assumed to be constant for von
Mises plasticity to overcome volumetric locking because of the possible large and incompressible plastic deformation.
Marc can also use a procedure (LARGE STRAIN parameter with option 2) based on a multiplicative decomposition of
deformation gradient into an elastic part and a plastic part together with a radial return-mapping algorithm for large
strain elastic-plastic analysis. A mixed formulation is used to deal with the problem associated with volume constraints.
This procedure is only available for plane strain, axisymmetric, and 3-D continuum elements.
Herrmann elements in Marc support additive plasticity. Herrmann elements have additional pressure degrees of freedom
which increase numerical costs but do a better job at modeling of incompressible or nearly incompressible materials in an
elasto-plastic analysis. Moreover Herrmann triangular and tetrahedral elements 155, 156, 157 and 247 when used with
additive plasticity are less expensive than their special purpose counterparts - strain smoothening elements 239, 240 and
241.
Marc uses Cauchy stress (true stress) and logarithmic strain with Updated Lagrange formulation.
It is instructive to derive the stiffness matrices for the updated Lagrangian formulation starting from the virtual work
principle in Equation (4-9).
Direct linearization of the left-hand side of Equation (4-9) yields:
 S ij  d E ij   dV =   i k  k j u i j dv (4-19)
V0 Vn + 1
where u and  are actual incremental and virtual displacements respectively, and  k j is Cauchy stress tensor.
s s
 dS i j E ij dV =    i j L i j k l   u k l  dv
(4-20)
V0 Vn + 1
s
 denotes the symmetric part of  , which represents the gradient operator in the current configuration. Also, in
Equation (4-19) and Equation (4-20), three identities are used:
1
 ij = --- F im S mn F jn
J
s
E i j = F mi   mn F n j (4-21)
and
1
L i j k l = --- F im F jn F kp F l q D m n pq
J
in which D mnpq represents the material moduli tensor in the reference configuration which is convected to the current
configuration, L ijk . This yields:
 K 1 + K 2 u = F – R (4-22)
where K 1 is the material stiffness matrix written as:
 K1 i j =   imn L m n pq  p qj
Vn + 1
in which  imn is the symmetric gradient operator-evaluated in the current configuration and  k l is the Cauchy stresses
and K 2 is the geometric stiffness matrix written as:
 K2 i j =   kl N i k N j l dv
Vn + 1
CHAPTER 4 77
while F and R are the external and internal forces, respectively.

Keeping in view that the reference state is the current state, a rate formulation analogous to Equation (4-14) can be obtained
by setting:
 
F ij =  ij E ij = d i j --------- = -------- Si j = i j (4-23)
X i x i
where F is the deformation tensor, and d is the rate of deformation. Hence,

v k  k ·
t·i  i da

 ij d ij +  ij --------- ------------- dv =
 x i x j  b i  i dv +  (4-24)
Vn + 1 Vn + 1 An + 1
in which b i and t i is the body force and surface traction, respectively, in the current configuration

In this equation,  ij is the Truesdell rate of Cauchy stress which is essentially a Lie derivative of Cauchy stress obtained as:
·
 –1 –1
 i j = F in  JF nk  kl F m l  F m j (4-25)
The Truesdell rate of Cauchy stress is materially objective implying that if a rigid rotation is imposed on the material, the
Truesdell rate vanishes, whereas the usual material rate does not vanish. This fact has important consequences in the large
deformation problems where large rotations are involved. The constitutive equations can be formulated in terms of the
Truesdell rate of Cauchy stress as:

 ij = L ijk d k (4-26)
Eulerian Formulation
In analysis of fluid flow processes, the Lagrangian approach results in highly distorted meshes since the mesh convects with
the material. Hence, an alternative formulation, namely Eulerian, is used to describe the motion of the body. In this
method, the finite element mesh is fixed in space and the material flows through the mesh. This approach is particularly
suitable for the analysis of steady-state processes, such as the steady-state extrusion or rolling processes shown in Figure 4-11.
Figure 4-11 Rolling Analysis
The governing differential equations of equilibrium for fluid flow through an enclosed volume are now written as:
D  v i  
- = b i + ---------i-j
----------------- (4-27)
Dt x j
D
where, ------ is the material time derivative of a quantity and v is the velocity of the particle flowing through the mesh. For
Dt
an incompressible fluid, Equation (4-27) along with continuity equation (mass conservation) yields:
v i v i 
 -------- + v j -------- = b i + ---------i-j (4-28)
t x j x j
The left-hand side of Equation (4-28) represents the local rate of change augmented by the convection effects. The same
principle can be called to physically explain the material time derivative of Cauchy stress; that is, Truesdell rate of Cauchy
stress. It can be seen from Equation (4-25) that:
 · v i v j v k
 ij =  ij – ---------  kj –  i k --------- +  i j --------- (4-29)
x k x k x k
The second and third terms on the right-hand side represent the convection effects. The last term vanishes for a completely
incompressible material; a condition enforced in the rigid-plastic flow of solids.
Nonlinear Boundary Conditions

There are three types of problems associated with nonlinear boundary conditions: contact, nonlinear support, and
nonlinear loading. The contact problem might be solved through the use of special gap elements or the CONTACT option.
Nonlinear support might involve nonlinear springs and/or foundations. Sometimes nonlinearities due to rigid links that
become activated or deactivated during an analysis can be modeled through adaptive linear constraints. Nonlinear loading
is present if the loading system is nonconservative, as is the case with follower forces or frictional slip effects.
CHAPTER 4 79
Discontinuities are inherent in the nature of many of these nonlinearities, making the solution by means of incremental
linear approximations difficult. Some of the most severe nonlinearities in mechanics are introduced by nonlinear boundary
conditions. It is, therefore, very important to be aware of potential problem areas and to have a good understanding of the
underlying principles. This awareness and understanding enables you to validate numerical answers and to take alternative
approaches if an initial attempt fails.
Arbitrary Eulerian-Lagrangian (AEL) Formulation

In the AEL formulation referential system, the grid moves independently from the material, yet in a way that is spans the
material at any time. Hence, a relationship between derivative with respect to the material and grid derivative is expressed
as:
·
  =  *  + c i   i (4-30)
p m
where c i is the relative velocity between the material particle, v and the mesh velocity, v ; for example,
i i
p m
ci = vi – vi (4-31)
p m
The second or latter term represents the convective effect between the grid and the material. Note that for v = vi , a
i
purely Eulerian formulation is obtained. The equation of momentum, for instance, can be represented as:
* p m v i  i j
v --------
i +   v j – v j  x = ---------
- + bi (4-32)
j x j
Due to its strong resemblance to the pure Eulerian formulation, AEL is also called quasi-Eulerian formulation.
Shell Thickness Update

The thickness of shells and other plane stress elements can be updated due to in-plane stretching. This is available with the
updated Lagrange formulation for standard isotropic and orthotropic materials. For elasto-plastic materials, the plastic
strain is also taken into account. The thickness update is optional and is controlled by an option in the material definition.
The thickness update is done separately for each layer in a multilayered element. For a composite shell, it is possible to have
the thickness update active in some layers and inactive in others. This is controlled by the settings in the respective material
used in the layer.
For an isotropic material, we have for the change of thickness strain,

 z = – ---   x +  y  +  pz (4-33)
E
The change in plastic thickness strain  pz is obtained from the in-plane plastic strain components assuming:
incompressibility:
 px +  py +  pz = 0 (4-34)
For an orthotropic material, we obtain:
 31  23
 z = – --------  x – --------  y +  pz (4-35)
Ez Ey
It should be noted that the Young’s modulus in the thickness direction, Ez, must be provided. Otherwise, the program stops
with an error message.
Nonlinear Boundary Conditions

There are three types of problems associated with nonlinear boundary conditions:
1. Contact:
The contact problem might be solved through the use of special gap elements of the CONTACT option.
2. Nonlinear support:
Nonlinear support might involve nonlinear springs and/or foundations. Sometimes nonlinearities due to rigid links
that become activated or deactivated during an analysis can be modeled through adaptive linear constraints.
3. Nonlinear loading:
Nonlinear loading is present if the loading system is nonconservative, as is the case with follower forces or frictional
slip effects.
Discontinuities are inherent in the nature of many of these nonlinearities, making the solution by means of incremental
linear approximations difficult. Some of the most severe nonlinearities in mechanics are introduced by nonlinear boundary
conditions. It is, therefore, very important to be aware of potential problem areas and to have a good understanding of the
underlying principles. This awareness and understanding enables you to validate numerical answers and to take alternative
approaches if an initial attempt fails.The three problems are discussed in detail in the following sections:
Contact Problems
Contact problems are commonly encountered in physical systems. Some examples of contact problems are the interface
between the metal workpiece and the die in metal forming processes, pipe whip in piping systems, and crash simulation in
automobile designs.
Contact problems are characterized by two important phenomena: gap opening and closing and friction. As shown in
Figure 4-12, the gap describes the contact (gap closed) and separation (gap open) conditions of two objects (structures).
Friction influences the interface relations of the objects after they are in contact. The gap condition is dependent on the
movement (displacement) of the objects, and friction is dependent on the contact force as well as the coefficient of
(Coulomb) friction at contact surfaces. The analysis involving gap and friction must be carried out incrementally. Iterations
can also be required in each (load/time) increment to stabilize the gap-friction behavior.
CHAPTER 4 81
B
A n
Figure 4-12 Normal Gap Between Potentially Contacting Bodies
Two options are available in Marc for the simulation of a contact problem. A detailed description of these options (gap-
friction element and the CONTACT option) is given in Chapter 6 Contact of this manual.
Nonlinear Support
Marc provides two options for the modeling of support conditions: springs and elastic foundations. Both linear and
nonlinear springs can be specified in the input. In a nonlinear problem, the spring stiffness and the equivalent spring
stiffness of the elastic foundation can also be modified through a user subroutine. In the nonlinear spring option, the
incremental force in the spring is
F = K  u 2 – u 1  (4-36)
where K is the spring stiffness, u 2 is the displacement increment of the degree of freedom at the second end of the spring,
and u 1 is the displacement increment of the degree of freedom at the first end of the spring.
Use the SPRINGS model definition option for the input of linear and nonlinear spring data. The USPRNG user subroutine may
also be used to specify the value of K based on the amount of previous deformation for nonlinear springs. In dynamic
analysis, the SPRINGS option can also be used to define a dashpot. In thermal analysis or electrical analysis (heat transfer,
Joule heating, heat transfer pass of a coupled analysis), the SPRINGS option can be used to define a thermal or electrical link.
In the elastic nonlinear FOUNDATION option, the elements in Marc can be specified as being supported on a frictionless
(nonlinear) foundation. The foundation supports the structure with an increment force per unit area given by
P n = K  u n u n (4-37)
where K is the equivalent spring stiffness of the foundation (per unit surface area), and u n is the incremental
displacement of the surface at a point in the same direction as P n .
To input nonlinear foundation data, use the FOUNDATION model definition option.
To specify the value of K for the nonlinear equivalent spring stiffness based on the amount of previous deformation of the
foundation, use the USPRNG user subroutine.
Nonlinear Loading
When the structure is deformed, the directions and the areas of the surface loads are changed. For most deformed
structures, such changes are so small that the effect on the equilibrium equation can be ignored. But for some structures
such as flexible shell structure with large pressure loads, the effects on the results can be quite significant so that the surface
load effects have to be included in the finite element equations.
Marc forms both pressure stiffness and pressure terms based on current deformed configuration with the FOLLOW FOR
parameter. The FOLLOW FOR parameter should be used with the LARGE DISP or LARGE STRAIN parameters. The CENTROID
parameter should not be included due to the use of the residual load correction.
Follower force point loads may be applied by either specifying this in the POINT LOAD option or by applying a transformation
which rotates with the displacements of the nodes specified in the COORD SYSTEM option.
A special case of nonlinear loading is the resultant pressure due to a gas in an enclosed cavity. In such problems, the pressure
changes as the volume changes based upon the ideal gas law. This is discussed in Chapter 9, in the Cavity Pressure Loading
section of this manual.
Buckling Analysis
Buckling analysis allows you to determine at what load the structure will collapse. You can detect the buckling of a structure
when the structure’s stiffness matrix approaches a singular value. You can extract the eigenvalue in a linear analyses to obtain
the linear buckling load. You can also perform eigenvalue analysis for buckling load in a nonlinear problem based on the
incremental stiffness matrices.
The buckling option estimates the maximum load that can be applied to a geometrically nonlinear structure before
instability sets in. To activate the buckling option in the program, use the BUCKLE parameter. If a nonlinear buckling analysis
is performed, also use the LARGE DISP parameter.
Use the BUCKLE history definition option to input control tolerances for buckling load estimation (eigenvalue extraction by
a power sweep or Lanczos method). You can estimate the buckling load after every load increment. The BUCKLE INCREMENT
option can be used if a collapse load calculation is required at multiple increments.
The linear buckling load analysis is correct when you take a very small load step in increment zero, or make sure the solution
has converged before buckling load analysis (if multiple increments are taken). Linear buckling (after increment zero) can
be done without using the LARGE DISP parameter, in which case the restriction on the load step size no longer applies. This
value should be used with caution, as it is not conservative in predicting the actual collapse of structures.
In a buckling problem that involves material nonlinearity (for example, plasticity), the nonlinear problem must be solved
incrementally. During the analysis, a failure to converge in the iteration process or nonpositive definite stiffness signals the
plastic collapse.
For extremely nonlinear problems, the BUCKLE option cannot produce accurate results. In that case, the AUTO INCREMENT
history definition option allows automatic load stepping in a quasi-static fashion for both geometric large displacement and
material (elastic-plastic) nonlinear problems. The option can handle elastic-plastic snap-through phenomena. Therefore,
the post-buckling behavior of structures can be analyzed.
The buckling option solves the following eigenvalue problem by either the inverse power sweep or the Lanczos method:
 K + K G  u u    = 0 (4-38)
where K G is assumed to be a linear function of the load increment P to cause buckling.

CHAPTER 4 83
The geometric stiffness K G used for the buckling load calculation is based on the stress and displacement state change
at the start of the last increment. However, the stress and strain states are not updated during the buckling analysis. The
buckling load is therefore estimated by:
P  beginning  + P (4-39)
where for increments greater than 1, P  beginning  is the load applied at the beginning of the increment prior to the
buckling analyses, and  is the value obtained by the power sweep or Lanczos method.
The control tolerances for the inverse power sweep method are the maximum number of iterations in the power sweep and
the convergence tolerance. The power sweep terminates when the difference between the eigenvalues in two consecutive
sweeps divided by the eigenvalue is less than the tolerance. The Lanczos method concludes when the normalized difference
between all eigenvalues satisfies the tolerance. The maximum number of iterations and the tolerance are specified through
the BUCKLE history definition option.
Perturbation Analysis
The buckling mode can be used to perform a perturbation analysis of the structure. In the manual mode, a buckling
increment is performed upon request and the coordinates are perturbed by a fraction of the buckling mode or eigenvector.
You can enter an individual eigenvector number and the fraction or can request that a combination of modes be used. In
the subsequent increments, the coordinates are:
 i
X = X + f ------ or X = X +  f i -------
- (4-40)
 i
The manual mode can be activated by using the BUCKLE INCREMENT model definition, or BUCKLE load incrementation option.
In the automatic mode, the program checks for a nonpositive definite system during the solution phase. When this occurs,
it automatically performs a buckle analysis during the next increment and updates the coordinates. The automatic mode
can be activated by using the BUCKLE INCREMENT option. Also, be sure to force the solution of the nonpositive definite system
through the CONTROL option or PRINT parameter.
Material Nonlinearities
In a large strain analysis, it is usually difficult to separate the kinematics from the material description. Table 4-5 lists the
characteristics of some common materials.
Table 4-5 Common Material Characteristics

Material Characteristics Examples Marc Models
Composites Anisotropic: Bearings, aircraft panels Composite continuum
1) layered, ds i j = C i j k d k elements
21 constants
2)Fiber reinforced, Tires, glass/epoxy Rebars
E t
S = ---  T CT – 1 
2
one-dimensional strain in fibers

Creep Strains increasing with time under Metals at high temperatures, ORNL
constant load. polymide films, Norton
Stresses decreasing with time under semiconductor materials Maxwell
constant deformations.
Creep strains are non-instantaneous.
Elastic Stress functions of instantaneous strain Small deformation (below Hookes Law
only. Linear load-displacement relation. yield) for most materials:
metals, glass, wood
Elasto-plasticity Yield condition flow rule and hardening Metals von Mises Isotropic
rule necessary to calculate stress, plastic Soils Cam -Clay
strain. Permanent deformation Snow Hill’s Anisotropic
upon unloading. Wood Generalized Mohr-
Coulomb
Hyperelastic Stress function of instantaneous strain. Rubber Mooney
Nonlinear load-displacement relation. Ogden
Unloading path same as loading. Arruda-Boyce
Gent
Bergstrom Boyce
Anisopic Hyperelastic
NLELAST
Hypoelastic Rate form of stress-strain law Concrete Buyukozturk
Viscoelastic Time dependence of stresses in elastic Rubber, Glass, industrial Simo Model
material under loads. Full recovery after plastics Narayanaswamy
unloading.
Viscoplastic Combined plasticity and Metals Power law
creep phenomenon Powder Shima Model
Shape Memory Superelastic and shape memory effect Biomedical stents, Satellite Auricchio, Thermo-
with phase transformations. antennae mechanical
CHAPTER 4 85
A complete description of the material types mentioned is given in Chapter 5 of this manual. However, some no
characteristics and procedural considerations of some commonly encountered materials behavior are listed next.
Inaccuracies in experimental data, misinterpretation of material model parameters and errors in user-defined material law
are some common sources of error in the analysis from the materials viewpoint. It is useful to check the material behavior
by running a single element test with prescribed displacement and load boundary conditions in uniaxial tension and shear.
Large Strain Elasticity

Structures composed of elastomers, such as tires and bushings, are typically subjected to large deformation and large strain.
An elastomer is a polymer, such as rubber, which shows a nonlinear elastic stress-strain behavior. The large strain elasticity
capability in Marc deals primarily with elastomeric materials. These materials are characterized by the form of their elastic
strain energy function. For a more detailed description of elastomeric material, see Elastomer in Chapter 5 of this manual.
For the finite element analysis of elastomers, there are some special considerations which do not apply for linear elastic
analysis. These considerations, discussed below, include:
 Large Deformations
 Incompressible Behavior
 Instabilities
 Existence of Multiple Solutions
Large Deformations
The formulation is complete for arbitrarily large displacements and strains.
When extremely large deformations occur, the element mesh should be designed so that it can follow these deformations
without complete degeneration of elements. For problems involving extreme distortions, rezoning must be done. Rezoning
can be used with the formulation in the updated Lagrangian framework using conventional displacement based elements.
Incompressible Behavior
One of the most frequent causes of problems analyzing elastomers is the incompressible material behavior. Lagrangian
multipliers (pressure variables) are used to apply the incompressibility constraint. The result is that the volume is kept
constant in a generalized sense, over an element.
Both the total, as well as updated Lagrange formulations, are implemented with appropriate constraint ratios for lower-
and higher-order elements in 2-D and 3-D. For many practical analysis, the LBB (Ladyszhenskaya-Babuska-Brezzi)
condition does not have to be satisfied in the strictest sense; for example, four node quadrilateral based on Herrmann
principle.
For elements that satisfy the LBB condition, error estimates of the following form can be established
h h m i n  k + 1 
u –u 1 + p –p 0 = Oh  (4-41)
where k and are the orders of displacements and pressure interpolations, respectively. If K = min  k + 1  , the rate
of convergence is said to be optimal, and elements satisfying the LBB condition would not lock.
The large strain elasticity formulation may also be used with conventional plane stress, membrane, and shell elements when
using the total Lagrance method. Because of the plane stress conditions, the incompressibility constraint can be satisfied
without the use of Lagrange multipliers.
Instabilities
Under some circumstances, materials can become unstable. This instability can be real or can be due to the mathematical
formulation used in calculation.
Instability can also result from the approximate satisfaction of incompressibility constraints. If the number of Lagrangian
multipliers is insufficient, local volume changes can occur. Under some circumstances, these volume changes can be
associated with a decrease in total energy. This type of instability usually occurs only if there is a large tensile hydrostatic
stress. Similarly, overconstraints give rise to mesh locking and inordinate increase in total energy under large
compressive stresses.
Existence of Multiple Solutions

It is possible that more than one stable solution exists (due to nonlinearity) for a given set of boundary conditions. An
example of such multiple solutions is a hollow hemisphere with zero prescribed loads. Two equilibrium solutions exist: the
undeformed stress-free state and the inverted self-equilibrating state. An example of these solutions is shown in Figure 4-13 and
Figure 4-14. If the equilibrium solution remains stable, no problems should occur; however, if the equilibrium becomes
unstable at some point in the analysis, problems can occur.
x
Figure 4-13 Rubber Hemisphere
CHAPTER 4 87
Figure 4-14 Inverted Rubber Hemisphere
When incompressible material is being modeled, the basic linearized incremental procedure is used in conjunction with
mixed variational principles similar in form to the Herrmann incompressible elastic formulation. These formulations are
incorporated in plane strain, axisymmetric, generalized plane strain, and three-dimensional elements. These mixed
elements may be used in combination with other elements in the library (suitable tying may be necessary) and with each
other. Where different materials are joined, the pressure variable at the corner nodes must be uncoupled to allow for mean
pressure discontinuity. Tying must be used to couple the displacements only.
Large Strain Plasticity

In recent years there has been a tremendous growth in the analysis of metal forming problems by the finite element method.
Although an Eulerian flow-type approach has been used for steady-state and transient problems, the updated Lagrangian
procedure, pioneered by McMeeking and Rice, is most suitable for analysis of large strain plasticity problems. The main
reasons for this are:
 Its ability to trace free boundaries.
 The flexibility of taking elasticity and history effects into account.
Also, the residual stresses can be accurately calculated.
The large strain plasticity capability in Marc allows you to analyze problems of large-strain, elastic-plastic material behavior.
These problems can include manufacturing processes such as forging, upsetting, extension or deep drawing, and/or large
deformation of structures that occur during plastic collapse. The analysis involves both material, geometric and boundary
nonlinearities.
In addition to the options required for plasticity analysis, the LARGE STRAIN parameter is needed for large strain plasticity
analysis.
In performing finite deformation elastic-plastic analysis, there are some special considerations which do not apply for linear
elastic analysis include:
 Choice of Finite Element Types
 Nearly Incompressible Behavior
 Treatment of Boundary Conditions
 Severe Mesh Distortion
 Instabilities
Choice of Finite Element Types

Accurate calculation of large strain plasticity problems depends on the selection of adequate finite element types. In
addition to the usual criteria for selection, two aspects need to be given special consideration: the element types selected
need to be insensitive to (strong) distortion; for plane strain, axisymmetric, and three-dimensional problems, the element
mesh must be able to represent nondilatational (incompressible) deformation modes.
Nearly Incompressible Behavior

Most finite element types tend to lock during fully plastic (incompressible) material behavior. A remedy is to introduce a
modified variational principle which effectively reduces the number of independent dilatational modes (constraints) in the
mesh. This procedure is successful for plasticity problems in the conventional “small” strain formulation. Zienkiewicz
pointed out the positive effect of reduced integration for this type of problem and demonstrates the similarity between
modified variational procedures and reduced integration. The lower-order elements, invoking the constant volumetric
strain or the lower-order elements, using reduced integration and hourglass control, behave well for nearly incompressible
materials. Higher order elements in Marc are formulated to be used in large strain analysis including contact.
Treatment of Boundary Conditions

In many large strain plasticity problems, specifically in the analysis of manufacturing processes, the material slides with or
without friction over curved surfaces. This results in a severely nonlinear boundary condition. The Marc gap-friction
element and CONTACT option can model such sliding boundary conditions.
Severe Mesh Distortion

Because the mesh is attached to the deforming material, severe distortion of the element mesh often occurs, which leads to
a degeneration of the results in many problems. The ERROR ESTIMATE option can be used to monitor this distortion. To avoid
this degeneration, generate a new finite element mesh for the problem and then transfer the current deformation state to
the new finite element mesh. The global adaptive and rezoning procedure in the program is specifically designed for
this purpose.
Instabilities
Elastic-plastic structures are often unstable due to necking phenomena. Consider a rod of a rigid-plastic incompressible
·
workhardening material. With  the current true uniaxial strain rate and h the current workhardening, the rate of true
uniaxial stress  is equal to:
· ·
 = H (4-42)
The applied force is equal to F = A , where A is the current area of the rod. The rate of the force is therefore equal to:
·
F· =  A + A· (4-43)
On the other hand, conservation of volume requires that:
·
A + A· = 0 (4-44)
CHAPTER 4 89
Hence, the force rate can be calculated as:

·
F· =  H –  A (4-45)
Instability clearly occurs if   H . For applied loads (as opposed to applied boundary conditions), the stiffness matrix
becomes singular (nonpositive definite).
For the large strain plasticity option, the workhardening slope for plasticity is the rate of true stress versus the true plastic
strain rate. The workhardening curve must, therefore, be entered as the true stress versus the logarithmic plastic strain in a
uniaxial tension test.
Computational Procedures for Elastic-Plastic Analysis

Three basic procedures for plasticity exist in Marc. In this section, the variational form of equilibrium equations and
constitutive relations, and incompressibility are summarized. Issues regarding return mapping procedures for stress
calculation in three-dimensional and plane-stress conditions are also discussed.
For notational purpose, three configurations are considered at any point, original ( t = 0 ), previous ( t = n ) and current
( t = n + 1 ). An iterative procedure, full or modified Newton-Raphson or arc-length is used to solve for the equilibrium
at t = n + 1 .
1. Small Strain Plasticity (reference configuration: t = 0 ):

In this approach, the basis of variational formulation are 2nd Piola-Kirchhoff stress, S and Green-Lagrange strain,
E . Equilibrium of the current state can thus be represented by the following virtual work principle:
 S : E dV =  t :  dA +  b :  dV (4-46)
Vn An Vn
During the increment, all state variables are defined with respect to the state at t = n and are updated at the end
of increment upon convergence.
The linearized form of constitutive equations is given as:
ep
dS = L : dE – .dE – dE. (4-47)
ep
in which S is the second Piola-Kirchhoff stress,  is the Cauchy stress, and L is the elasto-plastic moduli. This
linearization has the advantage that it is fully independent of the rotation increment, but the disadvantage is that
the linearization causes errors equal to the square of the strain increment. Moreover, imposing the incompressibility
condition in terms of the trace of Green-Lagrange strains leads to errors in the form of fictitious volume changes in
fully developed plastic flow.
The above procedure works well in the context of small strain plasticity. However, in many large deformation
problems including metal forming processes, the plastic strain increments can be very large and the above procedure
can lead to large errors in the results. The two finite strain plasticity formulations to model the large inelastic strains
are: rate based (hypoelastic) and total (hyperelastic).
2. Finite Strain Plasticity with additive decomposition of strain rates (reference configuration: t = n + 1 ).
This formulation is based on the integration of the constitutive equations in the current configuration. To maintain
objectivity, the notion of rotation neutralized stress and strain measures is introduced. All objective stress rates, are
manifestation of Lie derivative:
·
–1 –T T
L v    =       (4-48)
where  is a deformation measure; for example, deformation gradient or rotation tensor, R while  is a stress
measure in the current configuration. The general form for an objective stress rate is:

 · T
=  –  –  + tr  d  (4-49)
From Table 4-6, it can be seen that there is a possibility of a number of stress rates. It can be observed that while all
the above stress rates are objective, the Truesdell and Durban-Baruch rates would not yield symmetric matrices.
Table 4-6 Objective Stress Rates
Stress Rates  
Truesdell L 1
Cotter-Rivlin L 0
Oldroyd -LT 0
Jaumann-Zaremba-Noll W 0
–1
Green-McInnis-Nagdhi R· R 0
1
Durban Baruch ---  D + W  1
2
Marc’s implementation of rate formulation involves the use of Jaumann rate of Cauchy stress which is obtained as
an average of the Oldroyd and Cotter-Rivlin stress rates. Thus, the Jaumann rate can be written as:
 ·
 =  – W + W (4-50)
 · T
 =  +    +    + tr  d  (4-51)

where ( · ) is the ordinary rate and ( ) is the objective rate with W is the spin or the antisymmetric part of the
velocity gradient, L . The last term is neglected because of the incompressible nature of plasticity.
The equilibrium in the current state is given by the virtual work principle at t = n + 1 :
  :  dv =  t . da +  b .  dv (4-52)
V n+1 A n+1 V n+1
CHAPTER 4 91
Linearization of the above form leads to the variational statement:

T
  d :  – 2  d    :  + tr  d  :  +  :    u         dv =
V n+1
(4-53)
 
d   b .  dv +  t .  da –   :  dv
V A n+1
 V n+1
n+1
where,
RN T
d = R  d R (4-54)
RN
Within the context of rotation neutralized form of constitutive relations, d is defined as:
RN RN RN
d = L : d (4-55)
RN T
where, d = R  d  R . Also, the rotation neutralized strain can be calculated by:
RN –T MID –1
 = U + I  F + I  F + IU + I
M ID
with  being the mid-increment strain, a good approximation for incremental logarithmic strain measure.
Where e = U – I is the engineering strain and U is the stretch tensor obtained by the polar decomposition of
F . Admitting an error of the order of e 2 in the approximation of the logarithmic strain, one obtains:
ep
d = L : d (4-56)
ep RN T T
where, L = R  R  L R R (4-57)
During computations, the third term of Equation (4-53), tr  d    :  , is neglected due to its nonsymmetric
nature.
In a fully developed plastic flow, the volumetric part of the energy can become extremely large and lead to
volumetric locking. Hence, a special treatment of incompressibility is done to relax the volumetric constraint in an
assumed strain format. The volumetric part of deformation gradient is modified such that, the assumed
deformation gradient is:
1---
– 1---
3 3 (4-58)
F = J J F
Linearization of F in the original state relates it to the displacement gradients in the current state as:
DF =   u  F (4-59)
where
1 1
  u  =  u  i j – --- div  u   i j + --- div  u   i j (4-60)
3 3
where the first two terms are evaluated at each integration point and the last term is averaged over the element.
For a lower order element, the procedure leads to mean or constant dilatation approach. Considering Equations
(4-53) to (4-60) the resulting system can be expressed as:
s ep s s s T
   : L :   u  – 2    :   u  +  :   u       dv =
V n+1
(4-61)
  s T
d   b .  dv +  t .  da –      :  dv
V A n+1
 Vn + 1
n+1
In the event, the elastic strains become large in an elastic-plastic analysis the rate based constitutive equations do
not accurately model the material response. This results from the fact that the elasticity matrix are assumed to have
constant coefficients in the current deformed configuration.
In the next formulation, the rate of deformation tensor is decomposed multiplicatively into the elastic and plastic
parts to resolve these problems.
3. Finite strain plasticity with multiplicative decomposition of deformation gradient.
An alternative formulation, based on the multiplicative decomposition of the deformation gradient has been
implemented in Marc, namely:
e  p
F = F F F (4-62)
e  p
where F , F , and F are (elastic, thermal, and plastic) deformation gradients, respectively. The thermo-
mechanical coupling is implemented using the staggered approach.
The above decomposition has a physical basis to it as the stresses are derived from quadratic - logarithmic strain
energy density function:
1 e 2 1 e 2
W = ---   ln J  +  tr  --- ln b  (4-63)
2 2 
This function has been chosen due to the availability of the material coefficients from material testing in a small
strain case.
In the metal forming applications, the elastic strains are negligible and the rate-based (or hypoelastic) as well as the
total (or hyperelastic) form of constitutive equations give virtually the same results. However, many polymers,
metals subjected to large hydrostatic pressures, high velocity impact loading of metals and processes, where shape
changes after deformation need to be evaluated precisely, the hyperelastic formulation yields physically more
meaningful results.
CHAPTER 4 93
The return mapping procedure for the calculation of stresses is based on the radial return procedure. With the use
of exponential mapping algorithm, the incompressibility condition is imposed exactly:
p
det F = 1 (4-64)
The strain energy is separated into deviatoric and volumetric parts in the framework of mixed formulation. The
general form of three-field variational principle is:
e e e
  u p J  =  [W  b  + U  J  + p  J – J ] dV (4-65)
V0
where, – 2--- (4-66)

3
b = J b
e e e
W  b  and U  J  are the deviatoric and elastic volumetric parts of the free energy, p is the pressure, J and J
are elastic pointwise and average elastic Jacobian of the element, respectively. The volumetric free energy can be of
any form in Equation (4-64). However, if the incompressibility is enforced in a pointwise fashion and the volumetric
free energy is assumed to be of the form:
1--- 2
e 9 e 3
U  J  = --- K  J  – 1 (4-67)
2
the perturbed Lagrangian form of the variational principle can be cast in a two-field framework as:
 1--- 
 e 3  P2
  u P  =  W  b  + 3P   J  – 1  – ------- dV
  2K
(4-68)
V0
 
2---
e 3
Note that P = p  J  , hence, P is not the true pressure in Equation (4-68). Choice of the volumetric free energy
in Equation (4-67) is not arbitrary and is described in more detail in Chapter 7 in the Elastomer section of this
manual.
Linearization of the equilibrium condition arising from the stationary of the variational principle Equation (4-68),
yields:
  1---

s 1 e 31  s
 
  :  --- C d e v + P  J  --- 1  1 – 2I   :  u +
J   3  
V n+1   
 
1---
dP e 3T (4-69)
 :     u  .     + ------  J  1 :  dv =
J
  s T
d   b .  dv +  t .  da –      :  dv
V A n+1
 V n+1
n+1
The linearization of the constraint equation defines the term dP in Equation (4-69).
The derivation of the spatial tangent C d e v follows by a push forward of the material tangent into the current
configuration:
– 1 d ev –T
Hence, S = F  F (4-70)
where S is the pull back of the deviatoric Kirchhoff stress tensor. Assuming the entire incremental deformation to
p
be elastic, a symmetric stress tensor, Ŝ can be defined with respect to a fixed plastic intermediate configuration F n
as:
tr – 1 d e v t r –T
Ŝ =  F e    Fe  (4-71)
also noting that,

3
A
Ŝ =  S A N A  N A where, Ŝ A = ----------------
tr 2
- (4-72)
A = 1   Ae 
3 3 3
dŜ =  dŜ A N A  N A +    Ŝ B – Ŝ A   B A N A  N B (4-73)
A = 1 A = 1B  A
1   A  tr tr
where, dŜ A = -----------------------------------  -----------
- – 2 A  A B  B e d B e (4-74)
tr 2 tr 2  tr 
  Ae    Be   Be
and dN A =  AB N B (4-75)
CHAPTER 4 95
3
tr T tr tr tr
similarly, Ĉ =  F e  F e with F e =  A e nA  NA (4-76)
A = 1
3
tr tr
dĈ =  2 Be d Be N B  N B +
B = 1
(4-77)
3 3
 tr 2 tr 2 
     Be  –   Be    B A N A  N B
 
A = 1 BA
The algorithmic elasto-plastic tangent can be obtained as:
S T T Ŝ T T
C dev = 2F  F ------- F  F = 2  F etr   F etr  -------  F etr   F etr  (4-78)
 C  Ĉ
Combining Equations (4-73), (4-78), and (4-79), the spatial form of deviatoric part of the tangent as:
3 3
  A 
C dev =    -----------
  Be
tr
- – 2 A  AB n A  n A  n B  n B +

A = 1 BA
3 3 tr 2 tr 2
  Ae   B –   Be  A
  --------------------------------------------------------
tr 2 tr 2
-  n A  nB  nA  nB + nA  nB  nB  nA 
  Be  –   Ae 
A = 1 BA (4-79)
The singularity for two- or three-equal stretch ratios is removed by repeated application of L’Hospital’s rule. Due
to separate interpolation of the pressure, a larger system of equations needs to be solved. A static condensation of
the element variables is carried out before going in the solver. This step renders the singularity of the system to be
inversely proportional to the bulk modulus of the material. Due to the multiplicative split of the deformation
gradient and an additive split of the free energy into deviatoric and volumetric parts, this framework is eminently
suitable for implementation of general nonlinear elastic as well as inelastic material models. Besides, compressible
and nearly incompressible material response can be naturally handled.
Finally, the stress calculation is done via return mapping to the yield surface. There are several methods to return
the stress. Among them, the popular ones include:
• Closest point projection (which reduces to radial return for von Mises)
• Mean normal approach
• Midpoint rule
• Trapezoidal rule
• Tangent cutting plane algorithm
• Multistage method
Krieg and Krieg (1977) investigated the radial return algorithm which has been proposed by Wilkins (1964).
The third algorithm analyzed was the so-called secant stiffness method proposed by Rice and Tracy (1973)
which represented a special case of the generalized Trapezoidal rule for nonhardening materials algorithm in
contrast to the fully implicit radial return algorithm and the fully explicit tangent stiffness approach. The radial
return method (shown in Figure 4-1) was found to be the most accurate among the three analyzed, particularly
when large strain increments were used. The radial return and mean normal method are available in Marc and
are described next.
 el
1 
 final *
 
2 3
Figure 4-1 Radial Return Method
1. Closest Point Projection procedure:

The closest point projection or the backward Euler algorithm reduces to a radial return scheme for cases where there
is no anisotropy or plane stress condition. The trial stress is determined as:
d d
 rr =  + 2G d (4-80)
The final stress state is obtained by simply scaling the trial stress by a scale factor, 
2G
where,  = 1 – --------------- (4-81)
d
r r
the plastic consistency parameter  , is obtained by solving for consistency condition iteratively.
An explicit form of the material tangent for the isotropic, von Mises yield surface can be given as:
ep 1
L = 2G   I – --- 1  1 – 2G  N̂  N̂ (4-82)
 3 
1
where,  = ------------------ – 1 +  (4-83)
H
1 + -------
3G
CHAPTER 4 97
ˆ
Here N is the return direction normal to the current yield surface. For the plane stress case, the zero normal stress
condition is explicitly imposed in the yield condition. Since the yield surface is not a circle in the plane, the return
direction is not radial anymore.
2. Mean-Normal method:
Assuming the entire deformation to be elastic in nature, the trial stress is evaluated as:
d d d d
 mn =  + 2G +  1 –  G (4-84)
Due to the use of pressure independent yield function, only the deviatoric stresses are evaluated. As shown in
Figure 4-2,  1 –   represents the fraction of strain increment for which the plastic flow occurs.
 el
* +  o +  el
d
1
 final
*
o
d d
2 3
Figure 4-2 Mean-Normal Method
d
 includes the deviatoric stresses at the previous increment scaled such that it satisfies the yield condition. The
yield criterion might not be exactly satisfied due to temperature effects, numerical integration of elastic-plastic
relationships, or accumulated numerical inaccuracy.
F    = 0 indicates that the stress state is exactly on the yield surface. If F     0 , scale  by a factor  such
that:
F    = 0 0    1 (4-85)
The equivalent plastic strain, plastic strain tensor and the elasto-plastic moduli are obtained as:
e
p  C : N̂  : ̂  1 –  
 = ----------------------------------------------------- (4-86)
e
N̂ : C : N̂ + H
p p
 =  N̂ (4-87)
e e
ep e  C : N̂    C : N̂ 
L = C –  1 –   --------------------------------------------------- (4-88)
e
N̂ : C : N̂ + H
e
where, N̂ is the mean-normal direction to the yield surface, H is the hardening modulus and C is the elastic
tangent moduli. For nonhardening materials, the mean normal algorithm is a special case of trapezoidal rule.
Marc plasticity algorithms are unconditionally stable and accurate for moderate strain increments. However, overall
global stability of the system can be dictated by other considerations like contact, buckling which then guide the
selection of appropriate load increment size. It should also be recognized that the algorithmically consistent tangent
in the closest point projection algorithm is based on the current state, and a lack of embedded directionality (unlike
secant methods) in the tangent can lead to divergent solutions unless line search or automatic time stepping
algorithms are used. However, when it converges it shows quadratic convergence as compared to linear convergence
for the mean-normal scheme.
3. Multistage Return Mapping:
The return mapping scheme described in this section is used for isotropic and anisotropic plasticity when LARGE
STRAIN is used to describe the Updated Lagrange Procedure. Marc uses this scheme for the Hill and Barlat yield
criteria. This scheme is not available for pressure dependent yield criteria like Mohr Coulomb. In a large strain
analysis, Marc automatically chooses the return mapping scheme that is most suited for the given material.
The increment of Cauchy stress for elasto-plasticity becomes
e e e p
̂ = Ĉ : ̂ = Ĉ :  ̂ – ̂  . (4-89)
ˆ
In Equation (4-89), superscript ‘ ’ means a materially embedded co-rotational coordinate system. For the numerical
implementation of Equation (4-89), the general midpoint rule can be expressed as follows:
e
̂ n + 1 = ̂ nT + 1 – Ĉ N̂ n +  (4-90)
e
where ̂ nT + 1 = ̂ n + Ĉ ̂
(Case 1) if F  ̂ nT + 1   0 ,  = 0 , ̂ n + 1 = ̂ nT + 1
(Case 2) if F  ̂ nT + 1   0 ,  =  such that F  ̂ nT + 1  = 0
where superscript T stands for a trial state. For Case 2, the condition stating that the updated stress stays on the
p
strain-hardening curve (  =     ) provides the following condition:
e
F    =   ̂ nT + 1 – Ĉ N̂ n +   –    np +   = 0 . (4-91)
Equation (4-91) is a nonlinear equation to solve for  .

CHAPTER 4 99
If the strain increment is not small enough, it is difficult to obtain the numerical solution of Equation (4-91) even
though it has a mathematical solution. Therefore, an iterative scheme, which utilizes the control of the potential
residual, is introduced. The method is applicable to a nonquadratic yield function and a general hardening law. For
the implementation of the algorithm, Equation (4-91) has been modified to the following relationship with
 = 1 ; that is,
F    k   =   ̂ T –   k  Ĉ e N̂  k   –    np +   k   = F  k  (4-92)
where
F    0  = 0  = F  0  ,  F  k  F  0   F  1   F  k     F  N   F  N  = 0  , k = 0  N  ,
F =  F k – 1 – F k    Y  =     p = 0    and F k = 1N–1 are prescribed values.
As shown in Equation (4-92) and Figure 4-3, the direction of the first substep N̂  1  is guessed from the direction
N̂  0  , which is normal to the yield surface at the trial stress ̂ T . Then, the exact direction N̂  1  can be obtained
from the first substep nonlinear solution based on Euler backward method. In general, the new direction for the
second substep N̂  i + 1  is guessed from the direction N̂  i  based on the previous substep stress ̂  i  . This
procedure is completed when F = F  N   = 0  . In fact, the equation for the Nth step is equivalent to Equation
(4-92) and ̂ n + 1 = ̂  N  . Finally, the proportional logarithmic plastic strain remains normal to the yield surface
at the final stress ̂ n + 1 ; that is,

̂ p =  p -------  ̂ n + 1   =   N  N̂  N   . (4-93)
̂
T
  ̂ 
  ̂  1  
ˆ
  ̂  2   N 1 
ˆ
N 2  ̂ T
ˆ
N N 
  ̂  n + 1   p
̂  N 
  ̂  n  
̂ p
̂  2 
̂ n + 1 p
̂  2 
̂ n
Figure 4-3 Schematic View for Multistage Return Mapping Method

Therefore, the normality condition of the incremental deformation theory is satisfied at the current state  n + 1 
for  = 1 .
In order to solve Equation (4-92), the Euler Backward method is employed to solve the iteration procedure for the
i
kth substep. At each iteration,  (at  k  substep and  i  iteration) becomes
k
i i i i i

g 1    – N̂ E – 1 g 2    + g 3    H
i   k  k k   k    k 
 = -------------------------------------------------------------------------------------------------------------------- (4-94)
k i i i
N̂ E – 1 N̂ +H
k k k
where
i i N̂- –1

E –1 = Ĉ e – 1 +  ------ ,
k  k  ̂
i i
g 1    =   ̂  –    np +   – F  k  ,
  k    k 
i i
g 2    = Ĉ e – 1  ̂ – ̂ T  +  N̂ ,
  k  k
i –1 i
g3    = H   – n  –  ,
  k  k
and H is the hardening modulus in stress-strain curve.

The detailed derivations of Equation (4-95) are shown in the work of Yoon at al. [Ref. 31].
In order to solve the equilibrium equation iteratively, the elasto-plastic tangent modulus consistent with the current
return mapping method is obtained as follows:
d̂ j = d̂ n + 1 = L e p d̂ n + 1 (4-95)
–1
 CN̂ n +   CN̂ n +  N̂ n + 
where L ep =  C – ------------------------------------------------
- and C = Ĉ e – 1 +  ------------------- .
 N̂ n +  CN̂ n +  + h ̂ n + 1
In Equation (4-95), h is instantaneous slope and  = 1 is used.

CHAPTER 4 101
Creep
Creep is a time-dependent inelastic behavior that can occur at any stress level, either below or above the yield stress of a
material. Creep is an important factor at elevated temperatures. In many cases, creep is also accompanied by plasticity,
which occurs above the yield stress of the material.
Marc offers two schemes for modeling creep in conjunction with plasticity:
a. Treating creep strains and plastic strains separately using an explicit procedure (where the creep is treated
explicitly) or an implicit procedure (where both creep and plasticity are treated implicitly). These procedures
are available with standard options via data input or with user-specified options via user subroutines. More
details are provided below.
b. Modeling creep strains and plastic strains in a unified fashion (viscoplasticity). Both explicit and implicit
procedures are again available for modeling unified viscoplasticity. More details are provided in the section
titled Viscoplasticity in this chapter.
The options offered by Marc for modeling creep are as follows:
 Creep data can be entered directly through data input or user subroutine. For explicit creep, the CRPLAW user
subroutine is to be used, and for implicit creep, the UCRPLW user subroutine is to be used.
 An automatic time stepping scheme can be used to maximize the time step size in the analysis.
 Eigenvalues can be extracted for the estimation of creep buckling time.
In addition, for explicit creep, the following additional options can be used:
 Creep behavior can be either isotropic or anisotropic.
 The Oak Ridge National Laboratory (ORNL) rules on creep can be activated.
The creep analysis option is activated in Marc through the CREEP parameter. The creep time period and control tolerance
information are input through the AUTO CREEP history definition option. This option can be used repeatedly to define a new
creep time period and new tolerances. These tolerances are defined in the section on Creep Control Tolerances.
Alternatively, a fixed time step can be specified through the CREEP INCREMENT history definition option. In this case, no
additional tolerances are checked for controlling the time step.
Creep analysis is often carried out in several runs using the RESTART option. Save restart files for continued analysis. The
REAUTO option allows you to reset the parameters defined in the AUTO CREEP option upon restart.
Adaptive Time Control

The AUTO CREEP option takes advantage of the diffusive characteristics of most creep solutions. Specifically, this option
controls the transient creep analysis. You specify a period of creep time and a suggested time increment. The program
automatically selects the largest possible time increment that is consistent with the tolerance set on stress and strain
increments (see Creep Control Tolerances in this chapter).
The algorithm is: for a given time step t , a solution is obtained. Marc then finds the largest values of stress change per
cr el
stress,    , and creep strain change per elastic strain,    . It compares these values to the tolerance values, T s
(stress change tolerance) and T e (strain change tolerance), for this period.
The value p is calculated as the larger of:
      T 
or (4-96)
cr
    el   T
If p  1 , the program resets the time step as:

t ne w = 0.8 t old  p (4-97)
The time increment is repeated until convergence is obtained or the maximum recycles control is exceeded. In the latter
case, the run is ended.
If the first repeat does not satisfy tolerances, the possible causes are:
 Excessive residual load correction
 Strong additional nonlinearities such as creep buckling-creep collapse
 Incorrect coding in the CRPLAW, VSWELL, or UVSCPL user subroutine.
Appropriate action should be taken before the solution is restarted.
If p  1 , the solution is stepped forward to t + t and the next step is begun. The time step used in the next increment
is chosen as:
t ne w = t old if 0.8 < p < 1 (4-98)
t ne w = 1.25 t old if 0.65 < p< 0.8 (4-99)
t ne w = 1.5 t old if p < 0.65 (4-100)
Since the time increment is adjusted to satisfy the tolerances, it is impossible to predetermine the total number of time
increments for a given total creep time.
Creep Control Tolerances

Marc performs a creep analysis under constant load or displacement conditions on the basis of a set of tolerances and
controls you provide.These are as follows:
1. Stress change tolerance – This tolerance controls the allowable stress change per time step during the creep
solution, as a fraction of the total stress at a point. Stress change tolerance governs the accuracy of the transient creep
response. If you need accurate tracking of the transient response, specify a tight tolerance of 1 percent or 2 percent
stress change per time step. If you need only the steady-state solution, supply a relatively loose tolerance of 10-20
percent. It is also possible to check the absolute rather than the relative stress.
CHAPTER 4 103
2. Creep strain increment per elastic strain – Marc uses either explicit or implicit integration of the creep rate
equation. When the explicit procedure is used, the creep strain increment per elastic strain is used to control
stability. In almost all cases, the default of 50 percent represents the stability limit, so that you need not provide any
entry for this value. It is also possible to check the absolute rather than the relative strain.
3. Maximum number of recycles for satisfaction of tolerances – During AUTO CREEP, Marc chooses its own time step.
In some cases, the program recycles to choose a time step that satisfies tolerances, but recycling rarely occurs more
than once per step. Excessive recycling can be caused by physical problems such as creep buckling, poor coding of
the CRPLAW, VSWELL, or UVSCPL user subroutine or excessive residual load correction that can occur when the creep
solution begins from a state that is not in equilibrium.
The maximum number of recycles allows you to avoid wasting machine time under such circumstances. If there is
no satisfaction of tolerances after the attempts at stepping forward, the program stops. The default of five recycles
is conservative in most cases.
4. Low stress cut-off – Low stress cut-off avoids excessive iteration and small time steps caused by tolerance checks
that are based on small (round off) stress states. A simple example is a beam in pure bending. The stress on the
neutral axis is a very small roundoff-number, so that automatic time stepping scheme should not base time step
choices on tolerance satisfaction at such points. The default of five percent of the maximum stress in the structure
is satisfactory for most cases.
5. Choice of element for tolerance checking – Creep tolerance checking occurs as a default for all integration points
in all elements. You might wish to check tolerances in only 1 element or in up to 14 elements of your choice.
Usually, the most highly stressed element is chosen.
When you enter the tolerances and controls, the following conventions apply:
 All stress and strain measures in tolerance checks are second invariants of the deviatoric state (that is, equivalent
von Mises uniaxial values).
 You can reset all tolerances and controls upon the completion of one AUTO CREEP sequence.
Background Information
Creep behavior is based on a von Mises creep potential with isotropic behavior described by the equivalent creep law:
· cr
 = f    cr T t  (4-101)
The material behavior is therefore described by:
· cr 
 cr =  --------- (4-102)
 d

where --------- is the outward normal to the current von Mises stress surface and · cr is the equivalent creep strain rate.
 d
There are two numerical procedures used in implementing creep behavior. The default is an explicit procedure in which
the above relationship is implemented in the program by an initial strain technique. In other words, a pseudo-load vector
due to the creep strain increment is added to the right-hand side of the stiffness equation.
Ku = P +   T D c r dv (4-103)

V
where K is the stiffness matrix, and u and P are incremental displacement and incremental nodal force vectors,
respectively. The integral,
  T D cr dv (4-104)
V
is the pseudo-load vector due to the creep strain increment in which  is the strain displacement relation and D is the
stress-strain relation. When plasticity is also specified through a suitably defined yield criterion and yield stress in Marc,
the plasticity is treated implicitly while the creep is treated explicitly.
As an alternative, an implicit creep procedure can be requested with the CREEP parameter. In this case, the inelastic strain
rate has an influence on the stiffness matrix. Using this technique, significantly larger steps in strain space can be used. In
Marc, this option is only to be used for isotropic materials with the creep strain rate defined by:
· cr n cr
 = A f   g  T h  t 
where A is the creep constant that can be defined through input data or through the UCRPLW user subroutine; power law
expression is always to be used for the effective stress with the coefficient provided through the input data or the UCRPLW
user subroutine; and power law coefficients or more general expressions can be provided for the creep strain, temperature,
and time through the input data or the UCRPLW user subroutine, respectively.
When plasticity is also specified through a suitably defined yield stress in conjunction with the von Mises yield criterion, a
sub-iterative scheme within each Newton-Raphson cycle is used to determine the plastic strains needed to keep the stress
state on the yield surface and the creep strains that develop due to the equivalent stress being greater than a user-defined
back stress. The yield stress for the plastic component can be varied as a function of the equivalent plastic strain,
temperature and spatial coordinates. Similarly, the back stress for the creep component can be varied as a function of the
equivalent creep strain, temperature and spatial coordinates.
Creep Buckling
Marc also predicts the creep time to buckling due to stress redistribution under given load or repeated cyclic load.
The buckling option solves the following equation for the eigenvalue:
 K + K G  = 0 (4-105)
The geometric stiffness matrix, K G , is a function of the increments of stress and displacement. These increments are
calculated during the last creep time step increment. To determine the creep time to buckle, perform a buckle step after a
converged creep increment. Note that the incremental time must be scaled by the calculated eigenvalue, and added to the
total (current) time to get an estimate as to when buckling occurs.
AUTO THERM CREEP (Automatic Thermally Loaded Elastic-Creep/Elastic-Plastic-Creep Stress Analysis)

The AUTO THERM CREEP option is intended to allow automatic, thermally loaded elastic-creep/elastic-plastic-creep stress
analysis, based on a set of temperatures defined throughout the mesh as a function of time. The temperatures and transient
times are presented to the program through the CHANGE STATE option, using input option 3 (post file), and the program
creates its own set of temperature steps (increments) based on a temperature change tolerance provided in this option. The
times at all temperature steps are calculated by the program for creep analyses.
CHAPTER 4 105
At each temperature step (increment), an elastic/elastic-plastic analysis is carried out first to establish stress levels in the
structure. A creep analysis is performed next on the structure for the time period between current and previous temperature
steps (increments). Both the elastic/elastic-plastic stress and the creep analyses are repeated until the total creep time
provided in this option is reached. Convergence controls are provided on the CONTROL option for elastic-plastic analysis and
in the AUTO THERM CREEP option for creep analysis. The analysis can be restarted at temperature steps (increments) or at
creep steps (subincrements). The results can be saved on a post file (POST option) for postprocessing.
If no DIST LOADS, POINT LOAD, or PROPORTIONAL INCREMENT option appears with the AUTO THERM CREEP set, all mechanical loads
and kinematic boundary conditions are held constant during the AUTO THERM CREEP. However, DIST LOADS, POINT LOAD,
PROPORTIONAL INCREMENT, or DISP CHANGE can be included in the set – the mechanical loads and kinematic boundary
conditions which are then defined are assumed to change in proportion to the time scale of the temperature history defined
by the CHANGE STATE option and are applied accordingly. This is based on the fact that the increments of load and
displacement correspond to the end of the transient time of the AUTO THERM CREEP input.
Viscoelasticity
In a certain class of problems, structural materials exhibit viscoelastic behavior. Two examples of these problems are
quenching of glass structures and time-dependent deformation of polymeric materials. The viscoelastic material retains
linearity between load and deformation; however, this linear relationship depends on time. Consequently, the current state
of deformation must be determined from the entire history of loading. Different models consisting of elastic elements
(spring) and viscous elements (dashpot) can be used to simulate the viscoelastic material behavior described in Chapter 5.
A special class of temperature dependence known as the Thermo-Rheologically Simple behavior (TRS) is also applicable
to a variety of thermal viscoelastic problems. Both the equation of state and the hereditary integral approaches can be used
for viscoelastic analysis.
To model the thermo-rheologically simple material, the SHIFT FUNCTION model definition option can be used to choose
between the Williams-Landel-Ferry equation, the power series expression, the Narayanaswamy model, or the Arrhenius
model.
In Marc, two options are available for small strain viscoelastic analysis. The first option uses the equation of state approach
and represents a Kelvin model. The second option is based on the hereditary integral approach and allows the selection of a
generalized Maxwell model. The thermo-rheologically simple behavior is also available in the second option for thermal
viscoelastic analysis. The Time-dependent Inelastic Behavior section in Chapter 7 discusses these models in detail. Automatic
time stepping schemes AUTO CREEP and AUTO STEP can be used in a viscoelastic analysis for first and second options,
respectively.
The first option for viscoelastic analysis uses the Kelvin model. To activate the generalized Kelvin model in Marc, use the
VISCO ELAS or CREEP parameter. To input the matrices [A] and [B] for the Kelvin strain rate computations, use the CRPVIS
user subroutine. To input creep time period and the tolerance control for the maximum strain in an increment, use the
AUTO CREEP history definition option.
The Simo model for large strain viscoelasticity can be used in conjunction with the damage and hyperelastic Mooney or
Ogden material model. The large strain viscoelastic material behavior can be simulated by incorporating the VISCELMOON,
VISCELOGDEN, or VISCELFOAM model definition option. Viscoelasticity for Mooney and Ogden materials is available in both
the total and updated Lagrangian framework. Viscoelasticity for foam materials is available only in the updated Lagrangian
framework.
A modified Bergström-Boyce model can be used to simulate nonlinear, large strain viscoelasticity for hyperelastic materials.
This is a two network model. The rate independent part can be modeled with any Marc hyperelastic models defined with
MOONEY, GENT, ARRUDBOYCE, OGDEN, MARLOW, or FOAM options. The viscous responses are represented by Arruda-Boyce
model. The viscous properties are defined by BERGBOYCE option. The modified Bergström-Boyce model is available only in
the updated Lagrange framework. It can be used along with DAMAGE option.
Nonlinear structural relaxation behavior of materials can be modeled by the Narayanaswamy model which accounts for
memory effect. This model allows simulation of evolution of physical properties of glass subjected to complex time
temperature histories. The thermal expansion behavior for the Narayanaswamy model is controlled via the VISCEL EXP
Viscoplasticity
There are two procedures in Marc for viscoplastic analysis: explicit and implicit. A brief description of each procedure
follows:
1. Explicit Method
2. Implicit Method
Explicit Method
The elasto-viscoplasticity model in Marc is a modified creep model to which a plastic element is added. The plastic element
is inactive when the stress is less than the yield stress of the material. You can use the elasto-viscoplasticity model to solve
time-dependent plasticity and creep as well as plasticity problems with a nonassociated flow law.
The CREEP option in Marc has been modified to enable solving problems with viscoplasticity. The method is modified to
allow solving elastic-plastic problems with nonassociated flow rules which result in nonsymmetric stress-strain relations if
the tangent modulus method is used.
The requirements for solving the viscoplastic problem are:
 CREEP parameter and creep controls
 Load incrementation immediately followed by a series of creep increments specified by AUTO CREEP
 Use of the CRPLAW user subroutine and/or the NASSOC user subroutine
The following load incrementation procedure enables a viscoplastic problem to be solved:
1. Apply an elastic load increment that exceeds the steady-state yield stress.
2. Relieve the high yield stresses by turning on the AUTO CREEP option.
You may repeat steps 1 and 2 as many times as necessary to achieve the required load history.
Note: The size of the load increments are not altered during the AUTO CREEP process so that further load increments
can be effected by using the PROPORTIONAL INCREMENT option.
The viscoplastic approach converts an iterative elastic-plastic method to one where a fraction of the initial force vector is
applied at each increment with the time step controls. The success of the method depends on the proper use of the
automatic creep time step controls. This means that it is necessary to select an initial time step that will satisfy the tolerances
placed on the allowable stress change.
CHAPTER 4 107
allowable stress change x 0.7

The initial time step t =
Maximum viscoplastic strain rate x Young’s modulus
The allowable stress change is specified in the creep controls. The most highly stressed element usually yields the maximum
strain rate. It is also important to select a total time that gives sufficient number of increments to work off the effects of the
initial force vector. A total time of 30 times the estimated t is usually sufficient.
Marc does not distinguish between viscoplastic and creep strains. A NASSOC user subroutine allows you to specify a
nonassociated flow rule for use with the equivalent creep strains (viscoplastic) that are calculated by the CRPLAW user
subroutine. A flag is set in the CREEP parameter in order to use the viscoplastic option with a nonassociated flow rule.
The viscoplasticity feature can be used to implement very general constitutive relations with the aid of the ZERO and YIEL
user subroutines.
Since the viscoplasticity model in Marc is a modified creep model, you should familiarize yourself with the creep analysis
procedure (see Nonlinear Analysis at the beginning of this chapter).
Implicit Method
A general unified viscoplastic material law can be implemented through the UVSCPL. user subroutine When using this
method, you are responsible for defining the inelastic strain increment and the current stress.
Fracture Mechanics
The fracture mechanics capabilities in Marc covers the evaluation of energy release rate and J-integral including automatic
crack propagation. Two methods are offered for the evaluation: the mode separation method through the LORENZI option
and the Virtual Crack Closure Technique (VCCT) with the VCCT option. The VCCT option also supports automatic crack
propagation.
Linear Fracture Mechanics

Linear fracture mechanics presupposes existence of a crack and examines the conditions under which crack growth occurs.
In particular, it determines the length at which a crack propagates rapidly for specified load and boundary conditions. The
concept of linear fracture mechanics stems from Griffith’s work on purely brittle materials. Griffith stated that, for crack
propagation, the rate of elastic energy release should at least equal the rate of energy needed for creation of a new crack
surface. This concept was extended by Irwin to include limited amounts of ductility. In Irwin’s considerations, the inelastic
deformations are confined to a very small zone near the tip of a crack.
The basic concept presented by Griffith and Irwin is an energy balance between the strain energy in the structure and the
work needed to create a new crack surface. This energy balance can be expressed using the energy release rate G as
G = Gc (4-106)
G is defined as:
d
G = – -------- (4-107)
da
where,  is the strain energy and a is the crack length. G depends on the geometry of the structure and the current
loading. G c is called the fracture toughness of the material. It is a material property which is determined from experiments.
Note that the energy release rate is not a time derivative but a rate of change in potential energy with crack length. An
important feature of Equation (4-106) is that it can be used as a fracture criterion; a crack starts to grow when G reaches
the critical value G c .
The stress and strain fields near the tip of a crack are singular for a linear elastic material model. The stresses and strains
have the principal form
K
 = ------ f   
r
(4-108)
K
 = ------ g   
r
in a polar coordinate system centered at the crack tip. Thus, a linear elastic material is said to have a 1  r singularity near
a crack tip. It is easy to demonstrate that in both Griffith’s and Irwin’s considerations, the elastic energy release rate is
determined by a single parameter: the strength of the singularity in the elastic stress field at the crack tip. This is the so-
called stress intensity factor, and is usually denoted by capital K . The magnitude of K depends on the crack length, the
distribution and intensity of applied loads, and the geometry of the structure. Crack propagation occurs when any
combination of these factors causes a stress intensity factor K to be equal to or greater than the experimentally determined
material property K c , which is equivalent to Equation (4-106). Hence, the objective of linear fracture mechanics
calculations is to determine the value of K .
There are three possible modes of crack extension in linear elastic fracture mechanics as follows:
1. The opening mode (Mode I):
The opening mode (see Figure 4-15a), is characterized by the symmetric separation of the crack surfaces with respect
to the plane, prior to extension (symmetric with respect to the X-Y and X-Z planes).
2. Sliding mode(Mode II):
The sliding mode,is characterized by displacements in which the crack surfaces slide over one another perpendicular
to the leading edge of the crack (symmetric with respect to the X-Y plane and skew-symmetric with respect to the
X-Z plane).
3. Tearing mode (Mode III.see Figure 4-15):
The tearing mode, finds the crack surfaces sliding with respect to one another parallel to the leading edge (skew-
symmetric with respect to the X-Y and X-Z planes).
CHAPTER 4 109
y
y
x
x
z
z
(a) Mode I: Opening (b) Mode II: Sliding

y
x
z
(c) Mode III: Tearing

Figure 4-15 Irwin’s Three Basic Modes of Crack Extension
It is customary to associate a stress intensity factor with each of these mode: K I , K I I , and K I I I . There is also an associated
fracture toughness associated with each mode: K I c , K I I c , and K I I I c . The most critical mode is usually mode I and in
many cases the other modes are not considered. The connection between the energy release rate and the stress intensity
factors is given by
2 2
K K II 1 +  2
G = ------I- + -------
- + ------------- K I I I (4-109)
E' E' E
where
E' = E for plane stress (4-110)

and
E
E' = --------------- for plane strain. (4-111)
1 – 2
Marc uses the so-called J-integral for evaluating the energy release rate, see below. The J-integral is similar to G but is more
general and is also used for nonlinear applications. J is equivalent to G when a linear elastic material model is used.
Nonlinear Fracture Mechanics

Nonlinear fracture mechanics is concerned with determining under which conditions crack propagation (growth) occurs.
In this sense, nonlinear fracture mechanics is similar to linear fracture mechanics. However, there are additional questions
addressed in nonlinear fracture mechanics. Is the crack propagation stable or unstable? If it is stable, at which speed does it
occur? After some propagation, is the crack arrested?
There also exists a singularity at the crack tip in fracture mechanics problems with nonlinear elastic-plastic material
behavior, though the singularity is of a different nature. If one takes an exponential hardening law of the form:
  1/n
------ =  ----- (4-112)
0   0
it can be shown that the singularities in the strain and the stresses at the crack tip are of the form:
 = f   r – n   n + 1 
(4-113)
 = g   r – 1   n + 1 
If n approaches infinity, the material behavior becomes perfectly plastic and the singularity in the stresses vanishes. The
singularity in the strains, however, takes the form of:
 = f   r – 1 (4-114)
It has not been possible to establish that the strength of the singularity is the only factor that influences initiation of crack
propagation for nonlinear situations. In fact, it is doubted that initiation of crack propagation is dependent on only a single
factor. The J-integral probably offers the best chance to have a single parameter to relate to the initiation of crack
propagation.
The J-integral was introduced by Rice as a path-independent contour integral for the analysis of cracks. As previously
mentioned, it is equivalent to the energy release rate for a linear elastic material model. It is defined in two dimensions as:
u j
J =    W + T n 1 –  i j n i -------
x 1
- d (4-115)

where W is the strain energy density, T is the kinetic energy density,  i j is the stress tensor and u i is the displacement
vector. The x1 direction is the same as the x direction in the local crack tip system in Figure 4-16. The integration path 
is a curve surrounding the crack tip, see Figure 4-16.The J-integral is independent of the path  as long as it starts and ends
at the two sides of the crack face and no other singularities are present within the path. This is an important feature for the
numerical evaluation since the integral can be evaluated using results away from the crack tip.
CHAPTER 4 111
ni
y
x

Figure 4-16 Definition of the J-integral
Numerical Evaluation of the J-integral

The J-integral evaluation in Marc is based upon the domain integration method as described in [Ref. 29]. It is available as
the LORENZI model definition option. A direct evaluation of Equation (4-115) is not very practical in a finite element analysis
due to the difficulties in defining the integration path  . In the domain integration method for two dimensions, the line
integral is converted into an area integration over the area inside the path  . This conversion is exact for the linear elastic
case and also for the nonlinear case if the loading is proportional, that is, if no unloading occurs. By choosing this area as
a set of elements, the integration is straightforward using the finite element solution. In two dimensions, the converted
expression is
u j q 1
J =    ij -------
x 1
- – W 1 i -------- dA
 x i
(4-116)
A
for the simplified case of no thermal strains, body forces or pressure on the crack faces. For the general expression, see
[Ref. 1]. A is the area inside  and q 1 is a function introduced in the conversion into an area integral. The function q 1 can
be chosen fairly generally, as long it is equal to one at the crack tip and zero on  . The form of the function chosen in Marc
is that it has the constant value of one at all nodes inside  , and decreases to zero over the outermost ring of elements in
A . It can be interpreted as a rigid translation of the nodes inside  while the nodes on  remain fixed. Thus, the
contribution to Equation (4-116) comes only from the elements in a ring away from the crack tip. This interpretation is that
of virtual crack extension and this method can be seen as a variant of such a technique, although it is extended with the
effects of thermal strains, body forces, and pressure on the crack faces.The set of nodes moved rigidly is referred to as the
rigid region and the function q 1 as the shift function or shift vector. For the evaluation of the J-integral the direction of the
shift vector is simply the x axis in the local crack tip system.
In three dimensions, the line integral becomes an area integral where the area is surrounding a part of the crack front. In
this case, the selection of the area is even more cumbersome than in two dimension. The converted integral becomes a
volume integral which is evaluated over a set of elements. The rigid region is a set of nodes which contains a part of the
crack front, and the contribution to the integral comes from the elements which have at least one but not all its nodes in
the rigid region.
Determination of the Rigid Region and the Shift Vector

For the evaluation of the J-integral, Marc requires the nodes along the crack front, the shift vector, and the nodes of the
rigid region. Marc allows three ways of defining the rigid region. The first is direct input; the nodes or elements in the rigid
region are listed explicitly. With this variant, the shift vector is also specified directly. The second and third variants use an
automatic search for the nodes of the rigid region. The second variant is based on the mesh topology (connectivity) where
a number of regions of increasing size are found by Marc. The first region for two dimensions consists of the nodes of all
elements connected to the crack tip node. The second region consists of all nodes in the first region and the nodes of all
elements connected to any node in the first region and so on for a given number of regions. This way, contours of increasing
size are determined. In the third way of determining the rigid region, a radius is given and all nodes within that radius are
part of the rigid region. For the automatic search methods the shift vector can also be determined automatically. It is then
determined using the first element edge on the crack face (for meshes with notches see below). Symmetry at the crack face
is automatically detected by Marc, also for the case that the symmetry condition is applied by means of a rigid contact body.
In three dimensions, it is a bit more complicated. The crack front is defined by an unsorted list of nodes. In order to obtain
the variation of the J-integral along the crack front, a disk of nodes with the normal of the disk directed along the tangent
to the crack front can be determined. This type of disk can readily be defined if the mesh around the crack front consists
of brick elements in a regular mesh. This mesh is typically created from a two-dimensional mesh which is extruded along
the crack front. The topology based determination of the rigid region assumes such a mesh and creates disks of increasing
size at each crack front node from element faces. The shift vector at each crack front node is automatically determined to
be perpendicular to both the tangent to the crack front and the normal to the crack face at each crack front node. At the
first and last crack front node, where a free surface is assumed to exist, the shift direction is projected to the tangent to the
free surface. This is important since the shift must not change the outer boundary of the model. The geometry based search
method here works with a cylinder with the axis aligned with the tangent of the crack front and centered at the crack front
node. The length of the cylinder is given as a fraction of the distance to the neighboring crack front nodes. All nodes within
the cylinder are part of the rigid region. This method is useful if, for instance, the mesh around the crack front is created
with an automatic mesh generator. The shift vector is determined in the same way as for the topology based search method.
In elasto-plastic analyses, it is often advantageous to use a mesh where there are several (multiple) nodes at the crack tip. If
collapsed elements are used at the crack tip, the nodes are kept separate and a notch is formed as the crack tip deforms. For
this kind of mesh, a “multiple nodes” distance is given, which should be smaller than the smallest element at the crack tip.
All nodes within that distance are then considered part of the crack tip, and the first contour for the topology based search
will consist of all elements connected to any of these nodes (in three-dimensions of element faces connected to any of these
nodes). The normal to the crack face used for determining the shift vector is taken as the first face outside the crack tip
nodes. Thus, it is also possible to model the crack with an initial notch.
Mode Separation
For the linear elastic case, Marc automatically calculates and prints the stress intensity factors K I , K I I , and K I I I for the
three modes I, II, and III. This evaluation is done using analytical functions for the stress field near the crack tip. The
implementation is based upon the procedure outlined in [Ref. 32]. Mode separation is not calculated when nonlinearities
like large deformations, nonlinear material, or general contact are present. However, the use of glued contact is allowed to
connect different parts of the mesh near the crack front. The material is also not allowed to be temperature dependent.
Orthotropic materials are allowed. The calculation of the stress intensity factors is automatically suppressed if unsupported
features are used.
CHAPTER 4 113
The stress intensity factors are defined in the local crack tip system, see Figure 4-16. K I will be positive when the crack is
opening and negative when it is overlapping. Similarly, K I I is positive when the shear stress is positive ahead of the crack
in the local system.
Supported Features
The LORENZI J-integral option supports the following features:
 Linear and nonlinear materials.
 Large deformations. Both total and updated Lagrangian formulation are supported. The crack tip system can
undergo finite rotations and translations.
 Thermal loads.
 Transient dynamics.
 Contact between crack faces including friction.
 Loads on the crack faces.
 Mode separation for linear elasticity and no loads in the crack region.
Modeling Considerations
The main difficulty in the finite element analysis of linear elastic fracture mechanics is representing the solution near the
crack front. Typically, a focused mesh is used with the elements in a spider web shape. This gives a mesh with rings of
elements in a radial fashion and is particularly useful for studying the variation of the values of J with contour radius. If
higher order elements are used for an elastic analysis, it is advantageous to use so-called quarter point elements at the crack
tip. These are regular elements where the midside nodes at element sides for which one node is part of the crack tip are
shifted towards the crack tip to the quarter point location along the side. This gives a more accurate description of the
singularity at the crack tip and is important to use if a coarse mesh is used.
In three-dimensions, the mesh along the crack front is typically created from a planar mesh which is extruded along the
crack front region as mentioned above. This allows the rigid regions to be created in a regular manner which is advantageous
for accuracy. Note that Marc allows rigid contact surfaces to be used for applying symmetry conditions. For a three-
dimensional example, see the Marc User’s Guide example e090 “Fracture Mechanics Analysis with the J-integral”.
Numerical Evaluation of the Energy Release Rate with the VCCT Method
The VCCT option offers a simpler but more general way for obtaining the energy release rate. The implementation follows
the description in R. Krueger [Ref. 33]. Consider again Equation (4-107).
d
G = – -------- (4-117)
da
It states that G is the change in potential energy by a change in crack length. Now, consider the simple finite element model
in Figure 4-17. The models A and B are the same, except that in B, the crack has grown by one element edge of length a.
Suppose we do one analysis for each one of the two models. We can now calculate the energy release rate G as:
Fu
G = ------ (4-118)
2a
a
u
Model A Model B
Figure 4-17 Mesh for Illustrating the CCT Method
Here, F is the force (obtained from Model A) that keeps the crack together, and the crack opening, u , is obtained from
Model B. In order to obtain these quantities, we would need to perform two analyses and this method is often referred to
as CCT (Crack Closure Technique). In the virtual crack closure technique, we only do the analysis with a closed crack
(Model A) and use the opening displacement at the closest nodes to the crack tip. Figure 4-18 shows the case of pure mode
I. The other modes are treated similarly and separately. The displacements and reactions are transformed into the local crack
tip system for this evaluation.
a
F
u
Figure 4-18 The VCCT Method

CHAPTER 4 115
With x. y, and z denoting the coordinate directions in the local crack tip system (see Figure 4-15), we obtain:
Fy uy Fx ux Fz uz
G I = -----------
- G II = -----------
- G I I I = ----------
- (4-119)
2a 2a 2a
and the total energy release rate as:
G t o t = G I + G II + G I I I (4-120)
For higher-order elements, we need to include the contributions from the midside nodes (see Figure 4-19).
F1 u1 + F2 u2
G I = ------------------------------
- (4-121)
2a
F1
F2
u1
u2
Figure 4-19 Higher-order Elements
For the case that the midside nodes are not at the middle of the element edges (for example, using the 1/4 point position
for increased accuracy, the displacements for these nodes are interpolated to the appropriate locations.
For 3-D solids, we have a situation as shown in Figure 4-20. The situation is similar to the 2-D case, and the evaluation is
done separately for each node along the crack front. The area is given by the shaded part in Figure 4-20.
For the case of higher-order elements, we obtain the following by using the notation in Figure 4-21.
1
F 1 u 1 + F 2 u 2 + ---  F 3 u 3 + F 4 u 4 
2
G = ------------------------------------------------------------------------------- (4-122)
2a
u is Crack Opening
Displacement At Crack Face
Crack Front
Figure 4-20 3-D Mesh for VCCT
u1 u3
u4
F1 F3
u2
F4 F2
Figure 4-21 3-D Mesh for VCCT for Higher-order Elements
The above figures show a regular mesh of hexahedral elements. While a mesh designed like this is advantageous for accuracy,
it is not strictly necessary. It is also possible to use a general tetrahedral mesh or a hexahedral mesh with the crack front as
defined in Figure 4-22. The latter is typically obtained in the case of crack propagation as described below. The program will
find the appropriate nodes to use for the forces and crack opening displacement and calculate an area a.
CHAPTER 4 117
Figure 4-22 Example of Crack Front Not Following Element Edges
The definition of the data involved in the VCCT calculation is done automatically. The user only specifies the crack tip or
crack front. In order to check what the program finds there is debug option available. If the PRINT parameter is used with a
value of 43, the program will print out the nodes it finds, the corresponding forces, crack opening displacements, and crack
areas.
Symmetry is automatically detected and accounted for. The symmetry condition can be enforced by boundary conditions
or by rigid contact.
The mesh in the crack region can contain user tyings or contact. The program automatically detects if the crack tip node
is connected to a node, and the node it is connected to is considered part of the crack tip. The supported connections are
glued contact, user tying, RBE2 and RROD. For the case that the connection is done node-to-node, it does not matter if the
tied or retained node of the tying is selected as the crack node. With glued contact, it is not necessary to connect the parts
node-to-node. If the meshes do not match up, the crack will be treated similarly as a case of symmetry. This non-matching
glued case is automatically detected and for this case, it is necessary to select the touching node as the crack tip node. A
useful option when defining a crack using glued contact is DEACT GLUE. Suppose two parts are glued together and the crack
is defined as part of the interface between the bodies. One can then use DEACT GLUE to identify the nodes that should be on
the crack faces (see Figure 4-23). These nodes will have regular contact but not be glued.
Crack Tip
Glued Contact
Nodes Identified
with DEACT GLUE
Contact Body 1
Contact Body 2
Figure 4-23 Using DEACT GLUE for Defining a Crack
The calculation of VCCT is done in the current geometry in case of large deformations. The updated crack coordinate
system is calculated at each increment and the calculations are done in this system. Thus, arbitrary rotations and
deformations are allowed. The current crack tip system is available on the post file. The three coordinate axes can be plotted
as vectors for verification.
Figure 4-24 shows some examples of supported crack configurations. There is a large flexibility in how parts can be tied or
glued together to form a crack.
Crack Tip
Crack Front
Line Crack in Shell Shell Glued to Shell 3-D Solid
Shell Glued to Solid (Line Crack) Shell Glued to Solid (Surface Crack)
Figure 4-24 Examples of Supported Crack Types
Remeshing is supported for 2-D, 3-D, solids, and shells. A structure containing a crack can be remeshed and the new crack
tip is automatically found after remeshing. The crack opening displacement is saved before the remesh, and applied
CHAPTER 4 119
afterwards to make sure that the results are also correct after remeshing takes place. Remeshing can also be used for
automatic crack propagation as described below.
Automatic Crack Propagation

Marc can automatically grow cracks in structures. The crack propagation functionality is based upon the VCCT procedure.
Cracks can grow in 2-D solids, shells or 3-D solids, using a number of mesh updating methods as described below.
There are two basic modes of growth: fatigue and direct. In fatigue crack propagation, a load sequence is repeated a number
of times. After each sequence, cracks are grown, and optionally, a cycle count is performed. In direct crack propagation, a
transient analysis is performed, and cracks will grow if a given crack growth criterion is fulfilled. The two modes of growth
can also be combined in the same analysis. Then the fatigue sequence is followed, and direct propagation will occur during
the analysis if the growth criterion is fulfilled. This can, for example, be used in a fatigue analysis where the remaining
structure becomes too weak and the final failure is an abrupt crack growth under constant load.
The functionality includes the possibility to determine the shape of a crack front during crack growth. If multiple cracks
are present, growth rate of each crack is scaled based upon the local fracture parameters.
Cracks can grow all the way through a structure and reach a boundary. This can be an external boundary, an internal hole,
or the face of another crack. When a crack tip of a line crack or the entire front of a 3-D crack has reached a boundary, the
crack is made inactive. For a 3-D crack, part of the front may reach a boundary and the crack propagation continues for
the remaining parts of the front.
Crack bifurcation can be modeled in shell structure, where a crack reaches an intersecting shell. The detection of the
intersection and creation of a new crack in the intersecting shell is done automatically. See below for more details.
Fatigue Crack Growth

In a fatigue analysis, a load sequence is typically repeated a number of times and the growth of cracks are determined from
these load repetitions. We can distinguish between low cycle fatigue and high cycle fatigue. In a low cycle fatigue case, the
number of cycles is relatively low, and it may be possible to model every load cycle in the analysis. In high cycle fatigue this
is typically not possible, or even wanted.
There can be different goals of a fatigue analysis. We may want to determine the crack path; i.e., if the crack grows straight
ahead or bends off. We may want to determine the shape of a 3-D crack front during growth. We may have multiple cracks
and want to determine which one grows fastest. The number of load repetitions it will take for a crack to grow a certain
length or reach a boundary is often of interest. Alternatively, we may want to determine how far a crack will grow after a
given number of load repetitions. We may have a spectrum of loads of varying magnitude and want to determine the crack
growth due to a number of repetitions of the load spectrum.
Figure 4-25 Typical Load Spectrum
The basic structure of a fatigue analysis is as follows. The user defines a so-called fatigue time period. This defines the time
period of the fatigue loading. In a typical fatigue analysis, loads are varied periodically (for instance a loading/unloading),
and the fatigue time period is the time period of this repeated load sequence. During the fatigue time period, Marc collects
the smallest and largest energy release rates for each crack front node of each crack. It also saves the estimated crack growth
direction corresponding to the largest energy release rate. At the end of the fatigue time period, we perform a crack growth.
If a cycle count calculation has been requested, we perform this using the collected smallest and largest energy release rates.
Then the next fatigue period starts and the same thing is repeated.
When the crack grows, the mesh is updated using one of the available mesh updating procedures. The Mesh Update Methods
section explains this in more detail. For the crack update, we need to know in which direction each crack tip or front node
will grow, and how much: the crack growth increment. Marc offers three methods for determining the crack growth
increment. Two of them are legacy methods from earlier releases. The first method specifies that the crack growth
increment should be constant. Every node along a 3-D crack front will have the same growth increment. In the second
method, the growth increments are calculated by Paris’ law. This requires that the Paris’ law parameters are selected such
that the calculated growth increments are appropriate for the analysis. This can be difficult to accomplish, in particular for
3-D cracks.
The third method is the default and the recommended procedure. Here we specify a maximum growth increment, which
is scaled between different cracks and along crack fronts. This scaling allows the determination of the shape of the crack
front during growth and gives the relative growth speed among multiple cracks. The specified growth increment should be
carefully selected for the analysis. A too large growth increment may lead to inaccurate results, in particular if there is a
curved crack path. A small growth increment requires a smaller element size in order to properly model the crack. The
scaling of the growth increments is a critical aspect of this procedure. It is based upon the local values of the change in
energy release rate that was collected during the fatigue cycle, G . These values can either be used directly or through the
fatigue law (Paris’ law or a user subroutine). Among all cracks in the model that use fatigue, the front or tip node with the
largest G will use the user-specified crack growth increment. With the largest G denoted G max and the used-defined
crack growth increment a 0 , we have the following expression for the growth increment of any crack node if the direct
options is used:
G n
a =  ----------------- a 0 (4-123)
 G max
Here n is a user-defined exponent.

With the second option the scaling is based on Paris’ law, we have the following expression:
a fat
a = --------------- a 0 (4-124)
a max fat
Here a fat is the growth increment calculated by the fatigue law using the respective G . With the same Paris’ law
parameters for all cracks and if the exponent n is the same as the one used in Paris’ law then the two variants give the same
result. They will give different results if different cracks use different parameters. It is also possible to use the
UCRACK_FATIGUE_LAW user subroutine as an alternative to Paris’ law. Then this routine will determine the scaling of the
growth increments via Equation (4-124).
The scaling of the growth increments can give very small growth increments, which may lead to problems when updating
the mesh. The user can specify a minimum growth increment. If the calculated growth is less than this number, then no
growth will be applied for this node. If no minimum value is provided, then no growth is applied if the calculated growth
increment is less than 10% of the user specified growth increment.
CHAPTER 4 121
In order to ensure that the crack fronts stay smooth, we use a smoothing scheme based upon running averages for the
growth increments along the front. For crack fronts that are split up due to reaching a boundary, the smoothing is applied
to each continuous segment.
So far, we have used the energy release rate as the quantity that controls the scaling of the crack growth increment. As an
alternative, there is an option to use the stress intensity factor, K to scale the crack growth increment. The stress intensity
factor is available for an isotropic linear elastic material and calculated from the following expressions. For plane stress, we
have
K2 K I2I K I2I I
G I = ------I- , G II = -------
- , G I I I =  1 +   ---------- (4-125)
E E E
For plane strain and 3-D:
K I2 K I2I K I2I I
G I =  1 –  2  ------- , G I I =  1 –  2  -------- , G I I I =  1 +   ---------- (4-126)
E E E
The next aspect of a fatigue analysis is the cycle count. If all load cycles are explicitly modeled, then there is no need for
doing a cycle count. The number of cycles is given by the number of fatigue periods done in the analysis. In many cases,
the real number of cycles is substantially larger than the number of cycles modeled in the analysis. The procedure for
calculating the number of cycles is to use a small number of representative fatigue cycles in the analysis, and then use Paris’
law to calculate the real number of cycles. The basic form of Paris’ law is:
da
------- = CK m (4-127)
dN
In our approach, we prescribe the crack length a and want to calculate the number of fatigue cycles N. K typically varies
with a . In simple cases covered by handbook solutions it is possible to obtain K  a  . The handbook solution cannot be
applied for complex structures. The approach here is to assume a piece wise linear variation of K with a and integrate
Paris’ law. At the end of the previous fatigue cycle, we have a crack length a 1 and a range of stress intensity factor K 1 .
At the the end of the current fatigue cycle, we have a 2 and K 2 . We obtain the contribution to the total number of fatigue
cycles as:
a2
da
N =  ----------------
CK m
- (4-128)
a1
with K given as:

a – a2 a – a1
----------------
- K 1 + ----------------
- K 2 (4-129)
a1 – a2 a2 – a1
The actual expression for Paris’ law used in Marc can be selected by the user. There are currently two methods available,
expressed in the change of crack length for a single cycle:
Basic Paris law:
a = C   G max – G m i n  m – G tmh  (4-130)
a = C   K max – K m i n  m – K tmh  (4-131)
Square root Paris law:
a = C   G max – Gm i n  m – Gt h m  (4-132)
a = C   K max – Km i n  m – Kt h m  (4-133)
This formula is based upon an expression given in T. L. Anderson [Ref. 30]. G th and K th are user specified threshold values.
The integration is done as a cycle by cycle count in the general case (including the case where UCRACK_FATIGUE_LAW user
subroutine is used instead of Paris’ law), and as a direct integration if the threshold value is zero. The cycle count is
interrupted if this is estimated to take a very long time. The count is then indicated to be for infinite life. An infinite life
is indicated with the largest integer the program can handle.
The K has so far been generically given as K m a x – K m i n . In case of mixed mode loading the actual expression used is
more complex. While collecting the smallest and largest K values (and G values) during the fatigue time period, the
respective values for each crack mode are stored. They are stored with a sign, where loading in one direction is positive and
loading in the other direction is negative. This is important in order to be able to calculate a proper K . For the respective
crack mode we have: K I , K I I K I I I and we calculate the generic value as
K = K I2 + K II
2 + K 2
III (4-134)
A note on the sign of K and G. Marc keeps track of the sign internally, but the results are always printed as positive. An
exception is for mode I. A negative KI can only occur if the crack faces are overlapping, and this can happen if self-contact
is not taken into account. For this situation Marc reports a negative KI and a zero GI. A negative KI will not impact crack
propagation, only the fatigue cycle count.
So far we have assumed that the load magnitude of the representative fatigue cycle solved by Marc and the load magnitude
for the real load cycle are the same. In many cases the real load consists of a spectrum load with varying load magnitudes.
This can currently not be modeled directly in Marc but it can be done with UCRACK_FATIGUE_LAW user subroutine.
Other important aspects in the fatigue cycle count are effects of crack closure, K min  K max and others. This is currently
not supported in Marc but can easily be coded in UCRACK_FATIGUE_LAW. Crack closure due to penetration of crack surfaces
is typically taken care of by using contact (self contact). What is not included is the effect of crack closure due to crack tip
plasticity (without explicitly modeling plasticity) which is commonly accounted for by the Neuber correction.
Direct Crack Growth

The direct crack growth procedure is quite different from the fatigue procedure. Here, the goal is to determine the load at
which a crack will grow and which path it will take when it grows. The crack may grow in an unstable manner at constant
load, possibly growing through the whole structure. The crack growth may also be limited, so further load increase is
CHAPTER 4 123
possible. A so-called r-curve behavior is also possible to model. This is when the crack growth resistance increases when the
crack length increases.
A central concept of the direct crack propagation is the crack growth criterion. The crack growth criterion is always based
upon the energy release rate G, and not the stress intensity factor.
There are currently five different crack growth criteria available:
1. Total G: The crack will grow if G  G c , where G c is the crack growth resistance (sometimes also called fracture
toughness).
2. Total G calculated with the J-integral: The J-integral calculation is done with the LORENZI option in Marc. For this
case, the user either defines a crack using the J-integral with the same crack tip (or front for 3-D) as used for VCCT,
or this is created internally with default settings (topology search, one rigid region). The value of G used for the
growth criterion is taken from the J-integral results, and the growth direction is determined based upon the VCCT
values.
3. Separately for each mode: Crack growth will take place if either of the following is true: G I  G I c , G I I  G I I c
or G III  G IIIc .
GI n1 GI I n2 GI I I n3
4. Mixed mode criterion 1 (Power Law):  -------- +  ---------- +  -------------  1
 G I c  G I I c  G I I I c
5. Mixed mode criterion 2 (Reeder):

GI + GI I + GI I I
---------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------- 1
GI I + GI I I n1 GI I I  G I I + GI I I  n1

G Ic +  G IIc – G Ic  ---------------------------------------  +  G I I I c – G I I c  -------------------------  ---------------------------------------
 G I + G I I + G I I I GI I + GI I I  G + G + G 
I II III
In addition, it is possible to code a growth criterion using user subroutine UCRACK_FATIGUE_LAW.

The r-curve behavior is modeled by using a table variation of the respective crack growth resistance. The accumulated
growth of each crack node is available as an independent table variable. This quantity can also be post processed.
The distance a crack will grow for a given load is, in general, unknown, and in order to find this distance, we use an iterative
procedure when growing a crack. The user specifies a maximum growth increment (which may be scaled, see below). When
crack growth is detected during an increment, the mesh is updated for the crack growth, and the increment is redone with
the modified mesh. The loading is kept the same only the mesh has been changed. If crack growth is again detected, this
will be repeated. This process is repeated until no further growth is detected, or all growing cracks have reached a boundary
and thus have become inactive. The specified growth increment must be chosen with care. A small value makes it necessary
to use a fine enough mesh. Too large of a value will lead to an overshoot in the determination of the grown distance. It will
also give inaccurate results if the crack path is curved.
Similar to what is done for fatigue, the growth increments between cracks and along crack fronts are scaled based upon the
local energy release values. The following expression is used.
G
a = ------------ a 0 (4-135)
G max
Here, a 0 is the user-specified growth increment and G max is the largest energy release rate among all cracks. This way
we can calculate a variation of crack front shape during growth and model the effect that cracks with different loading will
grow at different speeds.
Mesh Update Methods

When a crack is growing we need to modify the mesh in order to model the new crack. There are currently four methods
available in Marc:
1. Remeshing based
2. Release of glued or tied interfaces
3. Mesh cutting for shells and 2-D solids
4. Mesh split at element edges or faces
The remeshing based growth is the most general option. It is supported for 2-D solids, shells, and 3-D tetrahedral solids.
It allows fully general crack paths and change of crack surfaces. The limitation is that the region where the crack is located
must support global remeshing.
The mesh split method along edges or faces limits the crack growth to follow the current mesh. The main application is
for cases where the crack path is known, for instance when a crack grows at a material interface.
The mesh cutting method modifies the mesh locally in order to update the crack. It allows arbitrary crack paths but the
accuracy may suffer since the mesh tends to become rather irregular.
With the method to release a glued or tied interface you define the crack path explicitly through the bodies which are
connected.
Remeshing based growth

Crack growth by means of remeshing is available for 2-D solid elements, shells and 3-D solid elements. The model must
be set up to do global remeshing, but no specific remeshing criterion needs to be set for the crack propagation. Any specified
remeshing criterion will be treated independently from the remeshings due to crack growth. It is, for example, allowed to
do remeshings during the fatigue load cycle.
When a remeshing occurs in a 2-D analysis, the program first finds the outline of the remeshing body. When a crack is
growing, this outline is extended to form a new crack tip location. In order to assure that we obtain a mesh with an element
edge in the x direction of the local crack tip system, we extend the outline a little further than to the new crack tip. The
new mesh is modified such that the new mesh is correct. This way we avoid meshes as shown in Figure 4-26a. A similar thing
is also done for shells and 3-D solids, so we always get a good mesh at the crack tip after remeshing.
If a crack reaches the boundary of the model, the remeshing body automatically splits. Thus, a crack can grow into another
crack, an internal hole of the body, or the external boundary. Due to the outline extension mentioned above, the crack will
reach the boundary slightly earlier than one would otherwise expect.
The setting of target element length after remeshing will be overridden for the case that it is larger than the growth
increment. This ensures that the mesh resolution is large enough to model the new crack tip. The mesh around the crack
tip is automatically refined when growth occurs. An example of a remeshed 2-D crack is shown in Figure 4-26b. It should
be mentioned that it is currently not possible to specify a finer mesh density at the crack tip than what is given by the growth
CHAPTER 4 125
increment while keeping a coarser mesh outside the crack region. This is because general mesh density control is not yet
available for 2-D.
With shells and 3-D solids, it is recommended to use mesh density control in the remeshing criterion. There is an option
to give a mesh density which depends on the distance to a crack. This allows the specification of an arbitrarily fine mesh at
the crack tip or crack front while keeping a coarser mesh outside the crack region. Figure 4-27 shows an example of mesh
refinement around a crack in a shell example.
a) poor mesh which is avoided during b) refined mesh at current crack tip
crack growth
Figure 4-26 Meshes after Remeshing
Figure 4-27 Example of Mesh Refinement Around a Crack
The procedure for 3-D is different from the 2-D and shell cases. Here we have a crack front which is growing. The entire
front may grow or only part of it, and different parts of the front may grow in different directions. For every crack front
node we have an estimated growth direction and a growth increment defining a growth vector. Depending on the user
input, the growth increments can be the same for each front node, calculated from the fatigue law or obtained by scaling
with the current results. From the growth increments and the growth directions a surface is constructed. This surface is
made smooth enough by dealing with crossing growth vectors, shortening growth vectors that would pass the boundary
and similar. Figure 4-28 shows a simple case of an edge crack. In the left-hand side picture, the green part is the current
surface mesh of the remeshed body, where some elements on the top are taken out for better visibility. The elements in red
describe the crack surface extension due to growth, where here all front nodes are growing the same amount. The part of
the crack surface which is inside the surface mesh will eventually form the new updated crack front. In order to create a
mesh which gives good accuracy, we construct the mesh such that there is a regular mesh around the crack front. The
combined surface and crack extension mesh is modified such that there is a cavity around the updated crack front, as shown
in the middle picture in Figure 4-28. This geometry is meshed, and the cavity is filled with the special mesh used around the
crack front. It is ensured that these meshes are compatible without any need for transition elements of any kind. This is
important for the data mapping procedure, which now only has a continuous mesh to deal with. The final mesh is shown
in the right-hand side of Figure 4-28.
Figure 4-28 Construction of crack extension and special mesh for crack growth.
When there is a distributed load on surfaces adjacent to a crack in a 3-D solid, then this load will be expanded into the
updated crack. If a crack is initiated (see section on Crack Initiation below) at a surface on which there is a distributed load,
the new crack faces will have the same load. If the crack faces of a growing crack has a distributed load, then the updated
crack faces will also have the distributed load.
Growth by releasing tyings or glued contact

For this option, the user needs to specify the crack growth path. The crack can only grow in the direction where there are
tyings or glued contact to release. All crack configurations are supported: 2-D, shells and 3-D.
For the case of user tyings, there are three types of tyings supported: user tying type 100, RBE2 and RROD. Here, the nodes
need to be aligned on both sides of the crack faces. The user specifies only the crack tip node; the program automatically
finds the tyings to release. It does not matter if the tied or retained node is selected as the crack tip node or crack front nodes.
For glued contact, the meshes do not need to match up. As mentioned in the previous section, a special non-matching glue
symmetry case is automatically used if the nodes do not match up. For this case, the tied nodes must be selected as the crack
tip nodes. When nodes in glued contact are released due to crack growth, they switch from glued contact to regular contact.
Thus, a node which is released can still contact the new crack face, but it will not be glued back again.
If higher-order elements are used, the midside node will be released when the corresponding corner node is released.
CHAPTER 4 127
If a crack reaches a model boundary, the crack evaluation for this crack will be turned off and the analysis can continue.
The program avoids that a single node is tied at a boundary; if the last element edge is released, then the last node will also
be released.
If the mesh is designed such that there are multiple choices for which element edge to release, then the one closest to the
estimated crack growth direction is chosen.
The crack front for the 3-D case allows for irregular meshes. The front always follows element edges, but in a mesh with
hexahedral elements, a concave corner node is marked as inactive so the actively used front nodes cross the diagonal of this
element face. This inactive front node becomes active if the neighboring front nodes are released due to growth. Gaps in
the crack front are allowed. If part of the front reaches a boundary (for example, a hole in the structure), then the crack
front will contain a gap and this gap may disappear if the front closes again (for example, if the whole front has passed a
hole in the structure).
Growth by mesh cutting

For this option, the crack can grow arbitrarily through a 2-D or shell mesh without global remeshing. Nodes are shifted
and elements are split up in order to create the cut and to improve the mesh. Only the mesh in the vicinity of the crack is
affected when the crack grows. Lower- and higher-order elements are supported with this option. The data from the
previous mesh is mapped to the new or modified elements similar to the remeshing case, but it is only done in the affected
part of the mesh.
This option is similar to the XFEM approach, but new boundaries are explicitly generated. If contact is used, then the
contact boundary is automatically updated so if self contact is used in the part where growth occurs, then the new crack
faces will use contact and friction (if defined).
The mesh cutting procedure is explained below in the Mesh Splitting Along Edges or Faces section. It is done in two steps. The
first step creates a cut up to the new tip node. In the second step a “scratch” is created ahead of the new tip in order to
ensure an accurate VCCT calculation. This is similar to what is done with remeshing with the extended outline.
The accuracy of the energy release rate calculation is not always adequate due to the often irregular mesh at the crack tip.
Growth along element edges or faces

For this option, the crack can grow along element edges for 2-D and shell elements and along faces for 3-D elements. The
element edge or face closest to the crack growth direction is used. New nodes are automatically inserted and the element
connectivity is changed for elements around the crack tip in order to grow the crack. The new nodes inherit the properties
of the respective original node.
For the 3-D case there is a procedure in place to find a continuous surface based upon the growth direction of each crack
front node.
Both higher- and lower-order elements are supported with this procedure.
Calculation of Crack Growth Direction

The program calculates an estimated crack growth direction for each crack front node. Six methods for defining this are
supported:
1. The maximum principal stress criterion ([Ref. 34])
2. Along the pure mode with largest G i – G i c (where i is mode I, II or III)
3. Along mode I
4. Along a specified vector
5. Stay on interface
6. Mode I and crack normal.
The maximum principal stress criterion states that a crack will grow in the direction normal to the direction of greatest
tension. The crack growth direction  d x ,d y  in the local crack tip system can be expressed in terms of the stress intensity
factors in mode I and II as:
2 2 2
K II KI I KI I KI I KI I
3 -------- + 1 + 8 -------- -------- – 3 -------- 1 + 8 --------
KI
2
KI
2 KI KI KI
2
dx = ---------------------------------------------- dy = ------------------------------------------------------- (4-136)
2 2
K II KI I
1 + 9 -------- 1 + 9 --------
2 2
KI KI
For the 3-D case, mode III is assumed to have the same effect as mode I so K I + K I I I is used instead of K I in the above
equation. It is, thus, assumed that crack growth occurs perpendicular to the crack front tangent.
The option “stay on interface” states that the crack may only grow along a given interface. For growth by releasing tyings
or glued contact, this is the only available option, and the interface is implicitly given by the tyings or contact bodies. It is
also available for growth along element edges or faces. Here, the user specifies an interface between two materials or between
two element sets and growth will only occur along this interface.
Mode I and crack normal is an option for 3-D growth. The user specifies a crack normal which is enforced as the crack
grows. The growth direction for each front node is initialized to the local x direction of the crack (the mode I direction),
and this is then projected onto the plane defined by the given crack normal. It is the responsibility of the user to make sure
that the resulting growth direction gives a valid growth.
The user can also specify the crack growth direction through the UCRACKGROW user subroutine.
Crack Growth into Intersecting Shells

If a shell structure has shell intersections, then a crack can bifurcate at the intersection. The intersecting part can be
connected to the base part (where the crack is) by shared nodes or glued contact. For the glued contact case, the nodes of
the two parts do not need to match up. It is, however, necessary to use the contact option to ignore the shell thickness of
the intersecting shells. Otherwise, it is not possible to obtain the correct glued contact combination at the shell intersection.
The crack bifurcation is done in two steps as shown in Figure 4-29. The first step is when the growing crack would cross the
intersecting shell. At this point, the crack growth increment is decreased so that the updated crack tip ends up exactly at
the intersection. A “scratch” in the mesh is generated in the intersecting shell in order to insure an accurate energy release
rate calculation. The second step is when this crack is determined to grow again. Then the crack in the base part grows as
usual, and a new crack is introduced in the intersection. The length of the new crack is the same is the growth increment
of the original crack and it grows perpendicular to the base part. The new crack inherits the properties of the original crack
but grows independently. Any number of new cracks can be generated this way.
All mesh updating methods for crack growth are available, except release tying/glued contact.
CHAPTER 4 129
If remeshing is used, then the parts need to be connected by glued contact. The current version does not support remeshing
of intersecting shells. If the intersecting part is not set up to use remeshing, then mesh cutting will be used for this part.
All growth direction methods are supported. If the option to grow along a vector is used, then the growth in the intersecting
part will grow along mode I instead. The vector would, in most cases, be an inappropriate direction for the intersecting part.
This procedure is currently not supported by segment-to-segment style contact. Only the node-to-segment method is
supported.
Mesh Before Reaching Intersection Growth Adjusted to Stop at Intersection; New Crack Added in
Scratch Added in Intersecting Part Intersecting Part
Figure 4-29 Creation of New Crack at Shell Intersection
Dynamic Fracture Methodology

In complete similarity with static fracture mechanics concepts, it is assumed that dynamic crack growth processes for linear
materials are governed by the following condition:
K I  t  = R ID  a· T B  a·  0 (4-137)
where K I  t  is the dynamic stress intensity factor for mode I, a· is the crack velocity, and R I D is the dynamic crack
propagation toughness, which is assumed to be a material parameter that in general depends on crack velocity a· ,
temperature T , and specimen thickness B .
The dynamic stress intensity factor depends on crack length ( a ), applied loading (  ), time ( t ), specimen dimensions ( D ),
temperature ( T ), and initial stress fields (  i ) caused by residual stresses or by an initial strain field. The prediction of the
crack propagation history and crack arrest event demands complete knowledge of the R I D vs. a· relation. The Dynamic
Fracture Methodology procedure consists of the following two phases:
1. Generation phase – in this phase, a crack arrest experiment is performed yielding a crack propagation-versus-time
curve. In addition, a numerical simulation of the experiment is carried out by using the measured crack propagation
curve. This is used as input for the numerical model. This allows the calculation of dynamic stress intensity factors
as a function of time. Combination of the latter relation with the measured crack propagation curve results in a
curve, which can be considered the dynamic crack propagation toughness-versus-crack velocity relation.
2. Application phase – in order to predict the crack growth and possible crack arrest point in a structural component,
the inverse problem is solved. Now, the actual stress intensity factors are calculated for the structural component,
that is subjected to a particular loading history, by means of a dynamic finite element analysis. These calculated
values are compared to the fracture toughness curve obtained during the generation phase, Equation (4-137), and
from this the crack growth is predicted.
Dynamic Crack Propagation

The concepts of fracture mechanics discussed in previous sections have been applied to the prediction of crack initiation,
as well as slow stable crack growth of statically loaded structures and for the prediction of fatigue crack growth in cyclically
loaded structures. In problems where inertial effects cannot be ignored, application of quasi-static fracture mechanics
techniques can lead to erroneous conclusions. The use of dynamic fracture mechanics concepts for these problems is clearly
of necessity. The main emphasis on dynamic fracture mechanics is to predict the initiation of stationary cracks in structures,
which are subjected to impact loading. It also focuses on the conditions for the continuous growth of fast propagating
cracks, and on the conditions under which a crack is arrested.
The problem of predicting the growth rate and the possible crack arrest point is quite complicated. It is often treated by
means of a so-called dynamic fracture methodology, which requires the combined use of experimental measurements and of
detailed finite element analyses. An essential step in this approach is formed by the numerical simulation of propagating
cracks by means of the finite element method.
The J-integral as implemented in Marc takes into account the effect of inertial and body forces, thermal and mechanical
loading and initial strains.
The use of the DYNAMIC parameter and the LORENZI model definition option allows for the calculation of dynamic energy
release rates for cracked bodies which are subjected to arbitrary thermal and mechanical loadings including initial stresses.
Crack Initiation
Cracks can be added during the analysis in Marc using the CRACK INIT option. Adding a crack has the effect that the mesh
is modified to create a crack. In addition, one can also activate VCCT definitions at the new crack tips or crack fronts that
are generated due to the crack initiation.
The crack initiation can be activated at the start of the job or at the start of a loadcase. Adding the crack at the start of the
job and at the start of the first loadcase would be the same thing. Allowing a crack to be initiated at the start of a loadcase
allows a pre-loading without a crack followed by a crack initiation and then the rest of the analysis will be done with the
crack.
If the option to add a VCCT definition is used, then the properties of the crack is taken from the template VCCT crack
which is provided by the user. If for instance the template VCCT definition has crack growth defined, then the initiated
crack will have that too.
There are three options for defining the location of the initiated crack as described in the following sections:
Mesh Splitting
The location of the crack is defined by specifying a list of element edges for 2-D and shells, or a list of element faces for 3-
D solids. The mesh is then split up as described in the section Mesh Splitting (Mesh Splitting Along Edges or Faces) below.
CHAPTER 4 131
Mesh Cutting
The location of the crack is defined by a straight line and is available for 2-D and shells. The method used is described in
the section Mesh Splitting (Mesh Cutting) below.
Remeshing
This option is available for 3-D tetrahedral solids. The crack is defined by means of a faceted surface, which is basically a
mesh made out of triangles. The actual crack is defined by the intersection of the faceted surface and the mesh. Figure 4-30
shows an example of a case with a surface crack and a through crack. The picture on the top right-hand side shows an
outline plot in wire frame mode for visualizing the created cracks. In the picture on the lower left-hand side, we see the
refined mesh around the cracks. Lower right shows a close-up view of the mesh at the surface crack. Similar to what is done
for remeshing of growing cracks, Marc automatically generates a special mapped mesh around the new crack fronts. This
special mesh is compatible with the rest of the mesh.
Surface Crack Through Crack
Figure 4-30 Example of Crack Initiation with Faceted Surface
The faceted surface is used as is without any smoothing of curved edges. Hence, one should use a sufficiently fine mesh for
the faceted surface. The actual mesh density obtained in the mesh does not depend on the mesh of the faceted surface.
Any number of faceted surfaces can be used for inserting cracks in the same body. A restriction is that they should not
overlap.
In order to obtain results with sufficient accuracy, it is typically necessary to use a fine mesh near the crack front of the
inserted crack. It is, in general, recommended to use the mesh density control in the remeshing settings in order to get a
refined mesh around the crack front. There is an option to specify a mesh density which depends on the distance to all
cracks in the body. If this option is used, the program will figure out that this should also be applied to the new crack front
coming from a crack initiation with a faceted surface.
Mesh Splitting
Marc has the capability to split up the mesh automatically during the analysis. The splitting can be done in two ways: split
at element edges or faces and by cutting through the mesh.
The first option can be done at nodes and along element edges (2-D and shells) or element faces (3-D). A new node is
created where the split is done. This node inherits the current properties of the original node. The elements at one side of
the split are modified to use the new node to create the opening in the mesh.
The second option cuts through the mesh along a straight line. It is currently supported for 2-D solids and 3-D shells.
Mesh Splitting Along Edges or Faces

The way the mesh splitting works is illustrated in the following. Suppose we have the 2-D mesh shown in Figure 4-31a. The
edges shown in red are targeted for mesh split and this results in the mesh shown in Figure 4-31b. Figure 4-32 illustrates some
different cases and how the mesh is then split. In general, we need to specify edges for the split so we know how to redefine
the element connectivity in order to split the mesh. The examples are shown in 2-D for simplicity of illustration. The mesh
split also applies to shells and 3-D solids. For 3-D solids, it is based upon faces instead of edges.
a b
Figure 4-31 Mesh Split at Red Edges
The mesh splitting technique is used in a number of features in Marc. The VCCT crack propagation along element edges
(see Direct Crack Growth) makes use of this feature.
The mesh can also be split by means of the USPLIT_MESH user subroutine (see Marc Volume D: User Subroutines and Special
Routines for details).
Crack initiation makes use of this option when initiation at faces or edges is used.
The DELAMIN model definition option allows the specification of mesh splitting between regions and within a region using
a stress criterion. The mesh is split if the following criterion is fulfilled:
 m  n
 -----n- +  -----t  1 if   0 (4-138)
 S n  S t n
CHAPTER 4 133
t n
otherwise  --------  1 . (4-139)
 St 
No split possible
Figure 4-32 Examples of Mesh Split

Here  n is the normal stress,  t is the tangential stress and S n , S t , m and n are user-defined parameters. Hence, no
splitting takes place due to compressive stresses. If S n or S t is zero, the respective term is set to zero. It is also possible to
replace the default stress criterion with a user defined criterion via the UDELAM user subroutine (see Marc Volume D: User
Subroutines and Special Routines for details).
The region where the split is considered can be a material, a composite definition or a PSHELL definition. With the option
to split between regions, the program checks all element edges (faces for 3-D solids) in the interface between the specified
regions. The stresses are extrapolated to the nodes and transformed to a coordinate system aligned with the region interface.
See Figure 4-33. The normal direction is perpendicular to the region interface. With the option to split within a region, it
checks all edges/faces inside a region. The normal direction is perpendicular to each element edge or face.
region 1
region 2
t
n
t
Figure 4-33 Coordinate System for Delamination between Regions
For layered elements like shells and solid composites, the delamination criterion is calculated separately for each layer. A
mesh split is triggered if all layers fulfill the delamination criterion. The same criterion is applied to each individual layer.
An alternative approach to model material failure in each layer is the Progressive Failure Analysis approach (PFA) available
through the FAIL DATA option (also see Progressive Composite Failure in Chapter 7 of this manual).
If contact is used, the program recalculates the contact boundary after mesh splitting takes place. This is important in order
to avoid penetration between the newly generated free surfaces.
Nodal post codes are available for postprocessing the separate terms in Equation (4) as well as the sum. The first has a nodal
post quantity labeled Delamination Index (Normal), the second is called Delamination Index (Tangential) and the sum
Delamination Index.
An option is available in the DELAMINATION option for inserting interface elements using a cohesive zone material model
where the mesh is split. This is automatically done whenever a split takes place. The new elements use the nodes of the
surrounding elements and the appropriate type of the interface element is automatically selected. For example, if a split
occurs between two 10-noded tetrahedral elements, then a 15-noded pentahedral element is inserted. If the split-up zone
grows, more interface elements are added. Collapsed interface elements are used where needed, and these elements change
to regular elements if the split zone grows further. There is no corresponding interface element for shells connected edge-
to-edge, so for this case no interface elements are added.The cohesive material properties of the interface elements must be
defined in the model. The ELEMENTS parameter defining the element type for the interface elements must be included.
The option is supported for lower- and higher-order 2-D solid, shell and 3-D solid elements. For the case of higher-order
elements, only the corner nodes are considered, and when corner nodes are split, so are the involved midside nodes. The
midside nodes are assigned the average delamination index value of the corner nodes for postprocessing.
Mesh Cutting
With this option, the mesh is modified by making a cut along a line. Nodes are shifted and elements subdivided in order
to create the cut. The cut is done along a straight line and the starting and ending points can be either inside or outside the
mesh. The shifting of nodes is done to avoid thin elements as far as possible. Results are transferred from the original mesh
to the modified mesh using the rezoning capabilities in Marc. This data transfer is only done in the affected region, as
opposed to remeshing where this is done in the whole remeshing body.
CHAPTER 4 135
The elements in the region of the cut must be of the same type. It is, for example, not possible to combine triangles and
quadrilaterals. This is due to a limitation in the rezoning. If triangles are needed together with quadrilateral elements, the
triangles should be defined as collapsed quadrilaterals.
The updated Lagrange procedure must be used in the element group where mesh cutting takes place.
Mesh cutting is currently supported for 2-D continuum elements that support the updated Lagrange procedure and shell
elements except solid shell elements. Both lower- and higher-order elements are supported.
The mesh splitting technique is used in a number of features in Marc. The VCCT crack propagation by mesh cutting makes
use of this feature (see Growth by mesh cutting). The mesh can also be split by means of the UCUT_MESH user subroutine (see
Marc Volume D: User Subroutines and Special Routines for details). Some examples of mesh cutting can be found in the description
of the UT_SPLIT2D utility routine in Marc Volume D: User Subroutines and Special Routines.
Dynamics
Marc’s dynamic analysis capability allows you to perform the following calculations:
 Modal (Eigenvalue) Analysis
 Harmonic Response
 Spectrum Response
 Transient Analysis
The program contains two methods for eigenvalue extraction and three time integration operators. Nonlinear effects,
including material nonlinearity, geometric nonlinearity, and boundary nonlinearity, can be incorporated.
Linear problems can be analyzed using modal superposition or direct integration. All nonlinear problems should be
analyzed using direct integration methods.
In addition to distributed mass, you can also attach concentrated masses associated with each degree of freedom of the
system. You can include damping in either the modal superposition or the direct integration methods. You can also include
(nonuniform) displacement and/or velocity as an initial condition, and apply time-dependent forces and/or displacements
as boundary conditions.
Modal (Eigenvalue) Analysis

Marc uses either the inverse power sweep method or the Lanczos method to extract eigenvalues (natural frequencies) and
eigenvectors (mode shapes). The DYNAMIC parameter is used to determine which procedure to use, and how many modes
are to be extracted. The inverse power sweep method is typically used for extracting a few modes while the Lanczos method
is optimal for several modes. After the modes are extracted, they can be used in a transient analysis or spectrum response
calculation.
In dynamic eigenvalue analysis, we find the solution to an undamped linear dynamics problem:
 K –  2 M  = 0
where K is the stiffness matrix, M is the mass matrix,  are the eigenvalues (frequencies) and  are the eigenvectors. In
Marc, if the extraction is performed after increment zero, K is the tangent stiffness matrix, which can include material and
geometrically nonlinear contributions. The mass matrix is formed from both distributed mass and point masses.
Inverse Power Sweep

Marc creates an initial trial vector. To obtain a new vector, the program multiplies the initial vector by the mass matrix and
the inverse (factorized) stiffness matrix. This process is repeated until convergence is reached according to either of the
following criteria: single eigenvalue convergence or double eigenvalue convergence. In single eigenvalue convergence, the
program computes an eigenvalue at each iteration. Convergence is assumed when the values of two successive iterations are
within a prescribed tolerance. In double eigenvalue convergence, the program assumes that the trial vector is a linear
combination of two eigenvectors.
Using the three latest vectors, the program calculates two eigenvalues. It compares these two values with the two values
calculated in the previous step; convergence is assumed if they are within the prescribed tolerance.
When an eigenvalue has been calculated, the program either exits from the extraction loop (if a sufficient number of vectors
has been extracted) or it creates a new trial vector for the next calculation. If a single eigenvalue was obtained, Marc uses
the double eigenvalue routine to obtain the best trial vector for the next eigenvalue. If two eigenvalues were obtained, the
program creates an arbitrary trial vector orthogonal to the previously obtained vectors.
After Marc has calculated the first eigenvalue, it orthogonalizes the trial vector at each iteration to previously extracted
vectors (using the Gram-Schmidt orthogonalization procedure). Note that the power shift procedure is available with the
inverse power sweep method.
To select the power shift, set the following parameters:
 Initial shift frequency – This is normally set to zero (unless the structure has rigid body modes, preventing a
decomposition around the zeroth frequency).
 Number of modes to be extracted between each shift – A value smaller than five is probably not economical
because a shift requires a new decomposition of the stiffness matrix.
 Auto shift parameter – When you decide to do a shift, the new shift point is set to
Highest frequency2 + scalar x (highest frequency - next highest frequency)2
You can define the value of the scalar through the MODAL SHAPE option.
The Lanczos Method

The Lanczos algorithm converts the original eigenvalue problem into the determination of the eigenvalues of a tri-diagonal
matrix. The method can be used either for the determination of all modes or for the calculation of a small number of
modes. For the latter case, the Lanczos method is the most efficient eigenvalue extraction algorithm. A simple description
of the algorithm is as follows.
Consider the eigenvalue problem:
–2 M u + K u = 0 (4-140)
Equation (4-140) can be rewritten as:
1
------ M u = M K – 1 M u (4-141)
2
Consider the transformation:
u = Q  (4-142)
CHAPTER 4 137
T
Substituting Equation (4-142) into Equation (4-141) and premultiplying by the matrix Q on both sides of the equation,
we have
1
------ Q T M Q  = Q T M K – 1 M Q  (4-143)
2
The Lanczos algorithm results in a transformation matrix Q such that:
QT M Q = I (4-144)
Q T M K – 1 MQ = T (4-145)
where the matrix T is a symmetrical tri-diagonal matrix of the form:
1 2 0
2 2 3
T = (4-146)
m
0 m m
Consequently, the original eigenvalue problem, Equation (4-141), is reduced to the following new eigenvalue problem:
1
------  = T  (4-147)
2
The eigenvalues in Equation (4-147) can be calculated by the standard QL-method.
Through the MODAL SHAPE option, you can either select the number of modes to be extracted, or a range of modes to be
extracted. The Sturm sequence check can be used to verify that all of the required eigenvalues have been found. In addition,
you can select the lowest frequency to be extracted to be greater than zero. If modes of an unconstrained model have to be
extracted including the rigid body modes, a negative lowest frequency should be entered.
The Lanczos procedure also allows you to restart the analysis at a later time and extract additional roots. It is unnecessary
to recalculate previously obtained roots using this option.
Convergence Controls
Eigenvalue extraction is controlled by:
The maximum number of iterations per mode in the power sweep method; or the maximum number of iterations for all
modes in the Lanczos iteration method.
Modal Stresses and Reactions

After the modal shapes (and frequencies) are extracted, the RECOVER history definition option allows for the recovery of
stresses and reactions at a specified mode. This option can be repeated for any of the extracted modes. The stresses are
2
computed from the modal displacement vector  ; the nodal reactions are calculated from F = K –  M . The nodal
vector of modal mass is calculated as m = M . The RECOVER option is also used to place eigenvectors on the post file.
Participation Factors and Effective Modal Masses

The participation factor for a given mode is defined as
c nj =    n i  T   M   T i j        n i  T   M     n i   (4-148)
where
c n j is the participation factor for mode n in the jth direction;
  n i  is the eigenvector value for mode n and degree of freedom i;
M is the mass matrix; and
 T i j  defines the magnitude of the rigid body response of degree of freedom i to imposed rigid body motion in the
jth direction and takes the following form
 1 0 0 0  Z – Z0  – Y – Y0  
 
 0 1 0 – Z – Z0  0  X – X0  
 
 0 0 1  Y – Y0  – X – X0  0  e 
  j
 0 0 0 1 0 
 
 0 0 0 0 1 0 
 0 0 0 0 0 1 
where
X , Y , and Z are the coordinates of the respective node;

X 0 , Y 0 , and Z 0 are the coordinates of center of rotation; and
ej is the unit vector (carrying 1 for row j and the rest being zeros).
The effective modal masses are calculated as squares of the participation factors.
ef f
m n ,j =  c n ,j  2
eff
where m n ,j is the effective modal mass for mode n in the jth direction.
CHAPTER 4 139
While the nodal vector of modal masses gives the significance of mass participation of the node for the given mode in the
given direction, the effective modal mass gives an idea about the mass contribution of the whole structure (or model) for
the mode in the given direction.
Harmonic Response
Harmonic response analysis allows you to analyze structures vibrating around an equilibrium state. This equilibrium state
can be unstressed or statically prestressed. Statically prestressed equilibrium states can include material and/or geometric
nonlinearities. You can compute the damped response for prestressed structures at various states.
In many practical applications, components are dynamically excited. These dynamic excitations are often harmonic and
usually cause only small amplitude vibrations. Marc linearizes the problem around the equilibrium state. If the equilibrium
state is a nonlinear, statically prestressed situation, Marc considers all effects of the nonlinear deformation on the dynamic
solution. These effects include the following:
 Initial stress
 Change of geometry
 Influence on constitutive law
The vibration problem can be solved as a linear problem using complex arithmetic.
The analytical procedure consists of the following steps:
1. Marc calculates the response of the structure to a static preload (which can be nonlinear) based on the constitutive
equation for the material response. In this portion of the analysis, the program ignores inertial effects.
2. Marc calculates the complex-valued amplitudes of the superimposed response for each given frequency, and
amplitude of the boundary tractions and/or displacements. In this portion of the analysis, the program considers
both material behavior and inertial effects.
3. You can apply different loads with different frequencies or change the static preload at your discretion. All data
relevant to the static response is stored during calculation of the complex response.
To initiate a harmonic response analysis, use the HARMONIC parameter. To define the excitation frequency, use the HARMONIC
history definition option. If you enter the HARMONIC history definition option with a set of incremental data, Marc assumes
those incremental data apply only for the harmonic excitation. This is true for applied boundary conditions as well as loads.
The small amplitude vibration problem can be written with complex arithmetic as follows
 K + iD –  2 M u = P (4-149)
where,
u = u re + iu im is the complex response vector,

P = P re + iP im is the complex load vector,
i = –1 ,
 is the excitation frequency.
The notation is further defined below:
K = K el + K s p (4-150)
where,
Ke l are element stiffness matrices,

K sp are the spring stiffness matrices.
M = M el + M mp (4-151)
where,
Me l are element mass matrices,

M mp are mass point contributions.
2
D = D el + D d +   M +   + ----- K (4-152)
   
where,
De l are element damping matrices,

Dd are damper contributions,
 is the mass damping coefficient,
 is the stiffness damping coefficient,
 is the numerical damping coefficient.
, ,  can be a function of the frequency and the temperature by referencing a table
If all external loads and forced displacements are in phase and the system is undamped, this equation reduces to
 K –  2 M u re = P re (4-153)
which could be solved without activating the complex arithmetic on the HARMONIC parameter.
The values of the damping coefficients (  ,  ,  ) are entered via the DAMPING model definition option. The spring and
damper contributions are entered in either the SPRINGS, PFAST, or PBUSH model definition options and mass points are
specified in the MASSES, CONM1, or CONM2 model definition options.
The element damping matrix ( D e l ) can be obtained for any material with the use of a material damping matrix which is
specified in the UCOMPL user subroutine. You specify the material response with the constitutive equation.
·
 = B + C (4-154)
where B and C can be functions of deformation and/or frequency.

CHAPTER 4 141
The global damping matrix is formed by the integrated triple product. The following equation is used:
T
D =    C dV el (4-155)
el v el
where  is the strain-displacement relation.
Similarly, the stiffness matrix K is based on the elastic material matrix B .

A special application of the harmonic excitation capability involves the use of the elastomeric analysis capability in Marc.
Here, the Mooney formulation (used in conjunction with the various Herrmann elements) is used to model the stress-strain
behavior of the elastomeric compound. In Marc, the behavior is derived from the third order invariant form of the strain
energy density function.
W  I 1 I 2  = C 10  I 1 – 3  + C 01  I 2 – 3  + C 11  I 1 – 3   I 2 – 3 
(4-156)
+ C 20  I 1 – 3  2 + C 30  I 1 – 3  3
with the incompressibility constraint

I3 = 1 (4-157)
where I 1 , I 2 , and I 3 are the invariants of the deformation. For the harmonic excitation, the constitutive equation has the
specific form
S i j =  D ijkl + 2i i j k l E k l (4-158)
with D ijkl as the quasi-static moduli following form the Mooney strain energy density function and
 i j k l =  0  C ik–1 C –1 + C –1 C –1  +    C –1 + C –1   + (4-159)
jl il jk 1 ik jl il jk
 2  C i k C jl– 1 + C il– 1 C jk  +  10  i j C k–l1 +  11  i j  k l +
 12  i j C kl +  20 C ij C k–l1 +  21 C i j  k l +  22 C i j C k l
The output of Marc consists of stresses, strains, displacements and reaction forces, all of which may be complex quantities.
The strains are given by
 =  u (4-160)
and the stresses by
·
 = B + C  (4-161)
The reaction forces are calculated with
R = Ku –  2 Mu + iD e l u + iD d u (4-162)
2
where –  Mu is only included if requested on the HARMONIC parameter.
The printout of the nodal values consists of the real and imaginary parts of the complex values, but you can request that
the amplitude and phase angle be printed. You do this with the PRINT CHOICE model definition option.
Spectrum Response
The spectrum response capability allows you to obtain maximum response of a structure subjected to known spectral base
excitation response. This is of particular importance in earthquake analysis and random vibration studies. You can use the
spectrum response option at any point in a nonlinear analysis and, therefore, ascertain the influence of material nonlinearity
or initial stress.
The spectrum response capability technique operates on the eigenmodes previously extracted to obtain the maximum
nodal displacements, velocities, accelerations, and reaction forces. You can choose a subset of the total modes extracted by
either specifying the lowest n modes or by selecting a range of frequencies. The modes can also be selected by specifying a
mass percentage cut-off value.
The spectral density input can be given with respect to frequency and damping ratio. The damping ratio is only applicable
with the Complete Quadratic Combination (CQC) method of mode combination. The spectral density input can be given
as displacement, velocity or acceleration. Stresses can be requested as output.
The participation factor for a given mode is defined as
c nj =    n i  T   M   T i j        n i  T   M     n i   (4-163)
where
c n j is the participation factor for mode n in the jth direction;
  n i  is the eigenvector value for mode n and degree of freedom i;
M is the mass matrix; and
 T i j  defines the magnitude of the rigid body response of degree of freedom i to imposed rigid body motion in the
jth direction and takes the following form
 1 0 0 0  Z – Z0  – Y – Y0  
 
 0 1 0 – Z – Z0  0  X – X0  
 
 0 0 1  Y – Y0  – X – X0  0  e 
  j
 0 0 0 1 0 
 
 0 0 0 0 1 0 
 0 0 0 0 0 1 
where
X , Y , and Z are the coordinates of the respective node

X 0 , Y 0 , and Z 0 are the coordinates of center of rotation which are given in the MODAL SHAPE option
ej is the unit vector (carrying 1 for row j and the rest being zeros)
CHAPTER 4 143
Options for Combination of Modes

The options for combination of modes are described below:
 Square Root of Sum of Squares (SRSS) Method:
N 12
R pI =   Rp i  2
i = 1
where
R pI is the combined periodic response for the Ith component of spatial input (I=1,2,3),
R pi is the periodic component of response of mode i, and
N is the number of modes considered in the combination of modal responses.
 Sum of Absolute values (ABS) Method:
N
R pI =  R pi
i = 1
 Naval Research Laboratory (NRL) Method [Ref. 52]:
N 12
R pI = R pi +   Rp i  2 for all i = j
i = 1
 Complete Quadratic Combination (CQC) Method:
N N 12
R pI =   i j Rp i Rp j
i = 1j = 1
where
 8   i  j  1  2   i + r j r 3  2  is the cross-modal coefficient

 ij = --------------------------------------------------------------------------------------------------------------
-
  1 – r 2  2 + 4 i  j r  1 + r 2  + 4   i2 +  j2 r 2 
i is the damping ratio for mode i,
j is the damping ratio for mode j, and
r the frequency ratio  j   i .
Options for Combination of Spatial Response

The two ways of combining the spatial components viz. the SRSS method and the linear combination method are
explained below.
 Square Root of Sum of Squares (SRSS) Method
3 12
R =   RI  2
I = 1
where
R is the combined response

RI is the Ith spatial component of response
 Linear Combination Method
3
R =  LI RI
I = 1
where L I is the combination coefficient of Ith spatial direction
 Weighted SRSS Method

This method is supported to maintain backward compatibility with older (than 2013) versions of Marc and is not
recommended.
ndeg 12
R =  CI  RI  2
I = 1
where
R is the combined response
RI is the Ith degree of freedom
CI is the coefficient of Ith degree of freedom
Note: Cumulative mass cutoff, modal combination methods, rigid response, and missing mass
response are not supported with the weighted SRSS method.
Missing Mass Response

The missing mass method is used to calculate the rigid response of the structure is explained below.
1. For each degree of freedom included in the periodic response calculation, the fraction of the degree of freedom mass
is determined as follows:
N
di =    c n j    n i  
I = 1
where,
CHAPTER 4 145
n is the mode number (1,2,3,…N)

N is the number of modes considered
 n i is the eigenvector value for mode n and degree of freedom i
j is the spatial input direction

c n j is the participation factor for mode n in the jth direction
2. The fraction of the degree of freedom mass not included in the summation of the modes is evaluated as follows:
e i = d i =  ij
3. The pseudo-static inertial forces associated with the summation of all higher modes for each degree of freedom i
are given by:
P i = S ai  M i   e i 
where
Pi is the force or moment to be applied at the degree of freedom i
S ai is the spectral acceleration of the cut-off frequency
Mi is the mass associated with degree of freedom i
ei is the fraction of mass not included
4. A static analysis is performed to obtain the missing mass response of the system by applying the above loads on the
model.
Rigid Modal Response

The rigid response of the analysis will be done as follows:
1. Gupta method is used for evaluating the rigid component of the modal response.
2. The rigid response component is defined as:
R ri =  i R i
where,
Ri is the modal response, and
i is the rigid response coefficient which is evaluated as
 i =  ln  f i  f l     ln  f 1  f 2  
for all f 1  f i  f 2 and takes the value of zero when f i  f 1 and value of 1.0 when f i  f 2 ;
while f 1 and f 2 are the bounding frequencies which are defined as follows:
f 1 = S a  m ax    2S v  m a x  
where
S a  max  is the maximum spectral acceleration
S v  max  is the maximum spectral velocity
f2 is taken as the frequency of the last mode, which is the highest frequency of all the modes
considered.
3. The periodic response component is defined as:
R pi =  1 –  i2 R i
The response from the “Missing mass method” will be combined in the following means:
n
R rI =   R ri  + R m m I
i = 1
2 2
R I =  R pI + R rI  1  2
where R mmI is the missing mass response for the Ith component of spatial input (I=1,2,3).
Note: The internal forces are identified as reaction forces on the post file.
Transient Analysis
Transient dynamic analysis deals with an initial-boundary value problem. In order to solve the equations of motion of a
structural system, it is important to specify proper initial and boundary conditions. You obtain the solution to the equations
of motion by using either modal superposition (for linear systems) or direct integration (for linear or nonlinear systems).
In direct integration, selecting a proper time step is very important. For both methods, you can include damping in the
system.
The following sections discuss the seven aspects of transient analysis listed below.
 Modal Superposition
 Direct Integration
 Time Step Definition
 Initial Conditions
 Time-Dependent Boundary Conditions
 Mass Matrix
 Damping
CHAPTER 4 147
Modal Superposition
The modal superposition method predicts the dynamic response of a linear structural system. In using this method, we
assume that the dynamic response of the system can be expressed as a linear combination of the mode shapes of the system.
For the principle of superposition to be valid, the structural system must be linear. Damping can be applied to each mode
used in the superposition procedure.
To select the eigenvalue extraction method and the number of modal shapes, use the DYNAMIC parameter. To select a fraction
of critical damping for each mode, use the DAMPING model definition option. The DYNAMIC CHANGE history definition option
can be used to select the time step.
Marc obtains the transient response on the basis of eigenmodes extracted. The number of modes extracted (rather than the
choice of time step) governs the accuracy of the solution.
Consider the general linear undamped problem
M u·· + K u = f  t  (4-164)
and suppose that n eigenvectors  1   n are known.
u =  i u i  t  (4-165)
The eigenmodes are orthogonal with respect to the M and K matrices. After substitution in Equation (4-164), we find a
set of uncoupled dynamic equations
m i u··i + k i u i = f i  t  (4-166)
where:
m i =  iT M i (4-167)
k i =  iT K i (4-168)
f i  t  =  iT f  t  (4-169)
We can introduce damping on each mode as a fraction of critical damping (  i ) for that mode. We can rewrite Equation
(4-166) in the form
m i u··i + 2m i  i  i u· i + k i u i = f i  t  (4-170)
or if we normalize  i such that m i = 1 :
u··i + 2 i  i u· i +  i2 u i = f i  t  (4-171)
The response of a particular mode is then given by the solution of Equation (4-171) which is:
t
ui  t  =  f i   h  t –   d (4-172)
0
with:
1 –i i t
h  t  = ------- e sin   id t  t0
 id
ht = 0 t0
 id =  1 –  i2    i
The evaluation of the Duhamel integral, Equation (4-172), can be performed exactly if the load changes linearly in a specific
time increment. Hence, if the load changes rapidly in a specific time period, small load steps have to be taken.
In case initial displacements u 0 or initial velocities u· 0 are given a transformation to the reduced modal system is needed
for those conditions:
u i0 =  iT  M  u 0 and u· i0 =  iT  M  u· 0
The solution of these initial conditions which must be added to the response given in Equation (4-172) is:
–i i t u· i0 + u i0  i  i
ui  t  = e -------------------------------  sin   id  t  + u i0  cos   id  t  (4-173)
 id
The initial accelerations due to given initial displacements u 0 and initial velocities u· 0 can be obtained by differentiation
of Equation (4-173):
u··i0 = –   i2  u i0  –  2   i   i  u· i0 
Direct Integration
Direct integration is a numerical method for solving the equations of motion of a dynamic system. It is used for both linear
and nonlinear problems. In nonlinear problems, the nonlinear effects can include geometric, material, and boundary
nonlinearities. For transient analysis, Marc offers four direct integration operators listed below.
 Newmark-beta Operator
 Houbolt Operator
 Generalized-Alpha Operator
 Central Difference Operator
To select the direct integration operator, use the DYNAMIC parameter. Specify the time step size through the DYNAMIC CHANGE
or AUTO STEP option. Direct integration techniques are imprecise; this is true regardless of which technique you use. Each
technique exhibits at least one of the following problems: conditional stability, artificial damping, and phase errors.
Newmark-beta Operator
This operator is probably the most popular direct integration method used in finite element analysis. For linear problems,
it is unconditionally stable and exhibits no numerical damping. The Newmark-beta operator can effectively obtain
solutions for linear and nonlinear problems for a wide range of loadings. The procedure allows for change of time step, so
CHAPTER 4 149
it can be used in problems where sudden impact makes a reduction of time step desirable. This operator can be used with
adaptive time step control. Although this method is stable for linear problems, instability can develop if nonlinearities
occur. By reducing the time step and/or adding (stiffness) damping, you can overcome these problems.
Houbolt Operator
This operator has the same unconditional stability as the Newmark-beta operator. In addition, it has strong numerical
damping characteristics, particularly for higher frequencies. This strong damping makes the method very stable for
nonlinear problems as well. In fact, stability increases with the time step size. The drawback of this high damping is that
the solution can become inaccurate for large time steps. Hence, the results obtained with the Houbolt operator usually have
a smooth appearance, but are not necessarily accurate. The Houbolt integration operator, implemented in Marc as a fixed
time step procedure, is particularly useful in obtaining a rough scoping solution to the problem.
Single Step Houbolt Operator

Two computational drawbacks of the Houbolt operator are the requirement of a special starting procedure and the
restriction to fixed time steps. In [Ref. 28], a Single Step Houbolt procedure has been presented, being unconditionally
stable, second order accurate and asymptotically annihilating. In this way, the algorithm is computationally more
convenient compared to the standard Houbolt method, but because of its damping properties, the time steps have to be
chosen carefully. This algorithm is recommended for implicit dynamic contact analyses.
Generalized Alpha Operator

One of the drawbacks of the existing implicit operators is the inability to easily control the numerical dissipation. While
the Newmark-Beta method has no dissipation and works well for regular vibration problems, the Single-Step Houbolt
method has numerical dissipation and works well for implicit dynamic contact problems. A single scheme that easily allows
zero/small dissipation for regular structural dynamic problems and high-frequency numerical dissipation for dynamic
contact problems is desirable. In [Ref. 35], a Generalized-alpha method has been presented as an unconditionally stable,
second-order algorithm that allows user-controllable numerical dissipation. The dissipation is controlled by choosing either
the spectral radius S of the operator or alternatively, two parameters fand m. The choice of the parameters provides a
family of time integration algorithms that encompasses the Newmark-Beta, Single-Step Houbolt and the Hilber-Hughes-
Taylor time integration methods as special cases.
Central Difference Operators

These explicit operators for IDYN = 4 and IDYN = 5 are only conditionally stable. The program automatically calculates
the maximum allowable time step. This method is not very useful for shell or beam structures because the high frequencies
result in a very small stability limit. This method is particularly useful for analysis of shock-type phenomena. In this
procedure, since the operator matrix is a diagonal mass matrix, no inverse of operator matrix is needed. However, this fact
also implies that you cannot use this method in problems having degrees of freedom with zero mass. This restriction
precludes use of the Herrmann elements, gap-friction elements, the pipe bend element, shell elements 72 and 89 and beam
elements 76 and 77. These shell and beam elements are precluded because they have a rotational degree of freedom, which
do not have an associated mass. The mass is updated only if the LARGE STRAIN or UPDATE parameter or the CONTACT option
is used. The elastomer capability can be used with explicit dynamics in an updated Lagrange framework where the pressure
variables are condensed out before going into the solver.
Consider the equations of motion of a structural system:
Ma + Cv + Ku = F (4-174)
where M , C , and K are mass, damping, and stiffness matrices, respectively, and a , v , u , and F are acceleration, velocity,
displacement, and force vectors. Various direct integration operators can be used to integrate the equations of motion to
obtain the dynamic response of the structural system. The technical background of the three direct integration operators
available in Marc is described below.
Newmark-beta Operator
The generalized form of the Newmark-beta operator is
u n + 1 = u n + tv n +  1  2 –  t 2 a n + t 2 a n + 1 (4-175)
v n + 1 = v n +  1 –  ta n + ta n + 1 (4-176)
n
where superscript denotes a value at the nth time step and u , v , and a take on their usual meanings.
The particular form of the dynamic equations corresponding to the trapezoidal rule
 = 12,  = 14
results in
4
 ------- 2 n 4
- M + ----- C + K u = F n + 1 – R + M  a n + ----- v n + Cv n (4-177)
 t 2 t   t 
where the internal force R is
R =   T dv (4-178)
V
Equation (4-177) allows implicit solution of the system
u n + 1 = u n + u (4-179)
Notice that the operator matrix includes K , the tangent stiffness matrix. Hence, any nonlinearity results in a reformulation
of the operator matrix. Additionally, if the time step changes, this matrix must be recalculated because the operator matrix
also depends on the time step. It is possible to change the values of  and  through the PARAMETERS option.
Houbolt Operator
The Houbolt operator is based on the use of a cubic fitted through three previous points and the current (unknown) in
time. This results in the equations
11 3 1
v n + 1 =  ------ u n + 1 – 3u n + --- u n – 1 – --- u n – 2   t (4-180)
6 2 3 
CHAPTER 4 151
and
a n + 1 =  2u n + 1 – 5u n + 4u n – 1 – u n – 2   t 2 (4-181)
Substituting this into the equation of motion results in

2
 ------- 11 1
- M + --------- C + K u = F n + 1 – R n + --------  3u n – 4u n – 1 + u n – 2 M +
 t 2 6t  t 2
(4-182)
1 7 3 1
-----  --- u n – --- u n – 1 + --- u n – 2 C
t  6 2 3 
This equation provides an implicit solution scheme. By solving Equation (4-179) for u , you obtain Equation (4-183), and
n+1 n+1
so obtain v and a .
u n + 1 = u n + u (4-183)
Equation (4-182) is based on uniform time steps – errors occur when the time step is changed. Also, a special starting
n–1 n–2
procedure is necessary since u and u appear in Equation (4-182).
Single Step Houbolt Operator

The Single Step Houbolt operator according to [Ref. 28] starts with the following equilibrium equation and expressions for
the velocity and acceleration:
m1 n+1 c1 n+1 k1 n+1 m n c n k n
 Ma +  Cv + Ku +  Ma +  Cv +  Ku =
(4-184)
f1 n + 1 f n
 F +aF
n+1 n n 2 n 1 2 n+1
u = u + tv + t a +  t a (4-185)
n+1 n n 1 n+1
v = v + ta +  ta (4-186)
Notice that in contrast to the Newmark and the standard Houbolt method, the equilibrium equation also contains terms
m1
corresponding to the beginning of the increment. Without loss of generality, the parameter  can be set to 1. Based on
asymptotic annihilation and second order accuracy, the remaining parameters can be shown to fulfill:
k 1 1 m k1 1
 = 0,  = ,  = + , = –1  2 ,  = 1  2 ,
c 1 12 c1 1 12 f k f1 k1
 = –  2 +    4 , =  2 + 3   4 , =  , = 
In this way, the number of unknown parameters has been reduced to two. Based on a Taylor series expansion of the
1 1
displacement about the nth time step,  and  should be related by  +  = 1  2 , which finally yields
1 1
 = 1  2  1  2 –   . According to [Ref. 28],  should be set to 3/2 (with  = – 1  2 ) to minimize the velocity error
1
and to 1/2 (with  = 0 ) to avoid velocity overshoot. The default values in Marc are  = 3  2 and  = – 1  2 , but
1
the user can modify  and  using the PARAMETERS model definition and history definition option.
Substitution of the velocity and acceleration into the equilibrium equation results in:
c1 1
 1    n+1 n
 -----------------------
1 2 k1
- M + ---------------------- C + K u = F
1 k1
– Ku + (4-187)
  t   t 
m
1 n 2 n  n
------------------------ M  tv + t a  – --------- Ma –
1 2 k1 k1
 t  
 c1  n 1 c
n   2   n
--------
- C  v + ta – ------------  tv n + t a n   – --------- Cv
k1 1 k1
   t   
Generalized Alpha Operator

From [Ref. 35], the equilibrium equations for the generalized alpha method can be expressed in the form
n + 1 + m n + 1 + f n + 1 + f n + 1 + f
Ma + Cv + Ku = F (4-188)
where
n + 1 + f n+1
u =  1 +  f u – f u n (4-189)
n + 1 + f n+1 n
v =  1 +  f v – f v (4-190)
n + 1 + m n+1 n
a =  1 +  m a – m a (4-191)
The displacement and velocity updates are identical to those of the Newmark algorithm
u n + 1 = u n + tv n +  1  2 –  t 2 a n + t 2 a n + 1 (4-192)
v n + 1 = v n +  1 –  ta n + ta n + 1
where, as shown in [Ref. 35], optimal values of the parameters  and  are related to  f and  m by
1 2
 = ---  1 +  m –  f  (4-193)
4
1
 = --- +  m –  f (4-194)
2
CHAPTER 4 153
In Equation (4-188),  f =  m =0 gives rise to the Newmark-beta method,  m =0, – 0.33   f  0 gives rise to the HHT
- a method [Ref. 36], and  f =0, 0   m  1 gives rise to the WBZ - a method [Ref. 37].
It is seen that the  f and  m parameters can be used to control the numerical dissipation of the operator. A simpler
measure is the spectral radius S. This is also a measure of the numerical dissipation; a smaller spectral radius value
corresponds to greater numerical dissipation. The spectral radius of the generalized alpha operator can be related to the  f
and  m parameters as follows
S
 f = – ------------- (4-195)
1+S
1 – 2S
 m = ---------------- (4-196)
1+S
S varies between 0 and 1. Accordingly, the ranges for the  f and  m parameters are given by – 0.5   f  0.0 and
– 0.5   m  1 .  = -0.5,  m = -0.5 corresponds to a spectral radius of 1.0 and  f = 0,  m = 1 corresponds to a spectral
f
radius of 0.0. By substituting the values in equation (5-145) and comparing with Equation (4-174), it can be noted that the
special case of S = 0 (  f = 0,  m = 1) for the generalized alpha operator is identical to the default single step houbolt
1
operator with  = 3  2 and  = – 1  2 . It can also be noted that the case of S = 1 has no dissipation and corresponds
to a mid-increment Newmark- beta operator.
In the Marc input file, the generalized alpha operator is flagged through the option 8 on the second field of the DYNAMIC
parameter. Options to define the associated generalized alpha parameters are available through the DYNAMIC parameter and
the PARAMETERS model definition and history definition options. On the DYNAMIC parameter, a default value of ‘0’ on the 8th
field indicates contact-optimized parameters (  f = 0,  m = 1) and a ‘1’ on the 8th field indicates non-contact optimized
parameters (  f = -0.05,  m = 0). ‘0’ would be recommended for dynamic contact problems or for any problem where
high-frequency numerical dissipation is desired. ‘1’ would be recommended for noncontact structural dynamic problems
and free-vibration problems where a small amount of numerical dissipation is desired. On the PARAMETERS model
definition and history definition options, the 4th, 5th, and 6th fields of the 5th data block are reserved for  f ,  m and S,
respectively. S is the main control parameter: 0  S  1 indicates that Equations (4-195) and (4-196) are used to calculate
 f and  m ; S = -1 indicates that  f and  m remain unchanged from previously set values; S = -2 indicates contact-
optimized parameters; S = -3 indicates noncontact optimized parameters; S = -4 indicates that the 4th and 5th fields will
be checked and if valid, used to set  f and  m directly.
Central Difference Operator

The central difference operator assumes a quadratic variation in the displacement with respect to time.
a n =  v n + 1  2 – v n – 1  2    t  (4-197)
v n =  u n + 1  2 – u n – 1  2    t  (4-198)
so that,
a n =  u n + 1 – u n    t 2  (4-199)
where,
u n = u n – u n – 1 (4-200)
for IDYN=4:
M M
-------- u n + 1 = F n – R n + -------- u n (4-201)
t 2 t 2
for IDYN=5:
n – 1---
M n+1 n n M n 2
-------- u = F – R + -------- u – Cv (4-202)
2 2
t t
Since the mass matrix is diagonal, no inverse of the operator matrix is needed. Also, since the operator is only conditionally
s, the critical time step is evaluated at the beginning of the analysis. For IDYN = 4, the critical time step is computed by a
power sweep for the highest mode in the system only at the beginning of the analysis. For IDYN = 4, no damping is
included. For IDYN = 5, an approximated method based on element geometry is used to compute the highest eigenvalue.
The critical time step is calculated at a user-specified frequency or every 100 steps. The variable time step can be used only
for IDYN = 5. Unless there is significant distortion in an element or material nonlinearity, the change of critical time step
is not significant.
Time Step Definition

In a transient dynamic analysis, time step parameters are required for integration in time. The DYNAMIC CHANGE option can
be used for either the modal superposition or the direct integration procedure. The AUTO STEP option can be used for the
Newmark-beta, the Single Step Houbolt, and the Generalized-alpha operator to invoke the adaptive time control. Enter
parameters to specify the time step size and period of time for this set of boundary conditions.
When using the Newmark-beta operator, decide which frequencies are important to the response. The time step in this
method should not exceed 10 percent of the period of the highest relevant frequency in the structure. Otherwise, large
phase errors will occur. The phenomenon usually associated with too large a time step is strong oscillatory accelerations.
With even larger time steps, the velocities start oscillating. With still larger steps, the displacement eventually oscillates. In
nonlinear problems, instability usually follows oscillation. When using adaptive dynamics, you should prescribe a
maximum time step.
As in the Newmark-beta operator, the time step in Houbolt integration should not exceed 10 percent of the period of the
highest frequency of interest. However, the Houbolt method not only causes phase errors, it also causes strong artificial
damping. Therefore, high frequencies are damped out quickly and no obvious oscillations occur. It is, therefore, completely
up to the engineer to determine whether the time step was adequate.
For the Generalized-alpha operator, depending on the chosen parameters, the integration scheme can vary between the
Newmark-beta operator and the Single-step houbolt operator. For spectral radii < 1, there is artificial damping in the
CHAPTER 4 155
system. Depending on the type of problem, the Generalized-alpha parameters and the associated time step should be
carefully chosen to reduce phase errors and effects of artificial damping.
In nonlinear problems, the mode shapes and frequencies are strong functions of time because of plasticity and large
displacement effects, so that the above guidelines can be only a coarse approximation. To obtain a more accurate estimate,
repeat the analysis with a significantly different time step (1/5 to 1/10 of the original) and compare responses.
The central difference integration method is only conditionally stable; the program automatically calculates the stable time
step. This step size yields accurate results for all practical problems.
Initial Conditions
In a transient dynamic analysis, you can specify initial conditions such as nodal displacements and/or nodal velocities. To
enter initial conditions, use the following model definition options: INITIAL DISP for specified nodal displacements, and
INITIAL VEL for specified nodal velocities. As an alternative, you can use the USINC user subroutine.
Time-Dependent Boundary Conditions

Simple time-dependent load or displacement histories can be entered on data blocks. However, in general cases with
complex load histories, it is often more convenient to enter the history through a user subroutine. Marc allows the use of
the FORCDT and FORCEM user subroutines for boundary conditions. the FORCDT user subroutine allows you to specify the
time-dependent incremental point loads and incremental displacements. The FORCEM user subroutine allows you to specify
the time-dependent magnitude of the distributed load.
Mass Matrix
The mass matrix is a discrete representation of the system mass. System mass can be defined through either distributed
masses and concentrated masses.
Distributed masses are defined for elements through the mass density material property. Marc offers both consistent and
lumped element mass matrices. The consistent mass matrix is given by
T
m =  N  N  dV
where  is the mass density and  N  is the shape function matrix. The lumped mass matrix is flagged through the LUMP
parameter. Marc uses the Hinton, Rock, Zienkiewicz lumping scheme to produce a diagonal mass matrix. The salient
features of this scheme are as follows: The diagonal coefficients of the consistent mass matrix are computed. The total mass
of the element m is also computed. A scale factor s is computed by adding the translational diagonal coefficients that are
mutually parallel and in the same direction. All diagonal coefficients are then scaled by m  s , thereby preserving the total
mass of the element.
Concentrated masses are defined through the MASSES, CONM1, and/or CONM2 model definition options. The MASSES option
allows the placement of particle masses m i at degree of freedom of i at specific nodes. Both translational and rotary masses
are allowed.
The CONM1 option allows the specification of a diagonal or 6 x 6 symmetric mass matrix at a node in the global cartesian
or in a local coordinate system. The local system is specified through a coordinate system ID. The fully coupled 3-D form
is given below:
M 11     
M 21 M 22  sym  
M 31 M 32 M 33   
 m c onm1  =
M 41 M 42 M 43 M 44  
M 51 M 52 M 53 M 54 M 55 
M 61 M 62 M 63 M 64 M 65 M 66
Each mass term can be varied using tables. Independent variables allowed include time, increment number, coordinates,
temperature. When the diagonal form is specified, only the M 11 to M 66 diagonal terms need to be specified. In 2-D, the
mass matrix reduces to 3 x 3 and only the M 11 , M 12 M 33 terms need to be specified. If the CONM1 mass matrix is
specified in a local system, then it is transformed to the global system before assembly.
The CONM2 option allows a concentrated diagonal mass and the mass moments of inertia with respect to the global cartesian
or a local coordinate system. The local system is specified through a coordinate system ID. The mass terms for each degree
of freedom at the node are a function of the concentrated mass, the moments of inertia and the distance of the center of
mass from the node. The offset distance from the node to the center of mass is given by the vector  V 1 V 2 V 3  . The 3-D
form of the CONM2 matrix is given below:
M     
0 M  sym  
0 0 M   
 m c onm2  = a
0 – MV 3 MV 2 I 11  
a a
MV 3 0 – MV 1 I 21 I 22 
a a a
– MV 2 MV 1 0 I 31 I 32 I 33
where the combined moments of inertia are given as follows:

a 2 2 a 2 2 a 2 2
I 11 = I 11 + M  V 2 + V 3  ; I 22 = I 22 + M  V 1 + V 3  ; I 33 = I 33 + M  V 1 + V 2  ;
a a a
I 21 = – I 21 – MV 1 V 2 ; I 31 = – I 31 – MV 1 V 3 ; I 32 = – I 32 – MV 2 V 3
Each mass coefficient can be varied using tables. Independent variables allowed include time, increment number,
coordinates, temperature. In 2-D, the mass matrix reduces to 3 x 3 and only M , I 33 , V 1 , and V 2 are required. If the
CONM2 mass matrix is specified in a local system, then it is converted to the global system before assembly.
CHAPTER 4 157
Damping
In a transient dynamic analysis, damping represents the dissipation of energy in the structural system. It also retards the
response of the structural system.
Marc allows you to enter two types of damping in a transient dynamic analysis: modal damping and Rayleigh damping.
Use modal damping for the modal superposition method and Rayleigh damping for the direct integration method. For
modal superposition, you can include damping associated with each mode. To do this, use the DAMPING option to enter the
fraction of critical damping to be used with each mode.
During time integration, Marc associates the corresponding damping fraction with each mode. The program bases
integration on the usual assumption that the damping matrix of the system is a linear combination of the mass and stiffness
matrices, so that damping does not change the modes of the system.
For direct integration damping, you can specify the damping matrix as a linear combination of the mass and stiffness
matrices of the system. You can specify damping coefficients on an element basis.
Stiffness damping should not be applied to either Herrmann elements or gap elements because of the presence of
Lagrangian multipliers.
Numerical damping is used to damp out unwanted high-frequency chatter in the structure. If the time step is decreased
(stiffness damping might cause too much damping), use the numerical damping option to make the damping (stiffness)
coefficient proportional to the time step. Thus, if the time step decreases, high-frequency response can still be accurately
represented. This type of damping is particularly useful in problems where the characteristics of the model and/or the
response change strongly during analysis (for example, problems involving opening or closing gaps).
Element damping uses coefficients on the element matrices and is represented by the equation:
n
  t 
C =    i M i +   i +  i ----- K
  i 
(4-203)
i = 1
where,
C is the global damping matrix.

Mi is the mass matrix of ith element.
Ki is the stiffness matrix of the ith element.
i is the mass damping coefficient on the ith element.
i is the usual stiffness damping coefficient on the ith element.
i is the numerical damping coefficient on the ith element.
t is the time increment.
, ,  can be a function of the time and the temperature by referencing a table.
If the same damping coefficients are used throughout the structure, Equation (4-203) is equivalent to Rayleigh damping.
The damping associated with springs and mass points can be controlled via the springs and masses input options. The
damping on elastic foundations is the same as the damping on the element on which the foundation is applied. For springs,
a dashpot can be added for nonlinear analysis.
Inertia Relief
Inertia relief refers to an analysis procedure that allows unconstrained systems to be subjected to a quasi-static analysis by
taking rigid body inertia forces into account. Examples of such systems include an aircraft in a steady turn, an underwater
structure in equilibrium under gravity and buoyancy. Conventional static analysis cannot be performed for such systems
since, in the absence of constraints, the stiffness matrix is singular. Inertia relief analysis is applicable to such free bodies.
The response is measured relative to a steady accelerating frame induced by the external loads.
Two important steps are followed by the solver in an analysis with inertia relief:
1. Rigid Body Mode evaluation: This is the central step in an inertia relief analysis since the rigid body modes are
extensively used in the load computations. The Support Method is currently provided in Marc to evaluate the rigid
body modes.
2. Inertia Relief Loads evaluation: The loads associated with the rigid body motion of the system are calculated using
the rigid body modes and the global lumped mass matrix. The loads are calculated such that they balance the
current external load vector in the system.
More details on each of the steps are provided below.
Rigid Body Mode Evaluation

Support Method
The Support Method requires the manual specification of nodal degrees of freedom which can be used to evaluate the rigid
body modes of the system. Any number of nodal entries can be provided, each with degrees of freedom ranging from 1 to
the maximum possible degrees of freedom at each node. These degrees of freedom are referred to as the r-degrees of freedom
set. It is the user’s responsibility to ensure that the r-degrees of freedom set is necessary and sufficient to evaluate the rigid
body modes of the system.
The rigid body modes are evaluated by sequentially setting each r-degrees of freedom to 1 with all other r-degrees of
freedom set to 0 and solving for the system of equations
 K l l   K lr     l  
  = 0 (4-204)
 K rl   K rr    I  
where  I  is the identity matrix for the r-degrees of freedom and   l  are the modal amplitudes for the remaining degrees
of freedom. Once the rigid body modes are evaluated, the r-degrees of freedom are set to 0 and used as fixed displacement
constraints. Other displacement boundary conditions, specified through normal options like FIXED DISP, DISP CHANGE, etc.
can be used in conjunction with the rigid body degrees of freedom constraints.
The Support Method is computationally inexpensive since the provided Support degrees of freedom are automatically
constrained and no additional ties/constraints are needed while calculating the response. It is available for single-processor
and multi-processor runs (both single-input and multi-input file formats).
The Support Method requires manual input to evaluate the rigid body modes of the system. For direct solvers, the rigid
body modes are calculated through repeated back-solves of the factorized matrix. The computation of rigid body modes is
CHAPTER 4 159
more expensive for the conjugate gradient iterative solver. Finally, the Support Method cannot be used in contact problems
with dynamically changing rigid body modes.
Rigid Body Mode Update

In a small displacement analysis, the rigid body modes are evaluated at the first cycle of the first increment of the inertia
relief loadcase and are retained through the loadcase. In a large displacement analysis, the rigid body modes are updated at
the beginning of each increment by using the current stiffness matrix while solving Equation (4-204).
Inertia Relief Loads Evaluation

A general derivation for the inertia relief load vector in a nonlinear analysis is presented here. At the i th iteration of the
 n + 1  th increment, the total response is given by
 u tn i+ 1  =  u rn  +  u rn +i –1 1  +  u rn +i 1  +  u n  +  u in–+11  +  u in + 1  (4-205)
where  u t  is the total response,  u r  is the rigid body response and  u  is the flexible response. Neglecting
accelerations associated with the flexible response, the governing equations are given as
n+1 n+1 n+1 n+1 n+1 n+1
 M i – 1   u··r i  +  K i – 1   u i  =  P n + 1  –  I i – 1  –  P  i – 1  (4-206)
n+1 n+1
where  M i – 1  and  K i – 1  are the mass and stiffness matrices at the  i – 1  th iteration of  n + 1  th increment,
n+1 n+1
 P n + 1  is the external load vector at the  n + 1  th increment, I i – 1 and  P  i – 1  are the internal load vector and
the inertia relief load vector at the  i – 1  th iteration of the  n + 1  th increment, respectively.
The change in the rigid body acceleration is computed as follows:
 u··rn +i 1  =   in–+11   A n +i 1  (4-207)
n+1
where   in–+11  is the matrix of rigid body modes at the  i – 1  th iteration of the  n + 1  th increment and  A  i 
represents the iterative change in the modal acceleration vector.
n+1
Pre-multiplying Equation (4-193) by individual rigid body modes,   k i – 1  T , and using the fact that the work done by
the rigid body modes is 0, one gets
n+1 n+1 n+1
 C i – 1   A  i  =  R i  (4-208)
n+1
where the kl th term of matrix  C i – 1  is given by
n+1 n+1 n+1 n+1

C i – 1 k l =   k i – 1  T  M i – 1    l i – 1 
n+1
and the k th term of  R i  is given by
n+1 n+1 T n+1 n+1 n+1

R i k =   k i – 1    P  –  I i – 1  –  P  i – 1   (4-209)
n+1
Note that if the rigid body modes are mass ortho-normalized, matrix  C i – 1  is an identity matrix. Solving Equation
n+1
(4-208) for  A  i  , the iterative change in the inertia relief load vector is given by
n+1 n+1 n+1 n+1

 P  i  =  M i – 1    i – 1   A  i  (4-210)
The governing equations in Equation 5-184 can then be rewritten in the form,
n+1 n+1 n+1 n+1 n+1
 K i – 1   u i  =  Pi  –  I i – 1  –  P  i 
n+1
where the inertia relief load vector at the i th iteration of the  n + 1  th increment,  P  i  , is given by
n+1 n+1 n+1

 P  i  =  P  i – 1  +  P  i 
Additional items to be kept in consideration for inertia relief analysis in Marc are as follows:
Convergence Checking
For systems that are unconstrained or just constrained against rigid body motion (statically determinate structures), the
reaction forces at nodes are 0 or very close to 0. In this case, using relative residual force checking may lead to unnecessary
recycling and possible non-convergence. It would be more appropriate to use displacement checking or absolute residual
force checking in such situations.
Time Stepping
Inertia relief can be used in conjunction with all time stepping procedures used for static analyses. Time step cutback
schemes are supported. Inputs specified in either table or non-table formats are also supported.
Parallel Processing
The Support Method is available for serial and parallel runs. For the multi-input file parallel mode, the Support nodes
should be available in all the domains. For the single-input file parallel mode, the inertia relief option should be present in
the model definition section itself in order to allow the Support nodes to be added to domains they are not naturally present
in.
Deactivating Inertia Relief

When inertia relief is active during an analysis, and is subsequently turned off, the user has three choices for treating the
pre-existing inertia relief load vector.
1. The inertia relief load vector can be retained. No new inertia relief loads are computed, but pre-existing inertia relief
loads are included in the right-hand side.
CHAPTER 4 161
2. The inertia relief load vector can be removed instantaneously. Note that this is the option used by Marc if the
INERTIA RELIEF option is present in one loadcase and is absent in the next loadcase. This could lead to convergence
difficulties at the beginning of the second loadcase due to the large change in load vectors.
3. The inertia relief load vector can be removed gradually. This allows the inertia relief load vector to be removed over
the course of the second loadcase.
Inertia Relief Ouput Options

The nodal inertia relief load and moment vectors are written to the post file by default using nodal post codes 51 and 52,
respectively. It should be noted that the inertia relief loads are always written out in the global cartesian system. Also, the
command INER can be used on the PRINT NODE option to obtain the inertia relief loads at a particular node. Also, when the
NO PRINT option is revoked, a summary of the total inertia relief load is printed in the out file.
Limitations for Inertia Relief Analyses

1. All inertia relief analyses provide the uninteresting solution of zero displacements when a uniform mesh is subjected
to uniform gravity loads or when a uniform shell mesh is subjected to uniform distributed loads. This is as expected
because the inertia relief equations place an inertia load on each node that is equal and opposite to the external load.
2. The Support Method is currently not supported for the CASI iterative solver.
3. The Support Method is capable of dealing with material and geometric non-linearities. However, boundary
condition non-linearities like contact between multiple bodies cause the number of rigid body modes to change in
the system. These require relative rigid body positions of the bodies to be continuously updated. These cannot be
handled by the Support Method.
Rigid-Plastic Flow
The rigid-plastic flow analysis is an approach to large deformation analysis which can be used for metal forming problems.
Two formulations are available: an Eulerian (steady state) and Lagrangian (transient) approach. The effects of elasticity are
not included. If these effects are important, this option should not be used.
In the steady state approach, the velocity field (and stress field) is obtained as the solution of a steady-state flow analysis.
The time period is considered as 1.0 and, hence, the velocity is equal to the deformation. In the transient formulation, the
incremental displacement is calculated.
The R-P FLOW parameter invokes the rigid-plastic procedure. This procedure needs to enforce the incompressibility
condition, which is inherent to the strictly plastic type of material response being considered.
Incompressibility can be imposed in three ways:
1. By means of Lagrange multipliers. Such procedure requires Herrmann elements which have a pressure variable as
the Lagrange multiplier.
2. By means of penalty functions. This procedure uses regular solid elements, and adds penalty terms to any
volumetric strain rate that develops. It is highly recommended that the constant dilatation formulation be used –
by entering a nonzero value in the second field of the GEOMETRY model definition option.
Penalty factor can be treated as constant or variable through the R-P FLOW parameter. The penalty value is entered
through the PARAMETERS option.
3. In plane stress analysis (shell and membrane elements), the incompressibility constraint is satisfied exactly by
updating the thickness. This capability is not available for steady state analysis.
In R-P flow analysis, several iterations are required at any given increment, the greatest number occurring in the first
increment. Subsequent increments require fewer iterations, since the initial iteration can make use of the solution from the
previous increment.
Due to the simplicity of the rigid-plastic formulation, it is possible to bypass stress recovery for all iterations but the last in
each increment, provided that displacement control is used. In such cases, considerable savings in execution time are
achieved. If nodal based friction is used in a contact analysis, then a stress recovery is always performed after each iteration.
Steady State Analysis

The steady state R-P flow formulation is based on an Eulerian reference system. For problems in which a steady-state
solution is not appropriate, an alternative method is available to update the coordinates. The UPNOD user subroutine is used
to update the nodal coordinates at the end of a step according to the relationship.
n
x in = x in – 1 + v t (4-211)
where n refers to the step number, v n is the nodal velocity components, and t is an arbitrary time step. t is selected in
such a way as to allow only a reasonable change in mesh shape while ensuring stability with each step.
Updating the mesh requires judicious selection of a time step. This requires some knowledge of the magnitude of the nodal
velocities that will be encountered. The time step should be selected such that the strain increment is never more than one
percent for any given increment.
The quantities under the title of ENGSTN in the printouts actually refer to the strain rate at an element integration point.
The reaction forces output by the program gives the limit loads on the structure.
Transient Analysis
In the transient procedure, there is an automatic updating of the mesh at the end of each increment. During the analysis,
the updated mesh can exhibit severe distortion and the solution might be unable to converge. Global adaptive meshing or
manual mesh rezoning can be used to overcome this difficulty.
The rigid-plastic flow capability is based on iteration for the velocity field in an incompressible, non-Newtonian fluid. The
normal flow condition for a nonzero strain rate can be expressed as:
 2  · · ·
 i j =  --- --·-  ij =      i j (4-212)
3
 
where
· 2· ·
 = ---  ij  ij (4-213)
3
CHAPTER 4 163
is the equivalent strain rate,  is the yield stress (which may be rate-dependent) and
1
 ij =  ij – ---  ij  kk (4-214)
3
gives the deviatoric stress. The effective viscosity is evaluated as:
2 
 = --- --·- (4-215)
3 
·
Note that as   0    . A cutoff value of strain rate is used in the program to avoid this difficulty. An initial value
·
for  is necessary to start the iterations. These values can be specified in the PARAMETERS option. The default cut-off value
is 10 – 6 , and the default initial strain rate value is 10-4.
The value of the flow stress is dependent upon both the equivalent strain, the equivalent strain rate, and the temperature.
This dependence can be given through the WORK HARD, STRAIN RATE, TEMPERATURE EFFECTS, or TABLE options, respectively.
For steady state analysis, the UNEWTN user subroutine can be used to define a viscosity. In this manner, a non-Newtonian
flow analysis can be performed. For the transient procedure, the URPFLO user subroutine can be used to define the
flow stress.
Superplasticity
Superplasticity is the ability of a material to undergo extensive deformation such as strains of 1000% without necking.
Superplastic behavior has been reported in numerous metal, alloys and ceramics. Every instance of superplasticity is
associated with: (1) A fine grain size ( 1 – 10 m ), (2) deformation temperatures  0.4T m , and (3) a strain-rate
sensitivity factor m  0.3 .
Using finite element analysis to simulate superplastic fabrication of complex parts used in the aerospace and automotive
industries requires this material behavior and contact with friction. Furthermore, the process pressure needs to be
automatically adjusted to keep the material within a target strain rate. The simulation can be used to predict
thinning, forming time, areas of void formation, and can ultimately be employed in shape optimization; thus, reducing the
number of prototypes of forming trials required to product an acceptable part.
Three mechanisms have been proposed to account for the high strain-rate sensitivity found in superplastic materials: (1)
Vacancy creep, (2) creep by grain boundary diffusion, and (3) grain boundary sliding. According to Ghosh and Hamilton,
the strain-rate sensitivity of metals arises from the viscous nature of the deformation process. The viscosity is a result of the
resistance offered by internal obstacles within the material. In dislocation glide and climb processes, the obstacles are a fine
dispersion of second phase particles within the grain interior, between which the dislocations are bent around and moved.
At high homologous temperatures ( T  0.4T m ; where T m is the melting temperature), the high diffusivities around
grain boundary regions can lead to grain boundary sliding. The overall rate sensitivity of a material is then a result of the
rate sensitivities of the grain boundary and the grain interior. The more the material behaves as a viscous liquid, the greater
its superplasticity.
The superplastic behavior is characterized by the dependence of the flow stress upon the strain-rate, which is usually
depicted by the logarithmic relationship shown in Figure 4-34.
As indicated in Figure 4-34, the stress-strain rate behavior of a superplastic material can be divided into three regions. Values
of strain-rate sensitivity, m (the slope of flow stress versus strain-rate curve) which is a measure of resistance to localized
necking, are relatively low in both the low stress-low strain rate region I and the high stress-high strain rate region III and
superplasticity is not manifested. Rather, superplasticity is found only in region II, a transition region in which stress
increases rapidly with increasing strain-rate. As temperature increases and/or grain size decreases, region II is displaced to
higher strain-rates. Moreover, the maximum observed values of m increase with similar changes in these parameters.
Certainly the forming process innovations evoked will need to be carefully studied and developed. Forming times are slow,
and there will be a critical need for optimizing forming pressures, stress strain-rate and deflection in sheet forming. Based
on the schematic flow stress-strain rate relationships given above, it is apparent that high values of m are requisite for
superplastic materials. Since, for a given material and forming temperature, m , usually varies with strain-rate, it is desirable
to control strain-rate during forming so that optimum or at least adequate strain-rate sensitivity is exhibited. Ductility is
also dependent upon forming temperature, which must lie within a narrow range. If forming temperatures and pressure
cycle are optimum, then unlike conventional ductile materials, superplastic materials are much less susceptible to localized
necking. Additionally, under such conditions, the flow stress occurring during forming is much lower than the mechanical
yield stress.
(a) Flow Stress

Region III
ln 
Region II
Decreasing grain size or

Region I increasing temperature
.
ln 
(b) Strain-rate
Sensitivity Decreasing grain size or
increasing temperature
d ln  Region I
m  = ------------·-
 
d ln 
Region II
Region III
.
ln 
Figure 4-34 Function of Strain-rate
CHAPTER 4 165
Thus, the superplastic materials may be viewed as exhibiting time-dependent inelastic behavior with the yield stress as a
function of time, temperature, strain-rate, total stress and total strain. Typical materials used in commercial superplastic
forming applications include Ti-6A1-4V titanium alloy and 5083 aluminium alloy.
The form of constitutive equation used to simulate superplasticity is given as:
·m
y =  (4-216)
The form in Equation (4-216) can be recovered by using appropriate constants in the ISOTROPIC model definition option to
define power law or rate power law. Thus,
m ·n
Power Law Model:  y = A   o +   + B (4-217)
m·n
Rate Power Law Model:  y = A  + B (4-218)
The superplastic forming process requires the use of the SPFLOW parameter. The use of this parameter automatically
activates the FOLLOW FOR and PROCESSOR parameters. The process parameters are controlled by the use of the SUPERPLASTIC
history definition option.
Typical outputs that are available from the superplastic forming simulation are the thickness distribution for the part, the
equivalent plastic strain rate, and the history of the process pressure.
The process pressure is automatically calculated during the analysis. The pressure magnitude is adjusted such that the
equivalent strain rate in the part is at or close to the user-specified target strain rate. The equivalent strain rate in the part
is an average value calculated by sampling a suitable subset of elements. The recommended scheme is one in which elements
with a strain rate greater than a cut-off factor times the maximum element strain rate are sampled. This maximum strain
rate is based on a smoothing algorithm described below. The cut-off factor can vary between 0 (all elements below the
maximum are sampled) and 1 (only the elements with maximum strain rate are sampled). The recommended value for the
cut-off factor is 0.7 to 0.9 (default value is 0.8). To reduce undesirable oscillations in the pressure-time history, a pressure
smoothing algorithm is incorporated. The basis for this algorithm is a smoothing of the maximum strain rate in the mesh
based on the fact that the maximum strain rate should be typically representative of a few elements in the mesh, rather than
an isolated individual value. The peak strain rates in a few elements are calculated. The number of elements that are used
in this calculation varies with the cut-off factor (for a cut-off factor of 1, only 1 element is used; for the default of 0.8, 10
elements are used; for a cut-off factor of 0, 50 elements are used). The strain rate values are successively disregarded in
descending order if the difference from the highest strain rate to the lowest differs by more than 10 percent from the mean.
The value of the cut-off factor has significant influence on the maximum strain rate control and on the smoothness of the
pressure-time curve. Larger the factor (that is, 0.9 or higher) provides more control on the maximum strain rate, but may
potentially cause oscillations in the pressure history. Smaller the factor (that is, 0.7 or lower) provides less control on the
maximum strain rate, but causes smoother pressure-time curves. The default of 0.8 should work in most cases - in situations
where physically realistic localized strain rates occur and one desires good control on these localized values, a higher value
could be used.
Soil Analysis
In Marc, the material models available for soil modeling are linear elasticity, nonlinear elasticity, the Cam-Clay model, von
Mises, linear Mohr-Coulomb, parabolic Mohr-Coulomb, orthotropic elastic, and exponential cap model.
This section has the solution procedure for fluid-soil analysis. In the current formulation, it is assumed that the fluid is of
a single phase, and only slightly compressible. This formulation will not be adequate if steam-fluid-solid analysis is
required.
There are three types of soil analysis available in Marc. In the first type, you perform an analysis to calculate the fluid
pressure in a porous medium. In such analyses, heat transfer elements 41, 42, or 44 are used. The SOIL model definition
option is used to define the permeability of the solid and the bulk modulus and dynamic viscosity of the fluid. The porosity
is given either through the INITIAL POROSITY or the INITIAL VOID RATIO options and does not change with time. The prescribed
pressures can be defined using the FIXED PRESSURE option, while input mass flow rates are given using either the POINT FLUX
or DIST FLUXES option.
In the third type of soil analysis, a fully-coupled approach is used. Element types 32, 33, or 35 are available. These elements
are “Herrmann” elements, which are conventionally used for incompressible analysis. In this case, the extra variable
represents the fluid pressure. The SOIL option is now used to define both the soil and fluid properties. The porosity is given
through the INITIAL POROSITY or the INITIAL VOID RATIO option. The prescribed nodal loads and mass flow rates are given
through the POINT LOAD option, while distributed loads and distributed mass flow rates are given through the DIST LOADS
option. The FIXED DISP option is used to prescribe either nodal displacements or pore pressures.
Technical Formulation
In soil mechanics, it is convenient to decompose the total stress into a pore pressure component p and the deviatoric or
effective stress  d .
 =  d – pI (4-219)
Note the sign convention used; a positive pore pressure results in a compressive stress. The momentum balance
(equilibrium) equations are with respect to the total stresses in the system.
 + f = u·· (4-220)
where  is the density, and f , u·· are the body force and the acceleration. The equilibrium equation can then be expressed
as:
 d – p + f = u·· (4-221)
The fluid flow behavior can be modeled using Darcy’s law, which states that the fluid’s velocity, relative to the soil’s skeleton,
is proportional to the total pressure gradient.
–K
u· f = -------  p +  f g  (4-222)

where,
u· f is the fluid’s bulk velocity

K is the soil permeability
 is the fluid viscosity
CHAPTER 4 167
f is the fluid density

g is the gravity vector.
The fluid is assumed to be slightly compressible.
·
f
p· = K ----- (4-223)
f f
where, K f is the bulk modulus of the fluid.
However, the compressibility is assumed small enough such that the following holds:
K K
   f ----  p +  f g    f   ----  p +  f g  (4-224)
 
It is also assumed that the bulk modulus of the fluid is constant, introducing the fluid’s compressibility  f .
f = 1  Kf (4-225)
The solid grains are assumed to be incompressible. Under these assumptions, the governing equations for fluid flow is:
K ·
  ----  p +  f g  =  f p + u· (4-226)

where,  is the medium’s porosity.

It is important to note that the medium’s porosity is only dependent upon the original porosity and the total strains. Letting
V f and V s stand for the fluid and solid’s volume
 = dV f   dV f + dV s  = 1 – J – 1  1 –  0  (4-227)
where J is the determinant of the deformation gradient and  0 is the original porosity, both with respect to the reference
configuration.
Introducing the weighting function  u and  p , the weak form, which is the basis for the finite element system, then
becomes:
   u f –  u u·· –  u  +  u p dv +  u t dA = 0 (4-228)

Vn + 1 A
where V and A are the conventional volumes and surface area and t is the applied tractions. Note that the applied
tractions is the combined tractions from both the effective stress  d and the pore pressure p .
and
K
  p ----  p +  f g  +  f  p p· +  p  u·  dv –   p q n dA = 0 (4-229)

Vn + 1 An + 1
where, the normal volumetric inflow, q n , is:
k
q n = ---  p +  f g   n (4-230)

The weak form of equilibrium can be written as:
ru =     u p  dv (4-231)
Vn + 1
rp =   p  K
----  p +  f g  +  f  p p· +  p   u· dv

(4-232)
Vn + 1
CHAPTER 4 169
Application of the directional derivative formula yields:
d
D u  r u   u = ------ r u  u + u  (4-233)
d  = 0
Hence,
D u  r u   u =    u     u p  +  u
T
 up  dv (4-234)
Vn + 1
Similarly,
D p  r u   p =     u p dv (4-235)
Vn + 1
D u  r p   u = 

  p  K
----  p +  f g  +  f  p p· +  p 

 u·   u –  p u·
T

Vn + 1 
(4-236)
u +  p   u· – u p  K

----  p +  f g  –  p

 K
----  u p dv

D p  r p   p =   p  K
---- p +  f  p p· dv

(4-237)
Vn + 1
with the displacement and pressures interpolated independently as:
u =  N i u i
and (4-238)
p =  Nj pj
we get a linearized system of equilibrium equation:
K uu K up  u   Ru 
  = F–  (4-239)
K pu K pp  p   Rp 
The resulting system of equations is highly nonlinear and nonsymmetric, and is solved by full Newton-Raphson solution
scheme.
Note that it is assumed that the permeability, porosity, viscosity, and the bulk modulus of the fluid are considered
independent of the state variables u and p . It is evident that, in general, this is not the case; however, in the analysis that
follows, these dependencies are ignored for tangent purposes. Note that they are included in the calculation of the residuals
R u and R p ; hence, convergence is always achieved at the true solution.
In the second type of soil analysis, the pore (fluid) pressure is directly defined, and the structural analysis is performed.
Element types 27, 28, or 21 are available. In such analyses, the pore pressure is prescribed using the INITIAL PORE and CHANGE
PORE options. The characteristics of the soil material are defined using the SOIL option. If an elastic model is used, the
Young’s moduli and Poison’s ratio are important. If the Cam-Clay model is used, the compression ratios and the slope of
the critical state line is important. For the Cam-Clay model, the preconsolidation pressure is defined using the INITIAL PC
option. For this model, it is also important to define an initial (compressive stress) to ensure a stable model.
Three types of analyses can be performed. The simplest is a solution for only the fluid pressure based upon the porosity of
the soil. In this case, a simple Poisson type analysis is performed and element types 41, 42, or 44 are used. In the second type
of analysis, the pore pressures are explicitly defined and the structural analysis is performed. In this case, the element types
21, 27, and 28 should be used. Finally, a fully-coupled analysis is performed; in which case, you should use element types
32, 33, or 35. Of course, the soil can be combined with any other element types, material properties to represent the
structure, such as the pilings.
Mechanical Wear
Mechanical wear is an important physical phenomenon in any structure subjected to repeated loadings. Often, this
behavior is modeled by determining the stress on the surface and using it in a subsequent fatigue calculation. For certain
applications, including manufacturing, disk brakes, bearings, gears, tires, and seals, it is important to know the amount of
wear and possibly the change in geometry which would influence the behavior. It should be noted that for some
applications, such as polishing, surface wear is the positive objective. Wear, which is defined as the removal of material from
the surface, may be due to mechanical processes or chemical processes. The latter are not considered but the UWEARINDEX
user subroutine may always be used to incorporate these effects. The wear due to mechanical behavior can be classified as:
1. Asperity deformation and removal
2. Plowing of the surface
3. Delamination
4. Adhesive
5. Abrasion
6. Fretting
7. Solid particle impingement
These processes lead to either mild wear or severe wear (rough, torn surfaces). The wear model implemented in Marc, based
upon Archard equation, is not applicable for severe wear.
There are several wear models available beginning with Archard’s equation:
Gt
w = KF  -----
H
where
K is the wear coefficient

F is the normal force
Gt is the sliding distance
H is the hardness
CHAPTER 4 171
This is converted into an incremental form. As the wear is calculated as a local quantity, we use normal stress instead of the
normal force (except for beam elements as discussed below). This leads to a series of models that are written as:
K Simple Archard
w· = ----    V r e l
H
K m n
w· = ----    V r e l Bayer exponential form
H
–T0  T
K m n
w· = ----  V rel exp Bayer exponential form with thermal activation
H
where
w· is the rate of change of wear in the direction normal to the surface

 is the normal stress
V re l is the relative sliding velocity
T0 is the activation temperature
T is the current temperature
A table may be associated with the coefficient to allow temperature dependence or other effects. The following quantities
can be used as independent variables for the wear coefficients: time, increment, current and original coordinate position,
normal stress and relative sliding velocity.
The incremental wear calculation is performed at the nodal points that are in contact. Hence, the CONTACT option is
required. Although friction is typically used in a wear analysis, it is not needed for the wear calculation itself since it only
uses the normal stress and the relative sliding velocity. The wear model and data is specified for each contact body that
should do wear calculation. The wear can occur on one or both sides of the bodies in contact. Each side uses the wear model
and properties of the respective body. A scaling factor can be used on the contact table allowing the wear to be different for
different contact combinations.
The incremental wear is calculated as:
w· t
and the wear is accumulated as:
w n + 1 = w n + w· t .
The total accumulated wear can be interpreted as the distance a node would move due to wear and indicates a volume loss.
The total volume loss due to wear is calculated per contact body and printed to the output file. The total accumulated wear
w t o t is available for postprocessing as nodal post code 77, Wear Index. The current wear rate w· is available as a vector in
the normal direction of the contact point as nodal post code 78, Wear Rate.
For continuum elements, the nodal coordinates can be updated using:

·
x n + 1 = x n + fw t  n
where, n is the normal to the surface and f is a scale factor. If large steps are taken such that the incremental wear becomes
large one might get inaccurate results due to nodes ending up outside the contact tolerance. New contact will be found
again, but small steps are often needed for accurate results. If the total amount of wear becomes large, then remeshing may
be required.
Wear calculation is supported for all continuum, shell, and beam elements. In addition, for thermo-mechanically coupled
analysis, wear calculation is supported on rigid (meshed) bodies with heat transfer. These also support coordinate update
due to wear. The coordinate update option is not supported for shell elements (since the shell thickness should be changed,
not the coordinate positions, but shell thickness update due to wear is not supported).
Beam elements can as mentioned be used in a wear calculation. This covers both the case that a beam element is touching
a solid or shell element and the case of beam-to-beam contact. Since we do not have a contact area available in beam
contact, we use the contact normal force instead of the contact normal stress in the wear calculation. This needs to be taken
into account when using wear on beam elements. The volume loss due to wear is calculated for each element as an area
obtained by the length of the element and the length offset by the amount of wear, and this area is multiplied with twice
the contact distance radius.
Design Sensitivity Analysis

Design sensitivity analysis is used to obtain the sensitivity of various aspects of a design model with respect to changes in
design parameters in order to facilitate structural modifications. The design parameters that are amenable to change are
called design variables. The two major aspects of the design model for which design sensitivity is considered herein are:
a. Design objective
b. Design model response
As a result, the design sensitivity analysis capability in Marc is currently limited to finite element models of structures with
linear response in the computation of:
1. Gradients for
a. An objective function (or the design objective), if one is defined by you (for example, minimizing the material
volume in the model).
b. Various types of design model responses under multiple cases of static mechanical loading, or free vibration, in
linear behavior.
2. Element contributions to the responses of the model.
The gradient of the objective function or of a response function is simply the set of derivatives of such a function with
respect to each of the design variables, at a given point in the design space (that is, for a given design).
For sensitivity analysis to proceed, the design model, the analysis requirements, the design variables, and the functions for
which the gradients are to be found have to be specified by you.
The existing design sensitivity analysis capability in Marc can be applied in one of two ways:
CHAPTER 4 173
1. As a stand-alone feature, where you are concerned only with obtaining sensitivity analysis results. Such an
application is completed with the output of the sensitivity analysis results.
2. Within a design optimization process, where you are concerned mainly with the optimization of a design objective
related to a finite element model. This type of an application of sensitivity analysis is transparent to you. The design
optimization process is completed with the output of design optimization related data, such as the optimized
objective function, related values of the design variables, and the analysis results for the optimized design.
These two procedures are described as follows:
1. Sensitivity analysis as a stand-alone feature
This type of solution usually aims at obtaining the derivatives of user prescribed response quantities at a given
design, with respect to each of multiple design variables specified by the user. This set of derivatives is therefore the
gradient of the response function at the given design in the design variable space (or, in short, in the design space).
For example, for a prescribed response function r, given the design variables x 1 , x 2 , and x 3 , the gradient is defined
as:
dr dr dr T
r = --------- --------- --------- (4-240)
˜ dx 1 dx 2 dx 3
The number of response quantities for which gradients are computed is limited either by the program default or
by a user-specified number. If you are interested in obtaining the sensitivity analysis results in order of criticality,
the option to sort them in this order is also available. The responses are currently prescribed as constraints with user-
defined bounds. If sorting is not required, the bounds can be mostly arbitrary, although they still have to conform
to the type of constraint prescribed. However, if sorting is required and is to be meaningful, it is important for you
to give realistic bounds on the response.
Element contributions to each response quantity are obtained as a by-product of the type of response sensitivity
analysis capability in Marc. Thus, the response r can be represented as a sum of these element contributions:
N
r =  re + C (4-241)
e = 1
where the second term, C , involves work done elsewhere, such as in support settlement, if any.
This is helpful for a visual understanding of which regions of the model contribute in what manner to each of the
response quantities at the given design, since it can be plotted in a manner similar to, say stress contours.
Finally, as an option, if you also prescribe an objective function, the gradient of the objective function with respect
to the design variables is also computed at the given design. Thus, for the objective function W , Marc obtains
dW dW dW T
W = --------- ---------  --------- (4-242)
˜ dx 1 x 2 dx n
2. Sensitivity analysis within a design optimization process:

The design optimization algorithm in Marc requires the utilization of gradients of the objective function and of the
constraint functions, which are very closely related to the response functions. The current algorithm, described
under “design optimization”, ignores your initial prescribed design, but instead begins by generating other designs
within the prescribed bounds for the design variables. Once the optimization algorithm is completed and the
optimized design is available, if a sensitivity analysis is required at the optimized state, it will be necessary for you
to modify the model accordingly and to use sensitivity analysis as a stand alone feature.
During design optimization, sensitivity analysis is performed for a maximum number of constraints either indicated
by the program default or prescribed by you.
Theoretical Considerations
The method currently employed in Marc for response sensitivity analysis is the “virtual load” method. For sensitivity
analysis of the objective function, finite differencing on the design variables is performed directly.
In the virtual load method, first a design is analyzed for the user-prescribed loadcases, and, if also prescribed, for
eigenfrequency response. The response of the structure having been evaluated for each of these analyses, the response
quantities for which sensitivity analysis is to be performed are then decided upon and collected in a database. In sequence,
a virtual loadcase is generated for each such response quantity. Reanalysis for a virtual loadcase leads to virtual
displacements.
The principle of virtual work is then invoked. This defines the element contributed part of the response quantity, for which
the virtual load was applied, as a dot product of the structural displacement vector, for the actual loadcase with which the
response is associated, and the virtual load vector itself.
The jth response r j can be expressed as the dot product of the actual load vector with the virtual displacement vector as:
T
rj = p uv (4-243)
By differentiation of this expression, you can show that the derivative of the response r j with respect to a given design
variable x i is given by:
T
dr dp v T dp dK
--------j = ---------- u + u v  -------- – -------- u
dx i dx i  dx dx i 
i
where K is the stiffness matrix of the structure. The response derivative above is evaluated on the element basis as:
T
dr dp v T dp dK
--------j =  ---------- u + u v  -------
- – -------- u (4-244)
dx i dx i  dx dx i 
e i e
where the vectors u and p v are the vector of element nodal displacements due to the actual loadcase and the vector of
element nodal forces due to the virtual load, respectively. The case of eigenfrequencies follows the same logic except that
an explicit solution for the virtual loadcase is not necessary. The derivatives are now evaluated at the element level via finite
differencing. This is known as the semi-analytical approach.
CHAPTER 4 175
Note that the derivative expression for the virtual load method is quite similar to that for the Adjoint Variable method. In
fact, while they are conceptually different approaches, for certain cases they reduce to exactly the same expressions.
However, for certain other cases, the terms take on different meanings although the end result is the same.
Design Optimization
Design optimization refers to the process of attempting to arrive at certain ideal design parameters, which, when used
within the model, satisfy prescribed conditions regarding the performance of the design and at the same time minimize (or
maximize) a measurable aspect of the design. In Marc, you can ask to minimize:
1. Total material volume
2. Total material mass
3. Total material cost.
When there is more than one material in the finite element model, the specification of different mass densities and unit
costs are taken into account in the computation of the objective function.
The performance requirements might not necessarily have to be related to response, but also to different concepts such as
packaging, design envelope, even maintenance. The current capability is based on optimization with constraints on
response. Also, the lower and upper bounds on the design variables themselves define the limits of design modifications.
Hence, the design optimization problem can be posed mathematically in the following format:
Minimize W
Subject to c j  0.0 j = 1  m
where W is the objective function, and c j is the jth constraint function, either specified as an inequality related to a
response quantity or as a limit on a design variable. For example, to limit the x-direction translation component at a node
k, the constraint can first be written as:
*
 u x  k  u xk
Assuming that the displacement is constrained for positive values, the normalized constraint expression c (dropping the
subscript j ) becomes:
 u – u   u  0.0 (4-245)
with its derivatives as:
dc – 1 du
------ = ------ ------ (4-246)
dx u dx
Within Marc, the constraints can be imposed on strain, stress, displacement, and eigenfrequency response quantities. For
stresses and strains, the constraints are defined as being on the elements, and for displacements, the constraints are defined
as being at nodes. Stress and strain components, as well as various functions of these components (the von Mises equivalent
stress and principal stresses, stresses on prescribed planes) and generalized stress quantities can be constrained. Similarly,
translation and rotation components of displacement, resultant and directed displacements as well as relative displacements
between nodes can be constrained. For free vibration response, constraints can be placed on the modal frequencies as well
as on differences between modal frequencies. A full listing of such constraints are given in Marc Volume C: Program Input.
The upper and lower bounds on the design variables are posed as:
xi  xu and xi  xl (4-247)
after which they can be transformed into expressions similar to Equation (4-245).
The response quantities associated with the model are implicit functions of the design variables. Analyses at sample design
points are used to build explicit approximations to the actual functions over the design space. This approach minimizes the
number of full scale analyses for problems which require long analysis times such as for nonlinear behavior. This method
is summarized next.
Approximation of Response Functions Over the Design Space

The design space for the optimization problem is bounded by limits on the design variables of a model to be optimized.
The simplest case is that of a single design variable, where the design space is a straight line, bounded at the two ends. For
2
higher number of design variables, say n, the design space can be visualized as a bounded hyperprism with n vertices. For
such a construct, you can build approximations to the constraint functions by way of analyses conducted at the vertices.
2
However, this requires n analyses.
We now note that the minimum geometrical construct spanning n-dimensional space is a simplex with n + 1 vertices.
Thus, an approximation based on analyses at vertices requires only  n + 1  analyses. The simplex has already been used
for first order response surface fitting based on only function values [Ref. 3]. However, the use here involves higher order
response functions.
Like the hyperprism, in one-dimension, the simplex degenerates into a straight line. However, in two dimensions it is a
triangle, and in general it is a hyper-tetrahedron. Figure 4-35 compares the simplex to the hyperprism in normalized two-
dimensional design space.
n=2
Hyperprism: 22 = 4 vertices
Simplex : 2 + 1 = 3 vertices
Figure 4-35 Comparison of the Simplex to the Hyperprism in Two-dimensions
While the orientation of the simplex in the design space appears to be a relatively arbitrary matter, once an origin and the
size of the simplex are prescribed, a simple formula will locate all vertices of a simplex in n-dimensions [Ref. 4]. The response
CHAPTER 4 177
gradient information at the simplex vertices is combined with the function values to achieve enhanced accuracy. Thus, an
analysis at a vertex can be utilized to yield both response function values and, by way of sensitivity analysis, the response
gradients at that vertex. The virtual load method employed in Marc for obtaining the response gradients is discussed under
Design Sensitivity Analysis. The response gradients can then easily be converted to constraint gradients for use in an
optimization algorithm.
As a result, the results of an analysis at a vertex of the simplex can be summarized as the vector of constraint function values
c j , and the gradient, c j , of each constraint function ( j ) at that vertex.
For the case of a one variable problem, the results of analyses at the two vertices are symbolized in Figure 4-36, for a
hypothetical constraint c j .
dc j1
------------ E1
dx
Possible Actual Function
c j1
2
1 c j2
E2 dc j2
x ------------
dx
Figure 4-36 Vertex Results for One-dimensional Design Space
From Figure 4-36, it appears almost natural to fit a cubic function to the four end conditions (two function values and two
slopes) depicted in the figure. However, this approach is too rigid, and is not easily generalizable to higher dimensions. On
the other hand, the use of two quadratics, which are then merged in a weighted manner gives higher flexibility and potential
for increased accuracy.
It can be seen that the two equations, E1 and E2 , are designed such that they both satisfy the function values at the two
vertices, but E1 satisfies the slope at vertex 1 only, and E2 satisfies the slope at vertex 2 only. Finally, at any design point
x , the response function c j can be approximated as:
c j =  W1 E1 + W2 E1    W1 + W2  (4-248)
where the weight functions W1 and W2 are normally functions of x and possibly some other parameters.
This type of approach has the further advantage that it is consistent with the use of the simplex for determining analysis
points and approximating constraint functions in the higher dimensional cases. Therefore, for an n-dimensional problem,
the simplex having n + 1 vertices, each equation needs to have  n + 1  + n = 2n + 1 unknown parameters. The
general quadratic polynomial without the cross-coupling terms satisfies this condition for all n .
Improvement of the Approximation

When approximations are used, the results of an optimization process need to be checked by means of a detailed analysis.
As a result, the approximations can be adjusted and the optimization algorithm can be reapplied. Depending on how
accurate the approximations prove to be and how many more detailed analyses are acceptable to you, this process can be
applied for a number of cycles in order to improve upon the results.
The Optimization Algorithm

The optimization algorithm implemented in Marc is the Sequential Quadratic Programming method [Ref. 5]. This method
is employed using the approximation described above. By obtaining response function and gradient values from the
approximate equations, the need for full scale analyses is reduced. The method is summarized below.
The quadratic programming technique is based on the approximation of the objective function by a quadratic function.
When nonlinear constraints exist, as is the case in most practical design optimization problems, the second order
approximation concept is extended to the Lagrangian which is a linear combination of the objective function and the
constraint functions. The solution method for a quadratic programming problem with nonlinear constraints can be
summarized as the following:
At each step, modify the design variables vector x by adding a scaled vector:
x  x + qs (4-249)
where, s is a search direction and q is the scale factor along the search direction. If the search direction has been
determined, the scale factor can be found by some type of line search along the search direction. The determination of the
search direction constitutes the major undertaking in the quadratic programming method. If H is the Hessian of the
Lagrangian and g is the gradient of the objective function, then the search direction s is that which minimizes the function
T T
g s +  s Hs   2 (4-250)
subject to the linearized constraints:
Js  – c (4-251)
where J is the Jacobian matrix of the constraints and c is the vector of constraint functions at the current iteration. Due
to lack of knowledge about which constraints will be active at the optimum, the Hessian of the Lagrangian is not always
readily available. Thus, some iterations take the form of a gradient projection step.
The coefficients of the constraint functions in the Lagrangian are the Lagrange multipliers which are unknown before
solution has started. At the optimum these multipliers are zero for inactive constraints. Normally, the above problem is
solved using only those constraints which appear to be active at or close to the current design point, with the assumption
that these constraints will be active at the optimum. The selection of these active constraints is done within the framework
of an active set strategy, the set being modified appropriately with the progression of the iterations. Similarly, the arrays g ,
H , and J are also modified as the iterations proceed.
CHAPTER 4 179
Marc User Interface for Sensitivity Analysis and Optimization

This is discussed under Input, Output, and Postprocessing in the following paragraphs.
Input
Input features related to design sensitivity and design optimization are similar. However, they differ in the parameter data
blocks and in the optional specification of an objective function for the case of sensitivity analysis. Therefore, other than
these two items, all discussion of input should be construed to refer to both procedures equally. All design sensitivity and
design optimization related parameters and options in a Marc input file start with the word DESIGN.
All loadcases and any eigenfrequency analysis request associated with sensitivity analysis or optimization should be input
as load increments in the history definition section after the END OPTION line. The maximum number of nodes with point
loads or the maximum number of distributed loadcases should be defined in the DIST LOADS parameter.
The type of solution requested from Marc can simply be indicated by the parameter input DESIGN SENSITIVITY or DESIGN
OPTIMIZATION. Only one of these lines can appear in the input. These parameters also let you change the sorting of
constraints, the maximum number of constraints in the active set, and the maximum number of cycles of design
optimization. The first two items are discussed in previous sections. A cycle of design optimization refers to a complete
solution employing the sequential quadratic programming technique followed by a detailed finite element analysis to
accurately evaluate the new design point reached by way of approximate solution. If required, the approximation is
improved using the results of the last finite element analysis, and a new cycle is started.
The specification of an objective function, being optional for design sensitivity, is made by use of the model definition
option DESIGN OBJECTIVE. This allows you to choose from one of the available design objectives.
The DESIGN VARIABLES option allows you to specify the design variables associated with the finite element model. You have
a choice of three major types of design variables:
1. Element cross-sectional properties as given by way of the GEOMETRY option.
2. Material properties as given by ISOTROPIC or ORTHOTROPIC options.
3. Composite element properties of layer thickness and ply angle, as given by the COMPOSITE option.
The properties which are supported are listed under the DESIGN VARIABLES option. The relevant elements, for which cross-
sectional properties can be specified as design variables, each has a section in Marc Volume B: Element Library , describing which
properties, if any, can be posed as design variables for that element.
Similarly, the list of material properties (currently all related to linear behavior) that can be design variables is given under
the DESIGN VARIABLES option. For composite groups, the layer thicknesses and ply angles can be given as design variables for
composite groups.
Design variables can be linked across finite element entities such that a given design variable controls several entities. An
example is the linking of the thicknesses of several plane stress elements by means of a single design variable. Thus, when
this variable changes, the thicknesses of all linked elements reflect this change. On the other hand, for the unlinked case,
all thicknesses are associated with separate design variables. This feature is controlled by the LINKED and UNLINKED
commands.
Response quantities or constraints on response quantities, are specified by means of the DESIGN DISPLACEMENT CONSTRAINTS,
DESIGN STRAIN CONSTRAINTS, DESIGN STRESS CONSTRAINTS, and DESIGN FREQUENCY CONSTRAINTS options. There is no limit to
the number of constraints. Displacement constraints are posed at nodes or groups of nodes, while the strain and stress
constraints are posed over elements or groups of elements, and frequency constraints are posed for free vibration modes. A
complete list can be found under the above mentioned options in Marc Volume C: Program Input.
When strain or stress constraints are prescribed, it is useful to know that the program evaluates such constraints at all
integration points of all layers of an element and proceeds to consider the most critical integration point at the related layer
for the element. Thus, a strain or stress constraint on an element normally refers to the most critical value the constraint
can attain within the element. During optimization, the most critical location within the element may change and any
necessary adjustment takes place internally.
For certain responses, the limiting values can be the same in absolute value for both the positive and negative values of the
response. For constraints on such response functions, you have the choice of prescribing either separate constraints on the
positive and negative values, or a combined constraint on the absolute value. The first approach is more accurate albeit at
a higher computational cost.
Output
For sensitivity analysis, the output file contains the following information:
 Echo of input, any warnings or error messages.
 A of numbers and definitions for your prescribed design variables. It is important to note that the variable values
are always output in the internal numbering sequence which is defined in this. Search for the words Design variable
definitions to reach this in the output file.
 Analysis results for your prescribed design.
 The value of the objective function and its gradient with respect to the design variables prescribed by you.
 Sensitivity analysis results for the response functions in the default or user-defined set, sorted in order of
criticalness (if specified). It should be noted that although the responses are sorted across multiple loadcases, the
sensitivity results are output for each loadcase. The related output file consists of a check on the actual response
value (obtained by sensitivity analysis) and the gradient of the response with respect to the design variables. For
eigenfrequency results, the check values on the response can be somewhat more accurate than the results from
eigenfrequency analysis since the latter is iterative, but the check uses the Rayleigh quotient on top of the iteration
results. A constraint reference number allows you to track the sensitivity analysis results plots when postprocessing.
Search for the word ‘Sensitivity’ to reach this output.
For design optimization, the following is written into the output file:
 Echo of input, any warnings or error messages.
 Certain indications that some analyses are being done, but no analysis results except for the ‘best’ design reached
during optimization.
 The objective function values and the vector of design variable values:
• At each of the simplex vertices.
• For the starting simplex vertex (from this point on, the information also includes whether the design is feasible
or not).
• At the end of each cycle of optimization.
• For the ‘best’ design found.
 Analysis results for the ‘best’ design.
CHAPTER 4 181
Postprocessing
This requires that you ask the program to create a post file. The following plots can be obtained by way of postprocessing.
For sensitivity analysis:
 Bar charts for gradients of response quantities with respect to the design variables.
 Contour plots of element contributions to response quantities. Thus, a finite element model contour plot gives
the element contributions to a specific response quantity which was posed in the form of a constraint in the input
file.
The increment number of the sensitivity analysis results is the highest increment number available in the post file. The
information for each response quantity is written out as for a subincrement. The zeroth subincrement corresponds to the
objective function information. Numbers for the other subincrements correspond to the constraint reference number(s) in
the output file.
For design optimization:
 Path plots showing the variation of the objective function and of the design variables over the history of the
optimization cycles. The best design (feasible or not) is not necessarily the last design point in the plot. The values
at the starting vertex are considered as belonging to the zeroth subincrement. Each optimization cycle is then
another subincrement. The increment number corresponding to these subincrements is taken as zero. The analysis
results for the ‘best’ design start from increment one.
 Bar charts where each chart gives the values of the design variables at the optimization cycle corresponding to that
bar chart.
Define Initial State with Results from a Previous Analysis

In many cases, it is necessary to analyze a nonlinear process in several stages. Each stage may involve different contact bodies
and boundary conditions, but history data such as temperatures, displacements, stresses and strains have to be carried over
for contact bodies to be passed from one stage to another. In Marc, it is possible to start a numerical simulation as an
axisymmetric or plane strain problem, and switch to a full 3-D analysis in a subsequent stage. This makes a multi-stage
analysis more efficient.
Basically, for multi-stage simulations there are two methods available to transfer data from one analysis to another. The first
method uses the PRE STATE option and is based on data transfer via the Marc results file. The second method uses Model
Sections and is based on data transfer via separate section files, which contents do not depend on the Marc results file. Both
methods will be discussed below.
Pre State
The PRE STATE option is designed to read data from a Marc results file and to use the data as initial conditions in the new
analysis. In a contact analysis, it allows the selection of contact bodies (by names) to easily apply the data to parts of the
current model. The PRE STATE option includes the capability of the obsolete AXITO3D option (as available in older Marc
releases).
The methodology of the PRE STATE option can be outlined as follows:
 Scan through the results file for the desired time step or increment.
 Collect the list of nodes and elements corresponding to the contact body transfer list.
 Collect node and element data as required (see the PRE STATE option in Marc Volume C: Program Input for the
required data types).
 Copy data to the new model as the initial conditions with axisymmetric or plane strain to 3-D expansion if
applicable.
The PRE STATE option takes several steps:
 Run first stage analysis (make sure that the relevant post file quantities have been selected).
 Open the results file.
 Extract and expand (if axisymmetric or plane strain to 3-D is required) the mesh at the increment for which the
data has to be transferred. In analyses without global remeshing, it is not required to base the mesh of the new
analysis on the deformed configuration of the previous analysis. Instead, the undeformed configuration of the
previous analysis can be used, but the displacements have to be transferred from the results file. Note that Mentat
offers the possibility to expand a model including boundary conditions.
 Complete the new model; for example, by adding contact bodies and other boundary conditions.
 Specify the PRE STATE option and select the data to be transferred.
 Run new analysis.
Most quadrilateral or hexahedral elements (including Herrmann elements) as well as most commonly used materials (such
as metal and rubber) can be used in the PRE STATE option. For large deformation problems, either the total or the updated
Lagrange formulation can be used. Thermal and dynamic effects can also be included.
For more detailed information on this feature, see the PRE STATE model definition option, in Marc Volume C: Program Input.
In general, the PRE STATE option can only transfer history data from the results file as the initial conditions in the new
model. Users are responsible for reapplying the boundary and load conditions as well as the material properties. If users use
Mentat, the boundary conditions defined in the previous axisymmetric or plane strain analysis automatically expand to 3-
D when the model is expanded to 3-D. However, if a load function is used and is required in the new model, curve shifting
is required. That is, if, in the previous model, a load is applied from 0 to 10, in the new model based on the last step of the
previous analysis, the load is applied starting from 10. For prescribed displacements, as nodal positions are updated to the
final configuration in the previous analysis, the load function should start from 0 in the new model.
Compatible Element Types for PRE STATE

It is important to select the correct element types compatible with the previous model for an axisymmetric or plane strain
to 3-D transfer. For the same element type transfer, many, but not all of the element types are supported.
For axisymmetric to 3-D transfer, supported element types are:
10  7 20  7
82  84 83  84
28  21 67  21
33  35 66  35
55  57 59  61
CHAPTER 4 183
48  23 142  23
144  146 145  146
166  147 167  147
169  148 170  148
40  43 172  174
For plane strain to 3-D solid elements, supported elements types are:
11  7 80  84
19  7 81  84
27  21 32  35
29  21 34  35
39  49
For 2-D to 2-D and 3-D to 3-D data transfer, the PRE STATE option supports most standard Marc element types for plane
stress, plane strain, axisymmetric, 3-D and (solid) shell applications. The supported structural element types are:
plane stress: 3, 26, 53, 124, 201

plane strain: 6, 11, 19, 27, 29, 32, 34, 54, 58, 60, 80, 81, 125, 128, 155
axisymmetric: 2, 10, 20, 28, 33, 55, 59, 66, 67, 82, 83, 119, 126, 129, 156
3-D solid: 7, 21, 35, 57, 61, 84, 127, 130, 134, 136, 157, 202, 247
membrane: 18, 30, 158, 200
(solid) shell: 22, 49, 72, 75, 138, 139, 140, 185
rebar: 23, 46, 48, 142-148, 165-170
The supported heat transfer element types are:
planar: 37, 39, 41, 69, 121, 131

axisymmetric: 38, 40, 42, 70, 122, 132
3-D solid: 43, 44, 71, 123, 133, 135, 137, 203
Model Sections
A fundamental limitation of the above mentioned PRE STATE option is the fact that the data transfer goes via results files.
This not only implies that all of the required data must be on the results file (which is not always possible and which can
result in large files), but also that in most cases, it is impossible to fully prepare the new analysis without having completed
the previous one. This limitation does not exist if use is made of model sections to transfer data from one analysis to
another. See Marc Volume C: Program Input, CREATE SEC option for a list of supported options. model sections can be used
in mechanical, thermal, or thermal-mechanically coupled analyses.
A model section can be seen as a self-contained part of a finite element model, which means that all information in terms
of nodal coordinates, element connectivity, material model, but also results data like displacements, temperatures, stresses,
and strains are available. When using model sections in a multi-stage simulation, one can include the model section
resulting from the analysis of the previous stage in the analysis of the current stage, even without, for example, knowing
how many elements are in the model section or which material model has been used.
The basic approach with model sections can be outlined as follows:
 Create a model section using the CREATE SEC option. A model section is stored in a file, which is written at the end
of a successful analysis. This can be an analysis without any applied loading (so just increment 0), or an analysis
with multiple increments. Note that multiple increments are not allowed if adaptive meshing options are present.
The situation corresponding to the end of the analysis is stored in the model section.
 Include a model section as defined in a previous analysis in the current analysis by using the IMPORT SEC option. As
mentioned above, all information about the material model and elements used, as well as the complete solution
state, is available in the model section. When defining contact bodies, one can directly assign a model section to a
meshed contact body. Moreover, boundary conditions like gravity loading and plastic heat generation can also be
applied to a model section. The imported model section can be repositioned, which can be e.g. convenient when a
different set of geometric (rigid) bodies is used compared to the previous stage.
 If in a multi-stage simulation the current analysis will be followed by another one, export the model section using
the EXPORT SEC option. During the export phase, it is possible to change the dimension of the axisymmetric or
plane strain model section to a full 3-D model section. It is also possible to duplicate a cyclic symmetry model
section into a full 360 degrees model section. For axisymmetric to 3-D transfer, the supported element types are:
10  7 20  7
28  21 67  21
33  35 66  35
55  57 59  61
82  84 83  84
For plane strain to 3-D solid elements, the supported elements types are:
11  7 27  21
32  35 80  84
Per model section, only a single material model can be used. Moreover, different element types, like solid and shell elements,
are not allowed. In case of multiple materials and/or element types, one should simply use multiple model sections, where
each of them has to be created using a separate Marc analysis.
A model section can be visualized in Mentat based on its outer boundary (edges for 2-D elements, faces for 3-D elements).
In Figure 4-37, an example of forming a channel in multiple stages has been given.
CHAPTER 4 185
(a) Initial plane strain model to create a Model Section
(b) Model set up for the first stage; the Model Section is used as a meshed contact body
(c) End of first stage (remeshing has been used)
(d) Model set up for the second stage; the Model Section is used as a meshed contact body
(e) End of second stage (remeshing has been used) (f) Expanded Model Section
Figure 4-37 Example of using Model Sections
Steady State Rolling Analysis

Rolling contact analysis of a cylindrical deformable body in a Lagrangian framework can be computationally expensive
because it may require not only time-dependent transient process but also a fine mesh in the entire body to accurately
capture contact characteristics. However, some problems involve only fully axisymmetric structures with constant
moving/spinning velocities. These problems can be considered steady state if a reference configuration, which moves with
the body but does not spin around the rolling axis, is used.
Marc provides the capability of steady state rolling analysis. The feature is characterized by a mixed Eulerian/Lagrangian
formulation with inertia effects in spinning/cornering deformable bodies. Using a non-spinning reference frame attached
to the wheel axel, the analysis becomes purely space dependent. It presents a better alternative to the unnecessary
computational burden of arriving at a steady state condition through a transient analysis. Furthermore, a finer mesh only
needs to be used in the contact region as opposed to the entire rolling surface.
Kinematics
We consider the axisymmetric body shown in Figure 4-38. The body spins at an angular velocity  s around the
axisymmetric axis T s at point P s and, simultaneously rotates with a cornering angular velocity  c around an axis T c at
point P c . Assume that a particle in the body has a location P 0 at time t = 0 . At time t , its motion contains three parts:
1. from P 0 to a location X because of the spinning
2. from X to Y because of a deformation Y = D  X  , where D is time independent function resulting from the
steady state condition
3. from Y to Z because of the cornering.
Ps
Ts
s
Tc
c
Pc
Figure 4-38 Kinematics
The three motions can be described as

X = Rs   Po – Ps  + Ps (4-252)
Y = DX (4-253)
Z = Rc   Y – Pc  + Pc (4-254)
with
R s = exp   s t  and R c = exp   c t  (4-255)
In Equation (4-255), ̃ s and  c are the skew-symmetric tensors associated with the rotation vectors  s T s and  c T c ,
respectively, with
 s  r =  s T c  r and  c  r =  c T c  r (4-256)
for any vector r .

Time derivative of Equation (4-255) gives:
· ·
R s =  s  R s and R c =  c  R c (4-257)
CHAPTER 4 187
Making use of Equations (4-252), (4-253), (4-256), and (4-257), the velocity of the particle can be obtained by the first time
derivative of Equation (4-254) as
· D
Z = R c   c T c   Y – P c  +  s ------- (4-258)

where  =  s t , is the spinning angle.
Similarly, the acceleration is obtained by the second derivative of Equation (4-254) with respect to time:
·· D 2D
Z = R c   c2  T c  T c – 1    Y – P c  + 2 s  c T c  ------- +  s2 ---------- (4-259)
  2
where,  denotes the tensor product and 1 is the unit tensor.
Transformation of Equations (4-258) and (4-259) into the reference configuration defined by X by premultiplying R cT , gives
D
v =  c T c   Y – P c  +  s ------- (4-260)

D 2D
a =  c2  T c  T c – 1    Y – P c  + 2 s  c T c  ------- +  s2 ---------- (4-261)
  2
where v and a are velocity and acceleration of the particle with respect to the reference frame.
Inertia Effect
The contribution of inertia effect into the right-hand-side of the system equation can be calculated using the weak form
 =  a  u dv
v
D
= –  c2   T c  T c – 1    Y – P c   u dv – 2 s  c  T c  -------  u dv (4-262)

V V
D u
+  s2  -------  --------- dv
 
V
where  is the density, v is the volume, and u is the displacement.

Linearization of Equation (4-262) gives:
u
 = –  c2  u   T c  T c – 1   u dv – 2 s  c  T c  ----------  u dv

v V
(4-263)
u u
+  s2  ----------  --------- dv
 
v
which can be used to calculate the contribution of the inertia effect to stiffness matrix.
Rolling Contact
To take into account the rolling effect in contact, the velocity vector in Equation (4-260) is decomposed into a normal and
a tangential component, with respect to the contact surface, for all nodes in contact with ground. The normal component
is then forced to become zero because of the contact conditions. The relative slipping velocity used in friction calculation
is the difference between the tangential component and the ground moving velocity.
Steady State Rolling with Marc

The capability of steady state rolling analysis in Marc has taken into account the effects of rolling frictions and the inertia
effects resulting from both spinning and cornering. The deformable rolling body can contact with multiple, flat or nonflat
rigid surfaces. A typical example is a tire model which is in contact with a rigid rim and a rigid road surface. The spinning
effects are involved only for selected contact body pair (for example, tire/road) which is defined with the SS-ROLLING history
model definition option. The steady state rolling analysis can follow a static stress analysis with various loadcases and can
also combine with any steady state loads. The feature is available for 3-D analysis only.
The element types supported for steady state rolling analysis include 7, 9, 18, 21, 35, 57, 61, 84, 117, 120, 146, 147, and 148.
The friction type supported in contact for steady state rolling analysis, based on nodal force, is Coulomb friction for rolling.
Because the system matrix becomes nonsymmetric in steady state rolling analysis, it is recommended that the Pardiso
parallel solver be used.
The feature requires the so-called streamline along the circumferential direction of the spinning body. Therefore, the 3-D
mesh must be generated by revolving its corresponding axisymmetric mesh. For a 3-D brick element, the circumferential
direction must be from the element face defined by nodes 1-2-3-4 to the element face defined by nodes 5-6-7-8.
Below is a list of options required in a steady state analysis, in addition to the others in a standard static stress analysis.
SS-ROLLING: Parameter activating the steady state rolling analysis

ROTATION A: Model definition option defining the spinning axis
CORNERING AXIS: Model definition option defining the cornering axis
SS-ROLLING: History definition option defining the spinning body, ground body, spinning body motion
(spinning/cornering/moving relative to the ground) and others. All velocities/forces defined in
the option are total values.
Structural Zooming Analysis

Local variations, such as the changes in model geometries or in the degrees of finite element refinement to achieve a better
evaluation of the local gradients in the solution, often need a complete re-analysis of the entire model. However, in cases
that these local changes have negligible influence on the solution a certain distance away from the changes, it is
computationally more efficient to model only the part with the local changes. It can be realized by applying the existing
loads or/and boundary conditions in the local model along with properly defined kinematic conditions to the local
boundaries connecting to the global model.
A typical structural zooming analysis contains two steps:
1. Global run to obtain a post file containing global results.
CHAPTER 4 189
2. Local run to define kinematic boundary conditions in the local model and to obtain refined results in the local
model.
This procedure can be repeated as many times as desired. Any local analysis can be the global analysis of next level
refinement.
The GLOBALLOCAL option (see Marc Volume C: Program Input for details) is used in the input of the local run to define the list
of nodes connecting to the global model. Marc calculates the deformation (temperature) history of these nodes, based on
their locations in the global model and on the solution of the global analysis. The obtained deformation (temperature)
history is then applied to the nodes as prescribed kinematic boundary conditions. Detailed steps include:
a. Reading in the GLOBALLOCAL option to get the list of connecting nodes.
b. Reading in the global model (element types, node coordinates, element connectivity, thickness if shell elements)
and solution of the model from the generated post file in global run.
c. Finding out the locations of local connecting nodes in the global model (the element each node is associated
with and its isoparametric location within the element).
d. Calculation of the deformation (temperature) of each connecting node for every increment available in the
global post file, based on its location in the global elements associated using the interpolation techniques,
and storage of the deformation (temperature) history in the format of time-dependent tables.
e. Applying the deformation (temperature) history of the connecting nodes to the local model as prescribed
kinematic boundary conditions.
All the steps above are performed automatically inside Marc once the GLOBALLOCAL option is used in the input file of the
local run.
If the global adaptive meshing is used in the first analysis, then the GLK model definition should also be used.
Element Types Supported

The global to local modeling can be used in the following four cases:
Global Model Local Model

2-D Solid 2-D Solid
3-D Solid 3-D Solid
3-D Shell/Membrane 3-D Shell/Membrane
3-D Shell/Membrane 3-D Solid
One can use composite continuum elements in the local model if the global model had composite continuum elements
with the same number of layers or was composed of shell elements
See Figure 4-39 for examples from 2-D solid elements to 2-D solid elements and from 3-D shell elements to 3-D solid
elements.
global model
local model
cracks
Cylinder
(a) 2-D Solid to 2-D Solid (b) 3-D Shell to 3-D Solid
Figure 4-39 Structural Zooming
Uncoupled Thermal Stress Analysis

A typical chaining analysis in Marc for thermal stress problems includes two steps:
 Heat transfer analysis to generate a post file containing temperature distribution on the model and history over the
time period considered.
 Thermal stress analysis based on the temperature distribution/history obtained from the previous heat transfer
analysis
To perform a chaining analysis for a thermal stress problem two options can be used in Marc. They are the CHANGE STATE
option and the MAP TEMP option.
Using the CHANGE STATE option, the mesh in stress analysis has to be the same as that used in the heat transfer analysis. The
structural element type has to correspond with the heat transfer element type. Please see Marc Volume B: Element Library for
this correspondence. The basic idea of the option is to read in the integration point temperatures for each element from
the post file of the heat transfer analysis and use them directly in the stress analysis. No spatial data mapping is required.
Since this option is integration point based, INITIAL STATE should be used if one needs to specify initial temperatures in the
stress analysis. This option can be used with AUTO THERM, AUTO THERM CREEP, and AUTO STEP.
The data mapping used with the MAP TEMP option contains the following steps:
1. Read in the MAP TEMP option to get the list of structural nodes. The temperature history of the nodes is expected to
be obtained from the heat transfer post file.
2. Read in heat transfer model (element types, node coordinates, element connectivity, etc.), and solution of the model
in the heat transfer analysis.
3. Find the location of the structural nodes in the heat transfer model (the element each node associated and its
isoparametric location within the element).
4. Calculate the temperature of each structural node for every increment available in the post file of the heat transfer
analysis, based on its location in the heat transfer element associated. Store the temperature history in the format
of a time-dependent table.
5. Apply the temperature history to the structural node in stress analysis.
CHAPTER 4 191
Since this option is node-based, INITIAL TEMP should be used if one needs to specify initial temperatures in the stress analysis.
This option cannot be used with AUTO THERM and AUTO THERM CREEP. It can be used with AUTO STEP.
Compared to the CHANGE STATE option, the MAP TEMP option provides more flexibility. It allows a temperature
distribution/history obtained from a previously heat transfer simulation to be used in the stress analysis even though the
meshes are dissimilar. The element types or even element classes used in heat transfer and stress analyses can be different.
The model or the time period covered in heat transfer and stress analyses can also be different. For time period outside the
range in heat transfer analysis, a time shift or end values or extrapolation option can be used. For the part of the structural
model or time period not covered by the post file of the heat transfer analysis, POINT TEMP can always be used to directly
input nodal temperatures.
Cure-Thermal-Mechanically Coupled Analysis

When resin is used in the manufacturing of composite parts, the curing process during and after the forming processes has
a significant effect on the thermal and mechanical behavior of the formed part. The shrinkage caused by curing may
severely distort the final geometry of the part. Marc has implemented a capability that allows you to perform cure-thermal-
mechanical coupled analysis to predict this behavior.
The curing analysis is incorporated into the existing staggered coupled thermal-mechanical analysis procedure (see Thermal
Mechanically Coupled Analysis in Chapter 6: Nonstructural and Coupled Procedure Library). Before the heat transfer analysis
pass takes place, the curing analysis is performed based on the estimated temperatures at the beginning and end of the
increment. The cure rate is then calculated according to the cure kinetics of the resin materials. Using the cure rate, a heat
flux due to cure reaction heat generation is calculated and added into the heat transfer system of equations. Volumetric
curing reaction heat flux is calculated according to the cure rate:
d
Q c = -------  1 – V f  r H r (4-264)
dt
where:
H r – resin cure reaction heat
 r – resin density
 – resin degree of cure
Vf – fiber volume fraction
The governing matrix equation for cure-thermal coupled analysis can be expressed as:
C  T T· + K  T T = Q + Q I + Q F + Q C (4-265)
In Equation (4-265), C  T  and K  T  are the temperature-dependent heat capacity and thermal conductivity matrices,
respectively. T is the nodal temperature vector. T· is the time derivative of the temperature vector, Q is the external heat
flux vector, Q I is the heat flux vector due to plastic work, Q F represents the heat generated due to friction, and Q C is the
heat generated due to curing.
With the CURING parameter included in the input data, Marc first activates the pass to calculate the degree of cure and curing
reaction heat flux at the beginning of each cycle of heat transfer analysis. Then, Marc considers the cure shrinkage strain in
the mechanical pass of the thermal-mechanical coupled analysis.
To include curing into a heat transfer or thermal-mechanical coupled analysis, it is necessary to add the CURING parameter
into the input data. In addition, the CURE RATE model definition option is also needed to define the curing properties,
usually called cure kinetics, of resin materials as one kind of property for heat transfer analysis.
To include cure shrinkage strain into the mechanical pass of a thermal-mechanical coupled analysis, in addition to having
both the CURING parameter and the CURE RATE model definition option in the input data, the CURE SHRINKAGE model
definition must be included to define the cure shrinkage properties of the resin material. Note that the cure shrinkage effect
is calculated based on the degree of cure, so the cure shrinkage effect is ignored if the degree of cure is not calculated during
the heat transfer pass.
Cure Kinetics
In Marc, the cure kinetics defines the cure rate as a function of the degree of cure and the temperature of resin materials.
There are different approaches to define the cure kinetics of resin materials: embedded cure kinetics models, table definition
and user subroutine.
CHAPTER 4 193
In a curing analysis, the cure rate is calculated for each time step. Assuming that the cure rate is defined as the function of
the degree of cure and temperature of:
d
------- = f   ,T  (4-266)
dt
The time integration of the degree of cure is done using backward Euler method:
 ni = t  f   ni – 1 ,T  +  n – 1 (4-267)
where,
i – iteration number for cure

n – current increment number
n–1 – previous increment number
f – function defined by cure kinetics model
t – time step size of the increment
The four cure kinetics models implemented in Marc are:

Cure kinetics model 1: by Lee, Loos and Springer (1982) [Ref. 42];
Cure kinetics model 2: by Scott(1991) [Ref. 49];
Cure kinetics model 3: by Lee, Chiu, and Lin (1992) [Ref. 43];
Cure kinetics model 4: by Johnston and Hubert (1996) [Ref. 51].
Tables 4-7 and 4-8 summarize the details of these cure kinetics models and the definition of the required parameters.
Table 4-7 Cure Kinetics Models Embedded in Marc

Model Equations Parameters
d A 1 A 2 A 3 E 1 E 2 E 3 B  C H R
------- =  K 1 + K 2    1 –    B –  
dt
Model 1   c
Lee, Loos and Springer d

------- = K 3  1 –  
(1982) [Ref. 42] dt
  c
–  E i   RT 
Ki = Ai e
(Included in Model 2) A E m n H R

d
------- = K m  1 –   n
Lee, Chiu, and Lin dt
(1992)[Ref. 43]; White and K = Ae –  E   RT 
Hahn (1992) [Ref. 47]
(Included in Model 2) d A E n H R
------- = K  1 –   n
Kenny (1992) [Ref. 48]; dt
Scott (1991)[Ref. 49] K = Ae –  E   RT 
d A 1 A 2 E 1 E 2 m n H R
Model 2 (combined Model) ------- =  K 1 + K 2  m   1 –   n
dt
Scott (1991)[Ref. 49] –  E i   RT 
Ki = Ai e
Model 3 d A 1 A 2 E 1 E 2 l m n H R
------- = K 1  1 –   l + K 2  m  1 –   n
Lee, Chiu, and Lin dt
–  E i   RT 
(1992)[Ref. 43] Ki = Ai e
d K m  1 –   n A E m n C  C 0  C T H R
Model 4 ------- = ---------------------------------------------------------------
dt C   –   C0 +  C T T  
Johnston and Hubert 1+e
(1995)[Ref. 51] –  E i   RT 
Ki = Ai e
CHAPTER 4 195
Table 4-8 Parameters used in the Embedded Resin Cure Kinetics Models
Variable Description Units
 Resin degree of cure. -
T Resin temperature K or R
HR Total resin heat of reaction ( = 0 to 1) J/kg or BTU/lbm
Ai Pre-exponential factor. /s
E i Activation energy. J/mol or BTU/mol
l Equation superscript. -
m Equation superscript. -
n Equation superscript. -
R Gas constant J/(mol K) or BTU/(mol R)
C Diffusion Constant -
C 0 Critical Resin Degree of Cure -
C T The Increase in Critical Resin Degree of Cure with Temperature -
In Marc, the table definition allows the user to define the cure rate as a function of the degree of cure and temperature if
the resin kinetics is not defined by the four embedded models.
In addition, the degree of cure and cure rate can also be calculated through the UCURE user subroutine.
Cure Shrinkage Strain

The cure shrinkage strain is calculated according to the volumetric shrinkage due to curing process. The resin degree of
S S
cure shrinkage is defined as the ratio of volumetric cure shrinkage  V r  and maximum volumetric cure shrinkage  V r 
S
Vr
of the resin material, as  S = ---------
- . Equation (4-268) then calculates the cure shrinkage strain:
S
Vr
S 13
 rS =  1 + V r  –1 (4-268)
Considering the anisotropic shrinkage behavior of composite with resin, the strain components of the composite are
calculated by using the directional Cure Shrinkage Coefficient (CSC) matrix by Equation 4-269.
 iSj = CS  C ij   rS  i j = 1 ,2 ,3 (4-269)
S
The calculation of volumetric cure shrinkage, V r , is conducted based on the cure shrinkage models defined by users. Marc
has two cure shrinkage models available. Note that a positive cure shrinkage strain indicates a reduction of volume.
Tables 4-9 and 4-10 summarize the details of the cure kinetics models and the definition of the requires parameters.
Table 4-9 Resin Cure Shrinkage Models

Model Equations Parameters
S   C 1
V r = 0.0
Model 1 S S 2
Vr = A * S +  Vr – A  * S C 1    C 2
S
Bogetti and Gillespie (1992) S S V r   C 1  C 2 A
Vr = Vr
[Ref. 41]   C 2
 –  C1
S = --------------------------
-
 C2 –  C1
S B   – c 
Model 2 V r = V rS  *10   C S
White and Hahn 1992 V r   C B
S S   C
[Ref. 47] Vr = Vr
Table 4-10 Parameters used in the Resin Cure Shrinkage Models

Variable Description Units
S
Vr Resin volumetric cure shrinkage -
S
Vr Total volumetric resin shrinkage from = 0 to 1. -
s The degree of cure shrinkage. -
C 1 Degree of cure after which the resin shrinkage begins (model 1). -
C 2 Degree of cure after which the resin shrinkage stops (model 1). -
A Linear cure shrinkage coefficient -
C Degree of cure after which the resin shrinkage begins (model 2). -
B Cure shrinkage model superscript. -
CHAPTER 4 197
References
1. Key, S. W. and R. D. Krieg, 1982, “On the Numerical Implementation of Inelastic Time-Dependent, Finite Strain
Constitutive Equations in Structural Mechanics”, Computer Methods in Applied Mechanics in Engineering, V.33, pp.
439–452, 1982.
2. Bathe, K. J., E. Ramm, and E. L. Wilson. “Finite Element formulation for Large Deformation Dynamic Analyses”,
International Journal for Numerical Methods in Engineering, V. 9, pp. 353-386, 1975.
3. Montgomery, D. C., Design and Analysis of Experiments, (2nd ed.) John Wiley and Sons, 1984.
4. Spendley, W., Hext, G. R., Himsworth, F. R., “Sequential Application of Simplex Designs in Optimisation and
Evolutionary Operation”, Technometrics, Vol. 4, No. 4,
pp. 441-461 (1962).
5. Gill, P. E., Murray, W., Saunders, M. A., Wright, M. H., “Sequential Quadratic Programming Methods for
Nonlinear Programming”, NATO-NSF-ARO, Advanced Study Inst. on Computer Aided Analysis and
Optimization of Mechanical System Dynamics, Iowa City, IA, Aug. 1-2, 1983, pp. 679-700.
6. Bathe, K. J. Finite Element Procedures Prentice-Hall, 1996.
7. Wilkinson, J. H. The Algebraic Eigenvalue Problem. Oxford: Clarendon Press, 1965.
8. Zienkiewicz, O. C., and Taylor, L. C. The Finite Element Method. Sixth Ed., Vol. 1 & 2. Oxford: Elsevier
Butterworth-Heinemann, 2005.
9. Barsoum, R. S. “On the Use of Isoparametric Finite Elements in Linear Fracture Mechanics.” Int. J. Num. Methods
in Engr. 10, 1976.
10. Dunham, R. S. and Nickell, R. E. “Finite Element Analysis of Axisymmetric Solids with Arbitrary Loadings.” No.
67-6. Structural Engineering Laboratory, University of California at Berkeley, June, 1967.
11. Hibbitt, H. D., Marcal, P. V. and Rice, J. R. “A Finite Element Formulation for Problems of Large Strain and Large
Displacement.” Int. J. Solids Structures 6 1069-1086, 1970.
12. Houbolt, J. C. “A Recurrence Matrix Solution for the Dynamic Response of Elastic Aircraft.” J. Aero. Sci. 17, 540-
550, 1950.
13. McMeeking, R. M., and Rice, J. R. “Finite-Element Formulations for Problems of Large Elastic-Plastic
Deformation.” Int. J. Solids Structures 11, 601-616, 1975.
14. Marc Update. “The Inverse Power Sweep Method in Marc.” U.S. Ed., vol. 3, no. 1, February, 1984.
15. Simo, J. C., Taylor, R. L., and Pister, K. S. “Variational and Projection Methods for the Volume Constraint in
Finite Deformation Elasto-Plasticity,” Comp. Meth. in App. Mech. Engg., 51, 1985.
16. Simo, J. C. and Taylor, R. L. “A Return Mapping Algorithm for Plane Stress Elasto-Plasticity,” Int. J. of Num.
Meth. Engg., V. 22, 1986.
17. Wriggers, P., Eberlein, R., and Reese, S. “A Comparison of Three-Dimensional Continuum and Shell Elements
for Finite Plasticity,” Int. J. Solids & Structures, V. 33, N. 20-22, 1996.
18. Simo, J. C. and Taylor, R. L. “Quasi-Incompressible Finite Elasticity in Principal Stretches, Continuum Basis and
Numerical Algorithms,” Comp. Meth. App. Mech. Engg., 85, 1991.
19. Marcal, P. V. “Finite Element Analysis of Combined Problems of Nonlinear Material and Geometric Behavior.”
in Proceedings of the ASME Computer Conference, Computational Approaches in Applied Mechanics,
Chicago, 1969.
20. Melosh, R. J., and Marcal, P. V. “An Energy Basis for Mesh Refinement of Structural Continua.” Int. J. Num.
Meth. Eng. 11, 1083-1091, 1971.
21. Morman, K. N., Jr., Kao, B. G., and Nagtegaal, J. C. “Finite Element Analysis of Viscoelastic Elastomeric
Structures Vibrating about Nonlinear Statically Stressed Configurations.” SAE Technical Papers Series 811309,
presented at 4th Int. Conference on Vehicle Structural Mechanics, Detroit, November 18-20, 1981.
22. Morman, K. N., Jr., and Nagtegaal, J. C. “Finite Element Analysis of Small Amplitude Vibrations in Pre-Stressed
Nonlinear Viscoelastic Solids.” Int. J. Num. Meth. Engng, 1983.
23. Nagtegaal, J. C. “Introduction in Geometrically Nonlinear Analysis.” Int. Seminar on New Developments in the
Finite Element Method, Santa Marherita Ligure, Italy, 1980.
24. Nagtegaal, J. C., and de Jong, J. E. “Some Computational Aspects of Elastic-Plastic Large Strain Analysis.” in
Computational Methods in Nonlinear Mechanics, edited by J. T. Oden. North-Holland Publishing Company,
1980.
25. Newmark, N. M. “A Method of Computation for Structural Dynamics.” ASCE of Eng. Mech. 85, 67-94, 1959.
26. Parks, D. M. “A Stiffness Derivative Finite Element Technique for Determination of Elastic Crack Tip Stress
Intensity Factors.” International Journal of Fractures 10, (4), 487-502, December 1974.
27. Timoshenko, S., Young, D. H., and Weaver, Jr., W. Vibration Problems in Engineering. John Wiley, New York:
1979.
28. Chung, J. and Hulbert, G.M., “A family of single-step Houbolt time integration algorithms for structural
dynamics”, Comp. Meth. in App. Mech. Engg., 118, 1994.
29. Shih, C.F., Moran B. and Nakamura K., “Energy release rate along a three-dimensional crack front in a thermally
stressed body”, International Journal of Fracture, vol. 30, pp. 79–102, 1986.
30. Anderson, T.L., Fracture Mechanics: Fundamentals and Applications (2nd ed.) CRC Press, 1995.
31. Yoon, J.W., Yang, D.Y. and Chung, K., “Elasto-plastic finite element method based on incremental deformation
theory and continuum based shell elements for planar anisotropic sheet materials”, Comp. Methods Appl. Mech.
Eng., 174, 23 (1999).
32. Shih, C.F. and Asaro, R.J., “Elastic-Plastic Analysis of Cracks on Bimaterial Interfaces: Part I – Small Scale
Yielding”, Journal of Applied Mechanics, Vol. 110, pp. 299–316, June 1988.
33. Krueger, R., “Virtual Crack Closure Technique: History, Approach and Applications”, Appl. Mech. Rev., Vol. 57:2,
pp. 109–143, March 2004.
34. Knops, B., “Numerical Simulation of Crack Growth in Pressurized Fuselages”, Doctoral thesis, Faculty of
Aerospace Engineering, Delft University of Technology, The Netherlands, 1994.
35. Chung, J. and Hulbert, G.M., “A time integration algorithm for structural dynamics with improved numerical
dissipation: The Generalized-Method”, Journal of Applied Mechanics, Vol. 60, pp. 371 - 375, June 1993.
36. Hilber, H.M., Hughes, T.J.R., and Taylor, R.L., “Improved Numerical Dissipation for Time Integration
Algorithms in Structural Dynamics”, Earthquake Engineering and Structural Dynamics, Vol. 5, pp. 283-292, 1977
37. Wood, W.L., Bossak, M., and Zienkiewicz, O.C., “An Alpha Modification of Newmark’s Method”, International
Journal for Numerical Methods in Engineering, Vol., 15, pp. 1562-1566, 1981.
38. J. Reeder, 3D Mixed-Mode Delamination Fracture Criteria - An Experimentalist's Perspective. Proceedings of
American Society for Composites, 21st Annual Technical Conference, Dearborn, Michigan, 2006.
CHAPTER 4 199
39. M.L. Benzeggagh and M. Kenane, Measurement of Mixed-Mode Delamination Fracture Toughness of
Unidirectional Glass/Epoxy Composites with Mixed-Mode Bending Apparatus, Composites Science and Technology,
vol. 56, pp. 439-449, 1996.
40. Yih-Farn Chen, Modeling Analysis of Defect Formation during Thick Laminate Compression Molding Process,
Presented at the American Helicopter Society 58th Annual Forum, Montreal, Canada, June 11-13, 2002.
41. Bogetti, T. A., Gillespie, J. W., 1992 Process-Induced Stress and Deformaton in Thick Section thermoset
Composite Laminates, Journal of Composite Materials, 26, 626-660.
42. Lee, W.I., A. Loos and G.S. Springer. 1982 Cure Kinetics and Viscosity of Fiberite 976 Resin, Journal of Composite
Materials, 21; 243-261.
43. S.N. Lee, M.T. Chiu, and H.S. Lin, 1992 Kinetic Model for the Curing Reaction of a Tetraglycidyl Diamino
Diphenyl Methane/Diamino Diphenyl Sulfone (TGDDM/DDS) Epoxy Resin System, Polymer Engineering and
Science 32 (15); pp. 1037-1046.
44. T. G. Gutowski, et. al. The consolidation of laminate composites, Journal of Composite Materials, Vol. 21, February
1987, pp. 172 – 187.
45. W. B. Young, Consolidation and cure simulations for laminated composites, Polymer Composites, February 1996,
Vol. 17, No. 1, pp. 142 – 148
46. Min Li and Charles Tucker, Optimal Curing for Thermoset Matrix Composites: Thermochemical and
Consolidation Considerations, Polymer Composites, Aug. 30, 2001
47. S. R. White and H.T. Hahn, 1992 Process Modeling of Composite Materials: Residual Stress Development during
Cure. Part I. Model Formulation, Journal of Composite Materials, 26, 2402-2422.
48. J.M. Kenny, 1992 Integration of Process Models with Control and Optimization of Polymer Composites
Fabrication, Proceedings of the Third Conference on Computer Aided Design in Composite Materials Technology, pp.
530-544.
49. E.P. Scott, 1991 Determination of Kinetic Parameters Associated with the Curing of Thermoset Resins Using
Dielectric and DSC Data, Composites: Design, Manufacture, and Application, ICCM/VIII, Honolulu, 1991, pp. 10-
0- 1-10.
50. George S. Springer, Resin Flow During the Cure of Fiber Reinforced Composites, Journal of Composite Materials
1982 16: 400-410.
51. P. Hubert, 1996 Aspects of Flow and Compaction of Laminated Composite Shapes During Cure, Ph.D. Thesis,
The University of British Columbia, Vancouver, B.C.
52. Regulatory Guide 1.92, Revision 2, U.S. Nuclear Regulatory Commission.
Chapter 5: Material Library
5 Material Library

Overview 201

Linear Elastic Material 201
 Composite Material 203

Gasket 241
 Nonlinear Hypoelastic Material 247
 Thermo-Mechanical Shape Memory Model - Type 1 261

Thermo-Mechanical Shape Memory Model - Type 2 271
 Mechanical Shape Memory Model 280
 Elastomer 286
 Time-independent Inelastic Behavior 306
 Time-dependent Inelastic Behavior 355
 Frequency-dependent Material Behavior 388
 Temperature Effects and Coefficient of Thermal Expansion 445
 Time-Temperature-Transformation 459
 Low Tension Material 461

Soil Model 463

Damage Models 475
 Nonstructural Materials 494
 References 496
CHAPTER 5 201
Material Library
Overview
This chapter describes the material models available in Marc. The models range from simple linear elastic materials to
complex time- and temperature-dependent materials. This chapter provides basic information on the behavior of various
types of engineering materials and specifies the data required by the program for each material. For example, to characterize
the behavior of an isotropic linear elastic material at constant temperatures, you need only specify Young's modulus and
Poisson's ratio. However, much more data is required to simulate the behavior of material that has either temperature or
rate effects. References to more detailed information are cited in this chapter.
Data for the materials is entered into Marc either directly through the input file or by user subroutines. Each section of this
chapter discusses various options for organizing material data for input. Each section also discusses the constitutive
(stress-strain) relation and graphic representation of the models and includes recommendations and cautions concerning
the use of the models.
Linear Elastic Material

Marc is capable of handling problems with either isotropic linear elastic material behavior or anisotropic linear elastic
material behavior.
The linear elastic model is the model most commonly used to represent engineering materials. This model, which has a
linear relationship between stresses and strains, is represented by Hooke’s Law. Figure 5-1 shows that stress is proportional
to strain in a uniaxial tension test. The ratio of stress to strain is the familiar definition of modulus of elasticity (Young’s
modulus) of the material.
E (modulus of elasticity) = (axial stress)/(axial strain) (5-1)
Stress
Strain
Figure 5-1 Uniaxial Stress-Strain Relation of Linear Elastic Material
Experiments show that axial elongation is always accompanied by lateral contraction of the bar. The ratio for a linear elastic
material is:
v = (lateral contraction)/(axial elongation) (5-2)
This is known as Poisson’s ratio. Similarly, the shear modulus (modulus of rigidity) is defined as:
G (shear modulus) = (shear stress)/(shear strain) (5-3)
It can be shown that for an isotropic material

G = E  21 + v (5-4)
The shear modulus G can be easily calculated if the modulus of elasticity E and Poisson’s ratio v are known.
Most linear elastic materials are assumed to be isotropic (their elastic properties are the same in all directions). Anisotropic
material exhibits different elastic properties in different directions. The significant directions of the material are labeled as
preferred directions, and it is easiest to express the material behavior with respect to these directions.
The stress-strain relationship for an isotropic linear elastic method is expressed as:
 ij =  ij  kk + 2G i j (5-5)
where  is the Lame constant and G (the shear modulus) is expressed as:
 = E    1 +    1 – 2   and (5-6)
G = E  21 + 
The material behavior can be completely defined by the two material constants E and v .
Use the ISOTROPIC model definition option for the input of isotropic linear elastic material constants E (Young’s modulus)
and v (Poisson’s ratio).
The effects of these parameters on the design can be determined by using the DESIGN SENSITIVITY parameter. The optimal
value of the elastic properties for linear elastic analysis can be determined using the DESIGN OPTIMIZATION parameter.
The stress-strain relationship for an anisotropic linear elastic material can be expressed as:
 ij = C ijkl  kl (5-7)
The values of C ijkl (the stress-strain relation) and the preferred directions (if necessary) must be defined for an anisotropic
material. For example, the orthotropic stress-strain relationship for a plane stress element is:
E1  21 E 1 0
1
C = -------------------------------  12 E 2 E2 0 (5-8)
 1 –  12  21 
0 0  1 –  12  21 G
There are only four independent constants in Equation (5-8).

To input anisotropic stress-strain relations, use the ORTHOTROPIC or ANISOTROPIC model definition option and the ANELAS or
HOOKLW user subroutine. The ORTHOTROPIC option allows as many as 9 elastic constants to be defined. The ANISOTROPIC
option allows as many as 21 elastic constants to be defined. If the anisotropic material has a preferred direction, use the
ORIENTATION model definition option or the ORIENT user subroutine to input a transformation matrix.
A Poisson’s ratio of 0.5, which would be appropriate for an incompressible material, can be used for the following elements:
Herrmann, plane stress, shell, truss, or beam. A Poisson’s ratio which is close (but not equal) to 0.5 can be used for constant
dilation elements and reduced integration elements in situations which do not include other severe kinematic constraints.
Using a Poisson’s ratio close to 0.5 for all other elements usually leads to behavior that is too stiff. A Poisson’s ratio of 0.5
CHAPTER 5 203
Material Library
can also be used with the updated Lagrangian formulation in the multiplicative decomposition framework using the
standard displacement elements. In these elements, the treatment for incompressibility is transparent to you.
Composite Material
Composite materials are composed of layers of different materials (or layers of the same anisotropic material) with various
layer thicknesses and different orientations. The material in each layer may be either linear or nonlinear. Tightly bonded
layers (layered materials) are often stacked in the thickness direction of beam, plate, shell structures, or solids.
Figure 5-2 identifies the locations of integration points through the thickness of beam and shell elements with/without the
COMPOSITE option.
Note that when the COMPOSITE option is used, as shown on the left, the layer points are positioned midway through each
layer. When the COMPOSITE option is not used, the layer points are equidistantly spaced between the top and bottom
surfaces. Marc forms a stress-strain law by performing numerical integration through the thickness. If the COMPOSITE option
is used, the trapezoidal method is employed; otherwise, Simpson’s rule is used.
*
* *
* *
* *
* *
Beams or Shells with Beams or Shells without
Composite Option Composite Option
Figure 5-2 Integration Points through the Thickness of Beam and Shell Elements
Figure 5-3 shows the location of integration points through the thickness of composite continuum elements. Marc forms
the element stiffness matrix by performing numerical integration based on the standard isoparametric concept.
* *
* *
* *
* *
Figure 5-3 Integration Points through the Thickness of Continuum Composite Elements
8
First Layer (5-6-7-8 face)
ζ Non Composite
Side View ζ
Composite
Side View
5
7
η 1 1
2
6 3 4
2
4 5
Bending
3 6 7 Strain
Continuous
1 8
4
ξ 9 10
3 11
5 12
Integration Point
Last Layer (1-2-3-4 face) 2
Figure 5-4 Element 185 Connectivity and Integration Points for Composite and Non Composite Materials
Layered Materials
To model layered materials including plates, shells, beams, and solids with Marc, use the COMPOSITE option. In this option,
three quantities are specified on a layer-by-layer basis: material identification number, layer thickness, and ply angle. The
entire set of data (a “composite group”) is then associated with a list of elements. For each individual layer, various
constitutive laws can be used. The layer thickness can be constant or variable (in the case of variable total thickness
elements), and the ply angle can change from one layer to the next. The orientation of the 0o ply angle within each element
is defined in the ORIENTATION option. The ply thickness and the ply angle can be used as design variables in a design
sensitivity analysis. The optimal values can be determined using the design optimization capability for linear
elastic analysis.
The material identification number specified in the COMPOSITE option, is cross-referenced with the material identification
number supplied in the ISOTROPIC, ORTHOTROPIC, ANISOTROPIC, NLELAST, TEMPERATURE EFFECTS, ORTHO TEMP, WORK HARD, and
STRAIN RATE options. The ISOTROPIC, ORTHOTROPIC, and ANISOTROPIC model definition options allow you to input material
constants such as Young’s modulus, Poisson’s ratio, shear modulus, etc. The TEMPERATURE EFFECTS and ORTHO TEMP options
allow for input of temperature dependency of these material constants. Material constants for a typical layer are as follows:
ti thickness of the ith layer

Young’s moduli E x x E y y E z z
Poisson’s ratios v x y v y z v z x
Shear moduli G x y G y z G z x
 density
 x x  yy coefficients of thermal expansion
y yield stress
Mat material identifier associated with temperature-dependent properties and workhardening data
The ANELAS, HOOKLW, ANEXP, and ANPLAS user subroutines can be used for the anisotropic behavior of elastic constants,
coefficient of thermal expansion, and yield condition.
CHAPTER 5 205
Material Library
In models where there is ply drop-off, it is possible to specify a user ply layer id to simplify postprocessing. In this way, the
top, middle, and bottom layers may all have the same id even if the elements have a different number of layers.
There are eight given classes of strain-stress relations. The class of a particular element depends on the number of direct
(NDI) and shear (NSHEAR) components of stress. Table 5-1 lists the eight classes of elements.
Table 5-1 Classes of Stress-Strain Relations

Class 1 NDI = 1, NSHEAR = 0
Beam Elements 5, 8, 13, 16, 23, 46, 47, 48, 52, 64, 77, 79 and Rebar Elements
   =  1  Ex x    
Axisymmetric Shells 15 and 17
  xx  1  Ex x –y x  Ey y  x x 
  =  
  yy  –x y  Ex x 1.  E y y   y y 
y x = x y Ey y  Ex x
Beam Elements 14, 45, 76, 78
  1  Ex x 0  
  =  
  0 1  Gx y  

Plane Stress, Plates and Thin Shells 49 and 72
  xx  1  Ex x –y x  Ey y 0  x x 
   
  yy  = –x y  Ex x 1  Ey y 0  y y 
   
  xy  0 0 1  Gx y  x y 
y x = x y  Ey y  Ex x 
Table 5-1 Classes of Stress-Strain Relations (continued)

Thick Axisymmetric Shells 1 and 89
  mm  1  Em m –  m  E  0  m m 
   
    = –m   Em m 1  E  0    
   
 T  0 0 1  Gm   T 

Plane Strain, Axisymmetric with No Twist, Elements 151-154.
 x x  1  Ex x –  y x  E y y –  zx  E zz 0  x x 
   
 y y  –x y  Ex x 1  Ey y –  zy  E zz 0  y y 
  =  
 z z  –  x z  E x x –  y z  E yy 1  E zz 0   zz 
   
 x y  0 0 0 1  Gx y  x y 
 yx =  x y E y y  E x x  z y =  y z E zz  E y y  x z =  zx E x x  E zz
Thick Shell, Elements 22, 75, and 140
  xx  1  Ex x –y x  Ey y 0 0 0  x x 
   
  yy  –x y  Ex x 1  Ey y 0 0 0  y y 
   
  xy  = 0 0 1  Gx y 0 0  x y 
   
  yz  0 0 0 1  Gy z 0  y z 
   
  zx  0 0 0 0 1  G zx   zx 
y x = x y Ey y  Ex x
CHAPTER 5 207
Material Library
Table 5-1 Classes of Stress-Strain Relations (continued)

Three-Dimensional Brick Elements, Elements 149, 150
  xx  1  Ex x –  y x  E y y –  z x  E zz 0 0 0  x x 
   
 yy  –x y  Ex x 1  Ey y –  z y  E zz 0 0 0  yy 
   
  zz  –x z  Ex x –y z  Ey y 1.  E zz 0 0 0   zz 
  =  
  xy  0 0 0 1  Gx y 0 0  x y 
   
  yz  0 0 0 0 1  Gy z 0  y z 
   
  zx  0 0 0 0 0 1  G zx   zx 
Classical Lamination Theory for Multi-Layered Shells

Basic CLT Theory
The stress tensor at a point inside a shell element is defined as:
 = G   – z  (5-9)
where G is the tangent matrix connecting stress and strain tensors; z is the coordinate of the point in the thickness
direction;  is the strain tensor on the middle surface of the shell and  is the tensor of curvature.
The membrane forces is given by:
f =   dz (5-10)
The bending moments is given by:

m =   – z  dz (5-11)
Substituting Equations (5-9) into Equations (5-10) and (5-11) and use the definitions:
hG 1 =  G dz (5-12)
h 2 G4 =   – z G dz (5-13)
IG 2 =  z 2 G dz (5-14)
we obtain:
 f  hG 1 h 2 G 4   
  =   (5-15)
 m  h 2 G 4 IG 2   
In Equations (5-12) though (5-15), h is the shell thickness and I = h 3  12 .

In the case where the composite material remains linear elastic, it is possible to choose alternative computational procedures
to improve computational time. This maybe activated using either the SHELL SECT parameter or the COMPOSITE option:
1. The first procedure is for linear elastic composite materials where no temperature dependent materials are present
and thermal strains occur. This substantially reduces the computational time and the memory requirements.
2. The second procedure is for linear elastic composite material, but either temperature dependent materials or
thermal strains are present. This substantially reduces the computational time.
PSHELL Option
PSHELL option is based on the classical lamination theory (also known as equivalent stiffness method). This option allows
you to define the membrane, bending, transverse shear, and coupling properties of the shell elements independently.
Shell made of homogeneous materials can be modeled with the PSHELL option by simply using the same material properties
for stiffness calculation of membrane, bending, and transverse shear deformations, and using nothing for the coupling
part if there is no offset of the shell middle surface. However, for homogeneous shell structures, it is more efficient to use
the standard shell technique which is relatively easy and inexpensive, generally more accurate, capable to deal with
nonlinear material behavior.
Shell structures with layered composite materials can be solved using a full integration technique or the PSHELL option. The
full layer integration technique is very general. It can simulate material behavior ranging from a simple linear material to a
very complex material with nonlinearity and failure mechanism. The use of the PSHELL option is limited to linear-elastic
behavior. By use of this option, composite layers are converted into one layer with equivalent stiffness behaviors. This way,
only one integration point is needed across the thickness which makes PSHELL very useful for shell structures with layered
composite materials. It is particularly attractive when the number of composite layers is large, because analysis of these
smeared shell structures uses less computer time and storage space.
The smeared material matrices G 1 , G 2 , and G 4 are defined in Equations (5-12), (5-13), and (5-14) representing membrane,
bending and coupling stiffness, respectively.
G 1 in Marc contains both membrane and transverse shear parts. Unless you want to adjust transverse shear stiffness, there
is no need to define new types of materials.
Generally G 1 , G 2 , and G 4 can be described by the ANISOTROPIC option.
Material Preferred Direction

Every element type in Marc has a default orientation (that is, a default coordinate system) within which element stress-
strain calculations take place. This system is also assumed to be the coordinate system of material symmetry. This is
especially important for non-isotropic materials (orthotropic, anisotropic, nlelast, or composite materials). With the
ORIENTATION option, you specify the orientation of the material axes of symmetry (relationship between the element
coordinate system and the global coordinate system, or the 0o ply angle line, if composite) in one of seven different ways:
1. As a specific angle offset from an element edge,
2. As a specific angle offset from the line created by two intersecting planes,
CHAPTER 5 209
Material Library
3. As a particular coordinate system specified by user-supplied unit vectors,

4. As specified by the ORIENT2 user subroutine,
5. By referencing a coordinate system defined by the COORD SYSTEM option.
6. By specifying rotations around the local element system,
7. By defining NURBS curves and obtaining the orientation as the tangent of the closest point on a curve.
For the first option (EDGE I-J orientation type), the intersecting plane is defined by the surface normal vector and a vector
parallel to the vector pointing from element node I to element node J. The intersection of this plane with the surface
tangent plane defines the 0o orientation axis. (See Figure 5-5.) The orientation angle is measured in the tangent plane
positive about the surface normal.
n = Normal to Surface Tangent Plane

Node I
Vector Parallel to Edge
I-J Projected onto
Surface Tangent Plane
Integration Point
 
Node J
00
Ply Angle Direction 1 of
preferred coordinate
system (fiber direction
in the ply)
Element Surface
Z
 = Orientation Angle (Positive

Y Right-hand Rotation About n)
X  = Ply angle (if COMPOSITE)
Figure 5-5 Edge I-J Orientation Type
For the second option (global plane orientation type), the intersecting plane is the chosen global coordinate plane. The
intersection of this plane with the surface tangent plane defines the 0o orientation axis. (See Figure 5-6.)
n - Normal to Surface Tangent Plane

Global ZX Plane
Surface 
Tangent 
Plane Direction 1
of preferred
coordinate
Int system
pre erse
ce
fer ctio
rfa
red n o
Su
pla f
nt
ne
me
Z

Ele Right-hand Rotation About n)
Y
X
 = Ply angle (if COMPOSITE)
Figure 5-6 Global ZX Plane Orientation Type
The third option (user-defined plane orientation type) makes use of one or two user-defined vectors to define the
intersecting plane. Using a single vector, the intersecting plane is that plane which contains the user vector and the chosen
coordinate axis. Using two user vectors, the intersecting plane is that plane which contains both of them. (See Figure 5-7.)
n = Normal to Surface Tangent Plane

Tangent Plane
u = user-defined vector
Surface Tangent
Plane
 
global
X Direction 1 of
Preferred
Intersection of Coordinate
Two Planes System
Element Surface
Z

Y Right-hand Rotation About n)
X  = Ply angle (if COMPOSITE)
Figure 5-7 User Defined XU Plane Orientation Type
Orientation type 3-D ANISO also makes use of two user-defined vectors, but in this case, the first vector defines the first
(1) principal direction and the second vector defines the second (2) principal direction. (See Figure 5-8.)
CHAPTER 5 211
Material Library
U2 = User Vector 2
U3 = U1 x U2
U1 = User Vector 1
Z
U1 = Direction 1 of Preferred Coordinate System

Y
U2 = Direction 2 of Preferred Coordinate System
X
Figure 5-8 3-D ANISO Orientation Type
In the fourth option, the ORIENT2 user subroutine is used for defining the material coordinate system.
In the fifth option, the material orientation is defined by the coordinate system defined with the COORD SYSTEM option. If
shell elements are used, the local x-axis is projected on the shell as shown in Figure 5-9.
x
MCID Coordinate System
z
G2
y G3
ymaterial
xmaterial
G4
G1
Figure 5-9 COORD SYSTEM when Shell Elements are used
The sixth option is available for hexahedral elements. The material orientation is defined as rotations around the three axes
of the element coordinate system. See Figure 5-10. The first preferred direction joins the centroids of faces 4-1-5-8 and 3-2-
6-7. A second vector joins faces 1-2-6-5 and 4-3-7-8. The third preferred direction is given by the cross product of the first
preferred direction and this vector. The second preferred direction is given by the cross product of the third and first
preferred directions. This system is then rotated around the three local axes by the three given angles.
8
7
5 6
4 3
2
Figure 5-10 Element Coordinate System Definition of Hexahedral Element
The seventh option method uses projection onto a curve. One or more NURBS curves are used for defining the preferred
system. A list of curves are given as input. These curves must be defined with the CURVES model definition option and only
the NURBS variant is allowed. Using the centroid of the element, the closest point on any of the given curves is found.
The first preferred direction is given by the tangent vector at this point. For 2-D elements, the second preferred direction
is given by the cross product of the global z direction and the first preferred direction. For 3-D elements, this option is only
supported for solid shell elements and solid composite elements. The third preferred direction is given by the thickness
direction and the second preferred direction by the cross product between the third and first preferred direction. The first
preferred direction is recalculated as the cross product between the second and third preferred directions to insure that we
have an orthogonal system.
Post codes 691 and 694 can be used to obtain the first and second orientation vectors of the material coordinate system. It
should be noted that these vectors are element based quantities and only account for the zero degree ply orientation. For
specific layer orientations, these vectors are further modified by the layer angle given through post code 697. Mentat can
be used to plot both the element orientations (zero degree ply angle) and the layer orientations. These postprocessing
features can be used to verify if the initial orientations are setup correctly (especially for the ORIENT2 user subroutine or the
COORD SYSTEM option) and to evaluate the evolution of the orientation vectors as the structure deforms. More details on
these orientation vectors are given under the POST option in Marc Volume C: Program Input.
For 3-D solid composite elements (types 149, 150, 175, 176) and for solid shell elements (type 185), there are two choices
available for the treatment of orientation options 3 (3D ANISO), 4 (USER SUB. ORIENT2), 5 (COORD SYSTEM), 6 (3D
LOCAL) and 7 (3D CURVE).
a. Use Orientations As Is – in this case, depending on the orientation option used, solid composites and solid shells
are treated differently. The various options are discussed in the following table.
CHAPTER 5 213
Material Library
Orientation Option Solid Composite Elements Solid Shell Elements Comments

3D ANISO No projection onto Element No Projection onto Given orientation vectors directly define
Plane (Figure 5-8). Element Plane material coordinate directions
(Figure 5-8).
USER SUB. Projects onto Element Plane Projects onto Element Ensures that third direction is always in ply
ORIENT2 (Figure 5-11). Plane (Figure 5-11). thickness direction
COORD SYSTEM No projection onto Element Projects onto Element For solid composites, given coordinate
Plane (Figure 5-8). Plane (Figure 5-11). system directions directly define material
coordinate directions
3D LOCAL Uses R-S-T directions of Uses direction cosines If 3D LOCAL directions are changed
element (from centroid of of element (same as through rotation angles, or if R/S
one face to another). See element coordinate directions for solid composite happen to
Figure 5-10. system) be in thickness direction, then the third
orientation direction is no longer along ply
thickness
3D CURVE Projects onto Element Plane Projects onto Element Ensures that third direction is always in ply
(Figure 5-11). Plane (Figure 5-11). thickness direction
Figure 5-11 Projection of Coordinate System to Element Plane. After Projection, the 1-2 Directions are in the Plane of the
Ply and the 3 Direction is in the Direction of the Ply Thickness.
b. Project Orientations to Element Plane – in this case, all five options (3D ANISO, USER SUB ORIENT2, COORD
SYSTEM, 3D LOCAL, and 3D CURVE) are always projected onto the element plane for both solid composites
and solid shells. The projection for COORD SYSTEM is demonstrated in Figure 5-11. For composites, projecting
orientations to the ply plane is expected to better simulate reality, wherein material properties are typically
defined in the ply plane and perpendicular to the ply plane.
Material Dependent Failure Criteria

Calculations of user specified failure criteria on a layer by layer basis are available in Marc. The available criteria are:
maximum stress (MX STRESS), maximum strain (MX STRAIN), TSAI-WU, HOFFMAN, HILL, HASHIN, HASHIN
FABRIC, HASHIN TAPE, PUCK, and the UFAIL user subroutine. During each analysis, up to three failure criteria can be
selected; failure indices and strength ratios are calculated and printed for every integration point. The FAIL DATA model
definition option is used for the input of failure criteria data. When the table driven input format is used, the failure
material parameters may reference a table to introduce temperature dependent behavior.
A simple description of these failure criteria is given below:
1. Maximum Stress Criterion
CHAPTER 5 215
Material Library
At each integration point, Marc calculates six failure indices FI and six strength ratios SR: The six failure indices
are given by:
  1
 ------- if 1  0
 X t
1. (5-16)
  1
 – ------- if 1  0
 X c
  2
 ------ if 2  0
 Yt 
2. (5-17)
  2
 – ------ if 2  0
 Y c

 -----3- if 3  0
 Zt 
3. (5-18)
–
-----3- if 3  0
 Z 
c
  12  (5-19)
4.  --------
S 12 
-
  23  (5-20)
5.  -------
S 23 
-
  31  (5-21)
6.  -------
S 31 
-
where ,
X t X c are the maximum allowable stresses in the 1-direction in tension and compression.
Y t Y c are maximum allowable stresses in the 2-direction in tension and compression.
Z t Z c are maximum allowed stresses in the 3-direction in tension and compression.
S 12 maximum allowable in-plane shear stress.
S 23 maximum allowable 23 shear stress.
S 31 maximum allowable 31 shear stress.
For the Maximum Stress Failure Criterion, the strength ratios SR are the reciprocals of the corresponding failure
indices FI.
1.0
SR = ---------
FI
For example, the sixth strength ratio for the Maximum Stress Failure Criterion is  S 31   31  . Note that if the
maximum allowable stresses are not defined or if the actual stresses are 0, then the strength ratio is set to 100.
2. Maximum Strain Failure Criterion
At each integration point, Marc calculates six failure indices and six strength ratios. The six failure indices are given
by:
1
 ------
- if 1  0
 e 1t
1. (5-22)
1 
 – ------- if 1  0
 e 
1c
2
 ------
- if 2  0
 e 2t
2. (5-23)
2 
 – ------- if 2  0
 e 
2c
3
 ------
- if 3  0
 e 3t
3. (5-24)
3 
 – ------- if 3  0
 e 
3c
 12 
 ------- (5-25)
4.  g -
12
 23 
 ------- (5-26)
5.  g -
23
 31 
 ------- (5-27)
6.  g -
31
CHAPTER 5 217
Material Library
where,
e 1t e 1c are the maximum allowable strains in the 1 direction in tension and compression.
g 12 is the maximum allowable shear strain in the 12 plane.
For the Maximum Strain Failure Criterion, the strength ratios at the integration points are the reciprocals of the
corresponding failure indices.
1.0
SR = ---------
FI
For example, the sixth strength ratio for the Maximum Strain Failure Criterion is  g 31   31  . Note that if the
maximum allowable strains are not defined or if the actual strains are 0, then the strength ratio is set to 100.
3. Hill Failure Criterion
Assumptions:
a. Orthotropic materials only
b. Tensile and compressive behavior are identical
At each integration point, Marc calculates the failure index FI as follows:
2 2 2 1 1 1 1 1 1
------1- + ------2- + -----3- –  ------
- + ------- – ------  1  2 –  ------- + ------ – -------  1  3
2 2 2  2 2 2  2 2 2
X Y Z X Y Z X Z Y
2
(5-28)
 122  132  23
1 1 1
–  ------- + ------ – -------  2  3 + -------- - + ---------  F
- + --------
 2 2 2 S 122 S 132 S 232
Y Z X
For plane stress condition, it becomes

2 2 2
  1  1  2  2  12 
 ------2- – ------------
- + ------- + --------- (5-29)
X X2 Y 2 S 12 2 
where
X is the maximum allowable stress in the 1 direction
Y is the maximum allowable stress in the 2 direction
Z is the maximum allowable stress in the 3 direction
S 12 S 23 S 31 are as before
F failure index scale factor
For the Hill Criterion, the strength ratio at each integration point is given by:
1.0
SR = ----------
FI
Note that if the failure index is 0 (either because no allowable stresses are prescribed or because actual stresses are
0), the strength ratio is set to 100.
4. Hoffman Failure Criterion
Note: Hoffman criterion is essentially Hill criterion modified to allow unequal maximum allowable stresses in
tension and compression.
At each integration point, Marc the failure index FI calculates:

2 2 2
 C1  2 – 3  + C2  3 – 1  + C3  1 – 2  + C4 1 + C5 2
2 + C 2 + C 2   F
+ C 6  3 + C 7  23 8 13 9 12
(5-30)
CHAPTER 5 219
Material Library
with:
1 1 1 1
C 1 = ---  ----------- + ------------- – -------------
2  Z t Z c Y t Y c X t X c
1 1 1 1
C 2 = ---  ------------- + ----------- – -------------
2  X t X c Z t Z c Y t Y c
1 1 1 1
C 3 = ---  ------------- + ------------- – -----------
2  X t X c Y t Y c Z t Z c
1 1
C 4 = ----- – ------
Xt Xc
1 1
C 5 = ----- – ------
Yt Yc (5-31)
1 1
C 6 = ----- – ------
Zt Zc
1
C 7 = --------
S 232
1
C 8 = --------
S 132
1
C 9 = --------
S 122
For plane stress condition, it becomes:

2 2 2
 1 1 1 1 1 2  12  1  2 
  ----- – ------  1 +  ----- – ------  2 + ------------- + ------------- + --------
- – -------------   F (5-32)
 Xt Xc Yt Yc Xt Xc Yt Yc S 2
12
Xt Xc 
where: X t X c Y t Y c Z t Z c S 12 S 23 S 31 F are as before.
For small ratios of, for example,  1 , the Hoffman criterion can become negative due to the presence
------
Note: of the linear terms. Xt
For the Hoffman Failure Criterion, the strength ratio SR is obtained by taking the smaller of the absolute values of
the two roots obtained by solving the quadratic equation:
A  SR  2 + B  SR  + C = 0
where for the general 3-D case:
2 2 2 2 + C 2 + C 2
C 1   2 –  3  + C 2   3 –  1  + C 3   1 –  2  + C 7  23 8 13 9 12
A = -------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------- ,
F
C 4 1 + C5 2 + C6 3
B = ------------------------------------------------------
- , and C = – 1.0 .
F
Note that if the allowable stresses are not defined or if the actual stresses are 0, then the strength ratio is set to 100.
5. Tsai-Wu Failure Criterion
Tsai-Wu is a tensor polynomial failure criterion. At each integration point, Marc calculates the failure index FI as
follows:
1
 ----- 1 1 1 1 1  12  22  32
– ------  1 +  ----- – ------  2 +  ----- – ------  3 + ------------
- + ------------
- + -----------
X X  Y Y  Z Z  Xt Xc Yt Yc Zt Zc
t c t c t c
2 2 2
(5-33)
 12  23  13
+ -------- + -------- + -------- + 2F 12  1  2 + 2F 23  2  3 + 2F 13  1  3   F
S 122 S 232 S 132
where X t X c Y t Y c Z t Z c S 12 S 23 S 31 F are as before.
F 12 Interactive strength constant for the 12 plane

CHAPTER 5 221
Material Library
For plane stress condition, it becomes:

2 2 2
 1 1  1 1 1 2 1 2 
  ----- – ------  1 +  ------- – ------  2 + ------------- + ------------- + --------
2
- + 2F 12  1  2   F (5-34)
 Xt Xc Y1 Yc Xt Xc Yt Yc S
12 
Note: In order for the Tsai-Wu failure surface to be closed,

1
2  ------------ 1 1 1 1 1
F 12 -  ------------- F 2  -------------  ----------- F 2  -------------  -----------
X t X c Y t Y c 23 Y t Y c Z t Z c 31 X t X c Z t Z c
See Wu, R.Y. and Stachurski, 2, “Evaluation of the Normal Stress Interaction Parameter in the Tensor
Polynomial Strength Theory for Anisotropic Materials”, Journal of Composite Materials, Vol. 18,
Sept. 1984, pp. 456-463.
For the Tsai-Wu failure criterion, the strength ratio SR is obtained by taking the smaller of the absolute values of
the two roots obtained by solving the quadratic equation:
A  SR  2 + B  SR  + C = 0
where for the general 3-D case:
 12  22  32 2
 12  232 2
 13
------------
- + ------------ - + ----------- + -------- + -------- + -------- + 2F 12  1  2 + 2F 23  2  3 + 2F 13  1  3
Xt Xc Yt Yc Zt Zc S 2 S 2
S 2
12 23 13
A = --------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------
-
F
1
 ----- 1 1 1 1 1
– ------  1 +  ----- – ------  2 +  ----- – ------  3
X X  Y Y   Z Z  , and C = –1.0 .
t c t c t c
B = --------------------------------------------------------------------------------------------------------------
-
F
Note that if the allowable stresses are not defined or if the actual stresses are 0, then the strength ratio is set to 100.
6. User-defined Failure Criteria
Using the UFAIL user subroutine, you can evaluate your own failure criterion as a function of stresses and strains at
each integration point. Two quantities can be returned by the routine:
A user-defined failure index
A user-defined strength ratio.
Up to six failure indices and strength ratios can be defined with UFAIL.
7. Hashin Failure Criterion
The Hashin failure criterion distinguishes between fiber failure and matrix failure. At each integration point, Marc
calculates the failure index FI for each mode as follows:
Tension fiber mode,  1  0
 11 2 1
 -------
- + --------   12
2 +  2  = 1 or (5-35)
 Xt  S 122 13
Compressive fiber mode,  1  0
 11
-----------
- = 1 (5-36)
Xc
Tensile matrix mode,  2 +  3  0
1 1 1
2 –    + --------
-------   2 +  3  2 + --------   23 2 3
2 + 2 
  12 13 (5-37)
Yt 2 S 232 S 122
Compressive matrix mode,  2 +  3  0
1 Yc 2 1 1
------   ----------- – 1   +   + ----------- 2 –   
-   2 +  3  2 + --------   23 (5-38)
Y c   2S 23  2 3
4S 23 2 S 232 2 3
1 2 + 2 
+ --------   12 13
S 122
For the Hashin failure criterion, the strength ratio SR for each mode is calculated as follows:
1.0
Tension fiber mode: SR = ---------- where FI is the failure index for the tension fiber mode
FI
1.0
Compressive fiber mode: SR = --------- where FI is the failure index for the compressive fiber mode
FI
1.0
Tensile matrix mode: SR = ---------- where FI is the failure index for the tensile matrix mode
FI
Compressive matrix mode: A  SR  2 + B  SR  + C = 0 where

1 1 2 1 2 2
-   2 +  3  2 + --------   23
A = ----------- –  2  3  + --------   12 +  13 
2 2 2
4S 23 S 23 S 12
1 Yc 2
B = ------   ----------- – 1   2 +  3  , and C = – 1.0 .
Y c   2S 23 
Note: Note that if the allowable stresses are not defined or if the actual stresses are 0, then the strength ratio
is set to 100.A fifth failure index is available for postprocessing. It is the maximum of the first four failure
criteria. Similarly, the minimum strength ratio is available as the 6th strength ratio.
CHAPTER 5 223
Material Library
8. Hashin Fabric Failure Criterion

The Hashin Fabric failure criterion is a variant of the Hashin criterion adapted for fabric type materials. The 1
direction is the first fiber direction, the 2 direction is the second fiber direction, and the 3 direction is through the
thickness direction. At each integration point, Marc calculates the failure indices as follows:
Tensile fiber 1 mode,  1  0
 2  12 2  13 2
 -----1- +  --------
- +  -------
- (5-39)
 X t  S 12   S 13
Compressive fiber 1 mode,  1  0
 2  12 2  13 2
 -----1- +  --------
- +  -------
- (5-40)
 X c  S 12   S 13
Tensile fiber 2 mode,  2  0
 2  12 2  23 2
 -----2- +  --------
- +  -------
- (5-41)
 Y t  S 12   S 23
Compressive fiber 2 mode,  2  0
 2  12 2  23 2
 -----2- +  --------
- +  -------
- (5-42)
 Y c  S 12   S 23
Tensile matrix mode,  3  0
 2  12 2  13 2  23 2
 -----3- +  --------
- +  -------
- +  -------
- (5-43)
 Zt   S 12   S 13  S 23
Compressive matrix mode,  3  0
 2  12 2  13 2  23 2
 -----3- +  --------
- +  -------
- +  -------
- (5-44)
 Z c  S 12   S 13  S 23
where: X t X c Y t Y c Z t Z c S 12 S 23 S 13 are as before.
For the Hashin Fabric Failure Criterion, the strength ratio SR for each of the six modes is calculated as:
1.0
SR = ----------
FI
where FI is the failure index for the corresponding mode. Note that if the maximum allowable stresses are not
defined or if the actual stresses are 0, then the associated strength ratios are set to 100.
9. Hashin Tape Failure Criterion

The Hashin Tape failure criterion is a variant of the Hashin criterion adapted for tape type of materials. The 1
direction is in the tape fiber direction, the 2 direction is perpendicular to the fiber direction in the plane of the tape,
and the 3 direction is through the thickness direction. At each integration point, Marc calculates the failure indices
as follows:
Tensile fiber mode,  1  0
 2  12 2  13 2
 -----1- +  --------
- +  -------
- (5-45)
 X t  S 12   S 13
 2  12 2  13 2
 -----1- +  --------
- +  -------
- (5-46)
 X c  S 12   S 13
Tensile matrix mode,  2 +  3  0
 2 +  3 2  22  33 1 2 2  13 2  23 2
 ------------------- – ----------------- +  --------
- +  -------
- +  -------
- (5-47)
 Yt  S 232  S 12   S 13  S 23
Compressive matrix mode,  2 +  3  0
Yc 2 2 + 3 2 + 3 2 2 3 1 2 2  13 2  23 2  2
  ----------- – 1  ------------------- +  ------------------- – ------------- +  --------
- +  -------
- +  -------
- + A 1  -----1-
  2S 23   Y c   2S 23  2  S 12   S 13  S 23  S 1 (5-48)
S 23
where: X t X c Y t Y c S 12 S 23 S 13 are as before. A 1 is equal to zero or one and is used to determine if the last
term should be used. S 1 is the maximum fiber stress for matrix compression.
1.0
For the Hashin Tape Failure Criterion, the strength ratio SR for the first three modes is calculated as SR = ---------
-
FI
where FI is the failure index for the corresponding mode.
For the compressive matrix mode, the strength ratio is calculated as follows:
A  SR  2 + B  SR  + C = 0
where
2 + 3 2 2 3  12 2  13 2  23 2 1 2
A =  ------------------- – ------------
- +  -------- +  -------- +  -------- + A 1  ------ ,
 2S 23  S 232  S 12   S 13   S 23   S 1
1 Yc 2
B = ------   ----------- – 1   2 +  3  , and C = – 1.0 .
Y c 2S 23
  
CHAPTER 5 225
Material Library
Note that if the maximum allowable stresses are not defined or if the actual stresses are 0, then the associated
strength ratios are set to 100.
10. Puck Failure Criterion
The Puck failure criterion distinguishes just as Hashin between fiber failure and matrix failure. The main difference
is in the matrix failure where the concept of a fracture failure angle is used. This angle is the angle at which the
fracture due to matrix failure will occur. See Figure 5-12 for the definition on how it is defined with respect to the
preferred coordinates system.
3 fp
t n
1
Figure 5-12 Definition of the Fracture Failure Angle  fp
The Puck failure criterion uses the following input parameters as defined above: X t X c Y t Y c S 12 . In addition,
four parameters are used for describing the slopes of the failure envelope: p 12c = p (–)|| , p 12t = p (+)|| ,
p 23c = p (–) , and p 23t = p 
(+) . The notation p
12c etc. is used in Marc Volume C: Program Input and in the
Marc output while the other notation is from Puck [Ref. 25]. The following parameters are calculated from the input
parameters and used in the following
Yc S 12  Yc 
R A = ----------------------------
- = --------------  1 + 2p 12c -------- – 1 (5-49)
2  1 + p 23c  2p 12c  S 12 
 21c = S 12 1 + 2p 23c (5-50)
The first two failure indices are given by:

Tensile fiber mode,  1  0

-----1- (5-51)
Xt
1
--------- (5-52)
Xc
The fracture failure angle  fp can be calculated analytically for the plane stress case provided that the following
relation between two of the failure envelope slopes is enforced:
RA
p 23c = p 12c -------- (5-53)
S 12
It is recommended that either p 12c or p 23c is specified in the input. The other is by default calculated using
Equations (5-49) and (5-53). This default setting is also used for other cases than plane stress although then this
relationship is not necessary.
For the plane stress case, we have three failure indices, denoted modes A, B, and C:
Mode A,  2  0 ,  fp = 0
 12 2 Y 2  2 2
 -------
- +  1 – p 12t -------t-  -----2- + p 12t -------
- (5-54)
 S 12  S 12  Y t S 12
 RA
Mode B,  2  0 and 0  -------2-  ----------- ,  fp = 0
 12  21c
1 2 + p 2
--------   12 12c  2  + p 12c  2  (5-55)
S 12
1 2  21c
-  ----------
Mode C,  2  0 and 0  -------- -.
2 RA
 12 2  2 Yc
  -------------------------------------
- +  -----2-  --------- (5-56)
  2  1 + p 23c S 12   Y c   2
For this plane stress case, the fracture failure angle is calculated from the following formula:
2
1   12 2  R A  
cos2  = -----------------------------   --------  -------- + 1 (5-57)
fp 2  1 + p 23c     2   S 12 
It is available as the sixth failure index but only for postprocessing.

For other than plane stress, the failure angle is not determined analytically. Instead, a numerical procedure is used
as outlined below.
The stresses in a failure plane defined by 1, n and t directions in Figure 5-12 are:
 n =  2 cos2  +  3 sin2  + 2 23 sin  cos  (5-58)
 nt =   3 –  2  sin  cos  +  23  cos2  – sin2   (5-59)
 n1 =  31 sin  +  12 cos  (5-60)

CHAPTER 5 227
Material Library
The failure index as a function of  is now formulated as follows,  n  0 :
1 2 n t 2 n 1 2
f =  ----- – p 1  n2 +  -------- +  --------- + p 1  n (5-61)
Y   R A  S 12 
t
and for  n  0
 nt 2 n 1 2
f =  -------
- +  --------
- +  p 2  n  2 + p 2  n (5-62)
 R A  S 12 
where,
p 23t  n2t p 12t  n21
p 1 = --------- ------------------------- + --------- ------------------------- (5-63)
R A  nt 2 + 2 S 12  2 +  2
n1 nt n1
p 23c  n2t p 12c  n21

p 2 = ---------- ------------------------- + ---------- ------------------------- (5-64)
R A  nt 2 + 2 S 12  2 +  2
n1 nt n1
The critical failure angle  f p and the corresponding failure index is obtained by evaluating equations (5-58) through
(5-64) with a sufficient number of angles between -90 and 90 degrees. For positive normal stress, it is saved as the
3rd failure index and as the 4th failure index. The 5th failure index is not used in this case. The critical failure angle
is stored as the 6th failure index for postprocessing just as for plane stress.
For the Puck Failure Criterion, the strength ratios for each mode are the reciprocals of the corresponding failure
1.0
indices, SR = --------- . Note that if the maximum allowable stresses are not defined or if the actual stresses are 0,
FI
then the associated strength ratios are set to 100.
A seventh failure criterion is available for postprocessing. It is the maximum of the first five failure criteria. Similarly,
the minimum strength ratio is available as the seventh strength ratio.
11. SIFT Failure Criterion
The Strain Invariant Failure Theory (SIFT) criterion is based on evaluating the failure indices using the strain
invariants. It accounts for micro-mechanical effects of the composite without having a micro-mechanical model.
The SIFT failure criterion also takes into effect, the thermo-mechanical influences on the composite material. The
Strain Invariant Failure Theory criterion is based on the work done by Gosse, J. H. et al, Boeing Company [Ref. 30].
Strain Invariants: The fundamental hypothesis of the SIFT criterion is that failure is based on the first invariant
and the second invariant (or equivalent strain) of the homogenous medium. In other words, if any of the two
quantities exceeds the critical limit, the failure is said to initiate.
The first invariant, which accounts for the dilatation nature of deformation, is given as:
J 1 =  11 +  22 +  33 (5-65)
where  11 ,  22 ,and  33 are the diagonal components of strain tensor. And the equivalent strain, which accounts
for the distortional (or deviatoric) nature of deformation, is given as:
12
 eq =  2  3   i j  i j   (5-66)
The critical values of the quantities are to be evaluated using experiments. Please refer to [Ref. 30] for further details.
Micro-mechanical enhancements: In a typical composite analysis, the deformation of the structure is not typically
linked to the stress and strain states of the heterogeneous phases of the composite and most predictions of composite
structure failure have been based upon stresses and strains within the homogenized system. The SIFT theory
considers the heterogeneous effects without having a homogeneous model. For a given state of strain in the
homogeneous medium, the SIFT criterion has a methodology called influence function formulation to evaluate the
heterogeneous states of strain for the fiber and matrix medium.
Representative volume element: The parameters required to account for the micro-mechanical effects of the model
is deduced using the representative volume element (RVE). The configuration of the RVE can be based on the
volume fraction and the fiber arrangement in the matrix of the composite. Two configurations of fiber arrangement
– square and hexagonal – are shown below.
Figure 5-13 Square and Hexagonal Array Representative Volume Elements for Finite Element Analysis [Ref. 29]
CHAPTER 5 229
Material Library
(a) Square Arrangement (b) Diagonal Arrangement

Figure 5-14 Representative Volume Elements for Determining Influence Parameters
Influence function formulation: Considering the representative volume element being subjected to an arbitrary
state of strain,  j –   j T   i j  = 1 – 6 including both thermal and mechanical components with  j the
effective coefficient of thermal expansion of the homogenized medium. The state of strain at a prescribed point
within the representative volume element can be determined from the following relationship:
k k k k
 i –  i T = M i j   j –  j T  + A i T i j = 1 – 6 (5-67)
k k
where M ij is the influence function matrix at the kth point in the RVE and A i is the thermal vector of strain, per
unit thermal change, induced at the point of interest within the boundaries of the unconstrained representative
volume element.
Essentially, the formulation helps to amplify the strain to account for the heterogeneous effects, which is then used
to calculate the strain invariant quantities. The enhanced strain invariants are then compared against the respective
critical values for predicting failure.
Methodology to obtain the influence matrices: The user is required to input the data required for this formulation
in the form of influence matrices. The influence matrices can be evaluated by performing a separate finite element
analysis of a model called representative volume element (RVE) – a heterogeneous model containing fiber and
matrix elements with respective materials properties. This can be achieved using the Mentat product as shown
below.
At a given point of interest, k , the components of the influence function matrix can be determined uniquely by
prescribing a canonical (simple) state of deformation of the representative volume element and carrying out three-
dimensional finite-element analyses within the representative volume element to determine the components of the
strain tensor at the specified point, k . For example, let  j = 0 ,  j = 2 – 6  and  1 = 1 and T = 0 then,
k k k k
 1 = M 11  1 or M 11 = 1
k k k k
 2 = M 21  1 or M 21 = 2
k k k k
 3 = M 31  1 or M 31 = 3
(5-68)
k k k k
4 = M 41  1 or M 41 = 4
k k k k
 5 = M 51  1 or M 51 = 5
k k k k
 6 = M 61  1 or M 61 = 6
It is to be noted that for any point, k , a single finite-element analysis with boundary conditions appropriate to the
condition,  1 = 1 yields 6 of the 36 coefficients in the influence function matrix. Five more finite element
k
analyses are required to determine the remaining terms of M i j . However, these same 6 finite element analyses can
k
produce the terms of M i j at another point, k , within the medium as well. Thus, only 6 finite element models need
to be constructed for a given representative volume element geometry to determine any number of influence
function matrices. Though, the user is open to choose specific points, SIFT specifies the points at which the
parameters needs to be determined. These points are called interrogation points (or locations). For the matrix phase,
the points chosen differ for the two representative volume elements. These points represent both the point the
greatest distance from the fiber phase and the two points where the fibers are closest to one another. For the
hexagonal array, the point where the least distance between fibers occurs is chosen along with two other
intermediate points.
Figure 5-15 Interrogation Locations within the Matrix Phase for Hexagonal and Square Array for Fiber Diameter, D. [Ref. 29]
CHAPTER 5 231
Material Library
The displacement boundary conditions for the prescribed extensional strain for the square array RVE in the x
direction are:
u = 0 at x = 0 and u = 10 – 6 at x = 1
v = 0 at y = 0 1
w = 0 at z = 0 1
The canonical shearing displacements at the RVE faces, for the rectangular array are shown as follows for the y–z
plane:
z y
y = 0 y = 1 and z = 0 z = 1 : u = 0 v = ---  10 – 6 w = ---  10 – 6
2 2
x = 0 x = 1 : u = 0
Thermal enhancement vectors: In the case of a uniform temperature change for the representative volume element,
boundary conditions were imposed to allow for free translation normal to the surfaces: x = 1 , y = 1 , and
z = 1 for the square array and for the hexagonal array, x = 1 , y = 1 , and z = 1 for the square array and
for the hexagonal array, x = 1 , y = 1 , and z = 3 , while holding deformations normal to the planes
x = 0 , y = 0 , and z = 0 to zero. This set of nodal boundary conditions is intended to yield a traction free
state of deformation of the RVE. In this case, the unit cell free thermal deformation is equal to that defined by the
coefficients of thermal expansion of the homogenized unidirectional lamina,  j T
k k k k
 i –  i T = M i j   j – x j T  – A i T  i j = 1 – 6 
Let  j =  j T j = 1 – 3
The thermal strain enhancement vector can be evaluated as:

k k
k  i –  i T
A i = --------------------------
- i = 1 – 6
T
Failure Indices and Strength Ratio

The failure index and the strength ratios are calculated as follows:
a. The strain is amplified using equation (5-67) for each of the MATRIX amplification matrices
b. The strain is amplified using equation (5-67) for each of the FIBER amplification matrices
c. The first invariants  J1  are calculated for each of amplified strains for the MATRIX material (The fiber
material is not considered for calculation of this invariant ([Ref. 29] and [Ref. 30]))
m
d. The equivalent strains   e  are calculated for each of amplified strains for the MATRIX material
f
e. The equivalent strains   e  are calculated for each of amplified strains for the FIBER material
m f
f. The maximum of J1 ,  e ,  e is compared against the respective user-specified critical values.
g. The failure indices are calculated as the ratio of the above max values with the respective critical values. The
m f
order of failure indices are corresponding to J1 ,  e ,  e . In other words, the first failure index is based on J1 ,
m f
the second based on point  e and the third based on point  e .
h. The strength ratio is the reciprocal of the failure index.
Mentat Procedure for SIFT Influence Matrices

In order to make it easy for the users to compute the strain enhancement coefficients, the Marc installation is equipped
with Python scripts to make the micro-mechanical model and to calculate the parameters. The procedure to obtain the
strain enhancement matrices and thermal enhancement vectors - given the properties of the constituent materials of the
composite and the volume fraction of fiber - is detailed below.
The Python scripting utility is used to run the scripts by choosing the following options from the Mentat GUI.
UTILS -> PYTHON -> RUN
The scripts are located in …\mentat\utilities\python directory under the Mentat installation directory. The scripts listed
below are to be copied to the user’s working directory.
sift_01_create_parameters.py
sift_02_compute_strain_enhancement.py
sift_03_apply_strain_enhancement.py
The steps are explained as follows:
1. Save any existing model.
2. Open a new model.
3. Run the Python script named sift_01_create_parameters.py by choosing UTILS->PYTHON->RUN. This script
creates the input parameter keywords required for the input of material parameters and volume fraction.
4. Modify the volume fraction of fiber and the material parameters of the constituent materials of the composite as
desired by choosing UTILS->PYTHON->PARAMETERS. The values available are representative in nature.
5. Run the Python script named sift_02_compute_strain_enhancemnt.py. This script builds the representative volume
element (RVE) model using the volume fraction and the material properties and runs seven simulations (jobs)
required for computation of the strain enhancement parameters - the influence matrices and thermal strain
enhancement vector.
6. After ensuring that all the seven simulations are completed, open any existing model (e7x38.mud for example) or
a new model and run the script named sift_03_apply_strain_enhancemnt.py. This script postprocesses the
simulation results of the RVE and computes the strain enhancement parameters. In addition, it creates a new
material in the user’s Mentat model and applies the computed properties in the DAMAGE EFFECTS -> STRAIN
INVARIANT -> INFLUENCE FUNCTION PROPERTIES section.
CHAPTER 5 233
Material Library
Nes:ot 1. The scripts are to be present in the current working directory.

2. It takes several minutes for the RVE model simulations to complete. Before running the third
script, the user needs to ensure that the result files named sift_job1.t16, sift_job2.t16 and so on
are available in the current working directory.
3. The RVE simulations are required to be run only once for a given set of constituent materials and
volume fraction of fiber. The results files and the third script, which are to be placed in the current
working directory, can be used for as many user models as required to create the influence matrices
and thermal enhancement vectors.
Interlaminar Shear for Thick Shell, Beam, Solid Shell, and 3-D Composite Brick Elements
Another addition made for composite analysis is the calculation of interlaminar shears. These interlaminar shears are
printed in the local coordinate system above and below each layer selected for printing by PRINT CHOICE or PRINT ELEMENT.
These values are also available for postprocessing. The TSHEAR parameter must be used for activating the parabolic shear
distribution calculations.
In Marc, the distribution of transverse shear strains through the thickness for thick shell and beam elements was assumed
to be constant. From basic strength of materials and the equilibrium of a beam cross section, it is known that the actual
distribution is more parabolic in nature. As an additional option, the formulations for elements 1, 22, 45, 75, 89, 140, 149,
150, and 185 (TSHEAR will be switched off for elements 185, 149 and 150 when the elements are stacked) have been modified
to include a parabolic distribution of transverse shear strain. The formulation is exact for beam element 45, but is
approximate for the other thick shell elements. Nevertheless, the approximation is expected to give improved results from
the previous constant shear distribution. Furthermore, interlaminar shear stresses for composite beams and shells can now
be easily calculated.
The generalized stiffness matrix for the complete section excluding transverse shear terms is given by:
F 11  11
F 22  22
F 12 X Y   12
= (5-69)
M 11 Y Z  11
M 22  22
M 12  12
where X, Y, and Z are 3 x 3 matrices
F = section forces
M = section moments
 = strain at mid plane of section
 = curvature
1,2 are in-plane directions
A unique X, Y direction in the plane of the section is defined by a rotation around the element normal which maximizes
the value of X 11 in the above equation.
We then assume that the stresses in the X and Y direction are uncoupled, this gives:
       
 F xx  X xx Y x x     Fy y  Xy y Yy y   
  =   xx  and   =   yy  (5-70)
 M xx  Y xx Z x x  x x   My y  Yy y Zy y  y y 
       
If we assume only bending and transverse shear in the section, all section forces are zero and inverting the above equation
gives:
   
  xx  H 1x    H 1y
  =   M x x  and  y y  =   My y  (5-71)
 x x  M 2x  y y  M 2y
   
For a point in the section, we can now define the stresses as:
 xx  z  = E  z    xx  z  = E  z     x x +  x x  z  = E  z    H 1 x + H 2 x  z M x x and
(5-72)
 yy  z  = E  z    yy  z  = E  z     y y +  y y  z  = E  z    H 1 y + H 2 y  z M y y
Since we assumed that all stresses in the X and Y direction are uncoupled, the equilibrium conditions through the thickness
are given by:
 z x  xx  z   z y  y y  z 
----------
- + -------------------- = 0 and ----------
- + -------------------- = 0 (5-73)
z x z y
where  zy and  zx are the transverse shear stresses.
From beam theory, we have:

M xx M y y
V x + -------------
- = 0 and V y + -------------
- = 0 (5-74)
x y
where M is the bending moments and V is the shear forces.

Combining the Equations (5-72), (5-73), and (5-74) gives:
 z x  z y
----------- = E  z    H 1x + H 2 x  z   V x and ----------
- = E  z    H1 y + H2 y  z   Vy (5-75)
z z
We can integrate this through the thickness giving:
CHAPTER 5 235
Material Library
t2 y
 zx  z  =  E  z   H 1 x + H 2 x  z V x dz and  z y  z  =  E  z   H 1 y + H 2 y  z V y dz (5-76)
–t  2 –t  2
with the boundary conditions that the shear stresses at the top and the bottom of the shell are zero.
We can now define the transverse shear stiffness by matching the shear strain energy over the section obtained with the
transverse shear stresses given in (5-76). This yields the flexibility matrix S:
t2
1   Vx  1   G x x  z  G x y  z    zx  z  
---  V V  S    = ---   z x  z  z y  z     dz (5-77)
2  x y   Vy  2   G y x  z  G y y  z    zy  z  
–t  2
where G is the transverse shear flexibilities of the material through the thickness.
Inversion of the flexibility matrix S gives the transverse shear stiffness of the section.
Interlaminar Stresses for Continuum Composite Elements

In Marc, the interlaminar shear and normal stresses are calculated by averaging the stresses in the stacked two layers. The
stresses are transformed into a component tangent to the interface and a component normal to the interface. The two
components, considered as shear stress and normal stress, respectively, are printed out in the output file.
By using POST code 501 or 511 (see Marc Volume C: Program Input) representing interlaminar normal and shear stresses
respectively, the interlaminar normal or shear stress can be written into a post file in the form of a stress tensor defined in
the global coordinate directions. Mentat can be used to plot the principal directions of the stress tensor which show the
magnitude and the direction of the stress, and the changes based on deformation.
Progressive Composite Failure

Marc supports progressive failure analysis for composites and other elastic materials. The material is assumed to be linear
elastic up to the point of failure. Failure is indicated by the failure criteria described in the previous section. When failure
occurs, the element stiffness is degraded. Marc offers three different methods for the material degradation as described
below. This is flagged through the FAIL DATA model definition option. While Marc allows up to three failure criteria to be
used for failure index calculation, only the first criterion is used for progressive failure. The material will not heal; the
damaged elements keep the degraded properties after unloading.
Model 1 – Selective Gradual Degradation

This model uses a selective degradation of the moduli depending on failure mode. The moduli are decreased gradually
when failure occurs. Within an increment, it attempts to keep the highest failure index less than or equal to one. Whenever
a failure index F larger than one occurs, stiffness reduction factors ri are calculated based upon the value of the failure
indices. The incremental contribution to the total reduction factor is calculated as:
r i = –  1 – e 1 – F  (5-78)
This is done differently for different failure criteria as described below. Six such reduction factors are stored and updated.
They are then used for scaling the respective material modulus according to:
ne w o rig
E 11 = r 1 E 11
ne w o rig
E 22 = r 2 E 22
ne w o rig
E 33 = r 3 E 33
new or ig
G 12 = r 4 G 12
new or ig
G 23 = r 5 G 23
new or ig
G 31 = r 6 G 31
The Poisson’s ratios are scaled in the same way as the corresponding shear modulus.
For the maximum stress and maximum strain criteria the reduction factors are calculated separately from each separate
failure index: r1 is calculated from the first failure index as given by Equation 5-16 above, r2 is calculated from the second
failure index from Equation 5-17 etc. Thus, there is no coupling of the different failure modes for these criteria.
For the failure criteria which only have one failure index: Tsai-Wu, Hoffman and Hill, all six reduction factors are decreased
in the same way, using the smallest of the ri:s.
For the criteria which distinguish between fiber and matrix failure (Hashin, Hashin-tape and Puck) there is a more complex
coupling between the failure modes. There is a default behavior which can be influenced by a number of input parameters.
The default is as follows.
 r1 depends on fiber failure (first and second failure index)
 r2 depends on matrix failure (third, fourth and fifth failure index)
 r3 behaves the same as r1 ( r 3 = r 1 )
 r4 behaves the same as r2 ( r 4 = r 2 )
 r5 and r6 behave the same as r4 ( r 5 = r 6 = r 4 )
In the FAIL DATA option, there are five parameters available for controlling the way the moduli are reduced. With the
exception of a1, they are only used for the Hashin variants and Puck.
a1 – Residual stiffness factor: The stiffness is never reduced to less than this factor. The default is 0.01.
a2 – Matrix compression factor. With this factor, r2 can reduce less due to failure in matrix compression. Experiments show
that certain materials show less degradation of the matrix properties in compression than in tension. See for example
[Ref. 27]. For the case that Fmc indicates failure in matrix compression, Equation 5-78 is modified into:
1 – Fm c
r 2 = –  1 – a 2   1 – e  (5-79)
CHAPTER 5 237
Material Library
a2 defaults to 0; so the default is full reduction

a3 – Shear stiffness factor. This factor is used for taking into account the effect that the shear stiffness G12 can reduce less
than the matrix stiffness E2. See [Ref. 28] for experimental data on this subject. With Fm indicating a matrix failure we have
1 – Fm
r 4 = –  1 – a 3   1 – e  (5-80)
The combined effect of a2 and a3 on the shear stiffness reduction for the case of matrix compression failure is then:
1 – Fm c
r 4 = –  1 – a 2   1 – a 3   1 – e 
a4 – E33 reduction from fiber failure: This factor controls the reduction of E33 due to fiber and matrix failure. The default
is as mentioned above that E33 reduces due to fiber failure. With this factor this can be changed to vary linearly with fiber
and matrix failure. With Ff indicating a fiber failure and Fm a matrix failure we have:
1 – Ff 1 – Fm
r 3 = –  1 – a 4   1 – e  – a4  1 – e  (5-81)
a5 – Shear reduction from fiber failure: With this factor it is possible to control the reduction of the shear stiffness due to
fiber failure. By default it only reduces due to matrix failure. With Ff and Fm as in the previous we have:
1 – Fm 1 – Ff
r 4 = –  1 – a 5   1 – e  – a5  1 – e  (5-82)
The Hashin-fabric failure criterion reduces the first three reduction factors from the respective failure index. The three shear
reduction factors are taken from the worst of the first three factors. The factors a2 through a5 are not used for this criterion.
In addition, it is also possible to use the UPROGFAIL user subroutine to explicitly define the reduction factors r1 through r6.
Model 2 – Selective Immediate Degradation

This model uses selective degradation just as Model 1, but the stiffness is abruptly decreased. As soon as failure is indicated,
the stiffnesses are set to a1 – the residual stiffness factor. The same rules as in Model 1 for how the different factors are
defined depending on the type of failure is applied here.
Model 3 – Original Marc Method

This is the original method in Marc.
1. Upon failure, the material moduli for orthotropic materials at the integration points are set to the smallest of the
original moduli, and the smallest is set to 10% of the original.
2. Upon failure, for isotropic materials, the failed moduli are taken as 10% of the original moduli.
3. If there is only one modulus, such as in a beam or truss problem, the failed modulus is taken as 10% of the original
one.
The different options are flagged through the FAIL DATA model definition option. The reduction factor is available for
postprocessing as the element post code number 80, Damage. This single value contains the smallest of the six reduction
factors.
Mixture Model
The mixture models allow the user to define a composite material that consists of multiple components for all element
types. Effective material properties are formed based upon the volume fraction of each material component. There are three
variations of the mixture model which have consequences on the type of material that may be used in each component.
The component material is defined using the conventional ISOTROPIC, ORTHOTROPIC, ANISOTROPIC, MOONEY, etc. options. The
mixture model acts as a continuum in the sense that debonding between the components is not considered in any of
the models.
Mixture Model 1
This model is applicable to linear elastic materials and heat transfer. An effective material property is formed as
N
ef f i
E =  Vf  E
i
i = 1
i
where E is the value of the ith component, V f is the volume fraction of the ith component, and N is the number of
i
components. The same summation is done for all properties. If an orthotropic component is used, it is done for all nine
material properties.
Note that for orthotropic or anisotropic properties, is assumed that all components have the same preferred directions.
If the component has temperature dependent properties, then the temperature dependence is evaluated for each
component which can be expressed as:
N
ef f · i
E T =  Vfi  E  T 
:
The stresses obtained are an effective stress associated with the mixture material and not the stress in each component.
Mixture Model 2
This model is applicable to linear elastic materials and heat transfer. An effective material stress-strain relationship is formed
as:
N
e ff i
D =  Vf D
i
i = 1
where
i i
 = D
e ff eff
 = D 
For heat transfer, analogous equations are used:
i i
q = D T
CHAPTER 5 239
Material Library
e ff eff
q = D T
Again, temperature dependent properties are dealt with on a component basis, and orthotropic or anisotropic mixtures
must have the same preferred orientations.
Mixture Model 3
This model is applicable to nonlinear material behavior. This requires additional allocation of memory to store the
associated state variables (plastic strain, shift tensors, etc.) for each component. It is similar to model 2 in that an effective
material stress-strain relationship is formed as:
N
ef f i
D =  Vf D
i
i = 1
i
but now, D includes material nonlinear effects. for the case of elastic-plastic materials, we now have two types of quantities
which are not completely consistent with one another. On a component level:
i i i i
 = D el   –  th –  p l 
where again the i indicates the component number which is the correct expression. We also calculate effective quantities.
These quantities are accumulated in the usual manner.
i i i
 n + 1 =  n + 
pl  i  pl  i  pl  i 
n + 1 = n + 
There are also effective quantities calculated as:

N
ef f
 =  Vf i
i
i = 1
N
pl  eff  pli
 =  V f 
i
i = 1
N
pl  eff  pli
 =  V f 
i
i = 1
which are also accumulated.
pl  eff  p l  eff  pleff
n + 1 = n + 
pl  eff  p l  eff  pleff

n + 1 = n + 
pl
where the  is the equivalent plastic strain. Note that the effective plastic strain is not used in a subsequent calculation
and is only provided for output.
Given a mixture of two identical nonlinear materials at a 50%-50% mix, the output of the effective equivalent plastic strain
will not be the same as for a solitary material.
Mixture model 3 can be used with most nonlinear material behavior, but there are a few restrictions in this release which
include:
 All damage models introduced via the DAMAGE option
 Gasket material
 Soils
 User-defined generalized stress-strain law.
 ORNL
 Rigid Plastic
 Viscoelasticity
 Cohesive
 Grain size effects
 Thermo-Pore
Mass Density
The effective mass density for all of the mixture models is calculated as:
N
ef f i
 =  Vf 
i
i = 1
Specific Heat
The effective specific heat for all of the mixture models is calculated as:
N
eff i
Cp =  Vf Cp
i
i = 1
Coefficient of Thermal Expansion

There are two ways to calculate the coefficient of thermal expansion; the first directly uses the rule of mixtures.
N
ef f i
 =  
i = 1
CHAPTER 5 241
Material Library
The second formulation:

i i
ef f E  Vf  
i
 = -----------------------------------
-
i
 E  Vf i
where, again, if temperature dependent material properties are used, these are applied at the component level.
i i
ef f  E  T   Vfi    T 
    = ------------------------------------------------------
i
 E  T   Vf i
Mixtures and Composites

It is possible to have a shell element or a composite brick element where the layer of a composite material may be a mixture
model that consists of multiple components.
Restrictions
If the VOID CHANGE or POROSITY CHANGE is used to define a state variable, and the material properties are a function of the
void ratio, then all components will use the same void ratio.
A mixture material cannot have as a component another mixture material.
Mixture materials should not be used with the PSHELL option.
Mixture materials cannot be used with design sensitivity or design optimization where the design variable is an elastic
property in the mixture material or one of the components.
Gasket
Engine gaskets are used to seal the metal parts of the engine to prevent steam or gas from escaping. They are complex (often
multi-layer) components, usually rather thin and typically made of several different materials of varying thickness. The
gaskets are carefully designed to have a specific behavior in the thickness direction. This is to ensure that the joints remain
sealed when the metal parts are loaded by thermal or mechanical loads. The through-thickness behavior, usually expressed
as a relation between the pressure on the gasket and the closure distance of the gasket, is highly nonlinear, often involves
large plastic deformations, and is difficult to capture with a standard material model. The alternative of modeling the gasket
in detail by taking every individual material into account in the finite element model of the engine is not feasible. It requires
a lot of elements which makes the model unacceptably large. Also, determining the material properties of the individual
materials might be cumbersome.
The gasket material model addresses these problems by allowing gaskets to be modeled with only one element through the
thickness, while the experimentally or analytically determined complex pressure-closure relationship in that direction can
be used directly as input for the material model. The material must be used together with the first-order solid composite
element types 149 (three-dimensional solid element), 151 (two-dimensional plane strain element) or 152 (two-dimensional
axisymmetric element). In that case, these elements consists of one layer and have only one integration point in the
thickness direction of the element.
Gaskets can be used in mechanical, thermal or thermo-mechanically coupled analyses. The usual staggered scheme of a heat
transfer pass followed by a structural pass is used for coupled analyses. For the heat transfer part, the elements used to model
the gaskets are type 175 (three-dimensional first-order solid element), type 177 (two-dimensional first-order planar
element), or type 178 (two-dimensional first-order axisymmetric element).
Constitutive Model
The behavior in the thickness direction, the transverse shear behavior, and the membrane behavior are fully uncoupled in
the gasket material model. In subsequent sections, these three deformation modes are discussed.
Local Coordinate System

The material model is most conveniently described in terms of a local coordinate system in the integration points of the
element (see Figure 5-16). For three-dimensional elements, the first and second directions of the coordinate system are
tangential to the midsurface of the element at the integration point. The third direction is the thickness direction of the
gasket and is perpendicular to the midsurface. For two-dimensional elements, the first direction of the coordinate system
is the direction of the midsurface at the integration point, the second direction is the thickness direction of the gasket and
is perpendicular to the midsurface, and the third direction coincides with the global three direction.
2 3
1
1
Midsurface
Integration Point Midsurface
Integration Point
Figure 5-16 The Location of the Integration Points and the Local Coordinate Systems in Two- and Three-dimensional
Gasket Elements
In a total Lagrange formulation, the orientation of the local coordinate system is determined in the undeformed
configuration and is fixed. In an updated Lagrange formulation, the orientation is determined in the current configuration
and is updated during the analysis.
In case an orthotropic base material and/or gasket material is used, it is highly recommended to specify a material
orientation using one of the following options:
 3D Aniso
 Coordinate system
 3D Local
 Curves
To allow easy interpretation, align the local z-axis (blue) with the element thickness direction.
CHAPTER 5 243
Material Library
Thickness Direction - Compression

In the thickness direction, the material exhibits the typical gasket behavior in compression, as depicted in Figure 5-17. After
an initial nonlinear elastic response (section AB), the gasket starts to yield if the pressure p on the gasket exceeds the initial
yield pressure py0. Upon further loading, plastic deformation increases, accompanied by (possibly nonlinear) hardening,
until the gasket is fully compressed (section BD). Unloading occurs in this stage along nonlinear elastic paths (section FG,
for example). When the gasket is fully compressed, loading and unloading occurs along a new nonlinear elastic path
(section CDE), while retaining the permanent deformation built up during compression. No additional plastic
deformation is developed once the gasket is fully compressed.
The loading and unloading paths of the gasket are usually established experimentally by compressing the gasket, unloading
it again, and repeating this cycle a number of times for increasing pressures. The resulting pressure-closure data can be used
as input for the material model. If the pressure also varies with temperature, then pressure-closure data at different
temperatures can be provided. The user must supply the loading path and may specify up to ten unloading paths. In
addition, the initial yield pressure py0 must be given. The initial yield pressure can also be varied with temperature and
spatial coordinates. The loading path should consist of both the elastic part of the loading path and the hardening part, if
present. If no unloading paths are supplied or if the yield pressure is not reached by the loading path, the gasket is assumed
to be elastic. In that case, loading and unloading occurs along the loading path.
loading path
py1 D
py G
py0 B
unloading path
Gasket Pressure p
A F C
cp0 cp cy0 cp1 cy cy1
Gasket Closure Distance c
Figure 5-17 Pressure-closure Relation of a Gasket
The loading and unloading paths must be defined using the TABLE model definition option (see Marc Volume C: Program
Input) and must relate the pressure on the gasket to the gasket closure. Optionally, at each closure value, the loading and
unloading paths can also be defined as functions of temperature and spatial coordinates using multi-variate TABLEs. The
unloading paths specify the elastic unloading of the gasket at different amounts of plastic deformation; the closure at zero
pressure is taken as the plastic closure on the unloading path. If unloading occurs at an amount of plastic deformation for
which no path has been specified, the unloading path is constructed automatically by linear interpolation between the two
nearest user supplied paths. The unloading path, supplied by the user, with the largest amount of plastic deformation is
taken as the elastic path at full compression of the gasket.
For example, in Figure 5-17, the loading path is given by the sections AB (elastic part) and BD (hardening part) and the
initial yield pressure is the pressure at point B. The (single) unloading path is curve CDE. The latter is also the elastic path
at full compression of the gasket. The amount of plastic closure on the unloading path is cp1. The dashed curve FG is the
unloading path at a certain plastic closure cp that is constructed by interpolation from the elastic part of the loading path
(section AB) and the unloading path CD.
The compressive behavior in the thickness direction is implemented by decomposing the incremental gasket closure into
an elastic and a plastic part:
c = c e + c p (5-83)
Of these two parts, only the elastic part contributes to the pressure. The constitutive equation is given by the following
equation:
p
p = D c c e = D c  c – c  .
Here, D c is the consistent tangent to the pressure-closure curve.
Plastic defomation develops when the pressure p equals the current yield pressure p y . The latter is a function of the
amount of plastic deformation developed so far and is given by the hardening part of the loading path (section BD in
Figure 5-17).
Initial Gap
The thickness of a gasket can vary considerably throughout the sealing region. Since the gasket is modeled with only one
element through the thickness, this can lead to meshing difficulties at the boundaries between thick regions and thin
regions. The initial gap parameter can be used to solve this. The parameter basically shifts the loading and unloading curves
in the positive closure direction. As long as the closure distance of the gasket elements is smaller than the initial gap, no
pressure is built up in the gasket. The sealing region can thus be modeled as a flat sheet of uniform thickness and the initial
gap parameter can be set for those regions where the gasket is actually thinner than the elements of the finite element mesh
used to model it. The initial gap can optionally be varied as a function of spatial coordinates by using a table.
Thickness Direction - Tension

The tensile behavior of the gasket in the thickness direction is linear elastic and is governed by a tensile modulus D t . The
latter is defined as a pressure per unit closure distance (that is, length). D t can optionally be varied as a function of
temperature and spatial coordinates by using a multi-variate table.
Transverse Shear and Membrane Behavior

The transverse shear is defined in the 2-3 and 3-1 planes of the local coordinate system (for three-dimensional elements)
or the 1-2 plane (for two-dimensional elements). It is linear elastic and characterized by a transverse shear modulus G t 1 .
In three-dimensional elements only, orthotropic transverse shear behavior may alternatively be characterized by the first
transverse shear modulus in the 2-3 plane G t 1 and by the second transverse shear modulus in the 3-1 plane G t 2 . The shear
moduli can optionally be varied as a function of temperature and spatial coordinates by using a multi-variate table.
CHAPTER 5 245
Material Library
The membrane behavior is defined in the local 1-2 plane (for three-dimensional elements) or the local 3-1 plane (for two-
dimensional elements) and is linear elastic and may be isotropic or orthotropic. Young’s modulus E m and Poisson’s
ratio  m that govern the membrane behavior are taken from an existing material that must be defined using the ISOTROPIC
or ORTHOTROPIC model definition option. The influence of these combinations of isotropic or orthotropic membrane and
isotropic or orthotropic gasket behavior on the constitutive relations is summarized in equation (5-85) and table Table 5-2
for three-dimensional elements and equation (5-86) and table Table 5-4 for two-dimensional elements. Multiple gasket
material can refer to the same isotropic or orthotropic material for their membrane properties (see also the GASKET model
definition option in Marc Volume C: Program Input).
Note: Although not all input from an Orthotropic base material is used, the base material still needs to be defined as
a valid material with all Young's moduli and shear moduli greater than zero.
Thermal Expansion
The thermal expansion of the gasket material is isotropic and the thermal expansion coefficient is taken from the isotropic
material that also describes the membrane behavior.
Constitutive Equations
As mentioned above, the behavior in the thickness direction of the gasket is formulated as a relation between the pressure p
on the gasket and the gasket closure distance c. In order to formulate the constitutive equations of the gasket material, this
relation (Equation (5-84)) must first be written in terms of stresses and strains. This depends heavily on the stress and strain
tensor employed in the analysis. For small strain analyses, for example, the engineering stress and strain are used. In that
case, the incremental gasket closure and the incremental pressure are related to the incremental strain and the incremental
stress by:
c = – h and p = –  (5-84)
in which h is the thickness of the gasket.
The resulting constitutive equation for three-dimensional elements, expressed in the local coordinate system of the
integration points, now reads
E1  21 E 1
-------------------------- -------------------------- 0 0 0 0  11
 11 1 –  12  21 1 –  12  21
 22  12 E 2 E2  22
-------------------------- -------------------------- 0 0 0 0 p
 33 1 –  12  21 1 –  12  21  33 –  33
= (5-85)
 12 0 0 C 0 0 0  12
 23 0 0 0 G 12 0 0
 23
 31 0 0 0 0 Gt 1 0
 31
0 0 0 0 0 Gt 2
in which C = hD c .
Table 5-2 Effective combinations of base and gasket materials for three-dimensional elements
Table 5-3
Isotropic Base + Isotropic Base + Orthotropic Base + Orthotropic Base +
Parameter Isotropic Gasket Orthotropic Gasket Isotropic Gasket Orthotropic Gasket
E1 Young's Modulus Base Young's Modulus Base E 1 Base E 1 Base
E2 Young's Modulus Base Young's Modulus Base E 2 Base E 2 Base
 12 Poisson's ratio Base Poisson's ratio Base  12 Base  12 Base
G 12 E1 E1 Modulus G 12 Base Modulus G 12 Base
--------------------------
- --------------------------
-
2  1 +  12  2  1 +  12 
G t1 Modulus Gasket First Modulus Gasket Modulus Gasket First Modulus Gasket
G t2 Modulus Gasket Second Modulus Modulus Gasket Second Modulus Gasket
Gasket
For two-dimensional elements, the equation is given by:
E1  31 E 1
-------------------------- 0 -------------------------- 0  11
 11 1 –  13  31 1 –  13  31
 22 p
0 C 0 0  22 –  22
= (5-86)
 33  13 E 3 E3  33
-------------------------- 0 -------------------------- 0
 12 1 –  13  31 1 –  13  31
 12
0 0 0 Gt
Table 5-4 Effective combinations of base and gasket materials for two-dimensional elements.
Table 5-5
Isotropic Base + Isotropic Orthotropic Base +
Parameter Gasket Isotropic Gasket
E1 Young's Modulus Base E 1 Base
E3 Young's Modulus Base E 3 Base
 31 Poisson's ratio Base  31 Base
G t1 Modulus Gasket Modulus Gasket
For large deformations in a total Lagrange formulation, in which the Green-Lagrange strains and the second Piola-
Kirchhoff stresses are employed, as well as in an updated Lagrange environment, in which the logarithmic strains and
Cauchy stresses are being used, similar but more complex relations can be derived.
CHAPTER 5 247
Material Library
Nonlinear Hypoelastic Material

The hypoelastic model is able to represent a nonlinear elastic (reversible) material behavior. For this constitutive theory,
Marc assumes that
· i j = L ijkl · kl + g ij (5-87)
where L is a function of the mechanical strain and g is a function of the temperature.

The stress and strains are true stresses and logarithmic strains, respectively, when used in conjunction with the LARGE STRAIN.
When used in conjunction with the LARGE DISP option only, Equation (5-87) is expressed as
· ·
S ij = L ijkl E kl + g ij (5-88)
where E S are the Green-Lagrangian strain and second Piola-Kirchhoff stress, respectively.
A HYPOELASTIC or NLELAST model definition option is necessary to invoke this model. These models can be used with any
stress element, including Herrmann formulation elements.
The tensors L and g are defined by you in the HYPELA2 user subroutine or via the NLELAST model definition option.
HYPELA2 User Subroutine

In order to provide an accurate solution, L should be a tangent stiffness evaluated at the beginning of the iteration. In
addition, the total stress should be defined as its exact value at the end of the increment. This allows the residual load
correction to work effectively.
In HYPELA2, additional information is available regarding the kinematics of deformation. In particular, the deformation
gradient ( F ), rotation tensor ( R ), and the eigenvalues (  ) and eigenvectors ( N ) to form the stretch tensor ( U ) are also
provided. This information is available only for the continuum elements namely: plane strain, generalized plane strain,
plane stress, axisymmetric, axisymmetric with twist, and three-dimensional cases.
For more information on the use of user subroutines, see Marc Volume D: User Subroutines and Special Routines.
NLELAST Model Definition

The NLELAST model definition in Marc supports the MSC Nastran NLELAST capability and simplified nonlinear elasticity
models. These models all have the form that a strain energy function does not exist but have been found to be useful in
solving engineering simulations, where only a limited amount of test data is available. The theory and algorithms are
adequate to trace the stress-strain curve accurately for the uniaxial loading cases. However, the theory is not based on the
classical theory of finite elasticity. Consequently, some of the constitutive relations may be violated in the multiaxial stress
state and, as a result, limit the use of this model to small strain.
Six models are implemented in Marc to support nonlinear elasticity based upon the input of a uniaxial stress strain law.
These models are summarized as:
1. MSC Nastran like NLELAST model based upon MATS1 input option.
2. Hypoelastic strain invariant model.
3. Model based upon working in principal strain space which results in induced anisotropic behavior.
4. Linear elasticity with tension and/or compression cut-off.

5. Bi-modulus elasticity with tension and/or compression cut-off.
6. Nonlinear orthotropic elasticity based upon non-rotating preferred directions.
The HYPELA2 user subroutine can be used if special material behavior is required.
Basic assumptions and definitions

A material model is called elastic if the stress is a function of the strain. This implies two important properties: the material
is reversible and the stress at a point in the material depends only on some measure of strain at that point. Elastic materials
implemented with this option are in the hypoelastic framework.
This group of nonlinear elastic materials is therefore grouped in Mentat under Materials -> HYPOELASTIC.
The nonlinearity in the material models introduced here results from the definition of the stress as a function of the strain
path. The models are anticipated to behave correctly in the small strain region only (less than 5%).
Input of uniaxial stress-strain data

In the next sections the theory and input is described for each of the different material models. As a basis it is assumed that
the user has a set of engineering stresses and engineering strains available. This means that in the uniaxial test:
 The measured elongations are compared to the original length of the test specimen, resulting in:
 engineering = L  L 0
 The measured forces are compared to the original cross-section of the test specimen, resulting in:
 engineering = F  A 0
For analysis without large incremental or total rotation or large incremental or total strains, the engineering values can be
used as is. In some models, certain characteristics still have to be derived. This is described in the sections per material type.
For analysis where the LARGE DISP, UPDATE, and FINITE, or PLASTICITY,3, or LARGE STRAIN parameters are used, the uniaxial test
data cannot be used ‘as is’. Engineering strain and stress are measures that are invalid once the strain in the analysis model
is no longer ‘small’ (approximately 5%). For these large strain simulations, the true stress and strain definition have been
adopted in Marc. This nonlinear strain measure is dependent upon the final length of the model. Since the NLELAST
material models are only valid in this small range it is not necessarily required to convert the engineering data.
Converting engineering strain/stress to true strain/stress

True strain is defined as:
 true =  L  L
By integration, it can be shown that this directly defines the true strain as:
 true = ln  1 +  engineering  (5-89)
CHAPTER 5 249
Material Library
Similarly, the uniaxial engineering stress data can be converted to true stress (also known as Cauchy stress) for
incompressible materials by means of:
 true =  engineering  1 +  engineering  (5-90)
and, in general, by  tr ue =  engineering   1 – v   engineering  2
By using Equations (5-89) and (5-90), the uniaxial test data can quickly be transformed into true stress/strain data. Figure 5-18
shows the deviation of engineering stress/strain compared to the engineering definition. In the remainder of this section on
nonlinear elasticity the engineering stresses and strains can be mutually exchanged with their true counterpart. The tables
can be either engineering or true data depending on the options activated in the analysis.
Comparison between engineering and true stress/strain
200
150
100
true or engineering stress
True Stress
Engineering Stress
50
0
-0.05 -0.04 -0.03 -0.02 -0.01 0 0.01 0.02 0.03 0.04 0.05
-50
-100
-150
-200
true or engineering strain
Figure 5-18 Comparison Between Engineering and True Stress/Strain
Type 1: MSC Nastran like nonlinear elasticity model

The stress curve is supplied in a table with independent variable mechanical strain (id=69). An additional independent
variable in the table may be temperature (id=12). Note that a table with independent variable mechanical strain is a
requirement for the analysis.
It is optional to define Poisson’s ratio as a function of temperature. The initial Young’s modulus is the slope of the equivalent
stress-strain curve is derived by Marc from the table at ? = 0. This imposes the continuous definition of the slope at ? = 0.
Theoretical Basis
The nonlinear elastic capability in MSC Nastran was designed to satisfy the equivalence of the deformation work per unit
volume in the simple tension to the strain energy per unit volume (conservation of energy), while the work done for
deformation may be defined by a stress-strain curve in simple tension, i.e.,
  d =  <>  d 
It is further assumed that the effective strain  may be defined by:

1 2 1
--- E = --- <>  D e   d 
2 2
From the total differential of the above equation, we obtain:

1
d = ------ <>  D e    
E
Substituting the latter in the first equation above, where stresses may be expressed in terms of total strains, i.e.

   = ------  D e    
E
The tangential matrix for such material may be obtained by differentiating the latter equation, i.e.
    
 D ne  = ------------- = ------  D e  + ------  ------- – ---  D e    <>  D e  T
 E E    
In the original MSC Nastran, implementation  is set to 1. In the Marc implementation, testing proved that setting this
value to 0.0 (default) improves convergence significantly.
Solution Algorithm
The user specifies the nonlinear stress-strain curve,     in a TABLE parameter along with a NLELAST model definition. The
NLELAST entry does require a table for Young’s modulus; whereas, strain dependency for Poisson’s ratio is optional.
The NLELAST material interfaces with the element routines with the following data:
 Input:    old ,    o l d ,    , E , 
 Output:    new ,    new ,  D n e 
The computational procedure is described below:

Step 1: Upon entry the new strain state is computed by:
   new =    old +   
Step 2: The effective strain is computed based on    new by

1
 2 = --- <>   e 
E
where
 e  =  De    
D e is simply formed using the Hookean law.

CHAPTER 5 251
Material Library
Step 3: The effective stress    is determined by looking-up the user-specified stress-strain curve for .  .
Step 4: The new stress state is determined by scaling the stress-strain law:

   new = ------   e 
E
Step 5: The tangential matrix is determined by:
   
 D ne  = ------  D e  + ---------  ------- – ---   e    e  T
E E 2    

Furthermore ------- is the slope at the current strain.

The tables for NLELAST must be supplied in the new table format.
Combining Uniaxial Tension and Compression Stress-strain Curves

Some materials exhibit appreciably different behavior in compression from that in tension even in the small strain range.
For uniaxial loading the magnitude of the strain in that direction becomes the effective strain, i.e.,
 =  x for uniaxial tension in x
 = –  x for uniaxial compression in x

In order to remain compatible with the MSC Nastran implementation of this material model, the user has the option to
ignore the uniaxial compression data, even if it is supplied on input.
In case of this unsymmetric material behavior, we need to be able to distinguish between a state of compression and a state
of tension. There are two known data points for one effective strain  , namely the effective stress for uniaxial tension   t 
and the effective stress for uniaxial compression   c  . Some method of interpolation or extrapolation is required to predict
the effective stress for the general stress state using two known data points.
The first stress invariant  I 1  has been adopted for interpolation/extrapolation as follows,
I1 = x + y + z
Considering that the pure shear is in the midway between simple tension and simple compression, it seems appropriate to
use the first stress invariant. Hydrostatic tension and compression cases will impose lower- and upper-bounds for
extrapolation, i.e.,
I 1 =  x for uniaxial tension/compression
I 1 = 0 for pure shear
I 1 = 3p for hydrostatic pressure

The instantaneous modulus  ------- should be interpolated or extrapolated in the same manner.
  
Solution Algorithm for Bilateral Stress-strain Relations

The new stress state is proportional in magnitude to the effective stress  , which should be determined as follows:
1. Compute the effective stress   e  based on   e  ; i.e.,
 e  =  De    
1 2 2 2
 = ---    x –  y  2 +   y –  z  2 +   z –  x  2  + 3   xy +  yz +  zx  for 3-D
2
2 2 2
 =  x –  x  y +  y + 3 x y for plane stress
1 2
 = ---    x –  y  2 +   y –  z  2 +   z –  x  2  + 3 xy for plane strain
2
2. Compute the first invariant of I =  x +  y +  z :
where  z = 0 for plane stress
3. Determine the ratio  r  by normalizing I I by  c , i.e.,
I1
r = -----
-
e
where r signifies the relative distance from the midpoint of  c and  t at  . It would be implausible to process a
large value of r (such is the case with a hydrostatic load). Therefore, r , will be confined to a plausible range,
– 1  r  1 . The value will be reset to the limit  r =  1  if r lies outside the range.
4. Look up the user specified stress-strain curve in the TABLE entry and determine  t and  c ; i.e.,
t =    
c =    
5. Determine based on  t ,  c . and r ; i.e.,
t + c t – c
- + r ------------------
 = -----------------
2 2

For the tangent matrix the instantaneous modulus  ------- should be determined using the same ratio (r) as follows:
  
1. Compute the instantaneous slope at  for tension, i.e.,

CHAPTER 5 253
Material Library
  yi + 1 – yi
------- = ----------------------
- for x i    x i + 1
   t xi + 1 – xi
where  x i y i  is the i-th data point in the TABLE entry
2. Compute the instantaneous slope at –  for compression; i.e.,
  yj + 1 – yj
------- = ----------------------
- for x i    x i + 1
   c xj + 1 – xj
    
3. Determine ------- based on ------- , ------- and r , i.e.,
    t    c
 1   r  
------- = ---  ------- +  ------- + ---  ------- –  -------
 2   t    c  2   t     c
Subsequently the values can now be used as in the tension-only algorithm.
Type 2: Strain Invariant Model

This model, like others in Marc is based upon the incremental theory where,
d = C el d e l
And C el is a Hooke’s law based upon the current values of the Young’s modulus and Poisson’s ratio. In this model these
quantities are a function of the strain invariants I 1 , I 2 , I 3 .
For this model, Marc computes the three strain invariants in each integration point of the element. The three strain
invariants are independent variables in the Young’s modulus vs strain-invariant table that is supplied on input. In principle
a table with all three strain invariants can be supplied if necessary, however, when converting the uniaxial data into Young
vs strain-invariant tables it is usually sufficient to choose one. The choice may depend on the element type used and the
need to model non-symmetric behavior in compression and tension. This is explained further below.
Marc looks up Young’s modulus in the user-supplied table and forms a Hookean stress-strain law.
In order to arrive at proper input, additional data conversion of the uni-axial data needs to be done. Starting from uniaxial
(engineering or true) stress strain data, the elasticity modulus (Young) needs to be derived for a number of engineering or
true strain data points. See Figure 5-19.
Uniaxial stress & Young vs Strain
160 10000
140
8000
Young's modulus
120
engineering strain
100 6000
Stress 80
Young's modulus
60 4000
40
2000
20
0 0
0 0.01 0.02 0.03 0.04 0.05
Strain
Figure 5-19 Uniaxial Stress and Young vs Strain
The definition of these invariants in Marc is as follows:

I 1 = trace    =  x x +  y y +  z z
2 2 2 2
I 2 = 1  2  trace         – I 1   =  x y +  yz +  zx   xx  yy +  yy  zz +  zz  xx 
2 2 2
I 3 = det    =  x x  y y  z z + 2 x y  y z  z x –  yy  yz +  yy  zx +  zz  xy
Because only two elastic moduli (Young’s modulus and Poisson ratio) which are functions of the invarients are defined, the
material will behave isotropically in the simulation. In the input table, the three strain invariants are indicated by means of
IDs 70, 71, and 72.
For uniaxial and plane stress elements, a special treatment is followed in Marc. The out-of-plane strains are not computed
internally. Therefore incompressibility (poisson = 0.5, see Figure 5-20) is assumed, such that I 1 = 0 : the first invariant is
very closely related to the hydrostatic strain  I 1  3  . The other invariants automatically follow from this assumption.
CHAPTER 5 255
Material Library
Influence of incompressibility on first strain invariant
0.02
0.015
First strain invariant
(poisson = 0.3) 0.01

0.005
(poisson = 0.5)
0
-0.05 -0.04 -0.03 -0.02 -0.01 0 0.01 0.02 0.03 0.04 0.05
-0.005
-0.01
-0.015
-0.02
Engineering strain
Figure 5-20 Influence of Incompressibility on First Strain Invariant
For a uniaxial element:

 xy =  yz = 0 and  y y =  z z
1
I 1 = 0 = > y y =  z z = – ---  x x
2
3 2
I 2 = ---  xx
4
1 3
I 3 = ---  xx
4
For a plane stress element:
 yz =  z x = 0
I 1 = 0 = > z z = –   x x +  y y 
2 2 2
I 2 =  xy +  x x +  y y +  x x  y y
2
I 3 =   xx +  y y    x y –  x x  y y 
It is important to realize that the incompressibility assumption for different elements has a direct impact on the material
data supplied in the input deck. For different element types, the input data derived from the uniaxial test data will be
different. We compare the behavior of a plain stress element (e.g., element 3), a hexahedral element (e.g., element 7), a plane
strain element (e.g., element 11) and a beam element (e.g., element 52).
For the materials, we define a Poisson’s ratio of 0.3 and we prescribe a fixed displacement which brings us into the nonlinear
part of the stress-strain curve. Note that Poisson’s ratio may also vary with the uniaxially measured strain, in this case this
has to be accounted for in the 2-D and 3-D elements.
3-D Element – Three Direct Components of Strain

For the hexahedral element, we may choose any of the invariants as independent variable. In order to get a corresponding
behavior with the uniaxial test, we have to supply the first invariant in the following form:
 xx =  engineering
 yy =  z z = –  engineering
I 1 =  xx +  y y +  z z =  1 – 2  engineering  or: I 1 =  1 – 2  true 
2-D Elements – Two Direct Components of Strain (= Plane Stress)

Although the plane stress elements in Marc have a third strain component unequal to zero, Marc does not compute it. It
has been chosen that the user input for the stress-strain data has to follow the incompressibility assumption  I 1 = 0  . In
this case, we may use the definition of the second invariant in order to define a dependency on strain, supposing a uniaxial
loading:
 xy =  yz =  z x = 0 (uniaxiality)
 yy = –  engineering
from I 1 = 0 , we deduce (see above):

2 2
I 2 =  xx +  y y +  x x  y y
2
I 2 =   2 –  + 1  engineering
2-D Elements – Three Direct Components of Strain (= Plane Strain)

For the plane strain elements, the third direct strain term is zero. In this case, incompressibility is not assumed and the strain
invariants can be easily derived from the conditions:  y y = 0 and  zz = 0 under uniaxial loading:
 xy =  yz =  z x =  z z = 0 (uniaxiality and plain strain)
 yy =     – 1   engineering
I 1 =  xx +  y y +  z z =  1 +     – 1   engineering
I 2 =  xx  yy
2
I 2 =     – 1   engineering
CHAPTER 5 257
Material Library
1-D Elements – One Direct Components of Strain

For the beam and truss elements, we cannot use the first invariant as independent variable either, since I 1 = 0 . We may
quickly derive that:
2 2
I 2 = 3  4 engineering (or I 2 = 3  4 true )
Loading Under Tension/Compression

It must be noted that for both beam and plane stress elements a nonsymmetric behavior in compression cannot be described
by means of the second invariant because I 2  0 (see Figure 5-21).
The third invariant can be used for this purpose; e.g., for the 1-D elements:
3 3
I 3 = 1  4 engineering (or I 3 = 1  4 true )
And finally for the 2-D plane stress elements:

 xy =  yz =  z x = 0 (uniaxiality)
 yy = –  engineering
Second strain invariant (> 0)
1.4000E-03
1.2000E-03
Second strain invariant
1.0000E-03
8.0000E-04
6.0000E-04
4.0000E-04
2.0000E-04
0.0000E+00
-0.05 -0.04 -0.03 -0.02 -0.01 0 0.01 0.02 0.03 0.04 0.05
Engineering strain
Figure 5-21 Second Strain Invariant (>0)
from I 1 = 0 , we deduce:
2
I 3 =   xx +  y y    x y –  x x  y y 
3
I 3    – 1  engineering
Data Approximation
Finally it is important to realize that the second and third invariants are nonlinear functions of the strain. As a result, even
a linear stress-strain curve will form a polynomial of the second or third order for the second and third strain invariant
respectively. It is, therefore, wise to apply the load in a sufficient number of increments, even though the actual behavior is
only linear. This is shown in Figure 5-22.
Type 3: Principal Strain Space Model

In the principal strain space theory, the strain tensor’s principal values are now computed. The compliance matrix is formed
using a total strain approach; i.e., the secant modulus is computed for each principal strain by looking up the stress in the
user defined table. This results in three values for the modulus. These values are then used to form an orthotropic stress-
strain law together with poisson’s ratio and a shear modulus, which are also user defined.
The definition of the orthotropic stress-strain law is described in Volume C (chapter 3). It must be noted that the constants
have to fulfill certain requirements in order to guarantee stability. These criteria are checked by Marc. However, if stability
cannot be guaranteed, Marc will continue to solve the increment at hand, and moreover that it will continue if the
increment converges. From this point on, the solution may not be trustworthy anymore and care should be taken.
Influence of # of datapoints on third strain invariant
2.0000E-05
1.5000E-05
1.0000E-05
Third strain invariant
5.0000E-06
0.0000E+00
-0.05 -0.04 -0.03 -0.02 -0.01 0 0.01 0.02 0.03 0.04 0.05
-5.0000E-06
-1.0000E-05
-1.5000E-05
-2.0000E-05
Engineering strain
Figure 5-22 Influence of Number of Data Points on the Third Strain Invariant
Theoretical Background
The Principal strain space theory can be defined as the incremental stress strain law which is based upon an orthotropic
Hooke’s law expressed with respect to a set of axis aligned with the principal axes based upon the strain tensor. The moduli
( E ,  , G ) are evaluated at the current values of the principal strains.
Consider E to be the reference material system of the element. For a continuum element, this is the global system, while
for a shell element this is the  1 ,  2 ,  3 system. Now, consider the total strain at the beginning of the increment  n . One
can obtain the principal strains and the eigenvectors of this system.
The principal strains are  1 ,  2 ,  3 and the eigenvectors are n 1 , n 2 , n 3 , The matrix that transforms the E system to the
principal system is R .
CHAPTER 5 259
Material Library
The state is then rotated into the principal, this includes the stresses, strains, thermal strains and creep strains.
 n + 1 = R T R , etc.
Based upon the principal strains, the material properties for the current value of the principal strains are determined.
The evaluated material constants are tested to make sure stability is not violated, also positive definiteness of the compliance
matrix is checked.
Then the constitutive law is evaluated based upon the current Young’s moduli, Poisson ratios, and shear modulus.
Based upon either the estimated or calculated incremental strain, the incremental stress in principal space is determined.
Subsequently the total stress in principal space is calculated and the quantities are rotated back to the E system.
p
 n + 1 = R n + 1 R T = R   p +  p R T
The constitutive law is also rotated back into the E system.
C = RC p R T
This is then used in the tangent stiffness matrix calculation.
Type 4: (Bimodulus) Elasticity Model

This is conventional Hooke’s law, where the Young’s modulus and Poisson ratio may be function of temperature or position,
but not strain or stress. Further below, a description is given of the treatment regarding tension and compression limits.
Linear Elasticity Bimodulus Elasticity
Type 5: (Bimodulus) Elasticity Model with Either No Tension, Limited Tension, No Compression, or Limited Compression Enhancements
In this model, the hydrostatic strain will be calculated and if the value is positive, the tensile values of Young’s modulus and
Poisson’s ratio will be used. If the hydrostatic strain is 0, as in the very beginning of the analysis, the tensile values will be
used.
For Model 4
Given total strains, the total stresses are calculated based upon tensile properties. This results in C and  n + 1 ,
For Model 5
To distinguish a tension or compression stress state the hydrostatic strain is computed.
If the hydrostatic strain is positive the total stresses are calculated based upon tensile properties. Resulting in C ten and
n + 1
If the hydrostatic strain is negative, the total stresses are calculated based upon compressive properties. This results in C com
and  n + 1 ,
No Tension Material No Compression Material
Bimodulus with both Tension and Compression Cut-off
For Either Model 4 or 5

The treatment of the cut-offs works as follows, if they are defined in the simulation.
The principal stresses  1 ,  2 ,  3 and the eigenvectors are n 1 , n 2 , n 3 are obtained.
If  I  Tensile stress limit, then principal direction i will be reduced.

CHAPTER 5 261
Material Library
If  I  Compressive stress limit, then principal direction i will be reduced.
The reduction implies that a new secant modulus is computed for the current reduced direction. The maximum secant
modulus in one of the reduced directions is used to form a new stress-strain law.
The transformation matrix: R e p is obtained between Element system E and Principal system P .
A stiffness matrix is recreated using the secant modulus of the current limit stress vs strain state.This is also done to improve
stability.
C z = R ep C p R e pT
Rotate stresses to principal system and maximize the stress to the limit in ith component, and then rotate back into element
system:
p
 n + 1 = R epT  n + 1 R e p
p–z
 nz + 1 = R ep  n + 1 R e p T
Note that there is no need to form C p because C is from isotropic elasticity, and R is a proper orthogonal matrix.
Therefore the stiffness matrix C can be formed in the global system with the limited secant modulus.
Type 6: Nonlinear Orthotropic Elasticity Model

This model is based upon orthotropic Hooke’s law, written with respect to a fixed preferred material orientation. The
difference is that the nine constants, E 11 , E 22 , E 33 ,  12 ,  23 ,  31 , G 12 , G 23 , G 31 may be a function of the spatial
position, temperature, and the strain in the associated direction. This can be expressed as:
E 11 = E 11  X ,T , 11  , etc.
 12 =  12  x ,T , 12  , etc.
G 12 = G 12  x ,T , 12  , etc.
The material properties are evaluated based upon the current value of the independent variables, and the Hooke’s law is
used to create the current tangent modulus. The incremental stress follows from these moduli and the incremental strains.
The evaluated material constants are evaluated to make sure stability is not violated, also it is determined if the compliance
matrix is positive definite.
As the material orientations do not change, the usual procedure for defining them is used.
Thermo-Mechanical Shape Memory Model - Type 1

NiTi alloys with near-equiatomic composition exhibit a reversible, thermoelastic transformation between a high-
temperature, ordered cubic (B2) austenitic phase and a low-temperature, monoclinic (B19) martensitic phase. The density
change and thus the volumetric are small and on the order of 0.003. The transformation strains are, thus mainly deviatoric,
of the order of 0.07-0.085. However, these small dilational strains do not necessarily lead to a lack of pressure sensitivity in
the phenomenology. The behavior of nitinol is different depending on whether the materials are subjected to hydrostatic
tension or compression.
Typical phenomenology is shown in Figure 5-23 taken from Miyazaki et al. (1981). The curves indicate that upon cooling,
the material transformation from austenite to martensite begins once the M s temperature is reached. Upon further cooling,
the volume fraction on martensite is a given function of temperature; the volume fraction becomes 100% martensite when
the M f temperature is reached. Upon heating, transformation from martensite to austenite begins only after A s
temperature is reached. This re-transformation is complete when the A f temperature is reached. Finally, note that the four
transformation temperatures are stress dependent. The experimental data indicate the M s , M f , A s , and A f may be
approximated from their stress-free values, M s0 , M f0 , A s0 , and A f0 by
e q
M s = M s0 + -------- ,
Cm
e q
M f = M f0 + -------- ; and
Cm
e q
A s = A s0 + -------- ,
Ca
e q
A f = A f0 + -------- .
Ca
where  eq is the von Mises equivalent stress. At a sufficiently high temperature, often called the M d temperature,
transformation to martensite does not occur at any level of stress.
The transformation characteristics such as the transformation temperatures depend sensitively on alloy composition and
heat treatment.
CHAPTER 5 263
Material Library
Mf As
1.0
Austenite to martensite &
0.9 martensite to austentie
decomposition
0.8
Stress = 0
0.7
Martensite Volume Fraction
Note:
0.6 600
After partial
transformation,
Tensile Stress (MPa)

0.5 decomposition
begins at As. 400
0.4
200
0.3
0
0.2 77 150 Ms 200 Af 250 300
Temperature (K)
0.1
MS Af
0.0
0 10 20 30 40 50 60 70 80 90 100
Temperature
Figure 5-23 Austenite to Martensite and Martensite to Austenite Decomposition
Transformation Induced Deformation

For the discussion of the thermo-mechanical response of NiTi, the data of Miyazaki et al. (1981) is shown in Figure 5-24.
Following this thermal history, it is observed that, when unstrained specimens with fully austenitic microstructures are
cooled, the transformation to martensite begins at a temperature of 190K; the transformation is complete at 128K. This
established the so-called martensite start ( M s0 ) and martensite finish ( M f0 ) temperatures at 190K and 128K, respectively.
With the imposition of an applied uniaxial tensile stress, the low temperature martensite is favored and the M s0 and M f0
temperatures increase. Upon heating a specimen with fully martensitic microstructure, the reverse transformation is
observed to begin at a temperature of 188K and to be complete at 221K. These define the austenite start ( A s0 ) and
austenite finish ( A f0 ) temperatures, respectively. Uniaxial tension tests are carried out in temperature ranges where
T  M s , M s  T  A f , and A f  T  T c where T c is defined as the temperature above which the yield strength of the
austenitic phase is lower than the stress required to induce the austenite-to-martensite transformation.
(a) 77K (b) 153K (c) 164K

300
200
100
0 0 0
400
(d) 224K (e) 232K (f) 241K
Tensile Stress (MPa)

300
200
100
0 0 0
(h) 273K (i) 276K
600 (g) 263K
400
200 Ms = 190K
AF = 221K
0 2 4 0 2 4 0 2 4
Strain (%)
Figure 5-24 Thermal History
In the temperature range where T  M f , the microstructures are all martensitic. The stress versus strain curves display a
smooth parabolic type of behavior which is consistent with deformation caused by the movement of defects such as twin
boundaries and the boundaries between variants. Note that unloading occurs nearly elastically and that the accumulated
deformation, caused by the reorientation of the existing martensite and the transformation of any pre-existing austenite,
remains after the specimen is completely unloaded. Note also that the accumulated deformation is entirely due to oriented
martensite and this would be recoverable upon heating to temperatures above the ( A s – A f ) range. This would, then,
display the shape memory effect.
Pseudoelastic behavior is displayed in the temperature range A f  T  T c . In this range, the initial microstructures are
essentially all austenitic, and stress induced martensite is formed, along with the associated deformation; upon unloading,
however, the martensite is unstable and reverts to austenite thereby undoing the accumulated deformation. Note that, as
expected, the stress levels rise with increasing temperature. In this range, the transformation induced deformation is nearly
all reversible upon unloading.
At temperatures where T  T c , plastic deformation appears to precede the formation of stress induced martensite. The
unloading part of the stress versus strain behavior displays nonlinearity and the unloading is now associated with permanent
(plastic) deformation. Permanent deformation, due to plastic deformation of the austenite, is nonrecoverable and, if such
deformation is large, shape memory behavior is lost.
CHAPTER 5 265
Material Library
Constitutive Theory
The model formulated below is based on the kinematics of small strains, although the extension to large strain is
straightforward. Accordingly, the incremental strain,  , is simply the sum of the following contributions:
 =  el +  th +  p l +  p h (5-91)
In Equation (5-91),  el is the incremental elastic, or lattice, strain rate;  t h is the incremental thermal strain,  p l is
the incremental visco-plastic strain, and  p h is the incremental strain associated with thermoelastic phase
transformations. The incremental elastic strain is taken to be simply related to a set of elastic moduli, L , and the
incremental Cauchy stress rate,  , as:
 = L el (5-92)
To calculate the coefficient of thermal expansion of the composite, the rule of mixtures is used as:
 =  1 – f  A + f M .
In the above equations, the superscripts A and M refer to the austenite and martensite values, and f is the volume fraction
of martensite.
Phase Transformation Strains

As noted earlier, the phase transformation induced strains are a result of the formation of oriented, stress induced,
martensite and the reorientation of randomly oriented thermally induced martensite. To account for this,  P h is
expressed as:
 Ph =  TR I P +  TW I N
where,
3 '
 TR IP = f  +  g   e q  eTq --- -------- + f  +   T I + f  -   P h . (5-93)
2 e q
and
3'
 TW I N = fg   eq  eTq ------------   e q    e q –  egf f  . (5-94)
2 e q
1 x+ x
where, f = f  +  + f  -  and   represents McCauley’s bracket where  x  = ---  ---------------- , x  o .
2 x 
T is the magnitude of the deviatoric part of

In Equation (5-93), f  +  represents the rate at which martensite is formed,  eq
the transformation, and  T is the volumetric part of the transformation strain. The function g   eq  is schematically
depicted in Figure 5-25, and is a measure of the extent to which the martensite transformation strains are aligned with the
deviatoric stress.  eq is the equivalent stress defined as:
3 d
e q = ---  :  d where  d is the deviatoric stress.
2
1.1
0.9
0.7
0.5
g
0.3
0.1
-1.0
0.0 0.5 1.0 1.5 2.0
stress/g0
Figure 5-25 Function g   eq 
The first two terms in Equation (5-93) describe the development of transformation induced strains due to the formation of
stress induced (partially oriented) martensite. f  -  is the change of formation of austenite; for example, the rate at which
the volume fraction of martensite decreases. The last term in Equation (5-93), therefore, represents the recovery of the
accumulated phase transformation strain.
Note that there is no dilatational contribution to  TW I N since f is fixed. Note that the twinning strain rate is zero when
g , or when the magnitude of the stress change is negative (  g
 e q is less than  eff eq  0 ). Hence,  ef f can be considered
as a stress below which no twinning is possible.
The function g represents the extent to which the transformation strains are coaxial with the applied deviatoric stress. This
function can be calibrated with the experimental data. Note for uniaxial stress-strain curves performed below the martensite
finish temperature, the material starts as 100% martensite, and that other than elastic strains, the deformation is dominated
by the “twinning” of the randomly oriented martensite.
CHAPTER 5 267
Material Library
A functional form that leads to sufficient fit to most experimental data has been implemented in Marc.
gb gd gf
e q e q  eq
g   eq  = 1 – exp g a  -------- + g c  -------- + g e  --------
 g0   g0   g0 
In most cases, the first term is sufficient, and a value of g a  0 and g b = 2 yields the best results. g 0 is a stress level used
to non-dimensionalize the constants, and can be chosen such that g  1 when the  eq  g 0 . In some cases, it is
necessary to include the higher powers of equivalent stress for better experimental fits. In these, cases suggested values for
g d = 2.55 or 2.75 and g e = 3 . However, depending on the values of g c and g e , this could lead to maxima or minima
values of g in the range of interest. Note that 0  g  1 and it should be a monotonically increasing function (an increase
in the stress level should lead to an increase in the increment of the phase strains). Thus, cut off values of g are provided
g g , it is held constant
in Marc, such that when g reaches its maximum value g = g m ax at the stress level  eq =  m ax 0
at the value g m ax . For the proper selection of g 0 , see the following section.
Experimental Data Fitting for Thermo-mechanical Shape Memory Alloy

The properties and transformation/retransformation behavior of Nititol depend upon alloy chemistry, microstructure and
the thermal processing applied to the specimens and eventually to the components built from them. Every time any of the
above change, it might be necessary to redo the calibration. Calibration of Nitinol experimental data works best if, in fact,
all specimens can be initially rendered as 100% austenite. The list of properties that require calibration is given as follows:
 The unstressed transformation temperatures, M s0 ,M f0 ,A s0 ,A f0 .
 The coefficients C M , and C A that provide the stress dependence of the transformation temperatures.
 The elastic constants ( E M ,E A ,v M and v A ).

 Coefficients of thermal expansion,  M , A .
 The calibration of the detwinning function, g   eq  that provides the description of the degree to which
martensite is co-axial with the deviatoric stress state.
 The yield stress of the pure martensite and austenite phase (  YM and  YA ), and their strain hardening properties.
 The calibration of the transformation strains,  eTq and  vT .
Transformation Temperatures and Their Stress Dependence ( M s0 ,M f0 ,A s0 ,A f0 ,C M , and C A )

For almost any use of shape memory alloy, it is highly desirable that one knows the Transformation Temperatures (TTRs)
of the alloy. The TTRs are those temperatures at which the alloy changes from the higher temperature austenite to the lower
temperature martensite or vice versa. There are in common use with NiTi alloys to provide helpful data to product
designers – Constant Load, DSC and Active A f . The detail procedures to obtain TTRs using the above methods are shown
in website (www.sma-inc.com). It is recommended that combined dilatometry and DSC tests be performed on unstressed
specimens of thermally processed material to establish both the unstressed transformation temperatures and the thermal
expansion coefficients. Those tests would provide a baseline set of values for M s0 ,M f0 ,A s0 , and A f0 . Note that the TTRs
are stress dependent parameters, but it is difficult, in practice, to prepare totally unstressed samples. In order to determine
the TTRs at zero stress, experimental data must be obtained at two or more stress levels. The particular transformation
point of interest can then be extrapolated to zero stress. The estimations of TTRs, C M , and C A are shown in Figure 5-26.
The typical range of TTRs is –200 to 100°C. So, it is difficult to recommend the default values. As references, there are
examples for two different SMA materials below.
SMA 1)
M s0 : – 50C , M f0 : – 100C , A s0 : 5C , A f0 : 20C , C M : 6.0Mpa/C , C A : 8.0Mpa/C
SMA 2)
M s0 : 190K , M f0 : 128K , A s0 : 188K , A f0 : 221K , C M : 5.33Mpa/K , C A : 6.25Mpa/K

Stress
CM
CA
M f0 M s0 A s0 A f0
Temperature
Figure 5-26 Typical Stress vs. Temperature Curve Showing the Stress Dependence of Martensite and Austenite Start and
Finish Temperature
Elastic Constants ( E M ,E A ,v M , and v A )

Literature estimates for the elastic moduli of martensite and austenite are typically in the range of
E M = 28000-41000 Mpa , E A = 60000-83000Mpa , v M = v A = 0.33 . However, most experimental data appears
to be significantly different than these. It is, therefore, suggested that estimates of these moduli should be made using actual
experimental data for the materials being calibrated. Initial loading from a state corresponding to 100% austenite produces
a linear elastic response from which E A can be readily estimated as in Figure 5-27. The modulus of martensite ( E M ) is also
estimated for the unloading line, again as illustrated in Figure 5-27. In this figure, the loading should be performed to
produce 100% martensite and thus the unloading occurs with the elastic response of martensite.
CHAPTER 5 269
Material Library
Typical pseudoelastic
response
T = T 1°C
EA
Stress
EM
Strain
Figure 5-27 Typical Stress-strain Curves in the Pseudo-elastic Regime, Depicting the Elastic Moduli
Thermal expansion coefficients (  M , A )

A recommended method for measuring thermal expansion coefficients is through the use of dilatometry whereby carefully
controlled cycles of temperature can be made. An alternative to this type of precise calibration, is to use literature values
that have been shown to be consistent with values measured on actual specimens.
These values are given as follows (See for example, TiNi Smart Sheet)
 M = 6.6  10 – 6  C = 3.67  10 – 6  F ; and
 A = 11.0  10 – 6  C = 6.11  10 – 6  F
Detwinning and the calibration of the g function g   e q 
The phenomenology of the NiTi phase transformation is such that the alignment of the martensite varies with the
prevailing deviatoric stress. This intensity is measured via the von Mises equivalent stress,  eq . As shown in Equations 5-93
and 5-94, the scaling function that provides the description of the degree to which the martensite is aligned is g   eq  . The
most direct path to calibrating this g function is to fit it to the uniaxial stress vs. strain curve for pure, randomly oriented
martensite conducted at a temperature below the M f0 temperature. Such a curve is shown as Figure 5-28. The solid curve
shown in Figure 5-28 is the actual measured record of uniaxial stress vs. total strain for a specimen of 100% martensite tested
at a temperature sufficiently low to ensure it remains 100% martensite. The dot curve is simply a convenient fit to it. The
T is the “equivalent deviatoric transformation strain”. Note that the function g is defined as it relates to the
parameter  eq
development of deviatoric strain due to the alignment of martensite variants. As mentioned in the previous section, in
general, the variables g a  0 , g b = 2f = 2.0 , g c  0 , g d = 2.25 and 2.75 , g e  0 and g f = 3.0 yield a good
match to many experimental results. It is often observed that there exists a threshold equivalent stress level below which
detwinning does not occur; this stress is referred to as  egf f . The value of g at this stress is g ef f = g   fgf  . Note that from
Equation (5-94), twinning strain is zero when  e q   egf f . In addition, it is also found, in practice, that the function g
g g
tends to approach unity at a finite equivalent stress level, called  0 . By definition, g   0  = 1 . Also, g a should be
chosen to match the general shape of the function. Since the ratio  eq  g 0 is less than one, the higher powers take effect
later, and thus g c can be added to lower the middle slope of the curve and g e to fix the final slope of the curve. However,
depending on the relative values of g a , g b and g c , this curve might reach a maximum in the range of interest, and
therefore, it should be cut-off at its maximum value g max . The value of g m ax which is reach at a stress value
g g g
e q  g0 = m ax are also supplied as an input to Marc. Usually g 0 = 2 ef f  10 ef f is a good approximation. But,
the selection of g 0 depends on the experimental measurement.
g0
 TW I N = g   e q  eTq Experimental
Model
Stress
g0
Strain
Figure 5-28 Typical Stress-Strain Curve of 100% Martensite Tested Below M f0 Temperature
Others
Yield stresses of the pure martensite and austenite phases for NiTi:  YM and  YA .
 YM = 70-140 Mpa
 YA = 195 – 690 Mpa

The calibration of the transformation strains for NiTi:
 eTq (deviatoric transformation strain): 0.05-0.085

 vT (volumetric transformation strain): 0 – 0.003.
 egff (detwinning stress): 100-150 Mpa
g
g0 2 eff  10 egf f
CHAPTER 5 271
Material Library
Note: The current model uses a nonsymmetric Jacobian matrix. It is recommended that the nonsymmetric solver be
used to improve convergence.
Thermo-Mechanical Shape Memory Model - Type 2

The Thermo-mechanical shape memory model type 2 is very similar to the thermo-mechanical shape memory model of
the previous section. The differences are in the enhanced definition of the equivalent stress used in the transformation po-
tential and in the treatment of plasticity. This section focuses on these differences only, for all other aspects of the model
the reader is referred to the previous section.
The equivalent stress used in the transformation potential of the previous section is always the equivalent Von Mises stress
which depends on J2 (second invariant of stress deviator) only. The enhanced definition may in addition depend on I1
(first invariant of stress) and J3 (third invariant of stress deviator) through the introduction of two extra material parameters
 t and  t that control the transformation behaviour in different states of stress.
The equivalent stress  in the thermo-mechanical shape memory model type 2 is given as:
1---
3--- 3
27 2    0 -1 +1
 = 3 t I 1 +  3J 2  +  t ------ J 3 -----   t  ------- (5-95)
2  t 2 2
Where,
I 1 =  ii
1
J 2 = --- s ij s ji
2
1
J 3 = --- s ij s jk s ki
3
The stress deviator s ij is the traceless part of the stress  ij and is defined as:
1
s i j =  ij – ---  kk  ij (5-96)
3
The two material constants  t and  t can e.g. be used to define distinct transformation behavior in pure tension and pure
compression. For  t = 0 and  t = 0 ,the expression for the equivalent stress reduces to the Von Mises equivalent stress.
The equivalent stress may be written as:
 = 3 t I 1 + Q t  J 2 J 3  (5-97)
with
1
3 ---
--- 3
2 27
Q t  J 2 J 3  =  3J 2  +  t ------ J 3 (5-98)
2
The total strain is decomposed additively in elastic strain, thermal strain, transformation strain and plastic strain
e th ph pl
 i j =  ij +  ij +  ij +  ij (5-99)
The stress is computed from the elastic strain by a linear isotropic Hooke’s law
e
 i j = L ijkl  f   kl (5-100)
where the fourth order isotropic elasticity tensor is given as:

2
L i j k l  f  = G  f    ik  jl +  il  jk  +  K  f  – --- G  f    ij  k l  (5-101)
 3 
Here f is the fraction of martensite in the multi-phase aggregate of austenite and martensite. The shear modulus G  f  and
the bulk modulus K  f  are computed by a linear mixture rule from the shear and bulk moduli of the two-phase fractions
M A
G  f  = fG +  1 – f G (5-102)
and
M A
K  f  = fK +  1 – f  K (5-103)
For both phase fractions the shear and bulk moduli follow from the entered Young’s modulus and Poisson’s ratio through
the relations
X
G
X E
= ------------------------
- (5-104)
X
21 +  
and
X
K
X E
= ---------------------------
- (5-105)
X
3  1 – 2 
where the superscript X denotes A for austenite and M for martensite.

The thermal strain rate · ij is given as:
th
· th
 i j =   f T·  ij (5-106)
The coefficient of thermal expansion is also computed by a linear mixture rule

M A
  f  = f +  1 – f  (5-107)
· ph
The rate of transformation strain  ij for the forward transformation from austenite to martensite is computed as:
· ph · T Q t T Q t · T ·
 ij = f g  Q t  eq ---------- + fg·  Q t  e q ---------- + f  V  i j f  0 (5-108)
 ij  i j
The g -function to model detwinning behavior is the same as in the previous section, but here it depends on the value of
the potential Q t . The rate of transformation strain · ij for the reverse transformation from martensite to austenite is com-
ph
puted as:
·
· ph f ph ·
 ij = ------------  ij f  0 (5-109)
f MAX MAX
CHAPTER 5 273
Material Library
where f MA X is the maximum martensite phase fraction reached during a forward transformation from austenite to marten-
site just before the reverse transformation sets in and  ph
ij MA X is the associated maximum transformation strain reached
during this forward transformation.
Classic flow theory of plasticity is used to model plastic deformation, which occurs for stress states on the yield surface de-
fined as:
–p
  i j  = 3 p  f I 1 + Q p  J 2 J 3  –  y    f  = 0 (5-110)
The potential Q p  J 2 J 3  is defined in a similar way as for the transformation behavior and is given as:
1
3 ---
--- 3
2 27
Q p  J 2 J 3  =  3 J 2  +  p  f  ------ J 3 (5-111)
2
For stress states inside the yield surface, i.e.    ij   0 , no plasticity occurs and stress states outside the yield surface, i.e.
· pl
   i j   0 , are unreachable. The rate of plastic strain  ij is proportional to the stress-gradient of the potential Q p and can
be written as:
· pl · Q p
 i j =  --------- (5-112)
 ij
where · is the plastic multiplier. This means that if the yield surface is I1-independent, the plasticity is associative, other-
wise it is non-associative. The material parameters  p and  p occurring in the expression for the yield surface are given by
a linear mixture law in terms of the parameters of the austenite and martensite phase fractions as:
M A
 p  f  = f p +  1 – f  p (5-113)
M A
 p  f  = f p +  1 – f  p (5-114)
The yield stress of the multi-phase aggregate is defined by a more general nonlinear mixture rule
– pl M –p l A –p l
 y    f  = m y  f  y    +  1 – m y  f   y    (5-115)
where the function my  f  can be given by a polynomial of degree up to five as:

5
i
my  f  =  Ci f (5-116)
i = 1
Note that my  1  = 1 , which means that the constants Ci must satisfy the condition
5
 Ci = 1
i = 1
The function m y  f  must be monotonically increasing from zero to one, i.e. m y'  f   0 .
pl
The yield stress of both phase fractions can be a function of the total equivalent plastic strain  which is defined as
t
pl 2--- · pl · pl
 =    dt (5-117)
3 ij ij
0
pl
and the hardening behavior is assumed to be isotropic. The total equivalent plastic strain  in both phase fractions is
assumed to be the same if plasticity and phase transformation occur simultaneously. If a yield stress is defined for one
phase fraction only (usually the martensite fraction to model plasticity after the forward phase transformation is complete)
the plasticity parameters of this particular phase fraction are used only and the mixture rules for  p ,  p and  y are not
used.
The note at the end of the previous section also applies to the thermo-mechanical shape memory model type 2, so the non-
symmetric solver is recommended for this model.
Calibration of the Parameters Related to Tension/Compression Sensitivity

The general form of the equivalent stress in the thermo-mechanical shape memory model type 2 is given in terms of the
stress invariants I1, J2 and J3 as
1---
3--- 3
2 27    0 1 1
 = 3 t I 1 +  3J 2  +  t ------ J 3 – ---   t  + --- 
2  t 2 2  (5-118)
The two material parameters t and t can be used to control the transformation behavior in different states of stress. In
particular, these parameters can be used to obtain asymmetric behavior in tension and compression of the alloy giving rise
to the tension/compression sensitivity also known as the strength differential effect. When stress-strain curves of a uniaxial
test in tension are compared with stress-strain curves of a uniaxial test in compression it is often observed that the behavior
is not entirely symmetric. When t=0 and t=0 the equivalent stress reduces to the Von Mises equivalent stress and this
results in symmetric behavior in tension and compression. However, for small positive values of t and t the stress where
the forward transformation from austenite to martensite is initiated in uniaxial tension is lower than the absolute stress
where this transformation is initiated in compression. Similarly, the stress where the reverse transformation from martensite
to austenite is initiated in tension is also lower than the absolute stress where this transformation is initiated in compression.
In addition, the maximum transformation strain in tension may be slightly higher than the absolute maximum transfor-
mation strain in compression. The pseudoelastic loop of an isothermal tensile test appears stretched in comparison to the
pseudoelastic loop of a compression test carried out at the same temperature as is illustrated in Figure 5-29. Note that the
tension and the compression curves in the figure represent the measurements carried out on two different specimen. The
curves in Figure 5-29 are idealized representations of the real measurements marking the four characteristic points where the
forward and reverse transformations start and end. The points are also shown in the phase diagram of Figure 5-30.
CHAPTER 5 275
Material Library
Figure 5-29 Pseudoelastic Behavior in Uniaxial Tension and Compression
The absolute stress values T and C where the transformation is initiated and the maximum absolute phase transforma-
tion strain values  p h and Cph can be used to evaluate both the parameters t and t in the phase transformation potential
T T
of Equation (5-118) as well as the deviatoric transformation strain e q . The t and t also have an effect on the evaluation
of the zero-stress transformation temperature pairs Ms , M f and As , Af0 and the slopes CM and CA of the lines separating
0 0 0
the phase transformation zones in the phase diagram of Figure 5-30.
Figure 5-30 Phase Diagram for Phase Transformations in a Shape Memory Alloy
In general, it is necessary to take into account if the analysis uses the small strain formulation or the large strain formulation.
It is assumed that the measurements of stress and strain record these values as engineering values. These are the defined
stresses and strains of the small strain formulation, but for the large strain formulation the engineering stresses and engi-
neering strains must be converted to true stresses and true strains. This assures that analysis results remain in close agree-
ment with experimental results should the experiments be simulated numerically with Marc. The conversion for uniaxial
loading is given by the following expressions
e
 =  1 + e 
and
 = ln  1 + e 
where the pairs ,  represent true stress, true strain and the pairs e, e represent engineering stress, engineering strain.
From these expressions it can be observed that for strain levels up to around 5% the differences between engineering and
true stress, strain values are still relatively small, so for calibrating the material data the conversion can be omitted if the
accuracy loss is acceptable. However, the large strain formulation is generally recommended, because it accounts for any
geometric nonlinear behavior, even if the strain levels remain small. In the following it is assumed that stresses and strains
represent true stresses and true strains so it is assumed that the four characteristic points of the pseudoelastic loops and the
maximum transformation strains, as shown in Figure 5-29, have been converted from measured engineering values to their
true counterparts. This can be done by converting the engineering stress-strain curves to true stress-strain curves and read-
ing the necessary stress and strain values from the converted curves.
It is assumed that the following experimental data is available.

 Isothermal stress-strain curves in tension with complete pseudoelastic loops measured for different temperatures.
 For at least one of the temperatures such a curve in compression.
The following material parameters need to be evaluated from experimental data:

1. The transformation potential parameters t and t.
T T
2. The deviatoric transformation strain e q and the volumetric transformation strain v .
0 0 0 0
3. The phase diagram temperature pairs Ms , Mf and As , Af and the slopes CM and CA.
4. The elastic material parameters EM, EA, M and A.
5. The coefficients of thermal expansion M and A.
6. The parameters of the g-function for the detwinning model.
7. The parameters of the plasticity model.
In this section the aspects related to points 1, 2, 3 and 7 are discussed. A discussion of the aspects related to points 4, 5 and
6 can be found in the sub-section Experimental Data Fitting for Thermo-mechanical Shape Memory Alloy in the section Thermo-Me-
chanical Shape Memory Model - Type 1.
For the calibration of t and t it is assumed that complete loading-unloading cycles have been traversed, i.e. that the for-
ward transformation from austenite to martensite and the reverse transformation from martensite to austenite both have
completed resulting in stress-strain curves with complete pseudoelastic loops as shown in Figure 5-29. It is also assumed that
CHAPTER 5 277
Material Library
the stress levels during the forward transformation are high enough to ignore the effects of the g-function. This justifies the
assumptions g(Qt)≈1 and g(Qt)≈0. It is furthermore assumed that the transformation deformations are volume preserving,
T
so it can also be assumed that  v ≈0. In this case the rate of transformation strain in the forward transformation can be
approximated as
(5-119)
· ph Q t ·
· T ---------
i j = f eq - f  0
 ij
where
Q Q t Q t 2
----------t = --------- s ij + ---------  s i j s k j – --- J 2  i j (5-120)
 ij J 2 J 3  3 
In above equation sij denotes the stress deviator of stress ij defined as
1 (5-121)
s ij =  ij – ---  k k  i j
3
and ij denotes the Kronecker delta (ij=1 when i=j, otherwise ij=0). The stress invariants are defined as
I 1 =  ii
1
J 2 = --- s ij s ji
2
1
J 3 = --- s ij s jk s k i
3
For the uniaxial state of stress there is one non-zero stress component only and the stress tensor becomes
 0 0
̃ =
0 0 0
0 0 0
The stress deviator tensor then becomes
2---
 0 0
3
s̃ = 1
0 – ---  0
3
1
0 0 – --- 
3
The expressions for the invariant stress quantities when the forward transformation starts and for the maximum transfor-
mation strains in the case of uniaxial tension and uniaxial compression, cf. Figure 5-29, are summarized in the following
Table 5-6.
Table 5-6 Invariant Stress, Strain Quantities for Uniaxial Tension and Uniaxial Compression
Uniaxial Tension Uniaxial Compression

I1= T I1= -C
J 2 = 1---  2 J 2 = 1---  2
3 T 3 T
2- 3
J 3 = ----- 2-  3
J 3 = – -----

27 T 27 C
1 1
--- ---
Qt =  1 + t  3 T Qt =  1 – t  3 C
1 0 0 1 0 0
1
 Qt
1
--- 1  Qt --- 1
3 0 – --- 0
---------- =  1 +  t  0 – --2- 0
3 ---------- = –  1 –  t 
  ij 2
 i j
0 0 – 1--- 0 0 – 1---
2 2
1 1
--- ---
3 3
 = 3 t +  1 +  t  T  = – 3 t +  1 – t  C
1 1
--- ---
ph T 3 ph T 3
T = eq  1 +  t  C = eq  1 –  t 
From Table 5-6 it can be observed that the ratio of the maximum transformation strains in tension and compression can be
given as
1---
ph
T  1 + t  3
----------
- = ---------------------
ph 1---
C
 1 – t  3 (5-122)
from which t can be solved as

  ph 
3
T - – 1
 -------------
 ph 
 C 
 t = -----------------------------
(5-123)
  ph 
3
T - + 1
 -------------
 ph 
 C 
T
Once t is known the deviatoric transformation strain eq is easily found from the expression for Tph as
ph (5-124)
T T
e q = -----------------------
1---
1 +  3 t
From Table 5-6 it can also be observed that the ratio of the stresses in tension and compression when the transformation
starts can be given as
CHAPTER 5 279
Material Library
1---
3
C 3 t +  1 +  t  (5-125)
------- = --------------------------------------------- -
T 1---
3
– 3 t +  1 – t 
from which t can be solved as

ph  1---
1--- 
1--- 
C 3  ------- – -------------  1 –   3
3
C T
-------  1 –   –  1 +    ph  t
 t t  T  
T
 = ----------------------------------------------------------------- C (5-126)
- = ------------------------------------------------------------ -
t
 C   C 
3  ------- + 1 3  ------- + 1
 T   T 
ph ph
C T  T
It is expected that t≥0 and t≥0, which means that it is expected that -  1 and ------C  ----------
------  1 ---------- -.
T C
ph  T p h
C
C
The ratio ------
T
for all four of the characteristic points in the pseudoelastic loops in uniaxial tension and compression should
be the same, so t could also be estimated from an averaged value of these ratios.
Once the parameters t and t have been estimated they can be used in estimating the parameters of the phase diagram of
i
Figure 5-30. The four characteristic points of the tensile tests are denoted as  T ,i=1,2,3,4. Likewise, the characteristic points
i
of the compressive tests are denoted as  C ,i=1,2,3,4. These stress values are converted to equivalent stress values by
1
---
3
multiplying the tensile stress values by the factor 3 t +  1 + t  and the compressive stress values by the factor
1
---
3
– 3 t +  1 – t  , i.e.
1
---
– i 3 i
T = 3 t +  1 +  t  T
and
1
---
– i 3 i
C = – 3 t +  1 – t  C
– i – i
The equivalent stresses are denoted here as  T and  C with i=1,2,3,4. The equivalent stresses can be plotted as a function
of temperature as illustrated by the different points in Figure 5-30. These points can be used to find the slopes CM and CA
of the line pairs marking the stress values at the start and end of the forward and reverse transformations and their inter-
sections with the temperature axis at zero stress resulting in the temperature pairs Ms0 , M f0 and A0s , Af0 .
When plasticity is present it is assumed to occur after the forward transformation has completed as shown by the dashed
continuation of the tensile loading curve in Figure 5-29. From this loading curve the initial yield stress and the hardening
data can be estimated. The hardening data defines the yield stress as a function of equivalent plastic strain, which is entered
in a TABLE. This means that the elastic part  e and the transformation part  ph of strain must be stripped off from the total
T
strain  as shown in Figure 5-29 to obtain the plastic part  p of strain as
p ph e (5-127)
 =  – T – 
It is also assumed that there is no tension/compression sensitivity in the plastic behavior, so J2 -plasticity can be applied
and the plastic data of the uniaxial tensile test suffices to extract the plastic material parameters. Should this sensitivity be
present in the material behavior it is also necessary to closely monitor the plastic behavior in compression and other defor-
mation modes like shear and by a similar approach as demonstrated for the transformation parameters an additional set of
p and p parameters for martensite can be extracted from the plasticity curves to define a yield surface that deviates from
the Von Mises yield surface. However, the experimental procedure to accurately define the shape of the yield surface can
become very elaborate.
Should simultaneous plasticity and transformation occur the procedure gets much more involved and much more cyclic
experimental data, including data with partial transformation from austenite to martensite at load reversal, is required. An
added difficulty may be the specification of plasticity parameters for both martensite and austenite and establishing the
right mixture rule for the yield stress. Simultaneous plasticity and transformation behavior may occur at elevated tempera-
tures, where the transformation stress levels rise, as can be observed from the phase diagram, but the yield stress usually
drops, so there may be stress and temperature levels, where both modes of deformation can occur simultaneously. However,
this is considered an exceptional case not generally encountered in Nitinol applications and the calibration of the material
parameters is not explored any further for this case.
Mechanical Shape Memory Model

Shape-memory alloys can undergo reversible changes in the crystallographic symmetry-point-group. Such changes can be
interpreted as martensitic phase transformations, that is, as solid-solid diffusion-less phase transformations between a
crystallographically more-ordered phase (the austenite or parent phase) and a crystallographically less-ordered phase (the
martensite). Typically, the austenite is stable at high temperatures and high values of the stress. For a stress-free state, we
indicate with the temperature above which only the austenite is stable and with the temperature below which only the
martensite is stable.
The phase transformations between austenite and martensite are the key to explain the superelasticity effect. For the simple
case of uniaxial tensile stress, a brief explanation follows (Figure 5-31). Consider a specimen in the austenitic state and at a
temperature greater than A; accordingly, only the austenite is stable at zero stress. If the specimen is loaded, while keeping
the temperature constant, the material presents a nonlinear behavior (ABC) due to stress-induced conversion of austenite
into martensite. Upon unloading, while again keeping the temperature constant, a reverse transformation from martensite
to austenite occurs (CDA) as a result of the instability of the martensite at zero stress. At the end of the loading/unloading
process, no permanent strains are present and the stress-strain path is a closed hysteresis loop.
 C
D
A 
Figure 5-31 Superelasticity

CHAPTER 5 281
Material Library
At the crystallographic level, if there is no preferred direction for the occurrence of the transformation, the martensite takes
advantage of the existence of different possible habit plates (the contact plane between the austenite and the martensite
during a single-crystal transformation), forming a series of crystallographically equivalent variants. The product phase is
then termed multiple-variant martensite and it is characterized by a twinned structure. However, if there is a preferred
direction for the occurrence of the transformation (often associated with a state of stress), all the martensite crystals trend
to be formed on the most favorable habit plane. The product is then termed single-variant martensite and is characterized
by a detwinned structure. According to the existence of different types of single-variant martensite species, the conversion
of each single-variant martensite into different single variants is possible. Such a process, known as a reorientation process,
can be interpreted as a family of martensite phase transformations and is associated with changes in the parameters
governing the single-variant martensite production (hence, it is often associated to nonproportional change of stress).
In addition to the thermo-mechanical shape memory model, a superelastic shape memory alloy model is also implemented
in Marc based on the work of Auricchio [Ref. 1] and [Ref. 2]. This work has been enhanced to allow different elastic
properties for the Austenite and Martensite phases.
The superelastic shape memory model has been implemented in Marc in the framework of multiplicative decomposition.
We assume the deformation gradient, F as the control variable, and the martensite fraction,  S as the only scalar internal
variable. We also introduce a multiplicative decomposition of F in the form:
F = F e F tr
where F e is the elastic part and F t r is the phase transition part.
Assuming an isotropic elastic response, the Kirchhoff stress  and the elastic left Cauchy-Green tensor b e , defined as:
b e = F e F eT , share the same principal directions. Accordingly, the following spectral decompositions can be introduced:
3
 =  A n A  n B
A = 1
3
d
d =  A n A  n B
A = 1
3
be =    Ae  2 n A  n B
A = 1
with  Ae the elastic principal stretches and  d the deviatoric part, according to the relation:
 = pI +  d (5-128)
where I is the second-order identity tensor, p is the pressure, defined as p = tr     3 , and tr  .  is the trace operator.
We can write Equation (5-128) with the following component form:
d
A = p + A (5-129)
with
d
p = K e ,  A = 2Ge Ae .
Phase Transformations and Activation Conditions

We consider two phase transformations: the conversion of austenite into martensite (A S) and the conversion of
martensite into austenite (S A). To model the possible phase-transformation pressure-dependence, we introduce a
Drucker-Prager-type loading function:
F =  d + 3p (5-130)
where  is a material parameter and

. indicates the Euclidean norm, such that:
3 12
d 2
d =   A  .
A = 1
Indicating variants in time with a superposed dot, we assume the following linear forms for the evolution of  S :
·
· F
 S = H AS  1 –  S  -------------------- for (A S) (5-131)
F – R fA S
·
· F
 S = H SA  S -------------------- for (S A) (5-132)
F – R fSA
where
2 2
R fAS =  fAS  --- +  , R fS A =  fS A  --- + 
 3   3 
SA
with  sAS ,  fAS ,  sS A , and  fS A material constants. The scalar quantities H A S and H embed the plastic-
transformation activation condition – hence, allowing a choice between Equations (5-131) and (5-132) – and they are defined
by the relations:
H AS = 1 , if R sA S  F  R fA S , or F·  0 . Otherwise, H A S = 0 .
H SA = 1 , if R fS A  F  R sS A , or F·  0 . Otherwise, H S A = 0 .
where
2 2
R sAS =  sAS  --- +  , R sS A =  sS A  --- +  .
 3   3 
CHAPTER 5 283
Material Library
Time-discrete Model
The time-discrete model is obtained by integrating the time-continuous model over the time interval [ t n t ]. In particular,
we use a backward-Euler integration formula for the rate-equations evaluating all the nonrate equations at time t . Written
in residual form and clearing fraction from Equations (5-131) and (5-132), the time-discrete evolutionary equations specialize
to:
R A S =  F – R fAS  s – H AS  1 –  S   F – F n  = 0 (5-133)
R S A =  F – R fSA  s – H S A  S  F – F n  = 0 (5-134)
where,
t
·
(5-135)
S =   dt =  S –  S ,n
tn
The quantity  S in Equation (5-135) can be computed expressing F as a function of  S and requiring the satisfaction of
the discrete equation relative to the corresponding active phase transition.
The detailed solution algorithm for stress update and consistent tangent modulus are given in the work of Auricchio [Ref. 2].
In the enhanced version of this model, the user enters different elastic constants for the two phases. In this case, the effective
elastic moduli are taken as:
A M
E = E  1 – S  + E S
A M
 =   1 – S  +  S
Experimental Data Fitting for Mechanical Shape Memory Alloy

The experiment for mechanical shape memory alloy is quite simple.
1. To determine the transformation stresses (  SA S , fA S , SSA , fSA ):
 SAS Initial Stress for Austenite to Martensite

 fAS Final Stress for Austenite to Martensite
 SSA Initial Stress for Martensite to Austenite
 fSA Initial Stress for Martensite to Austenite
Uniaxial tension test is performed at the same temperature at which the simulation is desired. Here is one example
set for SMA materials.
 SAS = 500Mpa ,  fA S = 600Mpa ,  SA S = 300Mpa ,  fSA = 200Mpa
2.  :It is measured from the difference between the response in tension and compression.
Case 1) if the behavior in tension and compression are the same, the value is set to 0.
Case 2) if the behavior in tension and compression have a difference as in the classical case of SMA, the value
is usually set to 0.1 if there is no compression data for the phase transformation. One value for the phase
transformation in compression, say  sA S– (sigAS_s_compression) is available,  is calculated as
follows:
 = sqrt  2  3    SA S – –  SA S     SA S – +  SA S 
CHAPTER 5 285
Material Library
3.  L : epsL is a scalar parameter representing the maximum deformation obtainable only by detwinning of the
multiple-variant martensite (or maximum strain obtainable by variant orientation). Classical values for epsL
are in the range 0.005 and 0.10. Marc sets the default value as 0.07.
Note: The mechanical shape memory model only supports ndi = 3 case (3-D, plane-strain and
axisymmetric elements). It does not support either ndi = 1 or ndi = 2 cases (1-D and plane-stress elements).
Conversion from Thermo-Mechanical to Mechanical SMA

Table 5-7 Conversion Table
Thermo-Mechanical SMA Mechanical SMA Enhanced Mechanical SMA
A A
E E = 0.5  E A + E M  E
A A
  = 0.5   A +  M  
M M
E E = 0.5  E A + E M  E
M M
  = 0.5   A +  M  
3 T 3 T
 L = sqrt  ---  eq  L = sqrt  ---  eq
T
 eq
 2  2
CM CM CM
CA CA CA
To To To
Table 5-8 Prediction from Linear Algebra

The Relationship between Mechanical Model
and Thermo-Mechanical Model
AS 0
s =  T o – M s C m
AS 0
f =  T o – M f C m
0
 sSA =  T o – A s C a
0
 fSA =  T o – A f C a
 AS
f
AS
s
CA
SA
s
SA
f
CM
T
Mf Ms As Af To
Elastomer
An elastomer is a polymer which shows nonlinear elastic stress-strain behavior. The term elastomer is often used to refer to
materials which show a rubber-like behavior, even though no rubbers exist which show a purely elastic behavior. Depending
upon the type of rubber, elastomers show a more or less strongly pronounced viscoelastic behavior. Marc considers both
the viscous effects and the elastic aspects of the materials behavior. These materials are characterized by their elastic strain
energy function.
Elastomeric materials are elastic in the classical sense. Upon unloading, the stress-strain curve is retraced and there is no
permanent deformation. Elastomeric materials are initially isotropic. Figure 5-32 shows a typical stress-strain curve for an
elastomeric material.
, Stress
100%
e, Strain
Figure 5-32 A Typical Stress-Strain Curve for an Elastomeric Material
Calculations of stresses in an elastomeric material requires an existence of a strain energy function which is usually defined
in terms of invariants or stretch ratios. Significance and calculation of these kinematic quantities is discussed next.
In the rectangular block in Figure 5-33,  1 ,  2 , and  3 are the principal stretch ratios along the edges of the block defined
by
CHAPTER 5 287
Material Library
i =  Li + ui   Li (5-136)
L3 3L3
Undeformed
Deformed
1L1 2L2
L2
L1
Figure 5-33 Rectangular Rubber Block
In practice, the material behavior is (approximately) incompressible, leading to the constraint equation
1 2 3 = 1 (5-137)
the strain invariants are defined as:

2 2 2
I1 =  + 2 + 3
1
2 2 2 2 2 2
I2 = 1  +   +  (5-138)
2 2 3 3 1
2 2 2
I3 = 1 2 3
Depending on the choice of configurations, for example, reference (at t = 0 ) or current ( t = n + 1 ), you obtain total
or updated Lagrange formulations for elasticity. The kinematic measures for the two formulations are discussed next.
A. Total Lagrangian Formulation
The strain measure is the Green-Lagrange strain defined as:
1
E i j = ---  C ij –  ij  (5-139)
2
where C ij is the right Cauchy-Green deformation tensor defined as:
C ij = F ki F kj (5-140)
in which F kj is the deformation gradient (a two-point tensor) written as:
x
F kj = -------k- (5-141)
X j
The Jacobian J is defined as:

1---
2
J =  1  2  3 =  det C i j  (5-142)
Thus, the invariants can be written as:
I 1 = C ii (implied sum on i)
2
 C ij C ij –  C ii  
I 2 = -----------------------------------------
- (5-143)
2
1
I 3 = --- e ijk e pqr C ip C j q C k r = det  C i j 
6
in which e ijk is the permutation tensor. Also, using spectral decomposition theorem,
2 A A
C ij =  A N i N j (5-144)
2
in which the stretches  A are the eigenvalues of the right Cauchy-Green deformation tensor, C i j and the eigenvectors are
A
Ni .
B. Updated Lagrange Formulation

The strain measure is the true or logarithmic measure defined as:
1
 ij = --- l n b i j (5-145)
2
where the left Cauchy-Green or finger tensor b i j is defined as:
b ij = F ik F j k (5-146)
Thus, using the spectral decomposition theorem, the true strains are written as:
3
A A
 ij =  ln   A  n i n j (5-147)
A = 1
A
where n i is the eigenvectors in the current configuration. It is noted that the true strains can also be approximated using
first Padé approximation, which is a rational expansion of the tensor, as:
–1
 ij = 2  V i j –  i j   V i j +  i j  (5-148)
where a polar decomposition of the deformation gradient F i j is done into the left stretch tensor V i j and rotation tensor
R i j as:
F ij = V ik R k j (5-149)
CHAPTER 5 289
Material Library
The Jacobian J is defined as:

1---
2
J =  1  2  3 =  det b i j  (5-150)
and the invariants are now defined as:

I 1 = b ii
1 2
I 2 = ---  b ij b i j –  b i i  
2 (5-151)
1
I 3 = --- e ijk e p qr b i p b j q b k r = det  b i j 
6
It is noted that either Equation (5-143) or Equation (5-151) gives the same strain energy since it is scalar and invariant. Also,
to account for the incompressibility condition, in both formulations, the strain energy is split into deviatoric and
volumetric parts as:
W = W deviatoric + W volumetric (5-152)
Thus, the generalized Mooney-Rivlin (gmr) and the Ogden models for nearly-incompressible elastomeric materials are
written as:
N N
gmr m n (5-153)
W deviator ic =   C mn  I 1 – 3   I 2 – 3 
m = 1 n = 1
where I 1 and I 2 are the first and second deviatoric invariants.

A particular form of the generalized Mooney-Rivlin model, namely the third order deformation (tod) model, is
implemented in Marc.
tod
W devratoric = C 10  I 1 – 3  + C 01  I 2 – 3  + C 11  I 1 – 3   I 2 – 3  + C 20  I 1 – 3  2 +
3 (5-154)
C 30  I 1 – 3 
where,
tod is the deviatoric third order deformation form strain energy function,
W deviatoric
C 10 C 01 C 11 C 20 C 30 are material constants obtained from experimental data
Simpler and popular forms of the above strain energy function are obtained as:
nh
W deviatoric = C 10  I 1 – 3  Neo-Hookean
(5-155)
mr
W deviatoric = C 10  I 1 – 3  + C 01  I 2 – 3  Mooney-Rivlin
Use the MOONEY model definition option to activate the elastomeric material option in Marc and enter the material
constants C 10 C 01 C 11 C 20 C 30 .
The TEMPERATURE EFFECTS model definition option can be used to input the temperature dependency of the constants C 10
and C 01 . The UMOONY user subroutine can be used to modify all five constants C 01 , C 10 , C 11 , C 20 , and C 30 . Using
the table driven input, all material parameters can reference tables to define temperature dependent behavior. For
viscoelastic, the additional VISCELMOON model definition option must be included.
The form of strain energy for the Ogden model in Marc is,
N
og d en  k k k
W de v iator ic =  -----k-   1 +  2 +  3 – 3 (5-156)
k  
k = 1
k
k – ------ 
3 k
where  i = J i are the deviatoric stretch ratios while C m n ,  k , and  k are the material constants obtained
from the curve fitting of experimental data. This capability is available in Mentat.
If no bulk modulus is given, it is taken to be virtually incompressible. This model is different from the Mooney model in
several respects. The Mooney material model is with respect to the invariants of the right or left Cauchy-Green strain tensor
and implicitly assumes that the material is incompressible. The Ogden formulation is with respect to the eigenvalues of the
right or left Cauchy-Green strain, and the presence of the bulk modulus implies some compressibility. Using a two-term
series results in identical behavior as the Mooney mode if:
 1 = 2C 10 and  1 = 2 and  2 = – 2C 01 and  2 = – 2
CHAPTER 5 291
Material Library
The material data is given through the OGDEN model definition option or the UOGDEN user subroutine. For viscoelastic
behavior, the additional VISCELOGDEN model definition option must be included.
In the Arruda-Boyce strain energy model, the underlying molecular structure of elastomer is represented by an eight-chain
model to simulate the non-Gaussian behavior of individual chains in the network. The two parameters, nk and N ( n
is the chain density, k is the Boltzmann constant,  is the temperature, and N is the number of statistical links of length
l in the chain between chemical crosslinks) representing initial modules and limiting chain extensibility and are related to
the molecular chain orientation thus representing the physics of network deformation.
As evident in most models describing rubber deformation, the strain energy function constructed by fitting experiment
data obtained from one state of deformation to another fails to accurately describe that deformation mode. The Arruda-
Boyce model ameliorates this defect and is unique since the standard tensile test data provides sufficient accuracy for
multiple modes of deformation.
The model is constructed using the eight chain network as follows [Ref. 3]:
Consider a cube of dimension  0 with an unstretched network including eight chains of length r 0 = Nl , where the
fully extended chain has an approximate length of Nl. A chain vector from the center of the cube to a corner can be
expressed as:
0 0 0
C 1 = ------  1 i + ------  2 j + ------  3 k (5-157)
2 2 2
Using geometrical considerations, the chain vector length can be written as:
1 12
r chain = ------- Nl   12 +  22 +  32  (5-158)
3
and
r chain 1 12
- = -------  I 1 
 chain = ------------ (5-159)
r0 3
2 0
C1
3 0
1 0
k
Figure 5-34 Eight Chain Network in Stretched Configuration
Using statistical mechanics considerations, the work of deformation is proportional to the entropy change on stretching
the chains from the unstretched state and may be written in terms of the chain length as:
r chain 
W = nkN  -------------  + ln --------------  – Ĉ (5-160)
 Nl sinh  
where n is the chain density and Ĉ is a constant.  is an inverse Langevin function correctly accounts for the limiting
chain extensibility and is defined as:
r chain
 = L – 1  ------------- (5-161)
 Nl 
where Langevin is defined as:

1
    = coth  – --- (5-162)

With Equations (5-159) through (5-162), the Arruda-Boyce model can be written
Arruda-Boyce 1 1 11
W dev = nk ---  I 1 – 3  + ----------  I 12 – 9  + -------------------  I 13 – 27 
2 20N 1050N
2
19 519 (5-163)
+ -------------------  I 14 – 81  + -------------------------  I 15 – 243  
3 4
7000N 673750N
Also, using the notion of limiting chain extensibility, Gent [Ref. 5] proposed the following constitutive relation:
e nt = – E
 I1 – 3 
W dGev ---  I m – 3  log  1 – --------------
- (5-164)
6  I m – 3
The constant EI m is independent of molecular length and, hence, of degree of cross linking. The model is attractive due
to its simplicity, but yet captures the main behavior of a network of extensible molecules over the entire range of possible
strains.
The Arruda-Boyce and Gent model can be invoked by using the ARRUDBOYCE and GENT model definition options,
respectively.
The anisotropic hyperelastic models can be defined using the ANISO HYPE model definition option. The detailed theory is
discussed in Anisotropic Hyperelasticity Model.
The volumetric part of the strain energy is for all the rubber models in Marc:
9 2
U = --- K  J 1  3 – 1  (5-165)
2
when K is the bulk modulus.

CHAPTER 5 293
Material Library
Marc will use the three field formulation for rubber materials (Mooney, Gent, Arruda Boyce, Ogden, Marlow, and
anisotropic hyperelastic model) if the bulk modulus is entered as a -1. A series form of the volumetric strain energy U to
capture the nonlinear pressure-volumetric strain relationship is implemented. The volumetric strain energy is expressed as:
5
U =  D i  J – 1  2i (5-166)
i = 1
where D i  i = 1 5  are material constants obtained from experimental data. This procedure should not be
used with Herrmann elements except for element types 155, 156, 157 or 247.
It can be noted that the particular form of volumetric strain energy is chosen such that:
1. The constraint condition is satisfied for incompressible deformations only; for example:

 > 0 if I 3  0

f  I 3   = 0 if I 3 = 1 (5-167)

 < 0 if I 3  0

2. The constraint condition does not contribute to the dilatational stiffness.
The pressure is obtained by taking the derivative of the volumetric strain energy function with respect to the
invariant yielding.
p = 3K  J – 1  (5-168)
The equation has a physical significance in that for small deformations, the pressure is linearly related to the
volumetric strains by the bulk modulus K .
The discontinuous or continuous damage models discussed in the models section on damage can be included with the
generalized Mooney-Rivlin, Ogden, Arruda-Boyce, and Gent models to simulate Mullins effect or fatigue of elastomers
when using the updated Lagrangian approach. In the total Lagrangian framework however, this is available for the Ogden
model only.
The rubber foam model which is based on Ogden formulation has a strain energy form as follows:
N N
     n
W =  -----n-   1 n +  2 n +  3 n – 3  +  -----n-  1 – J  (5-169)
n n  
n = 1 n = 1
where  n   n ,  n are material constants. When all terms  n are equal to zero, the second part of W in Equation (5-169)
is omitted.
You can define any other invariant based models through the use of the UENERG user subroutine when using the MOONEY
option in model definition for elasticity in total or Updated Lagrangian framework. A more general and easy-to-use the
UELASTOMER user subroutine (uelastomer.f) can be used to define a general strain energy function in the Updated
Lagrangian framework.
Once the strain energy function is defined, the stresses and material tangent can be evaluated for the total and Updated
Lagrangian formulations as:
A. Total Lagrangian Formulation:

The stress measure in the total Lagrangian formulation is the symmetric second Piola-Kirchhoff stress S i j ,
calculated as:
W W
S i j = ---------- = 2 ---------- (5-170)
E ij C ij
The material elasticity tangent is:

2 2
 W  W
D ijk l = ---------------------- = 4 ----------------------- (5-171)
E ij E kl C i j C k l
B. Updated Lagrangian Formulation:

The stress measure in the Updated Lagrangian formulation is the Cauchy or true stress calculated as:
2 W
 ij = --- ---------- b kj (5-172)
J b ik
The spatial elasticity tangent is:

2
4  W
L i j kl = --- b im ------------------------ b n l (5-173)
J b mj b k n
The material constants for the Mooney-Rivlin form can be obtained from experimental data. The Mooney-Rivlin form of
the strain energy density function is
mooney-rivlin
W deviatoric = C 10  I 1 – 3  + C 01  I 2 – 3  (5-174)
For the Mooney-Rivlin model, the force and deformation for a uniaxial test specimen can be related as
1
P = 2A 0  1 – ------   1 C 10 + C 01  (5-175)
 3
1
which can be written in the form:
P C 01
----------------------------------- = C 10 + --------
- (5-176)
1 1
2A 0   1 – ------
 2
1
where P is the force of the specimen, A 0 is the original area of the specimen, and  1 is the uniaxial stretch ratio. This
equation provides a simple way to determine the Mooney-Rivlin constants. The Mooney-Rivlin constitutive equation is
CHAPTER 5 295
Material Library
1 1
applicable if the plot of P  2A 0   1 – ------ versus ----- should yield a straight line of slope C 01 and intercept
 2 1
1
1
( C 01 + C 10 ) on the vertical axis ----- = 1 as shown in Figure 5-35.
1
0.4
G
0.3
/2(-1/2) (N/mn2)
E
D
C
A
0.2
B
0.1
0.5 0.6 0.7 0.8 0.9 1.0
1/
Figure 5-35 Plots for Various Rubbers in Simple Extension for Mooney-Rivlin Model
If only the Young’s modulus E is supplied, and full uniaxial data are not available then
C 01  0.25C 10 (5-177)
is a reasonable assumption. The constants then follow from the relation:

6  C 10 + C 01   E (5-178)
The material coefficients for the models can be obtained from Mentat. This allows you to select which model is most
appropriate for your data.
Anisotropic Hyperelasticity Model

A general, invariants-based strain energy function for fiber reinforced hyperelasticity can be described as:
ij ij
 =   I 1 I 2 I 3 I 4  I 5   i j  , i = 1 n and j = 1 i (5-179)
where n is the number of reinforcing fiber families. Each fiber family has a single direction defined by G i , a normalized
vector in reference configuration, or g i = F G i in current configuration. F is the deformation gradient and
 ij = G i  G j . In equation (5-179),
I 1 = trC = trb , (5-180)
1 1
I 2 = ---   trC  2 – trC 2  = ---   trb  2 – trb 2  , (5-181)
2 2
I 3 = detC = detb , (5-182)
ij
I 4 = G i  CG j = g i  g j , (5-183)
ij
I 5 = G i  C 2 G j = g i  bg j (5-184)
ij ij
with I 1 , I 2 , and I 3 are the three invariants of the right (or left) Cauchy-Green deformation tensor C (or b ); I 4 and I 5
are pseudo invariants measuring strains in the fiber directions.
If n = 1 , the material defined by equation (5-179) is transversely isotropic.
To facilitate the numerical treatment, in particular for the convenience of taking into account of volumetric constraints,
equation (5-179) can be split into a volumetric and a deviatoric part:
ij ij
 = U  J  + W  I 1 I 2 I 4  I 5   i j  , (5-185)
where U is the volumetric part of strain energy function and
J = I3 , (5-186)
I 1 = J – 2 / 3 I 1 = trb , (5-187)
1
I 1 = J – 4 / 3 I 2 = ---   trb  2 – trb 2  , (5-188)
2
ij ij
I4 = J –2 / 3 I4 = gi  gj , (5-189)
ij ij
I 5 = J – 4 / 3 I 5 = g i  bg j . (5-190)
In equation (5-186) through (5-190),
g i = J – 1 / 3 FG i = FG I (5-191)
b = J – 2 / 3 b = FF T (5-192)
The second Piola-Kirchhoff stress can be obtained by the first derivative of equation (5-185) with respect to C . Pushing
forward the second Piola-Kirchhof stress to the current configuration leads to the Kirchhoff stress and Cauchy stress if
divided by J . The hydrostatic pressure p and the deviatoric Cauchy stress s (or Kirchhoff stress  ) are obtained, after
some lengthy derivations, as:
CHAPTER 5 297
Material Library
dU  J 
p = ---------------- , (5-193)
dJ
 W W W
Js =  = dev  2  -------- + I 1 -------- b – -------- b 2
 I I  I 2
 1 2
n i
W
+   ---------  g i  g j + g j  g i  (5-194)
I 4ij
i = 1 j = 1
W
+ --------ij-  g i  g j + g i  g j + g j  g i + g j  g i   
I 5
where g i = bg i .
The tangent in reference configuration A v ol and A d e v , corresponding to the volumetric and deviatoric parts respectively,
can be obtained by the second derivative of equation (5-185) with respect to C .
Pushing forward A vol and A d e v into the current configurations leads to the tangent defined in deformed configuration
a v ol and a dev .
Marc uses mixed formulations, either on a two or three field variational principles, in the analysis for hyperelasticity. The
volumetric part has been built into the variational principles. The explicit form of a vo l or A vo l is not directly used.
Therefore, it will not be calculated.
Even though the derivation of a d ev or A d e v does not have any fundamental difficulties, it involves very lengthy and
complicated tensor operations. The final explicit form of a d ev or A d ev contains pages of formulations. In Marc, a d ev or
A d ev are calculated numerically. It facilitates users to define their owner models with user-defined subroutines since only
the first derivatives of energy function, equation (5-185), with respect to strain invariants, equations (5-186) through (5-190)
are required.
The following three specific forms of the volume preserving part of strain energy function W can be defined using the
ANISO HYPE option.
One is a simple extension of the so-called standard reinforcement model proposed by Qiu and Pence [Ref. 38]. The original
model has been proposed for transversely isotropic fiber reinforced elastomers. The extension makes it possible to take into
account multiple reinforcing fiber families. The strain energy function of the extended model is:
n
ii
W = C 10  I 1 – 3  + C 01  I 2 – 3  + k   I4 – 1 2 (5-195)
i = 1
where C 10 , C 01 , and k are material parameters.
The second model is proposed by Brown and Smith [Ref. 39]. Similar extension to take into account multiple reinforcing
fiber families has also applied to the model. The strain energy function is then given as:
2 if  I 4  1 
n 1
 i i --2-   kt
k = 
W = C 10  I 1 – 3  + C 01  I 2 – 3  + k   I 4 – 1
 
with
 kc if  I 4  1  (5-196)
i = 1
where C 10 , C 01 , k t , and k c are material parameters.
The third model by Gasser et al. [Ref. 40] is proposed to model arterial layers with distributed collagen fiber orientations:
k1
W = C 10  I 1 – 3  + --------  exp   k 2  E i 2  – 1   (5-197)
2k 2
With,
ii
E i =   I 1 – 3  +  1 – 3   I 4 – 1  (5-198)
1
C 10 , k 1 , k 2 , and  are material parameters. The parameter   0    --- defines the level of dispersion of fiber
 3
1
directions. The micro-buckling of fibers has been taken into account with  E i = ---  E i + E i  .
2
For the general form of W , the UELASTOMER_ANISO user subroutine should be used.
Marlow Model
The Marlow model is directly based on experimental data. It assumes that the strain energy function can be given as a
function of the first invariant only. The form of this function is not defined explicitly, but its derivatives with respect to the
first invariant, which are needed to compute stress and stiffness, are evaluated directly from test data. This means that no
curve fitting is required to obtain material parameters of a strain energy function and that the supplied test data can be
reproduced exactly by the material model. The test data can be obtained from either a uniaxial tension test, an equi-biaxial
tension test, or a pure shear (planar tension) test. Although the model reproduces the test data exactly, other modes of
deformation may only be reproduced in an approximate sense.
The Marlow model can be used to simulate incompressible or slightly compressible behavior. If compressibility is to be
included some information about a volumetric strain energy function must be entered. By default a bulk modulus of 5000
times the initial shear modulus is used in a volumetric strain energy function of the form given in Equation (5-165) and the
behavior is almost incompressible. The initial shear modulus is estimated from the test data.
The strain energy function is assumed to be the sum of a deviatoric and a volumetric part
W Ma rlow = W  I 1  + U  J  (5-199)
The information for the deviatoric behavior is derived directly from the test data. The possible forms of U to describe the
volumetric behavior are given in Equations (5-165) and (5-166).
The first invariant of deformation is defined as:
2 2 2
I1 = 1 + 2 + 3 (5-200)
CHAPTER 5 299
Material Library
where the  i are the three principal stretch ratios of the deformation. The volumetric deformation is defined as
J = 1 2 3 (5-201)
For incompressible material behavior we have J = 1 . When compressibility is accounted for, the stretch ratios can be
volume neutralized as  i = J – 1 / 3  i and the volume neutralized first invariant, based on these modified stretch ratios
becomes I 1 = J – 2 / 3 I 1 . This volume neutralized first invariant is used to describe the deviatoric material behavior.
The principal Cauchy stress  i corresponding to the principal stretch ratio  i can be given as: (5-202)
2 W 2 1
 i = --- --------   i – --- I 1 + U
------- (5-203)
J I 1  3  J
We can easily verify that,

1
---   1 +  2 +  3  = U
------- = p (5-204)
3 J
where p is the hydrostatic pressure.

The engineering stress in this direction is defined as:
J
 i e  = ----  i (5-205)
i
and its associated engineering strain is defined as:

ei = i – 1 (5-206)
We can thus write the engineering stress in each principal direction as:
e 2 W 2 1 J U
i = J – 2 / 3 ---- --------   i – --- I 1 + ---- ------- (5-207)
 i I 1  3   i J
U
For incompressible behavior, we have J = 1 and ------- must be replaced by a hydrostatic pressure p which is no longer
J
derivable from a strain energy function, but has to follow from the boundary conditions in the finite element model. We
may now apply these relations to the three test cases, the uniaxial case, the equi-biaxial case, and the pure shear case. The
three tests have to be carried out as tensile tests.
Uniaxial Case
e e e
In the uniaxial case we have stresses  1 =   e   0 and  2 = 3 = 0 . The corresponding deformations are
J
given as  1 =  = 1 + e  e  0  and  2 =  3 = --- .

The nonzero stress   e  for this case is given as:
W  3 – J
  e  = 2J – 2 / 3 --------  --------------- (5-208)
I 1   2 
The volume neutralized first invariant becomes:
2J
I 1 = J – 2 / 3   2 + ------ (5-209)
 
From the hydrostatic pressure, we find:

1 e U
---  = J ------- (5-210)
3 J
Given any value of I 1 , we can find corresponding  and J of the uniaxial test from Equations (5-209) and (5-210) and use
these values to compute the first derivative of W with respect to I 1 as:
W 1 2
-------- = --- J 2 / 3   e   --------------- (5-211)
I 1 2   3 – J
CHAPTER 5 301
Material Library
By differentiating this expression with respect to  , we can find the second derivative of W with respect to I 1 as:
J
 3 2 – ------ 
  e   ----- 2 J
- ------ – ----------------------- + --2-   e 

------------- +
  3J  3 – J 
2W 1 2/3 2  
---
2- = 2 J
----------  ---------------------------------------------------------------------------------------------------- (5-212)
I 1 I 1
  3 – J  --------

The derivatives of I 1 and J with respect to  can be found by differentiating the two relations that were used to find the
corresponding  and J of the uniaxial test.
Equi-biaxial Case
e e e e
In the equi-biaxial case, we have stresses  1 = 3 =   0 and  2 = 0 . The corresponding deformations
J
are given as  1 =  3 =  = 1 + e  e  0  and  2 = ------ .
2
W  6 – J 2
  e  = 2J – 2 / 3 --------  ------------------ (5-213)
I 1   5 
J2
I 1 = J – 2 / 3  2 2 + ------ (5-214)
  4
From the hydrostatic pressure, we find:
2 e U
---  = J ------- (5-215)
3 J
Given any value of I 1 , we can find corresponding  and J of the equi-biaxial test from Equations (5-214) and (5-215) and
use these values to compute the first derivative of W with respect to I 1 as
W 1 5
-------- = --- J 2 / 3   e   ------------------ (5-216)
I 1 2   6 – J 2
J
 6 5 – 2J ------ 
  e   ----- 2 J
- ------ – ------------------------------ + --5-   e 

------------- +
  3J  6 – J2 
2W 1 2/3 5  
---
2- = 2 J
----------  ----------------------------------------------------------------------------------------------------------- (5-217)
I 1 I 1
  6 – J 2  --------

corresponding  and J of the equi-biaxial test.
Pure Shear Case

e e e
In the pure shear (planar tension) case we have stresses  1 =   e   0 , 3  0 , and  2 = 0 . The corresponding
J
deformations are given as  1 =  = 1 + e  e  0  ,  3 = 1 , and  2 = --- .

W  4 – J 2
  e  = 2J – 2 / 3 --------  ------------------ (5-218)
I 1   3 
J2
I 1 = J – 2 / 3   2 + ------ + 1 (5-219)
 2 
From the hydrostatic pressure we find:
1 e 4 – J2 U
---  = ----------------------------------- J ------- (5-220)
3  +  2 – 2J 2 J
4
Given any value of I 1 , we can find corresponding  and J of the pure shear test from Equations (5-219) and (5-220) and
use these values to compute the first derivative of W with respect to I 1 as
W 1 3
-------- = --- J 2 / 3   e  ------------------ (5-221)
I 1 2  – J2
4
CHAPTER 5 303
Material Library
J
 4 3 – 2J ------ 
  e   ----- 2 J
- ------ – ------------------------------ + --3-   e 

------------- +
  3J  4 – J2 
2W 1 2/3 3  
---
2- = 2 J
----------  ----------------------------------------------------------------------------------------------------------- (5-222)
I 1 I 1
  4 – J 2  --------

corresponding  and J of the pure shear test.
The test data is entered in tabular form as the engineering stress   e  versus the engineering strain e =  – 1 in the TABLE
model definition option. This table is referenced in the MARLOW model definition option to associate it with one of the
three test cases. The independent variable in the table for the strain is of type “experimental data” (type 40). If the material
is also temperature dependent a multi-dimensional table can be entered where one independent variable is the strain and
 
the other independent variable is the temperature. We observe that ------ = ------ , so for any value of  , the stress   e  and
 e
  e 
the slope ------------- can easily be evaluated from the table.

If the material is incompressible we may substitute J = 1 in above expressions for the first and second derivatives of W ,
and there is no need for volume neutralization of stretch ratios and other invariant quantities. The derivative of J with
respect to  vanishes and the expressions for I 1 and its derivative in terms of  are greatly simplified. The equation for
I 1 is now used to evaluate the corresponding value of  of the test, which reduces to finding the roots of a polynomial
equation of order 2 or 3 depending on the test involved. The equation derived from the hydrostatic pressure is no longer
required.
The above relations break down in the undeformed state when  = 1 and J = 1 , since their denominators become
zero. We can make a limiting approach to obtain values of the first and second derivatives of W at e = 0 . If we assume
that the stress   e  near e = 0 can be approximated by a cubic polynomial as:
  e  = E1 e + E2 e 2 + E3 e 3 (5-223)
and that the derivative of W with respect to I 1 can be approximated as:
W
-------- = W 1 + W 2  I 1 – 3  (5-224)
I 1
where W 1 and W 2 are the first and second derivative of W with respect to I 1 at e = 0 (although not apparent from
above relation, this approximation is also of third order in e ), then we can express W 1 and W 2 in terms of E 1 , E 2 , and
E3 .
For the incompressible uniaxial case, we find:

E
W 1 = -----1-
6
– 4E 1 + 3E 3 (5-225)
W 2 = ------------------------------
54
For the incompressible equi-biaxial case, we find:
E
W 1 = -----1-
12
(5-226)
– 35E 1 + 6E 3
W 2 = --------------------------------
-
864
For the incompressible pure shear case, we find:
E
W 1 = -----1-
8 (5-227)
– 5E 1 + 2E 3
W 2 = -----------------------------
-
64
The values for E 1 , E 2 , and E 3 can be estimated by making a least squares fit through the first N data points of the table.
In Marc, the first N data points below a threshold of e = 0.025 are used, but this number is limited to 10. A minimum
of 3 is used regardless of the threshold. If the material is compressible, we can make similar polynomial expansions for J
U
and ------- up to order 3 in e and obtain values for W 1 and W 2 which include effects of compressibility. It may be observed
J
that in above relations for W 1 and W 2 , the coefficient E 2 is absent. Closer inspection of the polynomial expansions reveals
that for all three test cases, E 2 is proportional to W 1 , so the coefficients E 1 and E 2 are, in fact, not entirely independent.
5
For the incompressible uniaxial case, we find E 2 = – E 1 ; for the incompressible equi-biaxial case, we find E 2 = – --- E 1 ;
2
3
and for the incompressible pure shear case, we find E 2 = – --- E 1 . When compressibility is included, similar relations can
2
be derived.
In all cases, the initial shear modulus of the material can be estimated as:
G 0 = 2W 1 (5-228)
If no bulk modulus is entered on the MARLOW model definition option, Marc uses a bulk modulus of K = 5000G 0 .
If the test data contains a lot of noise, a smoothing procedure can be activated to reduce this noise. This procedure operates
on the table prior to the analysis and can optionally be invoked during the analysis. Prior to the analysis, a least squares fit
for a cubic polynomial around each table point is carried out using N data points on each side of the point under
CHAPTER 5 305
Material Library
consideration, so 2N + 1 data points are involved ( N is called the smoothing window half size). The number N can be
specified by the user, but cannot be less than 2. The function value of the table is adjusted according to this smoothing
procedure, but only after all data points have been processed. More precisely a three-pass smoothing procedure is carried
out using a smoothing window half size of 3N in the first pass, of 2N in the second pass and of N in the third pass
(provided the table has enough data points). During the analysis the table can be used in a piecewise linear sense, as is also
done when no smoothing is involved, or optionally a least squares procedure using a smoothing window half size of N can
be used again to locally approximate the table by a cubic polynomial. The latter has the advantage of continuous derivatives,
but is computationally more expensive. The smoothing procedure is only carried out for the strain dependency, but not for
the temperature dependency if present.
The equivalent Marlow strain may be placed on the post file using post code 495. The equivalent Marlow strain is the
engineering strain value computed from the deformation state in the model that corresponds to the engineering strain of
the test used to define the Marlow material and is the strain value used to interpolate in the table containing the test data.
Updated Lagrange Formulation for Nonlinear Elasticity

The total Lagrange nonlinear elasticity models in Marc have been augmented with a formulation in an Updated Lagrange
framework. Hence, Rezoning can be used for elastomeric materials based upon the current configuration. This is specially
useful in large deformation analysis since typically excessive element distortion in elastomeric materials can lead to
premature termination of analysis. The new formulation accommodates the generalized Mooney-Rivlin, Ogden, and foam
material models preserving the same format and strain energy functions as the total Lagrange formulation. In addition, the
Arruda-Boyce and the Gent models are only available in the updated Lagrange framework.
The updated Lagrangian rubber elasticity capability can be used in conjunction with both continuous as well as
discontinuous damage models. Thermal, as well as viscoelastic, effects can be modeled with the current formulation. Using
the table driven input, all of the material properties can reference tables, which facilitates the use of temperature dependent
material properties. The singularity ratio of the system is inversely proportional to the order of bulk modulus of the material
due to the condensation procedure.
A consistent linearization has been carried out to obtain the tangent modulus. The singularity for the case of two- or three-
equal stretch ratios is analytically removed by application of L’Hospital’s rule. The current framework with an exact
implementation of the finite strain kinematics along with the split of strain energy to handle compressible and nearly
incompressible response is eminently suitable for implementation of any nonlinear elastic as well as inelastic material
e  p
models. In fact, the finite deformation plasticity model based on the multiplicative decomposition, F = F F F is
implemented in the same framework.
To simulate elastomeric materials, incompressible element(s) are used for plane strain, axisymmetric, and three-dimensional
problems for elasticity in total Lagrangian framework. These elements can be used with each other or in combination with
other elements in the library. For plane stress, beam, plate or shell analysis, conventional elements can be used. For updated
Lagrangian elasticity, both conventional elements (as well as Herrmann elements) can be used for plane strain,
axisymmetric, and three-dimensional problems.
Time-independent Inelastic Behavior

In uniaxial tension tests of most metals (and many other materials), the following phenomena can be observed. If the stress
in the specimen is below the yield stress of the material, the material behaves elastically and the stress in the specimen is
proportional to the strain. If the stress in the specimen is greater than the yield stress, the material no longer exhibits elastic
behavior, and the stress-strain relationship becomes nonlinear. Figure 5-36 shows a typical uniaxial stress-strain curve. Both
the elastic and inelastic regions are indicated.
Stress Inelastic
Region
Yield
Stress
Strain
Elastic Region
Note: Stress and strain are total quantities.
Figure 5-36 Typical Uniaxial Stress-Strain Curve (Uniaxial Test)

Within the elastic region, the stress-strain relationship is unique. As illustrated in Figure 5-37, if the stress in the specimen is
increased (loading) from zero (point 0) to  1 (point 1), and then decreased (unloading) to zero, the strain in the specimen
is also increased from zero to  1 , and then returned to zero. The elastic strain is completely recovered upon the release of
stress in the specimen.
The loading-unloading situation in the inelastic region is different from the elastic behavior. If the specimen is loaded
beyond yield to point 2, where the stress in the specimen is  2 and the total strain is  2 , upon release of the stress in the
e p
specimen the elastic strain,  , is completely recovered. However, the inelastic (plastic) strain,  2 , remains in the
2
specimen. Figure 5-37 illustrates this relationship.
CHAPTER 5 307
Material Library
Total Strain = Strain and Elastic Strain

Stress
3 3
2 2 p
 3
Yield Stress  y
1 1
0 Strain
1 2 3
p e

p
 2e 2 = 2 + 2
2 p e
3 = 3 + 3
p e
 
3 3
Figure 5-37 Schematic of Simple Loading - Unloading (Uniaxial Test)
p
Similarly, if the specimen is loaded to point 3 and then unloaded to zero stress state, the plastic strain  3 remains in the
p p
specimen. It is obvious that  is not equal to  . We can conclude that in the inelastic region:
2 3
 Plastic strain permanently remains in the specimen upon removal of stress.

 The amount of plastic strain remaining in the specimen is dependent upon the stress level at which the unloading
starts (path-dependent behavior).
The uniaxial stress-strain curve is usually plotted for total quantities (total stress versus total strain). The total stress-strain
curve shown in Figure 5-36 can be replotted as a total stress versus plastic strain curve, as shown in Figure 5-38. The slope of
the total stress versus plastic strain curve is defined as the workhardening slope (H) of the material. The workhardening
slope is a function of plastic strain.
Total Stress
 
Plastic Strain
p
H = tan (Workhardening Slope)
= d/dp
Figure 5-38 Definition of Workhardening Slope (Uniaxial Test)
The stress-strain curve shown in Figure 5-36 is directly plotted from experimental data. It can be simplified for the purpose
of numerical modeling. A few simplifications are shown in Figure 5-39 and are listed below:
1. Bilinear representation – constant workhardening slope
2. Elastic perfectly-plastic material – no workhardening
3. Perfectly-plastic material – no workhardening and no elastic response
4. Piecewise linear representation – multiple constant workhardening slopes
5. Strain-softening material – negative workhardening slope
In addition to elastic material constants (Young’s modulus and Poisson’s ratio), it is essential to include yield stress and
workhardening slopes when dealing with inelastic (plastic) material behavior. These quantities can vary with parameters
such as temperature and strain rate. Since the yield stress is generally measured from uniaxial tests, and the stresses in real
structures are usually multiaxial, the yield condition of a multiaxial stress state must be considered. The conditions of
subsequent yield (workhardening rules) must also be studied.
 
 
(1) Bilinear Representation (2) Elastic-Perfectly Plastic
 
 
(3) Perfectly Plastic (4) Piecewise Linear Representation


(5) Strain Softening
Figure 5-39 Simplified Stress-Strain Curves (Uniaxial Test)
Yield Conditions
The yield stress of a material is a measured stress level that separates the elastic and inelastic behavior of the material. The
magnitude of the yield stress is generally obtained from a uniaxial test. However, the stresses in a structure are usually
multiaxial. A measurement of yielding for the multiaxial state of stress is called the yield condition. Depending on how the
multiaxial state of stress is represented, there can be many forms of yield conditions. For example, the yield condition can
be dependent on all stress components, on shear components only, or on hydrostatic stress. A number of yield conditions
are available in Marc, and are discussed in this section.
von Mises Yield Condition

Although many forms of yield conditions are available, the von Mises criterion is the most widely used. The success of the
von Mises criterion is due to the continuous nature of the function that defines this criterion and its agreement with
CHAPTER 5 309
Material Library
observed behavior for the commonly encountered ductile materials. The von Mises criterion states that yield occurs when
the effective (or equivalent) stress () equals the yield stress (y) as measured in a uniaxial test. Figure 5-40 shows the von
Mises yield surface in two-dimensional and three-dimensional stress space.
For an isotropic material:
 =   1 – 2  2 +  2 – 3  2 +  3 – 1  2  1  2  2 (5-229)
where 1, 2, and 3 are the principal Cauchy stresses.
d
2 3
Yield
Surface
Yield
Surface
Elastic
Region
1
Elastic 1
d
2
d
Region
(a) Two-dimensional Stress Space (b) p-Plane

Figure 5-40 von Mises Yield Surface
 can also be expressed in terms of nonprincipal Cauchy stresses.
 =    x –  y  2 +   y –  z  2 +   z –  x  2 + 6   x2y +  yz
2 2
+  zx  1  2  2 (5-230)
The yield condition can also be expressed in terms of the deviatoric stresses as:
3 d d
 = ---  ij  ij (5-231)
2
d
where  ij is the deviatoric Cauchy stress expressed as:
d 1
 i j =  ij – ---  kk  ij (5-232)
3
For isotropic material, the von Mises yield condition is the default condition in Marc. The initial yield stress  y is defined
in the ISOTROPIC and ORTHOTROPIC options. A user-defined finite strain, isotropic plasticity material model can be
implemented through the UFINITE user subroutine. In this case, the finite strain kinematics is taken care of in Marc. You
have to do the small strain return mapping only. See Marc Volume D: User Subroutines and Special Routines for more details.
Hill’s [1948] Yield Function

The anisotropic yield function (of Hill) and stress potential are assumed as:
2
 =  a 1   y –  z  2 + a 2   z –  x  2 + a 3   x –  y  2 + 3a 4  zx +
12
(5-233)
2 2
+ 3a 5  yz + 3a 6  x y   2
where  is the equivalent tensile yield stress for isotropic behavior.

Ratios of actual to isotropic yield (in the preferred orientation) are defined in the array YRDIR for direct tension yielding,
and in YRSHR for yield in a shear (the ratio of actual shear yield to   3 isotropic shear yield). Then the a 1 above are
defined by:
1 1 1
a 1 = ------------------------------------- + ------------------------------------- – ------------------------------------- (5-234)
YRDIR  2 **2 YRDIR  3 **2 YRDIR  1 **2
1 1 1
a 2 = ------------------------------------- + ------------------------------------- – ------------------------------------- (5-235)
1 1 1
a 3 = ------------------------------------- + ------------------------------------ – ------------------------------------ (5-236)
2
a 4 = -------------------------------------- (5-237)
YRSHR  3 **2
2
a 5 = ------------------------------------- (5-238)
YRSHR  2 **2
2
a 6 = -------------------------------------- (5-239)
YRSHR  1 **2
For anisotropic material, use the ISOTROPIC, ORTHOTROPIC, or ANISOTROPIC model definition options to indicate the
anisotropy. Use the ORTHOTROPIC option or the ANPLAS user subroutine for the specification of anisotropic yield condition
(constants a1 through a6, as defined above), and the ORIENTATION model definition option or the ORIENT user subroutine, if
necessary, to specify preferred orientations.
Hill’s (1948) Yield Criterion has been extensively used in sheet metal forming, especially for steel. The experimental data
can be related to the Marc input for the Hill’s Yield Criterion for the shells or plane stress case as given below.
CHAPTER 5 311
Material Library
N (Thickness Direction) R (Rolling Direction)
T (Transverse Direction)
Figure 5-41 Axes of Anisotropy
The sample of tensile coupon cut from a sheet in the three directions,  = 0 (rolling), 45° and 90° (transverse) is tested
to obtain  =  0 ,  45 , and  90 , respectively. Similarly, the anisotropy parameter defined as:
 width
r = -------------------
- (5-240)
 thickness
is obtained for the 0°, 45°, and 90° directions. The yield stress in the third (thickness) direction can be written as:
r 90  1 + r 0  r 0  1 + r 90 
 N =  0 ---------------------------
- =  90 ---------------------------
- (5-241)
r 0 + r 90 r 0 + r 90
The direct stress coefficients are now:

0
YRDIR  1  = -------- (5-242)
 av
 90
YRDIR  2  = -------- (5-243)
 av
N
YRDIR  3  = -------- (5-244)
 av
where  av is the initial yield stress on the stress-strain curve used. If the stress-strain curve is averaged from all directions,
 0 + 2 45 +  90
 a v is defined as  av = ----------------------------------------- in orthotropic plasticity. Similarly, the shear coefficients can be evaluated at:
4
3
YRSHR  1  = YRDIR  3  --------------------- (5-245)
2r 45 + 1
YRSHR  2  = YRSHR  3  = 1.0 (5-246)

It is noted that the transverse direction through the thickness is usually considered isotropic which is reasonable on physical
grounds. Also, notice that for complete isotropy,  0 =  45 =  90 and r 0 = r 45 = r 90 = 1 , which yields the von
Mises yield criterion.
Note that the use of the Hill Yield Function in Marc with all coefficients set to unity will give results varying from the von
Mises case due to differences in return mapping in both cases. By default, the von Mises case uses the mean normal return
scheme whereas the Hill case uses the multistage radial return scheme. One can refer to Chapter 5: Structural Procedure Library
for more details on these return mapping methods. The Hill results with unit ratio that can be reproduced by using the
von Mises yield criterion on conjunction with the Additive Decomposition with Radial Return option (PLASTICITY,4). In
this case, the von Mises yield criterion also uses the multi-stage radial return mapping procedure instead of the default
mean-normal scheme.
Barlat’s (1991) Yield Function

Barlat et al. [Ref. 6] proposed a general criterion for planar anisotropy that is particularly suitable for aluminum alloy sheets.
This criterion has been shown to be consistent with polycrystal-based yield surfaces which often exhibit small radii of
curvature near uniaxial and balanced biaxial tension stress states. An advantage of this criterion is that its formulation is
relatively simple as compared with the formulation for polycrystalline modeling and, therefore, it can be easily incorporated
into finite element codes for the analysis of metal forming problems.
For three dimensional deformation, the yield function f is defined as (Barlat et al. [Ref. 6])
m m m m
f = S1 – S2 + S2 – S3 + S3 – S1 = 2 (5-247)
where S i = 1 2 3 are principal values of a symmetric matrix S i j defined with respect to the components of the Cauchy
stress as
C 3   xx –  yy  – C 2   zz –  x x 
-----------------------------------------------------------------------------
- C 6 x y C5 z x
3
S = C 1   y y –  zz  – C 3   x x –  y y  (5-248)
C 6  xy ------------------------------------------------------------------------------ C 4 z y
3
C2  z z – x x  – C2  y y – z z 
C 5  zx C 4  zy -----------------------------------------------------------------------------
3
In Equation (5-248), the symmetry axes (x, y, z), which represent the mutually orthogonal axes of anisotropy, are aligned
with the initial rolling, transverse, and normal directions of the sheet. During deformation, the anisotropic yield surface
of each material element rotates so that the symmetry axes are all in different directions during deformation. Therefore, it
is necessary to trace the rotation of the yield surface during deformation in order to calculate the plastic strain increment
properly. The rotation of the anisotropy axes is carried out based on the polar decomposition method.
The material coefficients, C i = 16 in Equation (5-248) represent anisotropic properties. When C i = 16 = 1 , the
material is isotropic and Equation (5-247) reduces to the Tresca yield condition for m = 1 or  , and the von Mises yield
criterion for m = 2 or 4. The exponent “ m ” is mainly associated with the crystal structure of the material. A higher “ m ”
value has the effect of decreasing the radius of curvature of rounded vertices near the uniaxial and balanced biaxial tension
ranges of the yield surface, in agreement with polycrystal models. Values of m = 8 for FCC materials (like aluminum)
CHAPTER 5 313
Material Library
and m = 6 for BCC materials (like steel) are recommended. The yield surface has been proven to be convex for m  1 .
Figure 5-42 shows the yield surfaces obtained from von-Mises, Hill and Barlat yield functions for Aluminum 2008-T4 alloy.
1.5
σyy
Mises
σ
1.0
Hill
(1948)
0.5
Barlat’s 6D
σxx
σ
0.0
-0.5
-1.0
-1.0 -0.5 0.0 0.5 1.0 1.5
Figure 5-42 Comparison of Yield Surfaces Obtained from von Mises, Hill and Barlat Yield Functions
p
Utilizing the normality rule, the associated plastic strain increment  i j is obtained from the yield function f as
f
 ijp =  ---------- (5-249)
 ij
f
where  is a scalar function. The calculation of ---------- in Equation (5-249) is lengthy but straightforward. The stress
 i j
p f
integration to obtain  i j along with the calculation of ------------- are shown in the works by Chung and Shah, [Ref. 7], and
  
Yoon et al,.[Ref. 8].
In order to obtain four unknown independent coefficients  C 1 C 2 C 3 C 6  with the assumption of C 4 = C 5 = 1
(isotropic properties for transverse directions), it is necessary to use four stresses from the experimental data
  0  45  90  b  , where  0  45  90 are the tensile yield stresses at 0o, 45o, and 90o from the rolling direction, and
 b is the balanced biaxial yield stress measured from bulge test. A detailed procedure to calculate the coefficients of Barlat’s
yield function are summarized in the work of Yoon at el., [Ref. 9]. In Marc, Barlat’s coefficients are automatically calculated
from user inputs for  0  45  90  b in Mentat. If biaxial data is not available, generally Marc assumes  b =  0 .
Barlat 2005 Yield Functions

Based on the linear transformations of the deviatoric stress, Barlat et al.[Ref. 1] proposed two new yield criteria to describe
the anisotropic plastic behavior of metal and alloys for a full 3-D stress state. We refer the two models as Barlat’s Yld2004-
18p and Barlat’sYld2004-13p, respectively.
1. Barlat’s Yld2004-18p
The yield function of Barlat’s Yld2004-18p is proposed as:
m m m m m m m
f = S 1 – S 1 + S 1 – S 2 + S 1 – S 3 + S 2 – S 1 + S 2 – S 2 + S 2 – S 3 + S 3 – S 1
m m
+ S 3 – S 2 + S 3 – S 3 = 4 m
where S i and S i  i  = 1 3 are the principal values of s and s , respectively. s and s can be obtained by means of
a linear transformation as:
s = C  s = CT
s = Cs = CT
 is Cauchy stress tensor; s is the deviatoric Cauchy stress obtained by s = T ; C and C are two matrices containing
material properties.
For Barlat’s Yld2004-18p model, C and C are defined as:
0  –c 
– c 12 0 0 0
13

– c 21 0 
– c 23 0 0 0
 –c 
– c 31 0 0 0 0
32
C =
0 0 0 
– c 44 0 0
0 0 0 0 
– c 55 0
0 0 0 0 0 
– c 66
0  –c 
– c 12 0 0 0
13

– c 21 0 
– c 23 0 0 0
 –c 
– c 31 0 0 0 0
32
C =
0 0 0 
– c 44 0 0
0 0 0 0 
– c 55 0
0 0 0 0 0 
– c 66
The Barlat 1991 can become a special case of Barlat Yld2004-18p if the material parameters are defined appropriately.
Please see [Ref. 54] for details for material parameter conversion.
CHAPTER 5 315
Material Library
2. Barlat Yld2004-13p
The yield function of Barlat’s Yld2004-13p is proposed as
m m m m m m m m m
f = S 1 – S 2 + S 2 – S 3 + S 3 – S 1 –  S 1 + S 2 + S 3  + S 1 + S 2 + S 3 = 2 m
For Barlat’s Yld2004-13p model, C and C are defined as:
0 
– 1 – c 13 0 0 0

– c 21 
0 – c 23 0 0 0
–1 –1 0 0 0 0
C =
0 0 0 
– c 44 0 0
0 0 0 0 
– c 55 0
0 0 0 0 0 
– c 66
0  –c 
– c 12 0 0 0
13

– c 21 0 
– c 23 0 0 0
–1 –1 0 0 0 0
C =
0 0 0 
– c 44 0 0
0 0 0 0 
– c 55 0
0 0 0 0 0 
– c 66
This represents a more general version to model in [Ref. 55] (Barlat 2003).
Implementation and use of Barlat 2005 models are similar to those of Barlat 1991 model.
Barlat’s yield function can be accessed from ISOTROPIC, ORTHOTROPIC, or ANISOTROPIC model definition options and can be
used in conjunction with the ORIENTATION option.
Vegter Yield Criterion

The Vegter yield criterion has been developed for the modeling of the anisotropic plastic behavior of rolled metal sheets.
The yield function is interpolated with Bezier curves based on material parameters readily derived from mechanical tests
repeated along a number of directions with respect to the rolling direction.
Vegter Standard
The yield locus is defined using a piecewise second order Bezier interpolation according to the following equation:
2
 l o cu s =  i + 2   h –  i  +    i +  j – 2  h 
where  i and  j are reference points on the yield locus and  h is a hinge point (Figure 5-43). Hence, the yield locus is
divided in segments each of them denoted by reference points (i.e., stress points) derived experimentally. The main
advantage in adopting a second order Bezier interpolation is that the yield locus used in the analysis is smooth and
continuously differentiable.
Figure 5-43 Bezier Curve
For one quarter of the yield locus four characteristic stress states are selected as reference points: the equi-biaxial point, the
plane strain point, the uniaxial point and the pure shear point at yielding. During the experiments, not only the stresses
but also the strain rates at the yield points are recorded as they will be used to calculate the position of the hinge points h
(see Figure 5-43). These are derived as the intersection of the tangents at the two reference points  i and  j . The obtained
quarter of yield locus is shown in Figure 5-44. Accordingly, the Vegter criterion interpolates the experimental values and is
continuously differentiable.
Figure 5-44 Quarter of Yield Locus Determined on Four Reference Points and their Corresponding Reference Points
CHAPTER 5 317
Material Library
For planar orthotropic sheet materials, the yield locus depends on the angle  between the first principal stress and the
rolling direction. For the angular dependency of the reference points f k  and the strain ratios k    , Fourier series in
 are used:
2
f  =
k m = 0  k m cos  2m 
2
k    =  m = 0o m cos  2m 
where  k m (with k  ps, un,s h  ) and are Fourier coefficients (see [Ref. 76] for more details).
Usually, for orthotropic sheets the first three terms in the Fourier series are sufficient to provide accurate results.
Experimental tests (i.e., plane strain, uniaxial and shear tests) performed at 0°, 45°, and 90° with respect to the rolling
direction are typically employed. The results from bi-axial tests are independent of the angle with respect to which the
experiment is performed.
The results from experimental tests allow to derive only a quarter of the yield locus. To derive the complete yield locus, we
need to consider some properties valid for orthotropic material:
1. The yield locus is symmetric with respect to the origin since reverse loading leads yielding at equal absolute stresses;
2. The yield locus is symmetric with respect to the line 1 = 2 since the principal axis 1 and 2 can be exchanged;
3. Stresses on the yield locus are periodic with respect to  , namely f    = f   +  ;
4. The yield locus is symmetric with respect to  = 0 and  = 2 , thus f    = f  –   and
f   2 +  = f   2 – 
Accordingly, we can derive the full yield loci as illustrated in Figure 5-45.
Further, the complete yield locus at a generic angle  is obtained by means of the above Fourier equations of f k and
k    .
Vegter 2017
Compared to the standard formulation, the Vegter 2017 model consider only input parameters from uniaxial tensile tests
(i.e., tensile strength Rm, uniform elongation Ag) and the r-value. Thus, the number of input parameters is drastically
reduced compared to the ones required by the Vegter standard model, but with maintaining the accuracy due to the strong
empirical correlations between the tensile points and the shear, plane strain and biaxial points. All reference points needed
to interpolate the yield locus are derived from empirical relations detailed in see [Ref. 77][Ref. 78].
3D Extension
The extension of the Vegter plane stress criterion to the 3D case is derived using the Hill yield criterion.
3D 2 2 2 2 2 2 2
2   e q  = f   y –  z  + g   z –  x  + h   x –  y  + 2l yz + 2m zx + 2n xy
In case of plane stress, the above equation can be written as

ps g+h 2 f+h 2 2
e q = -------------  x + -----------  y – h x  y + n x y
2 2
(5-250)
Figure 5-45 Yield Locus Measurements for 0°-90° Direction and 45°-135° Direction [2].
Since the Hill yield criterion is independent of the hydrostatic pressure, accounting for a hydrostatic pressure equal to –z ,
the initial equation of the Hill criterion can be rewritten in the following form:
3D ps 2 2 2
e q =   eq   x –  z,  y –  z,  x y   + m x z + l  y z , (5-251)
ps
where  eq is replaced by the Vegter criterion. If the yield stresses in the out of plane direction z,  x z and y z , are known,
the coefficient m and l are calculated through
o 2 o 2
  un  0     un  0  
m = 4  --------------------- –  ------------------------ (5-252)
  xz    un  9 0 o  
and
o 2
  un  0  
l = 4  --------------------- – 1 ,
  yz 
where  un  0 o  and  un  9 0 o  are the uniaxial yield stresses in the rolling direction and in the transverse direction. If x z
and  y z are not known, their value can be calculated as
1 o
 x z = ---   un  0  +  bi 
2
and
CHAPTER 5 319
Material Library
1 o
 y z = ---   un  90  +  bi 
2
where  bi is the equi-biaxial yield stress.

Alternatively, the value of m and l can be set equal to 3 by assuming that the 3D extension of the Vegter yield criterion is
derived using the von Mises yield criterion.
Mohr-Coulomb Material (Hydrostatic Stress Dependence)

Marc includes options for elastic-plastic behavior based on a yield surface that exhibits hydrostatic stress dependence. Such
behavior is observed in a wide class of soil and rock-like materials. These materials are generally classified as Mohr-Coulomb
materials (generalized von Mises materials). Ice is also thought to be a Mohr-Coulomb material. The generalized Mohr-
Coulomb model developed by Drucker and Prager is implemented in Marc. There are two types of Mohr-Coulomb
materials: linear and parabolic. Each is discussed on the following pages.
Linear Mohr-Coulomb Material

The deviatoric yield function, as shown in Figure 5-46, is assumed to be a linear function of the hydrostatic stress.

f = J 1 + J 21  2 – ------- = 0 (5-253)
3
where, (5-254)
J 1 =  ii (5-255)
1 d d
J 2 = ---  ij  ij (5-256)
2
Figure 5-46 Yield Envelope of Plane Strain (Linear Mohr-Coulomb Material)
Analysis of linear Mohr-Coulomb material based on the constitutive description above is available in Marc through the
ISOTROPIC model definition option. Through the ISOTROPIC option, the values of  and  are entered. Note that,
throughout the program, the convention that the tensile direct stress is positive is maintained, contrary to its use in many
soil mechanics texts.
The constants  and  can be related to c and  by:

 3
c = ---------------------------------------------- ; ---------------------------------- = sin  (5-257)
12  1 – 3 2  1  2
 3  1 – 12 2  
where c is the cohesion and  is the angle of friction.
Parabolic Mohr-Coulomb Material

The hydrostatic dependence is generalized to give a yield envelope which is parabolic in the case of plane strain (see
Figure 5-47).
Yield Envelope
R
c 

x + y
------------------
-
2
f =  3J 2 + 3J 1  1  2 –  = 0 (5-258)
The parabolic yield surface is obtained in Marc through the ISOTROPIC model definition option. Enter the values  and 
through the ISOTROPIC model definition option.
2 
2 2 
 = 3  c – ------  = --------------------------------------------- (5-259)
 3  3  3c 2 –  2   1  2
where c is the cohesion.
Figure 5-47 Resultant Yield Condition of Plane Strain (Parabolic Mohr-Coulomb Material)
Buyukozturk Criterion (Hydrostatic Stress Dependence)

This yield criterion [Ref. 4], which originally has been proposed as a failure criterion, has the general form:
2 2
f =  3J 1 + J 1 + 3J 2 –  (5-260)
Through the ISOTROPIC model definition option, the user has to define  and the factor  , where  has a fixed value of
0.2 . The Buyukozturk criterion reduces to the parabolic Mohr-Coulomb criterion if  = 0 .
CHAPTER 5 321
Material Library
R c

x + y c2
------------------- -----
2 
Powder Material
Some materials, during certain stages of manufacturing are granular in nature. In particular, powder metals are often used
in certain forging operations and during hot isostatic pressing (HIP). These material properties are functions of both the
temperature and the densification. It should be noted that the soil model discussed in this chapter also exhibits some of
these characteristics.
There are two models for representing granular materials such as powders. The Shima-Oyane model and the exponential
cap model may be used to represent these types of materials. Both of these models are dependent upon the hydrostatic stress
and the relative density. It is felt that the newer exponential Cap model better represents the physical phenomena as shown
in the figure below.
Figure 5-48 Comparison with Experiments [Ref. 1]

In the Shima-Oyane model incorporated into Marc, a unified viscoplastic approach is used. The yield function is:
1 3 p2 1  2
F = ---  ---  d  d + ------ – y (5-261)
 2   2
where  y is the uniaxial yield stress,  d is the deviatoric stress tensor, and p is the hydrostatic pressure.  ,  are material
parameters.
 y can be a function of temperature and relative density,  ,  are functions only of relative density.
Typically, we allow:
q q4
 =  q1 + q2  3 
b b4
(5-262)
 =  b1 + b2  3 
where  is the relative density.
As the powder becomes more dense,  approaches 1 and the classical von Mises model is recovered.
It should be noted that the elastic properties are also functions of relative density. In particular, as the material becomes
fully dense, the Poisson’s ratio approaches 0.5.
The equivalent inelastic strain is defined as:
CHAPTER 5 323
Material Library
· 1 F
 eq = ---  ------
   y
where  is a viscosity defined as a function of temperature.

As most processes involving powder materials are both pressure and thermally driven, it can be necessary to perform a
coupled analysis. Marc also allows you to specify density effects for the thermal properties, conductivity and specific heat.
The basic input data is entered through the POWDER option. In addition to the TEMPERATURE EFFECTS option, there is a
DENSITY EFFECTS option. The initial relative density is entered through the RELATIVE DENSITY option.
--- 2 y
2 -
p-- 22 
-- 
 d : d
------------------
2 2 2
---   y
3
Figure 5-49 Yield Function of Shima Model
Exponential Cap Powder Model

The exponential cap powder model implemented into Marc is based upon the Sandia Geomodel [Ref. 2] and [Ref. 3], a
generalized plasticity model, wherein the combined effect of the two underlying mechanisms, porosity and microcracks, is
represented through a unified yield function given as:

f   , ,  = J 2  2 –  F s  I 1  – N  2 F c  I 1 , 
where
 is the stress tensor

 is the deviatoric backstress tensor
 is a internal state variable determining the position of the cap

J2 is the second invariant of the shifted stress tensor  = dev    – 
I1 is the trace of the stress tensor 
N is the maximum allowable translation of the yield surface when kinematic hardening is enabled
 is the Gudehus Lode angle function given as:
1 J3 3 3  2 1  J 3 3 3  2
  J 2 J 3  = --- 1 – -----  ----- + ----  1 + -----  ----- 
2 2 J2    2  J 2 
where J 2 ,J 3 are the second and third invariants of the stress tensor respectively and  is the strength ratio.
Shear Failure
Surface Fs
Fs Fc
J 2 = ---------------
-

Strain hardening
cap Fc
Direction of
plastic strain
on cap
I1 X X0 X0  
Cap Nose
Point Cap Branch
Point
Shear Failure Surface

The shear failure surface F s  I 1  is represented as:
A2 I1
Fs  I1  = A1 – A3 e – A4 I1
where, A 1 ,A 2 ,A 3 ,A 4 are all non negative material parameters.
fs
A1 – A3
I1
CHAPTER 5 325
Material Library
Cap Function
The cap function F c  I 1 ,  can be written as:
F c  I 1 ,  = 1 if I 1  
I1 –  2 otherwise
= 1 –  --------------
 X – 
where, X    =  – RF s    and R is the cap eccentricity parameter.
fc
1.0
I1 X 
The volumetric plastic strain  vp is determined from the expression:
D1  X – X0  – D2  X – X0  2
 vp = W e –1
where W , D 1 , D 2 , and X 0 are all material constants.
Evolution equations
The flow rule,
· f ·
 p = ------- 

·
where  is the consistency parameter.
Evolution of  is governed by:

· ·
 = h 
f I1 – 
3 ------- --------------
 R
where h  = min ----------------- , ----------------
-
 vp X  vp X
--------- ------- --------- -------
X  X 
When kinematic hardening is enabled, the backstress  evolves as:

· ·
 = H
 J  f
where, H  = H  1 – ---------2- dev  ------
 N   
Kinematic Hardening:
Kinematic hardening can be introduced by use of a shifted stress tensor  = dev    – 
The backstress tensor  is initialized to zero. Upon onset of yielding, the backstress  evolves in proportion to the
deviatoric part of the plastic strain rate.
In order to turn on kinematic hardening, the user needs to specify N , the maximum allowable translation of the yield
surface, and a material parameter H , which determines the speed at which the yield surface approaches the limit surface.
To suppress kinematic hardening, set both H and N to zero.
Non-associative flow rule

The exponential cap model supports non-associatively by allowing the user to specify a flow function  which has the same
PF
form as the yield function f . The flow potential can be made to differ from the yield function by assigning values to A 2 ,
PF PF PF
A4 , R , and  that differ from their counterpart parameters ( A 2 , A 4 , R , and  ) used to define the yield surface.
PF PF PF PF
To set an associative flow rule, set all potential flow coefficients ( A 2 , A 4 , R , and  ) to zero.
Simplification to standard models

The exponential cap powder model can be simplified to standard models such as von Mises, Linear Mohr-Coulomb etc.
by setting some of the parameters to zero. The following simplifications are documented here:
1. von Mises

The von Mises yield surface f  J 2  = J 2 – ------y- can be obtained by the following simplifications:
3
• Set A 2 and A 4 to 0
y
• Set A 1 – A 3 = -------
3
CHAPTER 5 327
Material Library
• Set  0 = – 1e30 (cap at infinity)
• Set R , W , D 1 ,and D 2 to 0
• Set  = 1
2. Linear Mohr-Coulomb

The Linear Mohr-Coulomb yield surface f  J 2  = I 1 + J 2 – ------y- can be obtained by the following
3
simplifications:
• Set A 2 = 0
• Set A 4 = 
y
• Set A 1 – A 3 = -------
3
• Set  0 = – 1e30 (cap at infinity)
• Set R , W , D 1 ,and D 2 to 0
• Set  = 1
User Input Parameters

Symbol Description SI Units
E Young’s modulus, supports table input Pa
 Poisson’s ratio, supports table input dimensionless
 Density Kg/m3
C th Coefficient of thermal expansion, to be used for thermal analysis 1/K
A1 First constant term in the function defining the shear failure surface Pa
A1  0
A2 Second constant term in the function defining the shear failure surface 1/Pa
A2  0
Set A 2 and A 4 to 0 for simplification to von Mises yield surface

A3 Third constant term in the function defining the shear failure surface A 3  0 Pa

Set A 1 – A 3 = ------y- to mimic the von Mises yield surface
3
f  J2  = J2 – y

A4 Fourth constant term in the function defining the shear failure surface dimensionless
A4  0
PF
A2 Potential function parameter for setting non-associative flow rule. 1/Pa
PF
A2 = A 2 for associative flow. Set to zero in input file for associative flow.
PF
A4 Potential function parameter for setting non-associative flow rule. Set dimensionless
PF
A4  A 2 for non-associative flow. Set to zero in input file for associative flow.
H Kinematic hardening parameter which governs how quickly the yield surface evolves dimensionless
towards the limit surface. Set both H and N to zero to suppress kinematic hardening.
N Kinematic hardening parameter which sets the maximum allowable translation of the Pa
yield surface. Set both H and N to zero to suppress kinematic hardening.
N0
R Cap eccentricity parameter. It is the ratio of the width divided by the height of the dimensionless
elliptical cap function. This parameter affects the stress level at which dilatation will occur
in triaxial compression.
R0
If not entered, R = 1 .
0 This is the initial value of I 1 at which the meridional yield profile branches away from the Pa
shear failure surface and morphs into the cap function.  0 must be negative. Set
 0 = – 1e30 (cap at infinity) to simulate a von Mises response.
W Limit value of the absolute plastic volumetric strain. This parameter is approximately dimensionless
equal to the initial porosity of the material and can be obtained from the hydrostatic crush
data.
D1 One third of the slope of the porosity versus pressure crush curve at the elastic limit. 1/Pa
D2 Extra fitting parameter for the hydrostatic crush curve data. Used only when the crush 1/Pa2
curve has an inflection point. Else set to zero.
 The triaxial extension to compression strength ratio (TXE/TXC) Convexity of the yield dimensionless
7 9
surface requires ---    --- .
9 7
If not entered,  = 1 .
PF
R Flow potential analog of the yield surface parameter R . Used only for non-associative dimensionless
flow.
PF
 Flow potential analog of the yield surface parameter S . Used only for non-associative flow. dimensionless
CHAPTER 5 329
Material Library
Plotable output parameters

Cap Softening Flag Set this flag to 0 to disallow contraction of the cap. Set to 1 to allow the powder Flag
cap to soften.
 The value of I 1 at which the meridional yield profile branches away from the Pa
shear failure surface and morphs into the cap function. This variable is referred
to as “Cap Branch Point” in Mentat.
X The value of I 1 at which the cap surface intersects the x-axis. This variable is Pa
referred to as “Cap Nose Point” in Mentat.
 v ol Current relative density based on total volumetric strain. dimensionless
 vpol Stress free relative density based on the plastic volumetric strain. dimensionless
Material Parameters
The fully generalized exponential cap powder model requires the user to specify 16 different material parameters. However,
simplifications of the exponential model can be run using as few as two material parameters. Material parameters for the
exponential cap powder model can be inferred by curve fitting the data obtained through a series of experiments. The
experiments needed and the method of curve fitting is detailed below.
Crush Curve Parameters

Step 1: Obtain hydrostatic pressure vs. total volumetric strain data by conducting experiments run to the point of total
pore collapse.
 vtotal
W
Use the elastic unloading curve to determine a shift distance that must be applied at any pressure to remove the elastic part
of the strain. This will give you a plot of pressure vs. plastic volumetric strain.
X
p
3
 vp
W
Use a curve fitting program to fit the above data to the following curve
p D1  X – X0  – D2  X – X0  2
v = W e –1
where
X 0 =  0 – Rf s   0  and R are the cap eccentricity parameter

W is approximately equal to the initial porosity of the material.
Shear Failure Parameters

Step 1: Conduct load-to-failure triaxial compression tests on the material and record all peak stress states. Each individual
experiment has precisely one stress state at which the second stress variant J 2 achieves a peak value.
Characterizing the shear failure surface requires numerous such experiments.
For every available load-to-failure experiment, find the stress state at which J 2 is larger than any of the other stress states
p e ak pe a k
in that experiment. Construct a table of data ( I 1 , J 2 ) where I 1 is the value of I 1 at the stress state at which J 2
is at its peak value.
Fit the data to the exponential shear failure function using a curve fitting program
A2 I1
fs  I1  = A1 – A3 e – A4 I1
More details on obtaining material parameters for the exponential cap powder model can be found in the following
reference:
A.F. Fossum and R.M. Brannon, “The Sandia GEOMODEL: Theory and User's Guide,” Technical Report SAND2004-
3226, Sandia Nat'l Laboratories, Albuquerque, N.M., 2004.
CHAPTER 5 331
Material Library
Obtaining Crush Curve and Shear Failure Parameters by Curve Fitting in Marc
Marc now supports curve fitting for the crush curve and shear failure parameters through the existing Mentat interface.
The user can input the experimental data via tables and Marc carries out the curve fitting and obtains both the shear failure
and the crush curve parameters. The following steps are used for curve fitting in Marc.
Crush Curve Parameters: (  0 ,W ,D 1 ,D 2 )

1. Set the value of W to – 1 in Mentat to indicate that you will be using curve fitting for the crush curve parameters.
The rest of the parameters can be set to 0.
p
2. Assign a table ID for W and include a table of pressure vs. the plastic volumetric strain (X vs.  v ). Data should be
sorted by X. Sample set of data and the corresponding plot are shown as follows:
Data p
Point X = 3*p v
1 -6.00E+09 -0.079988058
2 -5.00E+09 -0.079948061
3 -4.00E+09 -0.079774105
4 -3.00E+09 -0.079017528
5 -2.00E+09 -0.075726998
6 -1.50E+09 -0.071088731
7 -1.00E+09 -0.061415709
8 -8.00E+08 -0.055063898
9 -7.00E+08 -0.051115167
10 -6.00E+08 -0.04654114
11 -5.00E+08 -0.041242797
12 -4.00E+08 -0.03510544
13 -3.00E+08, -0.027996208
14 -2.00E+08 -0.019761202
15 -7.00E+07 -0.007076085
16 -7.00E+06 0
3. Marc internally uses nonlinear regression analysis to fit the data and to obtain the crush curve parameters
(  0 ,W ,D 1 ,D 2 ).
Shear Failure Parameters: ( A 1 ,A 2 ,A 3 ,A 4 )

1. Set the value of A 1 to – 1 in Mentat to indicate that you will be using curve fitting for the shear failure parameters.
The rest of the shear failure parameters can be set to 0.
2. Assign a table ID for A 1 and include a table of ( I 1 vs. J 2 ). Data should be sorted by I 1 . Sample set of data and
the corresponding plot are shown as follows:
Data
Point I1 J2
1 -600000000 631520775
2 -440000000 574163050
3 -400000000 553409562
4 -320000000 500886644
5 -280000000 467829871
6 -220000000 407501319
7 -190000000 371605878
8 -160000000 331244305
9 -130000000 285856240
10 -100000000 234811012
11 -80000000 197291022
12 -60000000 156711243
13 -40000000 112820812
14 -20000000 65348293
15 0 14000000
16 10000000 -13225871
3. Marc internally uses nonlinear regression analysis to fit the data and to obtain the shear failure parameters
( A 1 ,A 2 ,A 3 ,A 4 )
Limitations:
The following limitations should be noted.
1. The implementation treats the material as initially isotropic. Kinematic hardening is the only way to simulate
deformation induced anisotropy.
2. The elasticity model is hypoelastic rather than hyperelastic.
3. While the exponential shear failure curve is quite general, only an elliptic cap function is available for modeling
shear effects on pore collapse and other plastic volume reducing mechanisms.
4. The model is not extensively tested for tensile applications.
5. The triaxial extension/compression ratio is presumed constant. Neither does it change with pressure, nor does it
evolve with time.
CHAPTER 5 333
Material Library
6. Temperature dependent plasticity is not incorporated in the material model.

7. While the elastic parameters (Young’s modulus, Poisson’s ratio) can be made to vary with temperature and relative
density using tables, the same is not supported for the other material parameters.
8. Curve fitting programs used to fit experimental data to the required material parameters are not provided.
Tips and Tricks:

1. Getting the complete set of 16 material parameters for the exponential cap powder model is a daunting task which
requires extensive experimentation and curve fitting. But do remember that in most cases one can make do with
much lesser material parameters by running simplifications of the exponential cap model. Also the Sandia report,
referenced in the material parameters section above, has a handy set of material data for several geomaterials.
2. Make sure that the material properties satisfy the following constraints:
A1 – A3 –  N  0  (Unloaded virgin material must be below yield)

A2 A3 + A4  0 (Non-negative slope at low pressures)
A2  0 (Positive Lode radius)
A3  0 (Convexity condition)
A4  0 (Non-negative slope at low pressures)
0  0 (Initial cap position should be negative)
R0 (Positive cap radius)
7997
On violation of any of these conditions, an error message is displayed in the output file. Check output file to see
which condition has been violated.
3. If the cap radius R and the strength ratio  are not specified, then they are each set to a default value of 1.
PF PF
4. The exponential cap model assumes a associative flow rule as default when A 2 and A 4 are not specified (or
PF PF
specified to be 0) by internally setting A 2 = A 2 and A 4 = A 4 . To set a non-associative flow rule set values
PF PF
of A 2 and A 4 different than values of A 2 and A 4 .
5. Many times, powders are subjected to hydrostatic loading or close to hydrostatic loading. In such cases, J 2 is close
to zero and the yield stress is return mapped back to the nose of the cap. When running problems where such
hydrostatic loading is likely, better convergence can be achieved by using only the contribution of the deviatoric
stress to the initial stress stiffness matrix. This option can be found under LOADCASE  STATIC > SOLUTION CONTROL.
6. The exponential cap model has not been extensively tested under tension as most powder materials are subjected
to compression in HIP/CIP processes. The shear failure function and its derivatives become very large as the
exponent A 2  I 1 increases in the positive regime. We have internally set the limiting value of A 2  I 1 to be 10 and
issue the warning of large hydrostatic tension in element whenever this condition is violated. Small amounts of
tension are acceptable as A 2 is typically small. Only when one gets into large amounts of tension and I 1 becomes
really large the above condition will be violated which may make the return mapping numerically unstable. In such
cases a warning is printed in the output files and the simulation continues for as long as possible.
Oak Ridge National Laboratory Options

In Marc, the ORNL options are based on the definitions of ORNL-TM-3602 [Ref. 10] for stainless steels and ORNL
recommendations [Ref. 16] for 2 1/4 Cr-1 Mo steel.
The initial yield stress should be used for the initial inelastic loading calculations for both the stainless steels and 2 1/4 Cr-
1 Mo steel. The 10th-cycle yield stress should be used for the hardened material. The 100th-cycle yield stress must be used
in the following circumstances:
1. To accommodate cyclic softening of 2 1/4 Cr-1 Mo steel after many load cycles
2. After a long period of high temperature exposure
3. After the occurrence of creep strain
To enter initial and 10th-cycle yield stresses, use the ISOTROPIC or ORTHOTROPIC model definition option.
Effects of Temperature and Strain Rates on Yield Stress - not using table driven input
Marc allows you to input a temperature-dependent yield stress. To enter the yield stress at a reference temperature, use the
ISOTROPIC or ORTHOTROPIC model definition option. To enter variations of yield stress with temperatures, use the
TEMPERATURE EFFECTS and ORTHO TEMP model definition options. Repeat the TEMPERATURE EFFECTS and ORTHO TEMP model
definition options for each material, as necessary. The effect of temperatures on yielding is discussed further in Constitutive
Relations.
Marc allows you to enter a strain rate dependent yield stress, for use in dynamic and flow (for example, extrusion) problems.
To use the strain rate dependent yield stress in static analysis, enter a fictitious time using the TIME STEP option. The zero-
strain-rate yield stress is given on the ISOTROPIC or ORTHOTROPIC options. Repeat the STRAIN RATE model definition option for
each different material where strain rate data are necessary. Refer to Constitutive Relations for more information on the strain-
rate effect on yielding.
Work or Strain Hardening

In a uniaxial test, the workhardening slope is defined as the slope of the stress-plastic strain curve. The workhardening slope
relates the incremental stress to incremental plastic strain in the inelastic region and dictates the conditions of subsequent
yielding. The yield stress and the workhardening data must be compatible with the procedure used in the analysis. For small
strain analyses, the engineering stress and engineering strain are appropriate. If only the LARGE DISP parameter is used, the
yield stress should be entered as the second Piola-Kirchhoff stress, and the workhard data be given with respect to plastic
Green-Lagrange strains. If the LARGE STRAIN parameter is used, the yield stress must be defined as a true or Cauchy stress,
and the workhardening data with respect to logarithmic plastic strains.
CHAPTER 5 335
Material Library
Work hardening can be specified in Marc in the different ways listed below.
1. Table driven input
2. WORK HARD option
3. Chaboche
4. Power Law
5. Rate Power Law
6. Johnson-Cook
7. Kumar
8. User subroutine (WKSLP)
The first two methods enables the user to define the yield stress as a function of independent variables (like equivalent
plastic strain, temperature, and strain rate) through user-specified data. Apart from the above methods, Marc provides
commonly used hardening laws available in the literature like Chaboche, Power law, Rate Power Law, Johnson-Cook, and
Kumar models. The user can also customize the hardening law through the WKSLP user subroutine. Brief explanations of
all the above mentioned classes of models are given below.
Flow Stress Definition - using table driven input

When using the table driven input format, the yield (flow) stress dependence on the temperature, equivalent plastic strain,
and the strain rate is defined simultaneously through a single table. As this table may be either a piecewise linear description
or a mathematical equation based upon these independent variables, it allows a very general definition. The value of the
yield stress is obtained by evaluating the table, including the reference value given on the ISOTROPIC or ORTHOTROPIC option.
The strain hardening slope is obtained when necessary by numerically differentiating the values given.
Work Hardening Definition - not using table input

When using the nontable input format, the yield (flow) stress dependence on the equivalent plastic strain, temperature,
and strain rate are defined through the WORK HARD, TEMPERATURE EFFECTS, and STRAIN RATE options, respectively. The latter
two options were discussed in the previous section. The WORK HARD option is defined in more detail in this section.
There are two methods to enter this information, using the WORK HARD option. In the first method, you must enter
workhardening slopes for uniaxial stress data as a change in stress per unit of plastic strain (see Figure 5-50) and the plastic
strain at which these slopes become effective (breakpoint).
Stress
Slope Breakpoint
3
2  1
---------- 0.0
p
1 1
 2 p
---------- 1
p
 2

 3 p p
E E E E ---------- 1 +  2
p
 3
p p p
Strain
1  2  3 and so on.
Figure 5-50 Workhardening Slopes
Note: The slopes of the workhardening curves should be based on a plot of the stress versus plastic strain for a tensile
test. The elastic strain components of the stress-strain curve should not be included. The first breakpoint of the
workhardening curve should be 0.0.
In the second method, you enter a of yield stress, plastic strain points. This option is flagged by adding the word DATA to
the work hard statement.
Note: The data points should be based on a plot of the stress versus plastic strain for a tensile test. The elastic strain
components should not be included. The first plastic strain should equal 0.0 and the first stress should agree
with that given as the yield stress in the ISOTROPIC or ORTHOTROPIC options.
Chaboche Model
The Chaboche model is described in detail in the Time-dependent Inelastic Behavior section.
Power Law Model

m ·n
 y = A   0 +  p  + B p
·
where  y is the yield stress,  0 is the strain corresponding to initial yield stress,  p is equivalent plastic strain,  p is rate
of equivalent plastic strain rate and A , B , m , and n are material parameters.
Rate Power Law Model

·
m n
 y = A p  p + B
where the parameters are same quantities as that of Power law.
Johnson-Cook Model
·
n    p   T – T ro o m m
y =  A + B p   1+ C ln  ----
-   1 –  ----------------------------------
- 
·
   0    T me l t – T ro o m 
· ·
where  y is the yield stress,  p is the equivalent plastic strain,  p is the current equivalent plastic strain rate,  0 is strain
rate of material characterization and A , B , C , m and n are material parameters. T , T ro o m , T m el t are, respectively, the
current, room, and melting temperatures of the material in absolute scale.
The following conditions should be noted for the Johnson-Cook model.
 T should be between T r oo m and T me l t . If T  T ro o m , T is set to T ro o m . If T  T m el t , T = T m el t – 0.01 .
· · ·
  0  0 and    0 . If either condition is violated, the middle term in the above expression is set to 1.0.
CHAPTER 5 337
Material Library
Kumar Model
· 1n
  p
y = B0 * sin h – 1  ----- e Q   nR T 
 A
·
where  y is the yield stress,  p is equivalent plastic strain rate, Q is the activation energy, T is temperature, and A , B 0 ,
and n are material parameters.
Yada Law
Used for grain size predictions, is a generic variation of Kumar model. If the constants A and n in the Kumar model are
related to strain and the material constants B 1 , B 2 , B 3 , B 4 , B 5 , and B 6 as given below, we get Yada Law.
Formula Representations of Flow Stress

Some commonly used formulas may be used to define the flow stress. These include terms referencing the plastic strain,
strain rate, and temperature. The parameters can be functions of the conventional independent variables via the TABLE
option.
P5
------
GMT-0 P2 T P  ·P
y = A  P1  e  p 3  e p  p 4
P5
------ + P 7  p
GMT-1 P2 T P9 P  P6 T ·P + P T
y = A  P1  e T  p 3  e p   1 + p   p 4 8
P5 T + P6
-------------------------
GMT-2 P2 T P T+P p ·P + P T
y = A  P1  e  p 3 4  e  p 8 7
P
Hockett-Sherby – P 3   4
y = P2 –  P2 – P1   e  
WKSLP User Subroutine

The user-subroutine option can be employed to specific requirement of hardening law. The yield stress can be a function
of temperature, strain, strain-rate, etc. The user is required to return the work hardening slope and, optionally, the yield
stress at each integration point. If the latter is not defined, then the current yield stress is calculated internally. WKSLP or
USER_FSTRESS can be called multiple times within each cycle to capture the dependence of the slope and the yield stress on
temperature, strain, strain-rate, etc.
Work Hardening Rules

A number of workhardening rules (isotropic, kinematic, and combined) are available in Marc. A description of these
workhardening rules is given below.
Isotropic Hardening
The isotropic workhardening rule assumes that the center of the yield surface remains stationary in the stress space, but
that the size (radius) of the yield surface expands, due to workhardening. The change of the von Mises yield surface is
plotted in Figure 5-51b.
A review of the load path of a uniaxial test that involves both the loading and unloading of a specimen will assist in
describing the isotropic workhardening rule. The specimen is first loaded from stress free (point 0) to initial yield at point
1, as shown in Figure 5-51a. It is then continuously loaded to point 2. Then, unloading from 2 to 3 following the elastic
slope E (Young’s modulus) and then elastic reloading from 3 to 2 takes place. Finally, the specimen is plastically loaded
again from 2 to 4 and elastically unloaded from 4 to 5. Reverse plastic loading occurs between 5 and 6.
It is obvious that the stress at 1 is equal to the initial yield stress  y and stresses at points 2 and 4 are larger than  y , due
to workhardening. During unloading, the stress state can remain elastic (for example, point 3), or it can reach a subsequent
(reversed) yield point (for example, point 5). The isotropic workhardening rule states that the reverse yield occurs at current
stress level in the reversed direction. Let  4 be the stress level at point 4. Then, the reverse yield can only take place at a
stress level of –  4 (point 5).
The isotropic workhardening model (with a work slope of 0) is the default option in Marc.
For many materials, the isotropic workhardening model is inaccurate if unloading occurs (as in cyclic loading problems).
For these problems, the kinematic hardening model or the combined hardening model represents the material better.
 
3
4
2
1
y E 6
+4 5
E
E
3
0 0
3
4 21
4
5 1 
2
6
(a) Loading Path (b) von Mises Yield Surface
Figure 5-51 Schematic of Isotropic Hardening Rule (Uniaxial Test)
CHAPTER 5 339
Material Library
Kinematic Hardening
Under the kinematic hardening rule, the von Mises yield surface does not change in size or shape, but the center of the
yield surface can move in stress space. Figure 5-52b illustrates this condition. Prager’s law is used to define the translation of
the yield surface in the stress space.
The loading path of a uniaxial test is shown in Figure 5-52a. The specimen is loaded in the following order: from stress free
(point 0) to initial yield (point 1), 2 (loading), 3 (unloading), 2 (reloading), 4 (loading), 5 and 6 (unloading). As in isotropic
hardening, stress at 1 is equal to the initial yield stress  y , and stresses at 2 and 4 are higher than  y , due to
workhardening. Point 3 is elastic, and reverse yield takes place at point 5. Under the kinematic hardening rule, the reverse
yield occurs at the level of  5 =   4 – 2 y  , rather than at the stress level of –  4 . Similarly, if the specimen is loaded
to a higher stress level  7 (point 7), and then unloaded to the subsequent yield point 8, the stress at point 8 is
 8 =   7 – 2 y  . If the specimen is unloaded from a (tensile) stress state (such as point 4 and 7), the reverse yield can
occur at a stress state in either the reverse (point 5) or the same (point 8) direction.
To invoke the kinematic hardening in Marc, use the model definition options ISOTROPIC, ORTHOTROPIC, or ANISOTROPIC. To
input workhardening slope data, use the TABLE or WORK HARD option or the WKSLP user subroutine.
For many materials, the kinematic hardening model gives a better representation of loading/unloading behavior than the
isotropic hardening model. For cyclic loading, however, the kinematic hardening model can represent neither cyclic
hardening nor cyclic softening.
 3
7 7
4 4 2y
2
y 1
3 2y 8
8

0 5 5
6
(a) Loading Path 1

2
(b) von Mises Yield Surface
Figure 5-52 Schematic of Kinematic Hardening Rule (Uniaxial Test)
Combined Hardening
Figure 5-53 shows a material with highly nonlinear hardening. Here, the initial hardening is assumed to be almost entirely
isotropic, but after some plastic straining, the elastic range attains an essentially constant value (that is, pure kinematic
hardening). The basic assumption of the combined hardening model is that such behavior is reasonably approximated by
a classical constant kinematic hardening constraint, with the superposition of initial isotropic hardening.
The isotropic hardening rate eventually decays to zero as a function of the equivalent plastic strain measured by:
·p 2 p p 12
p =  dt =   --3- · i j · i j dt (5-263)
This implies a constant shift of the center of the elastic domain, with a growth of elastic domain around this center until
pure kinematic hardening is attained. In this model, there is a variable proportion between the isotropic and kinematic
p
contributions that depends on the extent of plastic deformation (as measured by  ).
Use the ISOTROPIC, ORTHOTROPIC, or ANISOTROPIC model definition option to activate the combined workhardening option
in Marc. Use the TABLE or WORK HARD option or the WKSLP user subroutine to input workhardening slope data.
There are two formulations in the Marc program. The older one occurs whenever PLASTICITY,4 is not included as a
parameter. The workhardening data at small strains governs the isotropic behavior, and the data at large strains
(  p  1000 ) governs the kinematic hardening behavior. If the last workhardening slope is zero, the behavior is the same
as the isotropic hardening model.

Initial Combined Fully Hardened
Elastic Hardening Pure Kinematic
Range Range Range
Stress
Initial
Yield
One-half Current
Elastic Range
3 d
Kinematic Slope, --- --------p
2 d

Strain
Figure 5-53 Basic Uniaxial Tension Behavior of the Combined Hardening Model
The newer formulation, which is invoked by including a PLASTICITY,4 parameter allows greater generality by introducing the
fractional contribution ( f h ) to kinematic hardening as a user input. As shown in Figure 5-54, f is the value between 0
and 1 .
Isotropic hardening f h = 0
Kinematic hardening f h = 1
Combined hardening 0  f h  1 (default f h = 0.5 )
Combined hardening model utilizing kinematic fraction factor is available for von Mises, Hill and Barlat models in
ISOTROPIC, ORTHOTROPIC, and ANISOTROPIC options.
CHAPTER 5 341
Material Library
Work Hardening Curve Isotropic Hardening Kinematic Hardening
  
y y
p p p
 = y +  1 – fh  *   – y   = f   – y 
Figure 5-54 Schematic Explanation for Kinematic Hardening Fraction
Flow Rule
Yield stress and workhardening rules are two experimentally related phenomena that characterize plastic material behavior.
The flow rule is also essential in establishing the incremental stress-strain relations for plastic material. The flow rule
p
describes the differential changes in the plastic strain components d as a function of the current stress state.
The Prandtl-Reuss representation of the flow rule is available in Marc. In conjunction with the von Mises yield function,
this can be represented as:
p 
d ij = d p ---------- (5-264)
 ij
where d p and  are equivalent plastic strain increment and equivalent stress, respectively.
The significance of this representation is illustrated in Figure 5-55. This figure illustrates the stress-space for the two-
dimensional case. The solid curve gives the yield surface (locus of all stress states causing yield) as defined by the
von Mises criterion.
Equation (5-264) expresses the condition that the direction of inelastic straining is normal to the yield surface. This
condition is called either the normality condition or the associated flow rule.
If the von Mises yield surface is used, then the normal is equal to the deviatoric stress.
2
dp p
d2
d1p
1
Yield Surface
Figure 5-55 Yield Surface and Normality Criterion 2-D Stress Space
Constitutive Relations
This section presents the constitutive relation that describes the incremental stress-strain relation for an elastic-plastic
material. The material behavior is governed by the incremental theory of plasticity, the von Mises yield criterion, and the
isotropic hardening rule.
Let the workhardening coefficient H be expressed as:
H = d  d p (5-265)
and the flow rule be expressed as:
p 
d = d p :  where  = ---------- (5-266)
 i j
Consider the differential form of the familiar stress-strain law, with the plastic strains interpreted as initial strains
d = C : d – C : d p (5-267)
where C is the elasticity matrix defined by Hooke’s law and “:” denotes the tensor contraction.
After substitution of Equation (5-266), this becomes
d = C : d – C : d p (5-268)
Contracting Equation (5-268) by 
 : d =  : C : d –  : C : d p (5-269)
and recognizing that
d =  : d (5-270)
with use of Equation (5-265) in place of the left-hand side,
Hd p =  : C : d –  : C : d p (5-271)

By rearrangement
 : C : d
d p = ------------------------------------------- (5-272)
H +  : C : 
Finally, by substitution of this expression into Equation (5-268), we obtain
d = L ep :d (5-273)
where L ep is the elasto-plastic, small strain tangent moduli expressed as:

ep  C :     C :  
L = C – ------------------------------------------------------ (5-274)
H +  : C : 
The case of perfect plasticity, where H = 0 , causes no difficulty.
CHAPTER 5 343
Material Library
Temperature Effects
This section discusses the effects of temperature-dependent plasticity on the constitutive relation.
The following constitutive relations for thermo-plasticity were developed by Naghdi. Temperature effects are discussed
using the isotropic hardening model and the von Mises yield condition.
The stress rate can be expressed in the form:
· ·
 i j = L ijkl  kl + h ij T· (5-275)
For elastic-plastic behavior, the moduli L i j k l are:
 
L i j k l = C ijkl –  C ijm n ------------- ------------ C p qk l  D (5-276)
  m n  p q 
and for purely elastic response:

L i j k l = C ijkl (5-277)
The term that relates the stress increment to the increment of temperature for elastic-plastic behavior is:
 2 
h ij = X ij – C ijkl  kl –  C i j k l -----------   p q X p q – ---  -------   D (5-278)
  k l  3 T 
and for purely elastic response:

H i j = X ij – C ijkl  kl (5-279)
where,
4   
D = ---  2 -------- + ---------- C i j k l ----------- (5-280)
9  p  ij  kl
and
C ijkl e
X i j = ---------------  (5-281)
T kl
and  kl are the coefficients of thermal expansion.
Strain Rate Effects

This section discusses the influence of strain rate on the elastic-plastic constitutive relation.
Strain rate effects cause the structural response of a body to change because they influence the material properties of the
body. These material changes lead to an instantaneous change in the strength of the material. Strain rate effects become
more pronounced for temperatures greater than half the melting temperature ( T m ). The following discussion explains the
effect of strain rate on the size of the yield surface.
Using the von Mises yield condition and normality rule, we obtain an expression for the stress rate of the form:
· · ·· p
 ij = L ijkl  kl + r ij  (5-282)
For elastic-plastic response:

 
L i j k l = C ijkl –  C ijm n ------------- ------------ C p qk l  D (5-283)
  m n  p q 
and
 2 
r ij = C ijmn ------------- ---  --------  D
 mn 3 ·p (5-284)

where,
4   
D = ---  2 -------- + ---------- C i j k l ----------- (5-285)
9  p  i j  k l
Time-independent Cyclic Plasticity (Chaboche Model)

The cyclic plasticity model is based on the work of Chaboche [Ref. 23]. The current version of Marc consists only of the
basic model and plastic-strain-range memorization. The associated time-dependent model is described on Time-dependent
Cyclic Plasticity (Chaboche Model).
The model combines the isotropic hardening rule, to describe the cyclic hardening (Figure 5-56a) or softening, and the
nonlinear kinematic hardening to capture the proper characteristic of cyclic plasticity like Bauschinger (Figure 5-56b),
ratchetting (Figure 5-56c), and mean-stress relaxation (Figure 5-56d) effect. The influence of the plastic strain range on the
stabilized cyclic response is taken into account by introducing the plastic-strain-range memorization variable (Figure 5-56e).
CHAPTER 5 345
Material Library
 
–

 
(a) Cyclic Hardening under (b) Bauschinger Effect (c) Ratchetting

Multiple Cyclic Loading

–

(d) Mean Stress Relaxation (e) Cyclic Hardening

Figure 5-56 Typical Behavior of Material that can be Simulated with Cyclic Plasticity Model
The von Mises yield function is now defined as follows:
f =  – R + k
1
---
3S ij S ij 2 1
where  =  ----------------- , s i j = ' i j – ---  i j ' k k , and ' =  – X
 2  3
X is the back stress tensor representing the center of the yield surface in stress space.
Isotropic Hardening/Softening
The isotropic hardening/softening determines the size of the elastic region during the plastic loading. In this model, it is
controlled by parameter R and k . The initial conditions of cyclic hardening are given as k =  y and R = 0 , while a
cyclic softening is initially described by k =  y – R 0 and R = R 0 .
The evolution equation for the variable R is described as follows:

· ·
R = b  R  – R 
where b and R  are material constants. R  represents the limit of the isotropic hardening/softening. In case of
–b ep s
hardening, then R = R   1 – e .
 
Nonlinear Kinematic Hardening

The nonlinear kinematic hardening is defined from the linear-Ziegler rule by adding the recall term as shown in the
evolution of the back stress tensor below:
· C ·
X = --------------   – X  – X 
R+k
where C and  are two material constants.  = 0 stands for linear-kinematic rule.
Plastic-strain-range Memorization
Several experimental observations show that the asymptotic stress value of cyclic hardening can depend on the prior history.
The influence of plastic-strain range on the stabilized cyclic response is evident from the comparison between the different
histories of loading used to obtained the cyclic curve. Therefore, an introduction of new internal variables that memorize
the prior maximum plastic range is introduced by defining a “memory” surface in the plastic strain space as follows:
2
F = ---  e   p –   – 
3
The evolution of the state variables are as follows:
· ·
 = H  F   n  n * 
· ·
 = 3  2  1 –  H  F   n  n * n * 
where n and n * are the unit normal to the yield surface f = 0 and to the memory surface F = 0 defined as follows:
2 p  p– 
2 -----------
---- *
n = ---
3 ·- and n = ---
3 
-

The coefficient  is introduced in order to induce a progressive memory. For  = 0.5 then, the memorization is
instantaneous and stabilization occurs after one cycle. A progressive memory is given by   0.5 .
The dependency between cyclic plastic flow and the plastic strain range is introduced by considering an asymptotic
isotropic state as follows:
– 2 
R  = Q M +  Q 0 – Q M e
where Q M , Q 0 and  are material constants.

CHAPTER 5 347
Material Library
Plastic Evolution Process and Elasto-plastic “Classical” Modular Matrix

·
The plastic evolution process must conform to the consistency condition, f = 0 . From this condition the plastic rate
multiplier can be derived as follows:
T
· a L ·
 = --------------------------------------------------------------------------------------------------------------------  (5-286)
T C T T
a La + -------------- a   – X  – a X + b  R  – R 
R+k
f 3 s
where a = ------ = --- -------------- .
 2R + k
Since the process involves nonlinear equation, iteration process using predictor-corrector technique is used. The predictor
is calculated using the “trial” elastic stresses as follows:
 B =  A + L  (5-287)
and then calculate f B based on  B and hardening parameter on A. Using the Taylor expansion at B, then
fB
 = --------------------------------------------------------------------------------------------------------------------
- (5-288)
C
a T La + -------------- a T   – X  – a T X + b  R  – R 
R+k
Having the global plasticity iteration converged, then the iteration to satisfy the plastic strain memorization is started. If
both iterations are converged, then the total plasticity iteration is considered completed.
Inserting Equation (5-319) into Equation and using Equation (5-286), the elasto-plastic classical tangent modular matrix
can be derived as follows:
 T 
EP aa L
L = L  I – --------------------------------------------------------------------------------------------------------------------
 T C T T 
 a La + -------------- a   – X  – a X + b  R  – R 
R+k
Hashiguchi Plasticity Model

As opposed to conventional plasticity methods, the Hashiguchi sub-loading surface model describes a smooth elastic-plastic
transition, by excluding a purely elastic domain and introducing a controlling function to attract the stress to the yield
surface. The concept of the sub-loading surface was advocated by Hashiguchi and Ueno (1977) and the elasto-plastic
constitutive equations based on this concept have been formulated by Hashiguchi (1980, 1989, 2016) and Hashiguchi and
Ueno (2017), and completed in the exact form by Hashiguchi (2017).
The formulation given below is based on the additive decomposition of the strain increment into an elastic and a plastic
part:
e p
d = d + d (5-289)
Moreover, the following Von Mises yield condition with isotropic and kinematic hardening is adopted:
f  ̂  = F  H  (5-290)
f  ̂  = ˆ
3  2 ' (5-291)
̂ =  –  (5-292)
dF
F  H  = F 0  1 + h 1  1 – exp  – h 2 H    , F' = ------- = F 0 h 1 h 2 exp  – h 2 H  (5-293)
dH
p' p p'
dH = 2  3 d  d = d  (5-294)
p 1 p
d = c k  d – ---------------------- d  (5-295)
 3  2 F 
where H is the isotropic hardening variable with the material constants h1 and h2 and  (= ' ) is the kinematic hardening
variable, i.e. the back stress which obeys the nonlinear kinematic hardening rule by Armstrong and Frederick (1966), while
ck and  are the material constants.
Here, note that f  ̂  in (5-291) is a homogeneous function of ̂ in degree-one and thus it follows by virtue of Euler’s
theorem for homogeneous functions that:
f  ̂ 
-------------- : ̂ = f  ̂  = F (5-296)
̂
which yields:
f  ̂ 
-------------- :̂
̂ F 1 n̂:̂
n̂ : ̂ = -------------------- = -------------------- -------------------- = --------- (5-297)
f  ̂  f  ̂  f  ̂  F
-------------- -------------- --------------
̂ ̂ ̂
where:
f  ̂  f  ̂  ˆ
'
n̂ = --------------  -------------- = --------- -
ˆ  n̂ = 1  (5-298)
̂ ̂ '
In order to describe the plastic strain increment induced by the increment of stress inside the yield surface, let the following
fundamental postulate be incorporated, which is referred to as the sub-loading surface concept: The stress approaches the
yield surface when the plastic strain increment is induced, exhibiting a continuous variation of the tangent modulus, but
it recedes from the yield surface when only the elastic strain increment is induced.
In this context, it is first required to incorporate the general measure which describes the approaching degree of the stress
to the yield surface, which is renamed to the normal-yield surface. Then, introduce the following sub-loading surface, which
always passes through the current stress and maintains a similar shape and orientation as the normal-yield surface (see
Figure 5-57):
f  ̂  = RF  H  (5-299)
where R  0  R  1  is the ratio of the size of the sub-loading surface to that of the normal-yield surface and is called the
normal-yield ratio. It provides a measure of the approaching degree of the stress to the normal-yield surface.
CHAPTER 5 349
Material Library
Figure 5-57 Normal-yield and sub-loading surfaces
Based on the above-mentioned fundamental postulate of elasto-plasticity, i.e. the sub-loading surface concept, the
increment of the normal-yield ratio must satisfy the following conditions.
    for R  R e 

  0  for R e  R  1 

dR  p (5-300)
 for d  0
 = 0  for R = 1 

  0  for R > 1 
where Re (<1) is the material parameter exhibiting the normal-yield ratio at the fatigue limit under which the plastic strain
increment is not induced. Here, the normal-yield ratio R evolves fulfilling (5-300) in the plastic loading process. On the
other hand, it evolves obeying the elastic constitutive relation in the elastic loading process. Then, it is calculated from the
equation of the sub-loading surface in (5-299), substituting a stress changing by the elastic constitutive relation under fixed
internal variables in the elastic loading precess. Consequently, it follows that:
p p
dR = U  R  d for d  0 (5-301)
where, U(R) is the monotonically decreasing function of R fulfilling the following conditions based on (5-300) as shown in
Figure 5-58.
    for R  R e   elastic state 


  0  for R e  R  1   sub-yield state 
UR (5-302)
 = 0 for R = 1 (normal-yield state)

  0 for R >1(over normal-yield state)
Figure 5-58 Function U(R) in increment of normal-yield ratio R
The explicit form of the function U(R) is given by:
 R – Re n
U  R  = u cot ---  ----------------- (5-303)
2 1 – Re
where   are the Macaulay’s brackets, defined by  s = 0 if s < 0, and  s = s if s  0 . Exponent n   1  is a

material parameter and variable u is defined by:
u = u exp  – u c  c C   (5-304)
in which in which u and u c are material parameters,  c is called the elastic-core yield ratio and C  ( – 1  C   1 ) is
defined by the inner product of the direction vector of the plastic strain rate and the outward normal of the elastic core
surface.
Time-differentiation of (5-299) for the sub-loading surface leads to the consistency condition:
f  ̂  f  ̂ 
-------------- :d – -------------- :d – RdF – dRF = 0 (5-305)
̂ ̂
Noting (5-297), (5-305) leads to:
dF dR
n̂ : d – d –  ------ + ------- ̂ = 0 (5-306)
 F R
Now, adopting the associated flow rule:
p p
d = d̂n̂  d̂ = d  0 (5-307)
CHAPTER 5 351
Material Library
ˆ holds leading to d p = d p' for the Mises yield function ((5-291)).

Here, n̂ = n'
The isotropic and the kinematic hardening rules in (5-294) and (5-295) are described by the following equations:
dH = 2  3 d̂ (5-308)
d = f̂ kn d̂ (5-309)
where,
1
f̂ k n = c k  n̂ – ----------------------  (5-310)
 3  2 F 
Substituting (5-307) with (5-294), (5-301) and (5-309) into (5-306), one gets:
2 F' UR
n̂ : d – f̂ kn d̂ –  --- ---- d̂ + ------------- d̂ ̂ = 0 (5-311)
 3F R 
It follows from (5-307) and (5-311) that

n̂:d p n̂:d
d̂ = ------------- , d = ------------- n̂ (5-312)
p p
M̂ M̂
where:
p 2 F' U  R 
M̂  n̂: f̂ kn +  --- ---- + ------------- ̂ (5-313)
 3F R 
p
The strain increment is described by using Hooke’s law, d = E  d – d  , and (5-312):
–1 n̂:d  – 1 n̂  n̂
d = E :d + ------------- n̂ =  E + --------------- :d (5-314)
p p
M̂  M̂ 
It follows from the (5-314) that,
p n̂:d p
n̂:E:d =  M̂ + n̂:E:n̂  ------------- =  M̂ + n̂:E:n̂ d̂ (5-315)
p
M̂
ˆ
from which the plastic multiplier d̂ . in terms of strain increment, denoted by the symbol d , is derived as:
ˆ n̂:E:d p n̂:E:d
d = ------------------------------- , d = ------------------------------- n̂ (5-316)
p p
M̂ + n̂:E:n̂ M̂ + n̂:E:n̂
The stress increment is given from the (5-314) with (5-316) as:
n̂:E:d  E:n̂  n̂:E 
d = E:d – ------------------------------- E:n̂ =  E – ------------------------------- :d (5-317)
p p
M̂ + n̂:E:n̂  M̂ + n̂:E:n̂
The loading criterion (Hashiguchi, 2000, 2017) is given by:

p
d  0:  n̂:E:d  0 
(5-318)
p
d = 0:  otherwise 
where the judgment whether or not the stress reaches the yield surface is not required since the plastic strain increment is
induced continuously as the stress approaches the normal-yield surface.
The smooth stress vs. strain curve predicted by the sub-loading surface model is schematically illustrated for the simplest
case of the perfectly-plastic material in Figure 5-59. Here, the response is reduced to that of conventional elasto-plasticity by
setting u   , exhibiting the abrupt transition from the elastic to the plastic state.
Figure 5-59 Smooth stress-strain curve predicted by the subloading surface (Simple illustration for the non-hardening
Mises yield condition).
The stress is always attracted to the normal-yield surface by virtue of (5-301) with (5-302). Therefore, it is automatically
pulled back to the normal-yield surface when it goes over the surface in a numerical calculation, because of dR < 0 for R >
1 from (5-301) with (5-302) (4th equation) as shown in Figure 5-60. Therefore, relatively large load increments can be applied
without numerical difficulties.
CHAPTER 5 353
Material Library
Figure 5-60 Stress is automatically controlled to be attracted to the normal-yield surface in the plastic loading process
Further details about the tangential-inelastic strain rate, the extended sub-loading surface model and the isotropic
hardening-stagnation can be found in Hashiguchi (2016, 2017), and Hasiguchi and Ueno (2017).
The following 17 material constants and the 5 initial values are contained in this model.
Material Constants:
Elastic moduli : E, 
Hardening isotropic: h 1 , h 2
kinematic: c k , 
Evolution of normal - yield ratio : u , u c , R e (<1), n   1 
Translation of elastic - core : c,  (<1)
Tangential inelasticity : c̃   1  , ñ   1 
Stagnation of isotropic hardening : C  0  C  1  ,  (> 1), 
Initial Values:
Normal - yield surface size: F 0
center:  0
Elastic - core : c0
Normal - isotropic hardening surface size: K̃ 0
center:  0
Annealing
During the manufacturing of metal pieces stresses accumulate due to the work hardening, decreasing the ductility of the
material. The heat treatment called annealing consists of heating the material beyond a so-called annealing temperature in
order to relieve these stresses. By heating the material, recrystallization can take place and the material becomes softer. It
can be used after work hardening, welding, casting, or after any deformation to relieve accumulated stresses. In the
aerospace industry, welded carbon steel components are an example of material that can be submitted to the annealing
process.
Two methods are available for annealing in Marc:
1. Annealing driven by temperature

The idea is to model the annealing effect as a material property which is activated by the temperature. Then annealing can
potentially occur at any time in a material. An annealing temperature is defined per material, and when the temperature at
the integration points exceeds the annealing temperature, annealing takes place.
The approach chosen here for annealing is to reset the plastic deformation history by resetting the total equivalent plastic
strain.
This is done via the ANNEAL PROP model definition option in Marc and in the Material Properties menu in Mentat. There
an annealing temperature is defined as well as a multiplication factor. The latter is applied to the total equivalent plastic
strain each time the temperature of the material exceeds the annealing temperature. The annealing temperature and the
multiplication factor can be entered as a table. A multiplication factor equal to zero means that the total equivalent plastic
strain will be set to zero.
In case of kinematic or combined hardening, the back stress tensor is also reset with the same multiplication factor.
The plot below illustrates how the total equivalent plastic strain drops to zero (multiplication factor equal to 0.0) at
increment 83, due to the material temperature that exceeds the annealing temperature. This plot is extracted from Marc
Volume E: Demonstration Problems, Problem 3.54.
For the ANNEAL PROP model definition option, the following characteristics are supported:
CHAPTER 5 355
Material Library
Properties Values
Hardening criteria von Mises, Hill, Barlat, Barlat Yld2004-18p, Barlat Yld2004-13p
Hardening method All
Hardening rule Isotropic, kinematic, combined
Decomposition rule Additive, multiplicative (FeFp)
Elements Solid, shell (including composite), plane stress available
Material Isotropic, orthotropic, anisotropic mixture
A temperature evolution is required during the analysis: it can be done in a structural analysis by using the temperature as
a state variable or in a coupled thermal-structural analysis.
Resetting the total equivalent plastic strain is the only action done by this option. Then the relieving of stresses is a
consequence of it. The first consequence is that the yield surface decreases to its value at the new (reset) total equivalent
plastic strain, at the current temperature (or any other independent variable) if the yield function is a function of it. Then
the stress analysis makes the strain tensors evolve so that they remain compatible with the new yield stress. This is especially
true for the elastic strain tensor.
2. Annealing activated by a loadcase

The second method is based on the idea that the annealing process is a specific step within a process chain and that the user
knows in advance when the annealing will take place. Therefore, this method is implemented in Marc as a loadcase.
This is done via the ANNEAL history definition option in Marc and is available in Mentat in the loadcase list. This option
can be applied to bodies.
There the user can:
 Zero out all stress tensors
 Zero out all strains (tensors and equivalent plastic strain)
 Reset the plastic deformation history, namely reset the equivalent plastic strain. Then the new equivalent plastic
strain is equal to its last value times the multiplication factor entered in this menu.
 Define the uniform temperature the body will have after the anneal loadcase.
In case of kinematic or combined hardening, the back stress tensor is also reset using the same multiplication factor.
Please note that during application of the ANNEAL option, no additional analysis is performed in this loadcase.
Time-dependent Inelastic Behavior

Force-displacement relationships vary in different material models. A perfectly elastic material and a perfectly viscous
material can be represented by a spring and a dashpot, respectively (as shown in Figure 5-61). In a perfectly elastic material,
the deformation is proportional to the applied load. In a perfectly viscous material, the rate of change of the deformation
over time is proportional to the load.
In the class of viscoelastic and creeping materials, the application of a constant load is followed by a deformation, which
can be made up of an instantaneous deformation (elastic effect) followed by a continual deformation with time (viscous
effect). Eventually, it can become pure viscous flow. Continued deformation under constant load is termed creep (see
Figure 5-62).
SPRING DASHPOT
f = ku f = u
f = force f = Force
u = Displacement u = Velocity (Time Rate of Change of Displacement
k = Spring Stiffness  = Viscosity of the Dashpot
Figure 5-61 Perfectly Elastic (Spring) and Viscous (Dashpot) Materials
 (Strain)
0 t (Time)
OA – Instantaneous Elastic Effect
AB – Delayed Elastic Effect
BC – Viscous Flow
Figure 5-62 The Creep Curve
A viscoelastic material can be subjected to sudden application of a constant deformation. This results in an instantaneous
proportional load (elastic effect), followed by a gradual reduction of the required load with time, until a limiting value of
the load is attained. The decreasing of load for a constant deformation, is termed relaxation (see Figure 5-63).
Viscoelastic and creeping materials can be represented by models consisting of both springs and dashpots because the
material displays both elastic effects and viscous effects. This implies that the material either continues to flow for a given
stress, or the stress decreases with time for a given strain. The measured relation between stress and strain is generally very
complex.
CHAPTER 5 357
Material Library
 (Stress)
0 t (Time)
Figure 5-63 The Relaxation Curve
Two models that are commonly used to relate stress and strain are the Maxwell and Kelvin (Voigt or Kelvin-Voigt) models.
A description of these models is given below.
The mathematical relation which holds for the Maxwell solid is:
· ·
 =  +  (5-319)
In the one-dimensional case for normal stress:
1
 = ---
E
(5-320)
1
 = ---

This relation can be depicted as a spring and dashpot in series, as shown in Figure 5-64. The integration of Equation (5-319)
yields:
 
 = --- +  --- dt (5-321)
E 
 
Figure 5-64 Maxwell Solid
The strain and stress responses of the Maxwell Solid model are shown in Figure 5-65 and Figure 5-66, respectively.
 

-----0-
E
t t
Constant Stress Applied Response to Constant Stress
Figure 5-65 Strain Response to Applied Constant Stress (Maxwell Solid)
 
t t

Constant Strain Applied Response to Constant Strain
Figure 5-66 Stress Response to Applied Constant Strain (Maxwell Solid)
The mathematical relation which holds for the Kelvin (Voigt or Kelvin-Voigt) solid is:
·
 =  +  (5-322)
This equation is depicted as a spring and dashpot in parallel. (See Figure 5-67). When  = 0 (no dashpot), the system is
a linearly elastic system in which  = E , the elastic modulus.
When E = 0 (no spring), the solid obeys Newton’s equation for a viscous fluid and  =  , the viscous coefficient.
Thus, we can rewrite Equation (5-322) in the form:
·
 = E +  (5-323)
In the above relation, we have considered one-dimensional normal stress and strain. The relation holds equally well for
shear stress  and shear strain  in which  = G , the shear modulus, and  =  , the viscous coefficient. Equation
(5-322) can be rewritten as:
·
 = G +  (5-324)
 

Figure 5-67 Kelvin (Voigt or Kelvin-Voigt) Solid
The strain responses of the Kelvin Solid model are depicted in Figure 5-68. For multiaxial situations, these equations can be
generalized to tensor quantities.
To invoke the Maxwell model, use the CREEP parameter. The creep strain can be specified as either deviatoric creep strain
(conventional creep) or dilatational creep strain (swelling). To invoke the Kelvin model, also use the CREEP parameter and
CRPVIS user subroutine.
CHAPTER 5 359
Material Library
0
t
t1
(a) Stress Pulse

 t
-----0-
E
t
(b) Strain Response to Stress of Infinite Domain

t1
t
(c) Strain Response to Stress Pulse of Finite Length
Figure 5-68 Strain Response to Applied Stress (Kelvin Solid)
Creep (Maxwell Model)

Creep is an important factor in elevated-temperature stress analysis. In Marc, creep is represented by a Maxwell model.
Creep is a time-dependent, inelastic behavior, and can occur at any stress level (that is, either below or above the yield stress
of a material). In many cases, creep is accompanied by plasticity which occurs above the yield stress of the material. The
creep behavior can be characterized as primary, secondary, and tertiary creep, as shown in Figure 5-69. Engineering analysis
is often limited to the primary and secondary creep regions. Tertiary creep in a uniaxial specimen is usually associated with
geometric instabilities, such as necking. The major difference between the primary and secondary creep is that the creep
strain rate is much larger in the primary creep region than it is in the secondary creep region. The creep strain rate is the
slope of the creep strain-time curve. The creep strain rate is generally dependent on stress, temperature, and time.
The creep data can be specified in either an exponent form or in a piecewise linear curve. To specify creep data, use the
CREEP model definition option. The CRPLAW user subroutine allows alternative forms of creep behavior to be programmed
directly.
·c d c
 = -------- (5-325)
dt
Creep Strain
C
Tertiary
Creep
Secondary
Creep
Primary
Creep
Time (t)
Note: Primary Creep:Fast decrease in creep strain rate
Secondary Creep:Slow decrease in creep strain rate Tertiary
Figure 5-69 Creep Strain Versus Time (Uniaxial Test at Constant Stress and Temperature)
Marc offers two schemes for modeling creep in conjunction with plasticity: (a) treating creep strains and plastic strains
separately; and (b) modeling creep strains and plastic strains in a unified fashion (viscoplasticity). Both schemes can be
treated using two different procedures: explicit and implicit.
Creep (Explicit Formulation)

There are six possible modes of input for creep constitutive data.
1. Express the dependence of equivalent creep strain rate on any independent parameter through a piecewise linear
relationship. The equivalent creep strain rate is then assumed to be a piecewise linear approximation to
·c dk  t 
 = A  f     g   c   h  T   ------------- (5-326)
dt
·c c
where A is a constant;  is equivalent creep strain rate; and  ,  , T , and t are equivalent stress, equivalent creep
strain, temperature and time, respectively. The functions f , g , h , and k are piecewise linear and entered in the
form as either slope-break point data or function-variable data. This representation is shown in Figure 5-70. Enter
functions f , g , h , and k through the CREEP model definition option. (Any of the functions f , g , h , or k can be
set to unity by setting the number of piecewise linear slopes for that relation to zero on the input data.)
2. The dependence of equivalent creep strain rate on any independent parameter can be given directly in power law
· n
form by the appropriate exponent. The equivalent creep strain rate is  c = A m    c   T p   qt q – 1 
Enter the constants A , m , n , p , and q directly through the CREEP model definition option. This is often adequate
for engineering metals at constant temperature where Norton’s rule is a good approximation.
·c
 = An (5-327)
CHAPTER 5 361
Material Library
3. Input procedure, one can define a table such that

·c
 = A  f  v 1 ,v 2 ,v 3 ,v 4  (5-328)
where v 1 ,v 2 ,v 3 ,v 4 are four variables that could include equivalent stress, temperature, time, strain, position, or
one of 30 variables. In this case, the function is piecewise linear.
4. Using the table driven input procedure, one can define the strain rate such that
· cr
 = A  equation  v 1 ,v 2 ,v 3 ,v 4  (5-329)
For example, one could enter the equation
A n e – QR  T (5-330)
The equation can take form as long as there is no conditional logic required. If this is not the case, the CRPLAW user
subroutine should be used.
5. Define the equivalent creep strain rate directly with the CRPLAW user subroutine.
6. Use the ISOTROPIC option to activate the ORNL (Oak Ridge National Laboratory rules) capability of the program.
Isotropic creep behavior is based on a von Mises creep potential described by the equivalent creep law
·
 = f    c T t  (5-331)
The material creep behavior is described by
· · c   
 icj =   ----------  (5-332)
  ij 
During creep, the creep strain rate usually decreases. This effect is called creep hardening and can be a function of time or
creep strain. The following section discusses the difference between these two types of hardening.
Consider a simple power law that illustrates the difference between time and strain-hardening rules for the calculation of
the creep strain rate.
c
 = t n (5-333)
c
where  is the creep strain,  and n are values obtained from experiments and t is time. The creep rate can be obtained
c
by taking the derivative  with respect to time
c
·c d
 = -------- = nt n – 1 (5-334)
dt
F4
F3
S3
Function F (X)
[for example, t    , F2 S2
c
g    , h (T),
k (t)]
S1
F1
X1 X2 X3 X4
Variable X (Such as , C, T, t)
(1) Slope-Break Point Data

Slope Break Point
S1 X1
S2 X2
S3 X3
(2) Function-Variable Data

Function Variable
F1 X1
F2 X2
F3 X3
F4 X4
Figure 5-70 Piecewise Linear Representation of Creep Data
However, t being greater than 0, we can compute the time t as:

c 1/n

t =  ----- (5-335)
 
Substituting Equation (5-330) into Equation (5-329) we have:

·c c n – 1  n
 = nt n – 1 = n   1  n     (5-336)
Equation (5-330) shows that the creep strain rate is a function of time (time hardening). Equation (5-336) indicates that the
creep strain rate is dependent on the creep strain (strain hardening). The creep strain rates calculated from these two
hardening rules generally are different. The selection of a hardening rule in creep analysis must be based on data obtained
from experimental results. Figure 5-71 and Figure 5-72 show time and strain hardening rules in a variable state of stress. It is
assumed that the stress in a structure varies from  1 to  2 to  3 ; depending upon the model chosen, different creep
strain rates are calculated accordingly at points 1, 2, 3, and 4. Obviously, creep strain rates obtained from the time
hardening rule are quite different from those obtained by the strain hardening rule.
c
1
2
3
3
1
4
2
0
t
Figure 5-71 Time Hardening
c
1
2
3
3
1 4
2
0
t
Figure 5-72 Strain Hardening
Oak Ridge National Laboratory Laws

Oak Ridge National Laboratory (ORNL) has performed a large number of creep tests on stainless and other alloy steels. It
has also set certain rules that characterize creep behavior for application in nuclear structures. A summary of the ORNL
rules on creep is given below. The references listed at the end of this section offer a more detailed discussion of the ORNL
rules.
1. Auxiliary Rules for Applying Strain-Hardening to Situations Involving Stress Reversals
CHAPTER 5 365
Material Library
The Blackburn Creep Law is required as the CRPLAW user subroutine. The parameter EQCP (first parameter in
CRPLAW) is defined as:
2 1/2
 c =  ---  ij  i j
c c
(5-337)
3 
when the ORNL constitutive option is flagged through use of the ISOTROPIC option. In all other cases, the definition
2 c c 1/2
 c =   ---  ij  i j (5-338)
3 
is retained. The equivalent primary creep strain passes into CRPLAW in EQCPNC, the second parameter. The second
parameter must be redefined in that routine as the equivalent (total) creep strain increment. The first parameter
(EQCP) must be redefined as the equivalent primary creep strain increment when the ORNL constitutive option
is flagged. During analysis with the ORNL option, equivalent creep strain stores the distance between the two
shifted origins in creep strain space (  in ORNL-TM-3602). The sign on this value indicates which origin is
currently active, so that a negative sign indicates use of the “negative” origin ( –  i j ).
2. Plasticity Effect on Creep

The effect of plastic strains on creep must be accommodated for the time-dependent creep behavior of 2 1/4 Cr -1
Mo Steel. Since plastic strains in one direction reduce the prior creep strain hardening accumulated in the reverse
direction, ORNL recommends that the softening influence due to plastic strains be treated much the same as when
reversed creep strain occurs. The following quantities are defined:
+ + +
N ij =   ijI –  i j   G (5-339)
– –
N ij =   ijI –  i–j   G (5-340)
where
 ijI is instantaneous creep strain components
 ij+ ; ij– positive and negative strain origins (5-341)
and
+
G = G   ijI –  i+j  =  2  3   iIj –  i+j    iIj –  i+j   1  2 (5-342)
G– = G   ijI –  i–j  =  2  3   iIj –  i–j   iIj –  i–j  1  2 (5-343)

Swelling
Marc allows pure swelling (dilatational creep) effect in a creep analysis. To use the swelling option, perform a regular creep
V
analysis as discussed earlier. Use the VSWELL user subroutine to define the increment of volumetric swelling  ------- . The
 V
increment of volumetric swelling is generally a function of neutron flux, time, and temperatures.
For example, radiation-induced swelling strain model for 20% C. W. Stainless Steel 316 can be expressed as:
V R 1 + exp     – t  
------- = Rt + --- ln ------------------------------------------------- (5-344)
V  1 + exp
where R , t , and  are functions of temperature,  is neutron flux, and t is time.
Creep (Implicit Formulation)

This formulation, as opposed to that described in the previous section, is fully implicit. A fully implicit formulation is
unconditionally stable for any choice of time step size; hence, allowing a larger time step than permissible using the explicit
method. Additionally, this method is more accurate than the explicit method. The disadvantage is that each increment may
be more computationally expensive. This model is activated using the CREEP parameter. There are two methods for defining
the inelastic strain rate. The CREEP model definition option can be used to define a Maxwell creep model. The back stress
must be specified through the field normally reserved for the yield stress in the ISOTROPIC or ORTHOTROPIC options. The yield
stress must be specified through the field normally reserved for the 10th cycle yield stress in the ISOTROPIC option. There is
no plastic strain when the stress is less than the yield stress and there is no creep strain when the stress is less than the back
stress.
The equivalent creep strain increment is expressed as:
·
 c = A m    c  n  T p   qt q – 1  (5-345)
Enter the constants A , m , n , p , and q directly through the CREEP model definition option. A more general expression
for the equivalent creep strain rate is given by:
·c dk  t 
 = A   m  g   c   h  T   ------------- (5-346)
dt
Enter the terms A and m and the functions g , h , and k through the UCRPLW user subroutine.
The creep strain components are given by:
d
3  ij
 ij = ---  -------
2 
d
where  ij is the deviatoric stress at the end of the increment. A is a function of temperature, time, etc. An algorithmic
tangent is used to form the stiffness matrix.
Based on a parameter defined in the CREEP parameter, one of three tangent matrices is formed. The first is using an elastic
tangent, which requires more iterations, but can be computationally efficient because re-assembly might not be required.
CHAPTER 5 367
Material Library
The second is an algorithmic tangent that provides the best behavior for small strain power law creep. The third is a secant
(approximate) tangent that gives the best behavior for general viscoplastic models. When creep is specified in conjunction
with plasticity, the elastic tangent option is not available.
Viscoplasticity
Explicit Formulation
The creep (Maxwell) model can be modified to include a plastic element (as shown in Figure 5-73). This plastic element is
inactive when the stress (  ) is less than the yield stress (  y ) of the material. The modified model is an elasto-viscoplasticity
model and is capable of producing some observed effects of creep and plasticity. In addition, the viscoplastic model can be
used to generate time-independent plasticity solutions when stationary conditions are reached. At the other extreme, the
viscoplastic model can reproduce standard creep phenomena. The model allows the treatment of nonassociated flow rules
and strain softening which present difficulties in conventional (tangent modulus) plasticity analyses.
The viscoplasticity option can be used to implement very general constitutive relations with the aid of the following user
subroutines: ZERO, YIEL, NASSOC, and CRPLAW. See Nonlinear Analysis in Chapter 5 for details on how to use these procedures.

e
 vp
p vp
 =  Plastic Element
Inactive if < y
Figure 5-73 Uniaxial Representation of Viscoplastic Material
Implicit Formulation
To allow for the implementation of general unified creep-plasticity or viscoplastic models, the UVSCPL user subroutine is
available. This routine requires you to define only the inelastic strain rate. The program automatically calculates a tangent
stiffness matrix (only elastic tangent or secant tangent can be used). This option is activated by indicating that the material
is VISCO PLAS in the ISOTROPIC or ORTHOTROPIC option.
Time-dependent Cyclic Plasticity (Chaboche Model)

The time-dependent effect of the model described in Time-independent Cyclic Plasticity (Chaboche Model) is modeled using the
unified viscoplastic framework. The viscoplastic potential is based on overstress quantity as follows:
K f n+1
 = -------------  ----
n+1 K
The viscoplastic strain rate is defined as follows:
·  3 · ' – X' 3·
v p = = ---  ---------------- = ---  f (5-347)
 2 ̃ e 2
where,
· f n
 =  ---- (5-348)
K
1---
·n
In this case, the viscoplastic stress is  v p = K .
The hardening rules are chosen to be identical to the time-independent case.
Viscoplastic Evolution Process and Visco-plastic “Classical” Modular Matrix

The iteration procedure, starting from the trial elastic stress as the predictor, is based on the implicit integration of Equation
(5-348) that can be expressed as follows:
 f n
r  = – ------ +  ---- (5-349)
t K
Using Newton iteration scheme, the iterative value of  (that is,  ) can be expressed as follows:
i
t r n
 = -------------------------------------------------------------------------------------------------------------------------------------
-
n – 1
n f
1 + ----  ----  a T La + 2--- Ca T a – a T X + b  R – R 
K K  3  
i
where r  is the i-th iteration of the residual of Equation (5-349).
The tangent modular matrix is based on the assumption that r  = 0 . Therefore the classical visco-plastic modular
matrix can be expressed as follows:
 
 
 T 
vp aa L
L = L  I – -----------------------------------------------------------------------------------------------------------------------------------
 T 2 T T 1 
 a La + --- Ca a – a X + b  R  – R  + -------------------------------
 3 n f n–1 
 ----  ---- t
K K
This matrix is in line with the consistent model derived in [Ref. 20].
Anand Solder Model

The Anand [Ref. 41] Solder model allows the user to model a solder material using the unified Anand model which consists
of a simple set of constitutive equations for large, isotropic, viscoplastic problems. There are two basic features in this Anand
CHAPTER 5 369
Material Library
model. First, this model needs no explicit yield condition and no loading/unloading criterion. The plastic strain is assumed
to take place at all nonzero stress values, although at low stresses the rate of plastic flow may be immeasurable. Second, this
model employs a single scalar as an internal variable to represent the isotropic resistance to plastic flow offered by the
internal state of the material.
The Anand Solder model uses a single scalar internal variable, s , which denotes the averaged isotropic resistance to
macroscopic plastic flow offered by the underlying isotropic strengthening mechanisms such as dislocation density, solid
solution strengthening, subgrain, and grain size effects, etc. The deformation resistance s is consequently proportional to
the equivalent stress. The flow equations and the evolution equations are given below:
Flow Equation:
1-
---
d p  m
Q
-------- = A sinh  ------- exp  – -------
dt  s  RT
where
s is the single internal variable representing deformation resistance s 0

A is the pre-exponential factor
 is the multiplier of stress
m is the strain rate sensitivity of stress
Q is the Activation Energy
R is the Universal Gas Constant
where the evolution equation is expressed as:

Evolution Equations:
ds  a B  d p
----- =  h 0 B -------  --------
dt  B  dt
s
B = 1 – -----
s*
1 d p Q n
s * = ŝ --- -------- exp  -------
A dt  RT
where,
h 0 is the hardening constant

ŝ is the deformation resistance saturation coefficient
n is the strain rate sensitivity of saturation
a is the strain rate sensitivity of hardening
s * represents the saturation value of s associated with a set of given temperatures and strain rates. Thus the Anand Solder
model has nine material parameters: A , Q ,  , m , h 0 , ŝ , n , a , where s 0 is the initial value of the deformation resistance
needed to determine the evolution of the deformation resistance.
Determination of Parameters for Anand Model

There are some standard test procedures for determination of the constants for the Anand Solder model. These involve a
steady state creep test followed by use of a nonlinear fitting method to determine the constants. More details about the
exact nature of tests required can be found in [Ref. 31]. The Anand material constants for some standard solder alloys are
included in the table below.
Material parameters of Anand model for some solders:
Solders
Material
Parameters 60Sn40Pb 62Sn36Pb2Ag 96.5Sn3.5Ag 97.5Pb2.5Sn
A  s –1  1.49e7 2.3e7 2.23e4 3.25e12
J
Q  --------- 90040 93632 73995 129557
 mol
 11 11 6 7
m 0.303 0.303 0.182 0.143
ŝ (MPa) 80.42 80.79 73.81 72.73
n 0.0231 0.0212 0.018 0.00437
h 0 (MPa) 2640.75 4121.31 3321.15 1787.02
a 1.34 1.38 1.82 3.73
s 0 (MPa) 56.33 43.32 39.09 15.09
Viscoelastic Material
Marc has two models that represent viscoelastic materials. The first can be defined as a Kelvin-Voigt model. The latter is a
general hereditary integral approach.
Kelvin-Voigt Model
The Kelvin model allows the rate of change of the inelastic strain to be a function of the total stress and previous strain. To
activate the Kelvin model in Marc, use the CREEP parameter.
k
The Kelvin material behavior (viscoelasticity) is modeled by assuming an additional creep strain  , governed by:
ij
d d
-----  k = A ijkl  kl – B i j k l  kkl (5-350)
dt ij
CHAPTER 5 371
Material Library
where A and B are defined in the CRPVIS user subroutine and the total strain is:
 ij =  ije +  ijp +  ijc +  ikj + ti hj (5-351)
th
ij
= thermal strain components (5-352)
 ije = elastic strain components (instantaneous response) (5-353)
 ipj = plastic strain components (5-354)
 icj = creep strains defined via the CRPLAW and VSWELL user subroutines (5-355)
 ikj = Kelvin model strain components as defined above (5-356)
The CRPVIS user subroutine is called at each integration point of each element when the Kelvin model is used.
Use the AUTO CREEP option to define the time step and to set the tolerance control for the maximum strain in any increment.
The CREEP option allows Maxwell models to be included in series with the Kelvin model.
Hereditary Integral Model

The stress-strain equations in viscoelasticity are not only dependent on the current stress and strain state (as represented in
the Kelvin model), but also on the entire history of development of these states. This constitutive behavior is most readily
expressed in terms of hereditary or Duhamel integrals. These integrals are formed by considering the stress or strain build-
up at successive times. Two equivalent integral forms exist: the stress relaxation form and the creep function form. In Marc,
the stress relaxation form is used.
The viscoelasticity option in Marc can be used for both the small strain and large strain Mooney or Ogden material stress-
relaxation problems for total Lagrange formulation. It can also be used with all hyperelastic models; i.e., Mooney, Ogden,
Gent, Arruda-Boyce, Foam and generalized hyperelastic materials if the Updated Lagragian formulation is used. A
description of these models is as follows:
Small Strain Viscoelasticity

In the stress relaxation form, the constitutive relation can be written as a hereditary integral formulation
t
d k l   
i j  t  =  G ijkl  t –   -----------------
d
- d + G i j k l  t  k l  0  (5-357)
0
The functions G ijkl are called stress relaxation functions. They represent the response to a unit applied strain and have
characteristic relaxation times associated with them. The relaxation functions for materials with a fading memory can be
expressed in terms of Prony or exponential series.
N
 n n
G ijk l  t  = G ijkl +  G i j k l exp  – t    (5-358)
n = 1
n n
in which G ijkl is a tensor of amplitudes and  is a positive time constant (relaxation time). In the current

implementation, it is assumed that the time constant is isotropic. In Equation (5-358), G i j kl represents the long term
modulus of the material.
The short term moduli (describing the instantaneous elastic effect) are then given by:
N
0  n
G ijk l = G ijkl  0  = Gi j k l +  Gi j k l (5-359)
n = 1
The stress can now be considered as the summation of the stresses in a generalized Maxwell model (Figure 5-74)
N
 n
i j  t  =  ij  t  +  i j  t  (5-360)
n = 1
where,
 
 ij = G ijkl  kl  t  (5-361)
t
n n n d k l   
i j =  G ijkl exp  –  t –      ------------------ d (5-362)
d
0
1 2 3 N
E E E E

E
1 2 3 N
   
i i i
 =  E
Figure 5-74 The Generalized Maxwell or Stress Relaxation Form
For integration of the constitutive equation, the total time interval is subdivided into a number of subintervals ( t m – 1 t m )
with time-step t = t m – t m – 1 . A recursive relation can now be derived expressing the stress increment in terms of the
values of the internal stresses  inj at the start of the interval.
CHAPTER 5 373
Material Library
With the assumption that the strain varies linearly during the time interval t , we obtain the increment stress-strain
relation as:
N N
 n n n n
 ij  t m  = G ijkl +    t G i j k l  k l –    t  i j  t m – t  (5-363)
n = 1 n = 1
where,
n
 n  t  = 1 – exp  –  t     (5-364)
and
n n n
  t  =   t    t  (5-365)
In Marc, the incremental equation for the total stress is expressed in terms of the short term moduli (See Equation (5-359)).
N N
0 n n n n
 ij  t m  = G ijkl –  1 –   t  G i j k l  k l  t m  –    t  i j  t m – t  (5-366)
n = 1 n = 1
In this way, the instantaneous elastic moduli can be specified through the ISOTROPIC or ORTHOTROPIC options. Moreover,
since the TEMPERATURE EFFECTS or TABLE option acts on the instantaneous elastic moduli, it is more straightforward to use
the short term values instead of the long term ones. Note that the set of equations given by (5-366) can directly be used for
both anisotropic and isotropic materials.
Isotropic Viscoelastic Material

For an isotropic viscoelastic material, Marc assumes that the deviatoric and volumetric behavior are fully decoupled and
that the behavior can be described by a time dependent shear and bulk modules. The bulk moduli is generally assumed to
be time independent; however, this is an unnecessary restriction of the general theory.
Both the shear and bulk moduli can be expressed in a Prony series
N
 n n
Gt = G +  G exp  – t   d  (5-367)
n = 1
N
 n n
Kt = K +  K exp  – t   v  (5-368)
n = 1
with short term values given by:
N
G0 = G +  Gn (5-369)
n = 1
N
K0 = K +  Kn (5-370)
n = 1
Let the deviatoric and volumetric component matrices  d and  v be given by:
4  3 –2  3 –2  3 0 0 0
–2  3 4  3 –2  3 0 0 0
–2  3 –2  3 43 0 0 0
d =
0 0 0 1 0 0
0 0 0 0 1 0
0 0 0 0 0 1 (5-371)
1 1 1 0 0 0
1 1 1 0 0 0
v = 1 1 1 0 0 0
0 0 0 0 0 0
0 0 0 0 0 0
0 0 0 0 0 0
The increment set of equations is then given by:
 Nd 
 0 n n
  t m  =  G –

  1 –  d  t  G  d   t m 

 n = 1 
 Nv  (5-372)
 0 n n
K –

 1 –  v  t  K  v   t m 

 n = 1 
Nd Nv
n n n n
–   d  t  d  t m – t  –   v  t  v  t m – t 
n = 1 n = 1
and
n n n n n
 d  t m  =  d  t G  d   t m  –  d  t  d  t m – t 
n n n n n (5-373)
 v  t m  =  v  t K  v   t m  –  v  t  v  t m – t 
Note that the deviatoric and volumetric response are fully decoupled.
The instantaneous moduli need to be given in the ISOTROPIC option. Time dependent values (shear moduli G n and time
n n n
constants  d ; bulk moduli K and time constants  v ) need to be entered in the VISCELPROP option.
Time-stepping is performed using the TIME STEP or AUTO STEP option in the history definition block.
CHAPTER 5 375
Material Library
Note that the algorithm is exact for linear variations of the strain during the increment. The algorithm is implicit; hence,
for each change in time-step, a new assembly of the stiffness matrix is required.
Anisotropic Viscoelastic Material

0
(5-363) can be used for the analysis of anisotropic viscoelastic materials. The tensor of amplitudes G i j kl must be entered
n n
through the ORTHOTROPIC option. However, both the G i j k l and the  must be entered using the VISCELORTH option.
Alternatively, a complete set of moduli (21 components) can be specified in the HOOKVI user subroutine.
The ORIENTATION option or ORIENT user subroutine can be used to define a preferred orientation both for the short time
0 n
moduli G ijkl and the amplitude functions G i j k l .
Incompressible Isotropic Viscoelastic Materials

Incompressible elements in Marc allow the analysis of incompressible and nearly incompressible materials in plane strain,
axisymmetric and three-dimensional problems. The incompressibility of the element is simulated through the use of an
perturbed Lagrangian variational principle based on the Herrmann formulation.
The constitutive equation for a material with no time dependence in the volumetric behavior can be expressed as
 N 
 0 n n  1
 ij  t m  = 2  G ijkl –  1 –   t  G i j k l  kl  t m  – ---  p p  t m  kl
  3
 n = 1  (5-374)
N
n n 1
–    t   i j   t m  + ---  kk  i j
3
n = 1
0
 p p  t m  = 3K  pp  t m  (5-375)
The hydrostatic pressure term is used as an independent variable in the variational principle. The Herrmann pressure
variable is now defined in the same way as in the formulation for time independent elastic materials.
 pp
H = -------------------------------
- (5-376)
2G 0  1 +  0 
The constitutive Equations (5-374) and (5-375) can then be rewritten
N
e n d n
 ij  t m  = 2G   i j +  H i j  –    t    i j   t m – t  (5-377)
n = 1
where,
N
Ge = G0 –   1 –  n  t G n  (5-378)
n = 1
0 0 e 0
G  1 +   – G  1 – 2 
 = ------------------------------------------------------------------- (5-379)
e
3G
Thermo-Rheologically Simple Behavior

The rate processes in many viscoelastic materials are known to be highly sensitive to temperature changes. Such
temperature-dependent properties cannot be neglected in the presence of any appreciable temperature variation. For
example, there is a large class of polymers which are adequately represented by linear viscoelastic laws at uniform
temperature. These polymers exhibit an approximate translational shift of all the characteristic response functions with a
change of temperature, along a logarithmic time axis. This shift occurs without a change of shape. These temperature-
sensitive viscoelastic materials are characterized as Thermo-Rheologically Simple.
A “reduced” or “pseudo” time can be defined for the materials of this type and for a given temperature field. This new
parameter is a function of both time and space variables. The viscoelastic law has the same form as one at constant
temperature in real time. If the shifted time is used, however, the transformed viscoelastic equilibrium and compatibility
equations are not equivalent to the corresponding elastic equations.
In the case where the temperature varies with time, the extended constitutive law implies a nonlinear dependence of the
instantaneous stress state at each material point of the body upon the entire local temperature history. In other words, the
functionals are linear in the strains but nonlinear in the temperature.
The time scale of experimental data is extended for Thermo-Rheologically Simple materials. All characteristic functions of
the material must obey the same property. The shift function is a basic property of the material and must be determined
experimentally. As a consequence of the shifting of the mechanical properties data parallel to the time axis (see Figure 5-75),
the values of the zero and infinite frequency complex moduli do not change due to shifting. Hence, elastic materials with
temperature-dependent characteristics neither belong to nor are consistent with the above hypothesis for the class of
Thermo-Rheologically Simple viscoelastic solids.
T0
f(T1 )
T1
T2 f(T2 )
GT
log10 t
Figure 5-75 Relaxation Modulus vs. Time at Different Temperatures
Let E(log10 t) be the relaxation modulus as a function of log10 t at uniform temperature, T . Then
E T  log10 t  = E T  log10 t+f  T  *  T   0 T 0  (5-380)

0
CHAPTER 5 377
Material Library
where f  T  is measured relative to some arbitrary temperature T . The modulus curve shifts towards shorter times with an
increase of temperature; f  T  is a positive increasing function for T  T 0 . If G T  t  denotes the relaxation modulus as a
function of time at uniform temperature T , so that,
G T  t  = E T  ln t  (5-381)
then
GT  t  = GT    (5-382)
0
where
 = aT(T)t (5-383)
The relaxation modulus (and the other characteristic functions) at an arbitrary uniform temperature is thus expressed by
the base temperature behavior related to a new time scale that depends on that temperature. In particular, any relaxation
time at temperature T is related to the relaxation time at the reference temperature T0 by the relation
  T0 
  T  = --------------
aT  T  (5-384)
There is some mapping of the time coordinate for nonuniform, nonconstant temperature, T  x t  , which depends on the
position.
For a nonuniform, nonconstant temperature, the shift function is a  T  x t   and the rate of change of reduced time
can be written as:
d = a T  T dt (5-385)
The reduced time  is now given as the integral over the temperature history of the point:
t
  x t  = 0a  T  x ,t  dt (5-386)
Marc offers four explicit forms for entering the shift function. The first is based on the familiar Williams-Landel-Ferry
(WLF) equation, which states that
C1  T – T0 
log 10aT  T  = -----------------------------
- = fT (5-387)
C2 + T – T0
and
t
f  T  t  
t =  10 dt (5-388)
0
Typically, the glassy transition point is taken as the reference temperature, in which case all relaxation curves at higher
temperatures make a left-shift on the log10 time scale as is illustrated in Figure 5-75. The second form of the logarithmic
shift is expressed by a polynomial expansion about an arbitrarily chosen reference temperature T0 as:
m
i
log 10 a T  T  =   i  T – T 0  where a =0
0 (5-389)
i = 0
The third form of the logarithmic shift is expressed by an Arrhenius equation as:
E 1 1
ln a T  T  = ---  ------ – --- (5-390)
R  T 0 T
where,
aT is the time based shift factor
E is the ratio of activation energy to the universal gas constant.

---
R
T is the temperature.
T0 is the glass transition temperature or reference temperature.
Figure 5-76 Three-Zone Combined WLF and Arrhenius Shift Function
The fourth form of the logarithmic shift combines the WLF and the Arrhenius equations. The shift function is defined in
three temperature zones as shown in Figure 5-76.
In the lower temperature zone T<TL it is defined by an Arrhenius function:
c1  TL – T0  cL 1 1
f  T  = ------------------------------- + ----------------  ------ – ---
c 2 + T L – T 0 ln  10   T L T
CHAPTER 5 379
Material Library
In the central temperature zone TL≤T≤TU it is defined by a WLF function:
c1  T – T0 
f  T  = ----------------------------
c2 + T – T0
In the upper temperature zone T>TU it is defined by an Arrhenius function:
c1  TU – T0  cU  1
- ------- – --1-
- + ---------------
f  T  = -------------------------------
c 2 + T U – T 0 ln  10   T U T
where,
T is the absolute Temperature.

f(T) is the logarithmic shift factor (base 10).
T0 is the reference temperature.
TL is the lower transition temperature.
TU is the upper transition temperature.
c1 is the WLF numerator constant.
c2 is the WLF denominator constant.
cL is the activation energy divided by the gas constant EL ⁄ R in the lower temperature zone.
cU is the activation energy divided by the gas constant EU ⁄ R in the upper temperature zone.
The WLF-like terms in the lower and upper temperature zones evaluate to constants which assure continuity with the
WLF-function of the central temperature zone.
All together the shift function is characterized by 7 material constants, the three temperatures T0, TL and TU, the two WLF
constants c1 and c2 and the two Arrhenius constants cL and cU. The reference temperature T0 should be between TL and
TU.
Enter the shift function parameters associated with Thermo-Rheologically Simple behavior through the SHIFT FUNCTION
model definition option. As one alternative to the above analytical expressions of the shift function Marc allows its
specification via the TRSFAC user subroutine. As a second alternative Marc allows the use of a TABLE, where log10 aT (T ) is
specified as a function of temperature and possibly other state variables. Also see the section Thermo-Rheologically Simple
Material Behavior in the Frequency Domain in the section on Frequency-dependent Material Behavior of this chapter”.
0
In addition to the Thermo-Rheologically Simple material behavior variations of initial stress-strain moduli G i j kl , the
temperature of the other mechanical properties (coefficient of thermal expansion, etc.) due to changes in temperature can
be specified via the TEMPERATURE EFFECTS option.
Note, however, that only the instantaneous moduli are effected by the TEMPERATURE EFFECTS option. Hence, the long term
moduli given by:
N
 0 n
G i j kl = G ijkl  T  –  Gi j k l (5-391)
n = 1
can easily become negative if the temperature effects are not defined properly. To avoid this physically incorrect behavior,
for isotropic viscoelastic materials it is possible to scale the time dependent moduli relative to Young’s modulus (assuming
that the shear and bulk moduli have an equivalent temperature dependency). This scaling has to be activated on the
VISCELPROP option. If this scaling is active, also a reference temperature for the scaling operation has to be provided on this
option. The applied modification involves the calculation of a scale factor f 1 according to:
ET GT KT

f 1 = ------------------- = ------------------- = ------------------- (5-392)
E  T r ef  G  Tr e f  K  Tr e f 
in which T is the current temperature and T r e f is the user-defined reference temperature. Using this factor, the shear and
bulk moduli are scaled as:
n n n n
G := f 1 G ; K := f 1 K (5-393)
which results in the following relaxation functions:
 N n 
 G n 
G  t T  = G0 T  1  –  ---------------------
-  1 – exp  – t   d    (5-394)
0
G  Tr e f 
 
 n = 1 
 N n 
 K n 
K  t T  = K0 T  1  –  ---------------------
-  1 – exp  – t   d    (5-395)
0
K  Tr e f 
 
 n = 1 
Large Strain Viscoelasticity

For an elastomeric time independent material, the constitutive equation is expressed in terms of an energy function W .
For a large strain viscoelastic material, Simo generalized the small strain viscoelasticity material behavior to a large strain
viscoelastic material. The energy functional then becomes
N N
n 0 n n n
  E ij Q ij  =   E ij  –  Qi j Ei j +  I  Qi j  (5-396)
n = 1 n = 1
n 0
where E ij are the components of the Green-Lagrange strain tensor, Q i j internal variables and  the elastic strain energy
0
density for instantaneous deformations. In Marc, it is assumed that  = W , meaning that the energy density for
instantaneous deformations is given by the third order James-Green-Simpson form or the Ogden form. When used with
Updated Lagrange, one can also use the Arruda-Boyce or Gent model.
CHAPTER 5 381
Material Library
The components of the second Piola-Kirchhoff stress then follow from:

N
  0 n
S ij = ---------- = ---------- –
E ij E ij  Qi j (5-397)
n = 1
n
The energy function can also be written in terms of the long term moduli resulting in a different set of internal variables T i j
N
n n
  E ij T ij  =  E ij  +  Ti j Ei j (5-398)
n = 1

where  is the elastic strain energy for long term deformations. Using this energy definition, the stresses are obtained
from:
N
   E  n
S ij = -------------------- +
E ij  Ti j (5-399)
n = 1
Let the total strain energy be expressed as a Prony series expansion

N

 =  +   n exp  – t   n  (5-400)
n = 1
If, in the energy function, each term in the series expansion has a similar form, Equation (5-400) can be rewritten as:
N

 =  +   n  0 exp  – t   n  (5-401)
n = 1
n
where  is a scalar multiplier for the energy function based on the short term values.
The stress-strain relation is now given by
N
 n
S ij = S ij +  T ij (5-402)
n = 1
 N 
    0
S i j = ----------- =  1 – n ---------
  E i-j
 (5-403)
E ij 
 n = 1 
Observing the similarity with the equations for small strain viscoelasticity, the internal variables can be obtained from a
convolution expression
t
n n ·0 n
T ij =  S ij   exp  –  t –     d (5-404)
0
Analogue to the derivation for small strain viscoelasticity, a recursive relation can be derived expressing the stress increment
in terms of values of the internal stresses at the start of the increment.
In Marc, the instantaneous values of the energy function are always given on the MOONEY, OGDEN, or FOAM option, the
equations are reformulated in terms of the short time values of the energy function
 N 
0 0
S i j  t m  = 1 –   1 –  n  t    n  S i j  t m  – S i j t m – t 

 
 n=1 
N (5-405)
n
–   n  t T i j  t m – t 
n = 1
n 0 0 n
T i j  t m  =  n  t  n  S i j  t m  – S i j  t m – t   –  n  t T i j  t m – t  (5-406)
It is assumed that the viscoelastic behavior in Marc acts only on the deviatoric behavior. The incompressible behavior is
taken into account using special Herrmann elements.
Large strain viscoelasticity is invoked by use of the VISCELMOON, VISCELOGDEN, or VISCELFOAM option in the constitutive
n n
option of the model definition block. The time dependent multipliers  and associated relaxation times  as defined
by Equation 5-404 are given in the VISCELMOON, VISCELOGDEN, or VISCELFOAM option. For the Ogden model, both deviatoric
and dilatational relaxation behavior is allowed.
Viscoelasticity can be modeled with Arruda-Boyce and Gent models using VISCELMOON option.
Time-stepping can be performed using the TIME STEP with AUTO LOAD or AUTO STEP option of the history definition block.
The free energy function versus time data being used for large strain viscoelasticity can be generated by fitting the
experimental data in Mentat or Patran provided the following two tests are done:
1. Standard quasi-static tests (tensile, planar-shear, simple-shear, equi-biaxial tension, volumertic) to determine the
0
elastomer free energy W constants.
2. Standard relaxation tests to obtain stress versus time.
Bergström-Boyce Model
The Bergström-Boyce viscoelastic model in small deformation theory is schematically shown in Figure 5-77. In Figure 5-77,
 ,  e , and  i are the total, elastic, and creep strains, respectively; E  is the Young’s modulus for the “time-infinity” spring
and E m is the Young’s modulus for the spring in the Maxwell element;  is the viscosity in the dashpot.
CHAPTER 5 383
Material Library
 = i + e
E


Em
i e
Figure 5-77 Rheological Model for Linear Viscoelastic Material Behavior
The total stress is then given as  = E   + E m  e and the strain energy density is:
1 1
    e  =  eq    =  n eq   e  = --- E   2 + --- E m  e2
2 2
where  eq and  neq represent the strain energy densities in the “time-infinity” spring and the spring in the Maxwell
element, respectively.
In contrast to the additive decomposition of deformations in small strain theory, in the case of large strain nonlinear
viscoelastic hyperelasticity, the deformation gradient F can be decomposed multiplicatively into a volumetric part J 1  3 1
and a volume-preserving part F . F can be further decomposed into an elastic part F e and an inelastic part F i , i.e.,
F = Fe Fi .
The strain energy function of the modified Bergström-Boyce model is then defined by:
  C C e  = U  J  + W  C  + W v  C e 
In the above equation, U  J  and W  C  are the strain energy functions in the time-infinity portion, representing the
volumetric and the volume-preserving parts, respectively; W v  C e  is the volume-preserving strain energy from the viscous
contribution; J is the determinant of deformation gradient; C is the volume-preserving part of the total right Cauchy-
Green deformation tensor, and C e is the elastic volume-preserving part of right Cauchy-Green deformation tensor from
viscous contribution. In Marc, U  J  and W  C  can be any of the Marc rubber or foam models; W v  C e  must be the
Arruda-Boyce model to be defined in this model definition option.
Bergström-Boyce defined the viscoelastic strain rate as:
m
· · c  v  2 
i
 =  0    chain – 1  2  ---------------------
iso
- 
 ̃ 
i
where  chain is the inelastic network stretch, and  ivs o is the deviatoric Kirchhoff stress from viscous contribution.
The value of ̃ is the user-defined normalization value with the same units as stress.
Marc defines the effective creep strain rate is defined by
· c
i
 = c 1   chain – 1  2  ivs o m
where c 1 , c 2 , and m are material parameters.
The value of c 1 is defined as
·
c 1 =  0   2̂  m
·
 0 is the effective creep rate.
The material parameters of the model are defined with BERGBOYCE option, along with other hyperelastic options available
in Marc.
This model is implemented within the updated Lagrange framework based on hyperelasticity and multiplicative
decompositions of deformation gradient.
Parallel Rheological Framework (PRF)

The parallel rheological framework can be used to model the response of filled rubber under cyclic loading. The model
consists of one principal network, n nonlinear visco-elastic networks, and m nonlinear elastic-plastic networks connected
in parallel as shown in [Ref. 55].

… …
0 1 2 n n+1 n+ 2 n+m
… …
Figure 5-78 Schematic Representation of the Parallel Rheological Framework (PRF)

CHAPTER 5 385
Material Library
Network 0 is the principal network already available in Marc. Any existing Marc rubber models along with damage models
(continuous or discontinuous) and visco-elastic model (Simo’s model) can be used in the network. This network can also
be the standard isotropic elastic-plasticity defined by the quadratic logarithmic strain energy function and by von Mises
yield criterion. The material properties of the principal network can be defined with any of the following model definition
options: MOONEY, OGDEN, GENT, ARRUDBOYCE, FOAM and ISOTROPIC.
Networks 1 to n are the nonlinear visco-elastic networks. At current release, the rubber model must be Arruda-Boyce
model and the viscosity is based on the Bergström-Boyce model. The material properties of the networks can be defined
with VISCO HYPE model definition option.
Networks n + 1 to n + m are the nonlinear elastic-plasticity networks to take into account permanent set effect of rubber.
At current release, the elastic parameters in the network can be defined by OGDEN, ARRUDBOYCE, or ISOTROPIC options. The
plasticity is based on von Mises criterion. The material properties of the networks can be defined with PERM SET model
definition option.
Multiplicative decompositions of deformation gradient are assumed in each network:
e cr i = 1 n
F = Fi Fi (5-407)
e p i = n + 1 n + m (5-408)
F = Fi Fi
e cr p
where F i , F i , and F i are elastic part, creep part and plastic part of deformation gradient, respectively.
The materials are defined by a strain energy function:

n+m
e
  C  = U  J  + W  C0  +  W  Ci  (5-409)
i = 1
e
C 0 and C i are the volume-preserving parts of elastic right Cauchy-Green deformation tensors, U  J  is the volumetric
strain energy, and J is the determinant of F .
The stress response can be calculated based on the hyperelasticity from equation (5-409).
This model is not available for trusses, beams, plane stress, or shell elements.
Narayanaswamy Model
The annealing of flat glass requires that the residual stresses be of an acceptable magnitude, while the specification for
optical glass components usually includes a homogenous refractive index. The design of heat treated processes (for example,
annealing) can be accomplished using the Narayanaswamy model. This allows you to study the time dependence of physical
properties (for example, volumes) of glass subjected to a change in temperature.
The glass transition is a region of temperature in which molecular rearrangements occur on a scale of minutes or hours, so
that the properties of a liquid change at a rate that is easily observed. Below the glass transition temperature T g , the material
is extremely viscous and a solidus state exists. Above T g , the equilibrium structure is arrived at easily and the material is in
liquidus state. Hence, the glass transition is revealed by a change in the temperature dependence of some property of a
liquid during cooling. If a mechanical stress is applied to a liquid in the transition region, a time-dependent change in
dimensions results due to the phenomenon of visco-elasticity.
If a liquid in the transition region is subjected to a sudden change in temperature, a time-dependent change in volume
occurs as shown in Figure 5-79. The latter process is called structural relaxation. Hence, structural relaxation governs the
time-dependent response of a liquid to a change of temperature.
Suppose a glass is equilibrated at temperature T 1 , and suddenly cooled to T 2 at t 0 . The instantaneous change in volume
is  g  T 2 – T 1  , followed by relaxation towards the equilibrium value V   T 2  . The total change in volume due to the
temperature change is  1  T 2 – T 1  as shown in Figure 5-79b. The rate of volume change depends on a characteristic time
called the relaxation time.
The slope of dV  dT changes from the high value characteristic of the fluid  1 to the low characteristic of the glass  g
as shown in Figure 5-80. The glass transition temperature T g is a point in the center of the transition region. The
low-temperature slope  g represents the change in volume V caused by vibration of the atoms in their potential wells. In
the (glassy) temperature range, the atoms are frozen into a particular configuration. As the temperature T increases, the
atoms acquire enough energy to break bonds and rearrange into new structures. That allows the volume to increase more
rapidly, so  1   g . The difference  =  1 –  g represents the structural contribution to the volume.
T1
T(t)
T2
t0 t
(a) Step Input for Temperature
V(0,T1)
g(T2-T1)
V(0,T2) l(T2-T1)
V(,T2)
t0 t
(b) Volume Change as Function of Temperature
Figure 5-79 Structural Relaxation Phenomenon

CHAPTER 5 387
Material Library
V(T)
V(T0) l
Liquid
State
V(T1)
g Transition Range
Solidus
State
T0
T2 T1 Tg Tf(T1) Tf (T1): Fictive Temperature
Figure 5-80 Property (Volume) – Temperature Plot
When a liquid is cooled and reheated, a hysteresis is observed as shown in Figure 5-81.
V
Equilibrium
Nonequilibrium
Tg T
Figure 5-81 Volume Change During Cyclic Temperature History
Unfortunately, the notion of a glass transition temperature is insufficient as real glassy materials generally exhibit a
temperature regime, called a transition range, across which their bulk properties gradually change from being solid-like to
liquid-like in nature.
As discussed earlier, properties have a time dependence in the transition range. An explanation for the strong time
dependence lies in that the material resides at a nonequilibrium temperature which lags behind the applied temperature
during the heating-cooling cycle. The nonequilibrium temperature is called the fictive temperature, T f , as shown in
Figure 5-80. The fictive temperature at T 1 T f  T 1  is found by extrapolating a line from V  T 1  with slope  g to intersect
a line extrapolated from V  T 0  with slope  1 (see Figure 5-80). For T  T 2 (well below the glass transition), T f reaches
a limiting value that is called T g . If the material were equilibrated at T f  T 1  , then instantaneously cooled to T 1 , it would
change along the line with slope  g because no structural rearrangement could occur. Therefore, it would have the same
volume as the continuously cooled sample.
The response of the volume change can be described by:

V  T 2 ,t  = V  T 1 ,  +  g  T 2 – T 1  +   l –  g   T f  t  – T 2  (5-410)
where T f  t  is the current value of the fictive temperature. The response function, M v , which dictates the value of the
fictive temperature is assumed to be linear in its argument and governs both the value of the fictive temperature as well as
the material property of interest.
Vt – V Tf  t  – T   
-------------------------------- = M v    t   = -------------------------------
- (5-411)
V0 – V T0 – T
By virtue of its linearity, Boltzmann’s superposition principle can be invoked to calculate the fictive temperature at any time:
t
d
Tf t  = T t  –  M v    t  –   t   -------(T  t )  dt' 
dt
(5-412)
–
The concept of reduced time,   t  is introduced in the spirit of Thermo-Rheologically Simple materials to capture the
disparate nonlinear response curves on a single master curve.
The reduced time used in Marc is given by the following expression:
t
 r ef
t =  -------------------
- dt (5-413)
  T  t  
–
Here  ref is the reference relaxation time of the material evaluated at a suitable reference temperature, T ref . The relaxation
time  at the given time and temperature can be represented as:
H 1 x 1 – x
 =  r ef  exp  – ---- --------- – --- – -----------------  (5-414)
 R T T Tf 
ref
The parameter x allows you to dictate how much of the fictive temperature participates in the prescription of the relaxation
time, and must, therefore, range between 0 and 1.
H is the activation energy for the particular process and R is the gas constant. A typical response function is:

M v    = exp  – -- (5-415)
 
Multiple structural relaxation times can exist. The response function has, therefore, been implemented as:
n

Mv    =   W g  i  exp  – ----
 
i
i = 1
For a complete description of the model, it is necessary to prescribe the following:
1. The weight  W g  i for each term in the series (usually   W g  i  1 ).
2. The reference relaxation times  i ref .
CHAPTER 5 389
Material Library
3. The fraction parameter x and the activation energy-gas constant ratio.

4. The solid and liquid coefficients of thermal of expansion,  g and  1 through the VISCEL EXP option.
A stable algorithm is employed to calculate the convolution integrals. For improved accuracy it is recommended that the
time steps used during the simulation be sufficiently small.
In Figure 5-82, the volume of cube of material, which is allowed to contract freely and is experiencing a 100oC quench, is
displayed.
T=100
Volume
gT
lT
Temperature
e
Tim
Figure 5-82 Volume-Temperature-Time Plot
Frequency-dependent Material Behavior

Many solid materials exhibit viscoelastic behavior when they are being deformed. If the load application is slow and the
loads are static and remain constant for a longer period of time, the material behavior relaxes to its long term equilibrium
behavior and there is no need to include rate effects in the material description. However, if the loading is dynamic, the
rate effects become noticeable even if inertia effects may be ignored. In particular, when the excitation is periodic a steady
state is reached where the stiffness and damping aspects of the material have to be considered.
The stiffness and damping aspects of a viscoelastic material, when observed in the frequency domain, are often
characterized by the storage and loss moduli, which in general are frequency dependent. The storage modulus characterizes
the frequency dependent stiffness and the loss modulus characterizes the frequency dependent damping behavior of the
material.
When the vibration amplitudes are small, the dynamic effects can be linearized around the static deformation state, and
harmonic sub-increments can be carried to study the dynamic behavior in the frequency domain as a linear perturbation
around this static equilibrium state. If the static deformation is nonlinear, this dynamic response also depends on the
nonlinear static deformation state.
The effects of frequency on linear elastic as well as nonlinear hyperelastic materials are discussed in the Viscoelastic Material
Behavior in the Frequency Domain section.
For filled rubbers, the effect of the vibration amplitude on storage and loss modulus, known as the Payne effect or Fletcher-
Gent effect, can be taken into account. The Payne effect is discussed in the Amplitude Dependent Relaxation in the Frequency
Domain section.
In general, viscoelastic behavior is temperature dependent. The effects of temperature on the storage and loss modulus can
be taken into account by defining a shift function assuming thermo-rheologically simple material behavior. This is
discussed in the Thermo-Rheologically Simple Material Behavior in the Frequency Domain section.
Sometimes the storage and loss modulus depend on the static pre-deformation. This can be taken into account by defining
a so called complex correction factor. This is discussed in the Deformation Dependent Relaxation in the Frequency Domain section.
In summary the most important factors that may influence the storage and loss modulus of a material are:
1. The vibration frequency
2. The vibration amplitude
3. The temperature
4. The amount of static pre-deformation
The finite element formulation of the harmonic behavior is summarized in the Harmonic Equations of Motion section. If there
is no Payne effect present, the harmonic equations are linear, but if the Payne effect is included in the formulation, they
become nonlinear and must be solved in an iterative fashion.
Choosing a certain material model for the storage and loss behavior of a material and evaluating its material constants
through a fitting procedure can be an involved process. It may often be circumvented by defining the storage and loss
modulus directly in tabular form. This is discussed in detail in the Using Tables to Define Storage and Loss Moduli section.
The steps needed to set up an analysis including frequency dependent material behavior are summarized in the Performing
Viscoelastic Analysis in the Frequency Domain section.
Viscoelastic Material Behavior in the Frequency Domain

This section focusses on the effects of frequency on the storage and loss modulus of a viscoeleastic material. It does not
include the effects of temperature, amplitude and static pre-deformation. It first discusses linear material behavior followed
by a discussion on nonlinear material behavior. It then focuses on isotropic material behavior where the deviatoric and
volumetric behavior can be uncoupled and touches on some aspects related to incompressible or nearly incompressible
material behavior. The section concludes by establishing the relation to the Morman Phi-coefficients model.
Linear Material Behavior

When the material behavior is linear, its viscoelastic behavior in the time domain can be represented by hereditary integrals
of the following form (summation is implied over repeated indices)
0  dg  s  0
 ij  t  = g    G ijkl  k l  t  + 0 -------------
- G i j k l   k l  t – s  –  kl  t   ds (5-416)
ds
where
i j is the stress tensor

 ij is the strain tensor
CHAPTER 5 391
Material Library
0
G i j kl is the instantaneous (short term) elasticity tensor
t is the time parameter
s is the history parameter
gt is the normalized relaxation function (i.e., g  0   1
g is the long term value of the relaxation function (i.e., g  t  for t   )
In this integral expression, we have made the assumption that  i j  t  = 0 , when t  0 . The history parameter s can be
read as “time units ago” extending from the present  s = 0  over the entire deformation history  s    .
The relaxation function is often represented by a Prony series:
N t-
----
k k
gt = g +  
g e  (5-417)
k = 1
where,
k is the relaxation time for the kth Prony term

gk is the factor of the kth Prony term
N is the number of terms in the Prony series
g is the long term value of the Prony series
N
and the factors have been normalized such that g +  gk = 1 .
k = 1
When there are harmonic excitations, the strain history can be represented with the help of complex arithmetic as:
 ij  t  = ̂ ij e it (5-418)
with,
̂ ij is the complex strain amplitude

 is the excitation frequency (radians per time)
t is the time
i is the imaginary unit  i 2 = – 1 
Substituting this strain history in the hereditary integral of equation (5-416), we obtain a harmonic stress response that, in
the same way as the strain history, can be written as:
 i j  t  = ̂ ij e it = G i j k l   ̂ k l e i  t (5-419)

with,
̂ i j is the complex stress amplitude

G i j kl    is the complex stiffness modulus tensor
The complex stiffness is given as

 
G i j kl    = g    +   h  s  sin  s  ds + i  h  s  cos  s  ds G ij0kl (5-420)
0 0
which follows from the hereditary integral in equation (5-416) when employing partial integration. Here, we have also
introduced the function h  s  = g  s  – g    for s  0 , and we have used the fact that e i  t = cos  t  + i sin  t  .
The real part of the complex stiffness modulus is called the storage modulus of the material and represents the frequency
dependent stiffness properties of the material:
s tor age  0

G i j kl    = Re  G i j k l     = g    +   h  s  sin  s  ds G i j kl (5-421)
0
The imaginary part of the complex stiffness modulus is called the loss modulus of the material and represents the frequency
dependent damping properties of the material:
lo ss  0
G i j kl    = Im  G ijkl     =   h  s  cos  s  ds G i j kl (5-422)
0
st o r ag e l o ss
The complex stiffness modulus, therefore, is G i j k l    = G i j kl    + iG i j kl    and the stress-strain relation in the
frequency domain is formulated as:
0
̂ i j = G ijkl   ̂ kl = g̃   G i j k l ̂ k l (5-423)
where g̃    is the transform to the frequency domain of the normalized relaxation function g  t  (i.e. g̃    = 1 for
   ; also note that for   0 , we obtain the long term static response, since g̃  0  = g    ), which can be given as
 
g̃    = g    +   h  s  sin  s  ds + i  h  s  cos  s  ds (5-424)
0 0
When the relaxation function is given by a Prony series, we obtain explicit expressions for the storage and loss modulus as:
N
st ora ge gk2 2 0
G ijk l  = g +  ----------------------k2- G i j k l (5-425)
1 +  2 k
k = 1
and
N
lo ss g  k   k 0
G i j kl    =  ----------------------
1 +  2  k2
- Gi j k l (5-426)
k = 1
The energy dissipated per cycle and unit volume of material is the work done by the stresses over one vibration cycle of the
deformation history, and with the above stress and strain history representations, this is computed as follows:
CHAPTER 5 393
Material Library
2
------
 d i j
WD = 0 Re   ij  t  Re --------- dt (5-427)
dt
When we write the complex strain amplitude as:
r i
̂ ij = ̂ ij + i̂ ij (5-428)
and the complex stress amplitude as:
r i
̂ i j = ̂ ij + i̂ ij (5-429)
the dissipated energy evaluates to

i r r i
W D =   ̂ ij ̂ ij – ̂ i j ̂ i j  (5-430)
The average heat production per unit time and unit volume is obtained when we divide this energy by the time period
2   of the cycle
i r r i
Q =   ̂ ij ̂ ij – ̂ ij ̂ i j   2 (5-431)
Linear isotropic viscoelastic material behavior in the frequency domain is available through the ISOTROPIC and VISCELPROP
model definition options in combination with the HARMONIC parameter. Linear orthotropic material behavior is discussed
in Linear Orthotropic Material Behavior. It is essential that complex damping gets activated on this parameter, since for any
nonzero frequency the material behavior will include damping contributions. The instantaneous (short term) stiffness
properties are entered on the ISOTROPIC model definition option. The viscoelastic properties are entered on the VISCELPROP
model definition option. The viscoelastic properties can be entered by Prony series, as for viscoelastic analysis in the time
domain. Marc will then evaluate the storage and loss modulus with the help of equations (5-425) and (5-426) as a function
of frequency. The storage and loss modulus can also be entered directly in tabular form using the TABLE option with either
the frequency or the log-frequency (with base 10) as the independent variable. Although the instantaneous properties are
entered on the ISOTROPIC option, Marc uses the long term properties in the static loadcases that precede harmonic loadcases.
Therefore, if tables are used for the storage and loss modulus, it is important that Marc can evaluate proper values in the
tables at zero frequency. When tables are used for the storage and loss modulus, the requirement that the static stiffness
properties must be the instantaneous stiffness properties can be relaxed. This is discussed in detail in the Using Tables to Define
Storage and Loss Moduli section.
The average heat production per unit time and unit volume as expressed by equation(5-431) is available on the post file by
selecting element post-code 620 on the POST model definition option.
Linear Orthotropic Material Behavior

Although the relations of the previous section not necessarily imply isotropic material behavior, they do not provide the
most general description of orthotropic material behavior. The orthotropic viscoelastic material tensor that can be entered
in the time domain is a Prony series expansion of following form:
N t
– -----
0
  n
G i j kl  t  = G ijkl –  G ijnkl  1 – e  (5-432)
 
n = 1
0
Here G ijkl is the short term orthotropic stress-strain law based on the material constants that are entered on the
n
ORTHOTROPIC option and G i j k l are the orthotropic contributions for the nth term in the Prony series expansion based on
the material constants that are entered on the VISCELORTH option. The complex frequency dependent orthotropic material
tensor of the previous section is the transformation of the following function in the time domain
0
G i j kl  t  = g  t G ijkl (5-433)
Here g  t  is the normalized relaxation function (i.e., g  0  = 1 ). If g  t  would be approximated by a Prony series this
would lead to the following expression in the time domain
N t
– -----
  n 0
G i j kl  t  = 1–  gn 1 – e  Gi j k l (5-434)
 
n = 1
It may be observed that equation (5-434) is of less general nature than equation (5-432).
The complex orthotropic stress-strain law in the frequency domain based on the time domain Prony series expansion of
equation (5-432) is
N N
0 n 1 n  n
G ijk l    = G ijkl –  G i j k l ----------------------2- + i  G i jk l ----------------------2- (5-435)
1 +  2 n 1 +  2 n
n = 1 n = 1
The complex orthotropic stress-strain law in the frequency domain based on the time domain expression of equation (5-433)
can be represented as:
0
G i j kl    =  g s torag e    + ig l o ss    G i j k l (5-436)
A general complex orthotropic stress-strain law in the frequency domain can be represented as:
sto ra ge l os s
G i j kl    = G ijkl    + iG i j k l    (5-437)
s t or a ge st o ra g e st o ra g e
The orthotropic storage part G i j k l    is computed from orthotropic Young’s moduli E xx    , E yy  ,
sto rag e s t o ra g e st o ra g e st o ra g e
and E zz    ; orthotropic Poisson’s ratios v x y    , v yz    , and v zx    ; and orthotropic shear
s tor age s t o ra g e s t o ra g e
moduli G xy    , Gy z    , and G z x    where these material parameters are specified as a function of
l o ss l os s
frequency. Similarly, the orthotropic loss part G i j k l    is computed from orthotropic Young’s moduli E xx    ,
CHAPTER 5 395
Material Library
loss lo ss l os s l os s l os s
E yy    , and E z z    ; orthotropic Poisson’s ratios v xy    , v yz    , and v zx    ; and orthotropic shear
lo ss lo s s loss
moduli G xy    , G yz    , and G z x    where these material parameters are specified as a function of frequency.
The precise form of the G i j k l tensors depends on the number of direct (NDI) and shear components (NSHEAR) of stress
and follows the classification given in Table 5-1. The inverted forms from this table for the direct components of the two
nontrivial cases are summarized in Table 5-9. The inverted forms for NDI=1 and the shear terms are trivial and are not listed
here.
Table 5-9 Direct Stress-direct Strain Orthotropic Relations

NDI=2
  xx  1 Ex x vy x Ex x  x x 
  = ---
-
D v E  
  yy  x y y y Ey y  y y 
with
D = 1 – vx y vy x
and the symmetry relation
Ey y
v yx = v xy --------
Ex x
NDI=3
   
  xx   1 – v y z v z y E x x  v y x + v yz v zx E xx  v zx + v zy v yx E xx   xx 
  1  
  yy  = ---
D
-  v + v v E
xy x z z y y y  1 – v zx v xz E yy  v zy + v zx v xy E yy   yy 
   
 z z   v x z + v x y v y z E z z  v y z + v yx v xz E zz  1 – v xy v yx E zz   zz 
   
with
D = 1 – v x y v y x – v y z v z y – v z x v x z – v x y v yz v zx – v yx v zy v xz
and the symmetry relations
Ey y Ez z , E xx
v yx = v xy -------- , v z y = v y z -------
- v x z = v zx --------
Ex x Ey y E zz
Linear orthotropic viscoelastic material behavior in the frequency domain is available through the ORTHOTROPIC and
VISCELORTH model definition options in combination with the HARMONIC parameter. It is essential that complex damping
gets activated on this parameter, since for any nonzero frequency the material behavior will include damping contributions.
The instantaneous (short term) stiffness properties are entered on the ORTHOTROPIC model definition option. The
viscoelastic properties are entered on the VISCELORTH model definition option. The viscoelastic properties can be entered by
Prony series, as for viscoelastic analysis in the time domain. Marc will then evaluate the complex stress-strain law with the
help of equation (5-435) as a function of frequency. The storage and loss information can also be entered directly in tabular
form using the TABLE option with either the frequency or the log-frequency (with base 10) as the independent variable. Two
different methods are available to use tabular input and the method for each material is selected on the VISCELORTH model
definition option. With the first method one table is entered for the normalized storage modulus and one table is entered
for the normalized loss modulus to obtain the complex stress-strain law representation of equation (5-436). With the second
method a storage and a loss table is entered for each Young’s modulus, Poisson’s ratio and shear modulus that appears in
the material description to obtain the general complex stress-strain law representation of equation (5-437). Although the
instantaneous properties are entered on the ORTHOTROPIC option, Marc uses the long term properties in the static loadcases
that precede harmonic loadcases. Therefore, if tables are used for the storage and loss modulus, it is important that Marc
can evaluate proper values in the tables at zero frequency. When tables are used for the storage and loss modulus, the
requirement that the static stiffness properties must be the instantaneous stiffness properties can be relaxed. This is
discussed in detail in the Using Tables to Define Storage and Loss Moduli section. The material parameters entered on the
ORTHOTROPIC and VISCELORTH options define the behavior in the material (preferred) directions. The precise alignment in
space of these material directions is defined with the ORIENTATION model definition option. The COMPOSITE model definition
option can be used to stack orthotropic materials using different orientations inside shell elements.
The average heat production per unit time and unit volume is again given by equation (5-431) and is available on the post
file by selecting element post-code 620 on the POST model definition option.
In addition to the three direct input methods, it is possible to enter a general orthotropic or anisotropic material behavior
with the help of user subroutines HOOKLW, ANELAS, and HOOKVI. The instantaneous (short term) behavior is defined with
HOOKLW and/or ANELAS. The frequency dependent behavior is defined with HOOKVI. The way the user subroutines are
processed depends on the input method selected on the VISCELORTH model definition option and is summarized in
Table 5-10.
CHAPTER 5 397
Material Library
Table 5-10 User Subroutine Processing when defining Frequency Dependent Orthotropic and
Anisotropic Materials
Input Method User Subroutine Processing
Prony series In static loadcases, HOOKLW and/or ANELAS are used to define the short term material behavior and
HOOKVI is used for each term (ITERM>0) in the Prony series expansion to facilitate the evaluation
of the long term (i.e. at zero frequency) material behavior.
In harmonic loadcases, HOOKLW and/or ANELAS are used to define the short term material behavior
and HOOKVI is used at each frequency for each term (ITERM>0) in the Prony series expansion. This
allows the definition of general storage and loss contributions as represented by equation (5-435).
First table input In static loadcases, HOOKLW and/or ANELAS are used to define the short term material behavior. The
method returned stress-strain law is scaled by the normalized storage table value at zero frequency and
HOOKVI is not used.
In harmonic loadcases, HOOKLW and/or ANELAS are used to define the short term material behavior.
The returned stress-strain law is scaled by the normalized storage table value at the current
frequency to obtain the storage contribution and by the normalized loss table value at the current
frequency to obtain the loss contribution. Again, HOOKVI is not used. This allows the definition of
general storage and loss contributions as represented by equation (5-436).
Second table input In static loadcases, HOOKVI is called once (ITERM=-1) to define the long term (i.e. at zero frequency)
method material behavior and HOOKLW and ANELAS are not used.
In harmonic loadcases, HOOKVI is called at each frequency once for the storage contribution
(ITERM=-1) and once for the loss contribution (ITERM=-2). Again, HOOKLW and ANELAS are not
used. This allows the definition of general storage and loss contributions as represented by equation
(5-437).
Nonlinear Material Behavior

When the material behavior is nonlinear, the viscoelastic model of Simo [Ref. 20] for finite strain viscoelasticity is used. The
dynamic behavior is obtained by applying this model to small harmonic dynamic motions around a nonlinear static
equilibrium state and linearizing the model around this static equilibrium state. The resulting model describing the
frequency response is referred to as the Linearized Simo Model [Ref. 37].
The stress response due to a given strain history can be given by a hereditary integral of the form:
0  dg  s  0 0
S i j  t  = g    S ij  t  + 0 -------------
-  S i j  t – s  – S i j  t   ds (5-438)
ds
where,
S ij is the second Piola Kirchhoff stress

0
S ij is the instantaneous second Piola Kirchhoff stress
t is the time parameter
s is the history parameter
gt is the normalized relaxation function (i.e., g  0  = 1 )
g is the long term value of the relaxation function (i.e., g  t  for t   )
The instantaneous second Piola Kirchhoff stress is derived from a strain energy function as:
0 W 0  E ab  t  
S ij  t  = --------------------------------
- (5-439)
E ij
where,
W0 is the instantaneous strain energy function

Ei j is the Green-Lagrange strain
When we consider small dynamic motions around a nonlinear equilibrium state, the strain history may be represented as:
E i j  t  = E ij* + E ij  t  (5-440)
where we make the convention that terms with an asterisk (*) denote terms referring to the static equilibrium state.
Substituting this strain history in the hereditary integral of equation (5-438) and linearizing, we obtain a stress response of
the form:
S ij  t  = S ij* + S ij  t  (5-441)
with,
0*  dg  s  0*
S i j  t  = g    K ijkl E k l  t  + 0 -------------
- K i j k l  E k l  t – s  – E kl  t   ds (5-442)
ds
and
0*  2 W 0  E ab * 
K ijk l = ----------------------------
- (5-443)
E ij E kl
CHAPTER 5 399
Material Library
0*
We observe that K ijkl is the instantaneous stiffness of the material in the static equilibrium state. If we now consider
harmonic deformations of the form:
E i j  t  = Ê ij e it (5-444)
we get a harmonic stress response of the form:
S ij  t  = Ŝ ij e it (5-445)
with the harmonic stress-strain relation:

0*
Ŝ ij = K ijkl   Ê kl = g̃   K i j k l Ê k l (5-446)
where Ê ij and Ŝ ij are the complex amplitudes of the Green-Lagrange strain increment and the second Piola Kirchhoff
stress increment, respectively, and g̃    is the transform of the normalized relaxation function as given in equation (5-424).
This type of viscoelastic behavior in the frequency domain is available for the following nonlinear elastomeric materials:
MOONEY, GENT, ARRUDBOYCE, MARLOW, OGDEN, and FOAM, where these model definition options specify the instantaneous
material response. Viscoelastic response in the frequency domain can be analyzed for these materials by specifying the
corresponding viscoelastic material data through either the VISCELMOON, VISCELOGDEN, or VISCELFOAM options in
combination with the HARMONIC parameter. It is essential that complex damping gets activated on this parameter; since for
any nonzero frequency, the material behavior will include damping contributions. The viscoelastic properties can be
entered by Prony series, as for viscoelastic analysis in the time domain. Marc then evaluates the storage and loss modulus
with the help of (5-425) and (5-426) as a function of frequency. The storage and loss modulus can also be entered directly in
tabular form using the TABLE option and choosing either the frequency or the log-frequency (with base 10) as the
independent variable. Although the instantaneous properties are entered on the aforementioned model definition options,
Marc uses the long term properties in the static loadcases that precede harmonic loadcases. Therefore, if tables are used for
the storage and loss modulus, it is important that Marc can evaluate proper values in the tables at zero frequency. When
tables are used for the storage and loss modulus, the requirement that the static stiffness properties must be the
instantaneous stiffness properties can be relaxed. This is discussed in detail in the Using Tables to Define Storage and Loss Moduli
section.
Alternatively, the user can define his own strain energy function through the user subroutines UENERG (Mooney material
only using the total Lagrange framework) or UELASTOMER in combination with viscoelastic material input through the
appropriate viscoelastic material model definition option and the HARMONIC parameter. In this case, the user defines the
instantaneous (short term) response in the user subroutine. Marc evaluates the long term static properties based on the
viscoelastic material data and uses these in static loadcases preceding the harmonic loadcases.
Nonlinear static loadcases preceding a harmonic loadcase can be carried out using either the total Lagrange or the updated
Lagrange framework and this is controlled by the settings made on the LARGE STRAIN parameter. The stress and strain results
of the harmonic loadcase depend on the framework being used in the analysis. Within the total Lagrange framework, the
harmonic strain amplitudes Ê i j are increments of the Green-Lagrange strain and their conjugate stress amplitudes Ŝ i j are
increments of the second Piola Kirchhoff stress. Within the updated Lagrange framework, the harmonic strain amplitudes
̂ ij are objective increments of the logarithmic strain and their conjugate stress amplitudes ̂ i j are objective increments of
the Cauchy stress.
These increments are related to each other by push forward and pull back operations as:
*–1 *–1
̂ ij = F pi F qj Ê pq
1 * * (5-447)
̂ i j = ------ F ir F js Ŝ r s
J*
The harmonic stress-strain law within the total Lagrange framework is given in equation (5-446). Its counterpart within the
updated Lagrange framework can be given as
0*
̂ i j = L ijkl   ̂ kl = g̃   L i j k l ̂ k l (5-448)
with,
0* 1 * * * * 0*
L i j kl = ----- F ip F jq F kr F l s K p qr s (5-449)
J*
0*
Here, F ij* is the deformation gradient of the static deformation, J * its determinant, and K i j kl is the short term stiffness
of the material in the static equilibrium state as given in equation (5-443).
The average heat production per unit time and unit volume is given by a similar expression as equation (5-431), but now
using the harmonic stress and strain amplitudes associated with the chosen analysis framework. It is available on the post
file by selecting element post-code 620 on the POST model definition option. The static pre-stress does not contribute to
this heat production and it should be noted that depending on the chosen analysis framework this heat production is per
unit undeformed volume (total Lagrange) or per unit deformed volume (updated Lagrange).
Uncoupled Deviatoric and Volumetric Response

When the deviatoric and volumetric response is fully uncoupled, it is possible to specify viscoelastic data for each
deformation mode, so the viscoelastic input consists of relaxation data for the deviatoric response and relaxation data for
the volumetric response. In this case, the frequency response for both deformation modes is also uncoupled.
For linear isotropic materials the harmonic stress-strain law reads
̂ ij = 2G   ê ij + K   ̂ v  i j (5-450)
where,
ê ij is the complex amplitude of the deviatoric strain

̂ v is the complex amplitude of the volumetric strain
G is the complex shear modulus
K is the complex bulk modulus
The complex shear and bulk modulus are given as:
G    = g̃ D   G  0  (5-451)
CHAPTER 5 401
Material Library
K    = g̃ V   K  0  (5-452)
where g̃ D    and g̃ V    are the transforms to the frequency domain of the deviatoric and volumetric normalized
relaxation functions, respectively, and G  0  and K  0  are the instantaneous shear and bulk moduli, respectively.
For nonlinear materials, the strain energy function is assumed to be the sum of a volumetric part  W 0 V  and a deviatoric
part  W 0 D  .
W 0 = W 0 V  J  + W 0 D  E i j  (5-453)
where,
J is the determinant of the deformation gradient

Ei j is the deviatoric part of the Green-Lagrange strain
The deviatoric part of the Green-Lagrange strain is based on the multiplicative split of the deformation gradient into a
volumetric and a deviatoric part as
F i j = J 1 / 3 F ij (5-454)
and is given as:

1
E i j = ---  F ki F kj –  ij  (5-455)
2
The total second Piola Kirchhoff stress can be written as:
V D
S ij  t  = S ij  t  + S ij  t  (5-456)
where,
V
S ij is the volumetric part of the second Piola Kirchhoff stress
D
S ij is the deviatoric part of the second Piola Kirchhoff stress
where the volumetric part of the second Piola Kirchhoff stress can be written further as
V –1
S ij  t  = – J  t p  t C ij  t  (5-457)
where,
p is the hydrostatic pressure

–1
C ij is the inverse of the right Cauchy-Green tensor
For both the relaxation of the hydrostatic pressure and the relaxation of the deviatoric part of the second Piola Kirchhoff
stress, we may now arrive at independent hereditary integrals describing the volumetric viscoelastic behavior
 dg V  s 
p  t  = g    V p 0  t  + -  p 0  t – s  – p 0  t   ds
0 ---------------- (5-458)
ds
and the deviatoric viscoelastic behavior
D 0  dg D  s  0 0
S ij  t  = J  t  – 2 / 3 DEV g    D S i j  t  + 0 -----------------
-  S i j  t – s  – S i j  t   ds (5-459)
ds
Here, g V  t  and g D  t  are the normalized relaxation functions for the volumetric and the deviatoric behavior,
respectively, and the DEV-operator used in equation (5-459) is defined as:
1 –1
DEV  S ij  = S ij – --- C p q S p q C i j . (5-460)
3
The instantaneous stress terms that arise in equations (5-458) and (5-459) are found from the volumetric and deviatoric parts
of the instantaneous strain energy function as:
W 0 V  t 
p 0  t  = ------------------------ (5-461)
J
and
0 W 0 D
t
S ij  t  = DEV  ------------------------ (5-462)
 E 
ij
For small dynamic motions around a static equilibrium state, we may linearize the stress expressions given in equations
(5-456), (5-457), (5-458), and (5-459) and when we assume harmonic deformations, we arrive at similar harmonic stress-strain
relations as in equation (5-446), but now one for the volumetric response and one for the deviatoric response. For the
volumetric response, we arrive at
V V* V  V p
S ij  t  = S ij + S ij  t  + S i j  t  (5-463)
with,
V  0 V *  dg V  s  0 V *
S i j  t  = g    V K i j k l E k l  t  + 0 ----------------
- Ki j k l  E kl  t – s  – E kl  t   ds (5-464)
ds
where,
0 V *  2 W 0 V * * – 1 * – 1
K ijk l = J *2 ---------------------- C i j C k l (5-465)
J 2
and
V p 0 V p*
S i j  t  = g    V K i j k l E k l  t  (5-466)
where,
W 0 V * * – 1 * – 1
0 V p * = J * -------------------
K ijk l  C i j C k l –  C i*k– 1 C j*l– 1 + C i*l– 1 C j*k– 1   (5-467)
J
CHAPTER 5 403
Material Library
For the deviatoric response, we arrive at:

D D* D  D S
S ij  t  = S ij + S ij  t  + S i j t (5-468)
with,
D  0 D *  dg D  s  0 D *
S ij  t  = g    D K i j k l E k l  t  + 0 -----------------
- K i j kl  E kl  t – s  – E kl  t   ds (5-469)
ds
where,
0 D * * – 4---  2 W 0 D* 1--- * W 0 D* *–1 *–1 *–1 *–1

K ijk l = J 3 ----------------------- + C p q --------------------  C i k C j l + Ci l Cj k  +
E ij E k l 3 E p q
1  *  2 W 0 D* *  2 W 0 D* * – 1 * – 1
--- C pq ------------------------ C r s – 2C p*q ----------------------- C ij C kl – (5-470)
9 E pq E r s E p q 
1  – 1 *  2 W 0 D* *–1  2 W 0 D *
--- C ij C pq ------------------------ + C k l C p*q -----------------------
3 E p q E k l E p q E i j
and
D S 0 D S*
S ij  t  = g    D K i j k l E k l  t  (5-471)
where
0 D S * 2 0 D* * – 1 * – 1 0 D*
K ijk l = – ---  S ij Ck l + Ci j Sk l  (5-472)
3
The complex stiffness modulus can now be given as
0 V * 0 D * 0 V p * 0 D S *
K ijk l    = g̃ V   K ij k l + g̃ D   K i j k l + g    V K i j kl + g    D K i j kl (5-473)
Many materials show viscoelastic response only for the deviatoric deformations and the volumetric deformations are time
independent. In that case only deviatoric viscoelastic data is specified. This is in particular the case for materials that are
incompressible or nearly incompressible (also see Incompressible and Near Incompressible Response) for which there is no
viscoelastic volumetric response.
A material specified through the FOAM model definition option does not have the uncoupling of deviatoric and volumetric
response as discussed in this section, but the other elastomeric materials and the linear elastic materials do have this
uncoupling. These materials support relaxation behavior in their deviatoric as well as in their volumetric response; and
therefore, they support uncoupled frequency dependent stiffness and damping behavior in their harmonic deviatoric and
harmonic volumetric response. The storage and loss factors are available on the post file by selecting post codes 861 for the
storage factor of the deviatoric or coupled response, post code 862 for its loss factor, post code 863 for the storage factor of
the volumetric response and post code 864 for its loss factor.
Incompressible and Near Incompressible Response

If an elastic material is incompressible or nearly incompressible its incremental stress-strain law becomes singular. For
example, for a linear elastic isotropic material the Poisson’s ratio approaches ½ and it is easily verified that the resulting
stress-strain law is singular. If it is attempted to enforce incompressibility in an approximate manner by using a high bulk
modulus, there is not only a risk of singularity, but there is also a risk of volume locking. Special techniques have to be
employed to overcome the difficulties that arise as a result of the incompressibility or near incompressibility constraint. In
Marc, special elements using the Herrmann formulation can be used in this case. These elements can also be employed in
a harmonic analysis. A harmonic analysis with viscoelastic material properties can thus be carried out using standard
elements (for example, element type 7) or using special elements with the Herrmann formulation (for example, element
type 84). It should be noted that the standard elements also use special formulations to overcome the difficulties associated
with incompressibility, but in contrast to the elements using the Herrmann formulation, they do not have special external
degrees of freedom.
Relation to the Morman Phi-coefficients Model

An alternative approach using the PHI-COEFFICIENTS model definition option is available in Marc based on a model derived
by Morman [Ref. 35], but this approach is available for MOONEY material behavior only using the total Lagrange framework
(also see Chapter 5, Structural Procedure Library, Dynamics, Harmonic Response).
The model of Morman is based on the Finite Linear Viscoelasticity model of Coleman and Noll [Ref. 36]. Morman used
this model to describe the viscoelastic small amplitude vibration response around static large deformation equilibrium states
for incompressible rubber materials where the time and deformation behavior was assumed to be separable. The general
form of the Coleman and Noll model for isotropic (not necessarily incompressible) material behavior can be formulated as
follows by specifying the temporal behavior of the Cauchy stress
 1 t
 i j  t  =  ij  t  + 0 --2- E i j k l  B p q  t  s G k l  s  ds (5-474)
where,
t –1 –1
G i j  s  = F ki  t C kl  t – s F l j  t  –  i j (5-475)

This specifies the Cauchy stress  i j as the sum of an equilibrium stress  i j  t  and a functional that is linear in the non-
t
equilibrium deformation history G i j  s  . The latter tensor defines a deformation measure as the difference between the
past deformation at time t – s and the current deformation at time t . The strain measure for the past deformation is based
–1
on the right Cauchy-Green tensor C i j  t – s  and F i j is the inverse of the deformation gradient. The tensor E i j kl is a
fourth order tensor being a function of the time history parameter s and the current deformation, which is characterized
by the left Cauchy-Green tensor B i j  t  . If we investigate small motions around a static equilibrium state, we may
approximate the stress of equation (5-474) as:
  dg  s 
 ij* +  ij  t  =  ij* +  i j  t  + 0 ------------- *     t – s  –   t   ds
- E i j k l  B pq kl kl (5-476)
ds
CHAPTER 5 405
Material Library
where we have assumed the separability of time and deformation response, which becomes necessary if linear superposition
is assumed to hold for small dynamic deformations around the equilibrium state. The strain increments are defined in terms
of the displacement increments as:
1  v i  t  v j  t 
 ij  t  = ---  ------------------
- + ------------------- (5-477)
2  x * x i* 
j
Here x i* are the coordinates of the material points in the deformed configuration of the static equilibrium state. The
equilibrium Cauchy stress increment can be approximated as
 ij  = g     –  kk  i0j* + w i k  k0j* –  i0k* w k j +  i k  kj

0*
+  i0* 0*
k  kj + L i j kl  kl  = g
    0
ij (5-478)
where,
1  v i  t  v j  t 
w i j  t  = ---  ------------------
- – ------------------- (5-479)
2  x * x i* 
j
is the incremental rotation or spin tensor and

2 0 *
1 * * * *  W  Ea b 
L i0j*kl = ------ F ip F jq F kr F l s ----------------------------- (5-480)
J* E p q E r s
is the material stiffness tensor, which is equation (5-443) pushed forward to the deformed configuration of the static
equilibrium state. We may now define an objective increment of the equilibrium Cauchy stress as
 ij  T R =  ij   + g      k k  i0j* – w i k  k0j* +  i0* 0* 0*

k w kj –  i k  kj –  i k  kj 
(5-481)
0*
= g    L ijkl  k l
This objective stress increment can be viewed as a stress increment based on the Truesdell rate of the Cauchy stress. The
objective stress increment that arises as a natural result of the linearization of the Simo model can, thus, be viewed as the
objective Truesdell stress increment where the fourth order tensor in the integral of equation (5-476) is taken as the material
stiffness tensor of equation (5-480); i.e., E i j k l = L i0*
jkl
 dg  s 
0* 0*
 ijTR  t  = g    L ijkl  k l + - L  B *    kl  t – s  –  kl  t   ds
0 ------------- (5-482)
ds i j k l p q
The approach taken by Morman (however, he applies his model to incompressible materials in which case the trace term
of the incremental strain is absent) is to define an objective equilibrium stress increment as
 ij  Mo =  ij   + g      k k  i0j* – w i k  k0j* +  i0*

k w kj 
(5-483)
= g      i k  k0j* +  i0k*  k j + L i0j*k l  k l  = g    M i0j*kl  kl
In order for the short term response to have the same (but scaled) elastic stiffness properties as the long term response,
Morman defines the fourth order tensor in the integral of equation (5-476) as E i j kl = M i0j*kl and arrives at
 dg  s 
 iMo    M 0 *  + 0*
- M i j k l  B p*q    kl  t – s  –  kl  t   ds
j t = g ij k l kl 0 -------------
ds
(5-484)
Morman applies his model to incompressible hyperelastic materials in which case his objective stress increment can be
viewed as a stress increment based on the Jaumann rate of the Cauchy stress. The relation between the fourth order tensors
in equations (5-482)and (5-484) is seen to be given as
0* 0* 1 0* 0* 0* 0*
M ijk ---
l = L ijkl + 2   i k  j l +  i l  j k +  j k  i l +  j l  i k  (5-485)
The model in Marc has been slightly generalized by defining the stiffness tensor used in equation (5-476) as
0* 0* 1 0* 0* 0* 0*
E ijk ---
l = L ijkl +  2   i k  j l +  i l  j k +  j k  i l +  j l  i k  (5-486)
where  is a parameter that can be chosen between zero and one and is called the objective stress increment factor. For
 = 0 , we obtain the Linearized Simo Model, and for  = 1 , we obtain at the Morman model. The value of  can
be given on the VISCELHARM parameter and by default its value is taken as zero. The model of Morman is available for all
hyperelastic materials that support viscoelastic material behavior in the frequency domain by defining the input as for the
Linearized Simo Model and using the VISCELHARM parameter setting the value of  to  = 1 .
Amplitude Dependent Relaxation in the Frequency Domain

Filled rubbers, in contrast to unfilled natural rubbers, show a pronounced effect of the vibration amplitude on the storage
and loss moduli of the rubber; i.e., the vibration amplitude has a pronounced effect on the stiffness and damping properties
of the material. This effect is known as the Payne effect or the Fletcher-Gent effect. It is generally observed that the storage
modulus is high for very low amplitudes and decreases for increasing amplitudes to reach some asymptotic minimum at
high amplitudes, whereas the loss modulus reaches a maximum at some intermediate amplitude as roughly illustrated in
Figure 5-83. Although at the microscopic level, the Payne effect is observed to be caused by certain damage effects in the
microstructure these damage effects are largely recovered when the dynamic loading stops, so the deformation processes
accompanying the Payne effect are considered to be reversible. Therefore, it is different from the Mullins effect which
occurs in filled as well as in unfilled rubbers and where the stiffness changes are attributed to irreversible damage effects.
G’ G’’
log() log()
Figure 5-83 Storage Modulus (left) and Loss Modulus (right) as a Function of Logarithmic Vibration Amplitude
CHAPTER 5 407
Material Library
Most of the models in Marc for the Payne effect are based on rheological models using spring, dashpot and friction/slip
elements. They give an adequate phenomenological description of the Payne effect without considering any details of the
processes that may take place in the material at the micro level. However, the Kraus/Ulmer model is also available, which
was one of the earliest models giving a quantitative description of the Payne effect based on considerations of the processes
taking place at the micro level during harmonic vibrations at a constant amplitude.
The different models are described in the subsections below. The first model is based on a thixotropic approach. It uses a
viscoelasticity formulation based on hereditary integrals, but it includes process dependent viscosities meaning that the
relaxation times become process dependent. The basic rheological elements in this model are spring and dashpot elements.
The second model is based on a triboelastic approach. It uses a cyclic plasticity formulation where the hysteresis loop
determines the storage and loss modulus when the material is subject to a harmonic excitation. The basic rheological
elements in this model are spring and friction/slip elements. The two approaches can be combined in an additive sense and
in a multiplicative sense resulting in models that use spring, dashpot, and friction/slip elements as their basic rheological
elements. The data accounting for the Payne effect can also be entered directly in tabular form and a UPAYNE user subroutine
is available for alternative formulations not covered by the standard models mentioned earlier.
The Payne effect is only supported for isotropic materials, i.e. for linear isotropic materials defined through the ISOTROPIC
model definition option or for hyperelastic isotropic materials defined through the MOONEY, GENT, ARRUDBOYCE, MARLOW,
OGDEN or FOAM model definition options. It is not supported for orthotropic or anisotropic materials. In material models
where the deviatoric and volumetric response is uncoupled, the Payne effect can be present independently in both
deformation modes, so it can also be activated in the volumetric response. However, it is expected that in most practical
applications it will be active in the deviatoric response only, since most rubber materials are incompressible or nearly
incompressible and show no viscoelastic behavior in their volumetric response.
Definition of Amplitude in the Payne Effect Formulations

The amplitude used in all models describing the Payne effect is a scaled 2-norm of the true strain increment tensor
2
---  ij  ij , where  i j denotes the complex conjugate of  i j . This amplitude definition results in an equivalent
3
tensile strain amplitude. Note that, in a harmonic analysis, we are dealing with complex numbers. In a purely linear analysis,
there is no difference between the engineering strain increment and the true strain increment. In a total Lagrange analysis
the Green Lagrange strain increment E i j is transformed to the current static equilibrium configuration with the
deformation gradient of this static equilibrium configuration to obtain the true strain increment as
* –1 –1 –T –1 –1
 ij = F ki F lj* E k l or in tensor form  = F * EF * from which we can find  = C * E . The
terms with an asterisk (*) denote terms related to the static equilibrium state. In an updated Lagrange analysis, the true
strain increment  ij is the direct outcome of the analysis.
In case the deviatoric and volumetric response of the material is uncoupled, an amplitude for both deformation modes
needs to be defined. For the deviatoric response, the scaled 2-norm as given above is applied to the deviatoric part of the
1 2
strain increment tensor e i j =  i j – ---  k k  i j giving --- e i j e i j . Here,  kk is the trace of the true strain
3 3
increment tensor representing its volumetric part. For the volumetric response, we take the magnitude of the trace of the
true strain increment tensor  k k  l l . In the UAMPLT user subroutine, the user can define an alternative norm based on
the complex strain increment tensor. This subroutine is always called in a nonlinear harmonic analysis, so the user can
override the norm computed by Marc. The amplitudes are computed at each integration point of each element that uses
amplitude dependent material data and they can be written to the post file through the POST model definition option
choosing post code 698 for the amplitude of the deviatoric or coupled response, and post code 699 for the amplitude of
the volumetric response. Table 5-11 lists the amplitude (and the nonzero strain components) for the two most common
measures used in tests for measuring storage and loss moduli.
Table 5-11 Amplitude Measures of Common Tests

Type of test 
Tension or compression assuming incompressibility    11 =   22 =  33 = –   2 
Shear   3   12 =  21 =   2 
The Thixotropic Model

This model was developed by Lion and coworkers [Ref. 42] and is based on a viscoelasticity formulation that can be given
in terms of hereditary integrals resulting in an instance of the Finite Linear Viscoelasticity model of Coleman and Noll
[Ref. 36]. In Marc, the hereditary integral formulation is based on the Simo model for finite strain viscoelasticity [Ref. 20],
which reduces to a special case of the Finite Linear Viscoelasticity model of Coleman and Noll when small perturbations
around a nonlinear static equilibrium state are being considered. The ideas developed by Lion and applied by him to a finite
linear viscoelasticity model can also be applied to the Simo viscoelasticity model to introduce amplitude dependent
behavior in dynamic loading histories. First, a short overview is given of the Simo model, and then the ideas developed by
Lion are applied to this model to introduce the amplitude dependence. In the following sections, reference is made to a
general strain energy function and its accompanying stress and strain measures. If the material formulation uncouples its
deviatoric and volumetric response, these references are to the strain energy function and the stress and strain measures of
the respective deformation mode.
Overview of the Simo Viscoelasticity Model

In rheological terms, the model of Simo can be viewed as an extension to the nonlinear regime of a linear generalized
Maxwell model as depicted in Figure 5-84, where a number of Maxwell elements (spring- and dashpot elements in series)
are combined in parallel to model the dissipative processes that take place during deformation. With a long term stiffness
G    and a stiffness G  k  of the kth Maxwell element, the instantaneous stiffness G 0 of this model becomes
N
G0 = G +  Gk (5-487)
k = 1
With a dashpot viscosity   k  of the kth Maxwell element, its relaxation time becomes:
k
 k = ---------- (5-488)
Gk
CHAPTER 5 409
Material Library
A stress relaxation   t  at constant strain  is given in this model as:
N t- N t-
 – ----

  – ----


  t  = G  t  =  g    +  g  k  e k G 0  =  g +  g k e k  0 (5-489)
   
 k = 1   k = 1 
where the term inside the parenthesis is known as a Prony series expansion. The Prony factors g  k  in this series expansion
are simply the ratios of the stiffness G  k  of the kth Maxwell element to the instantaneous stiffness G 0
Gk
g  k  = ---------- (5-490)
G0
G
G1 1
G2 2
GN N
Figure 5-84 Generalized Maxwell Model
In order to obtain a generalized Maxwell model that can be applied to multidimensional finite deformations, Simo
introduces a Helmholtz free energy function (per unit undeformed volume) of the form:
N
k k k
  E ij Q ij  = W0 E ij –  Qi j Ei j +   Qi j  (5-491)
 = 1
The material is characterized by an arbitrary instantaneous strain energy function W 0  E i j  t   , where E i j  t  is the current
strain at time t .
With this strain energy function, the instantaneous stress is evaluated as:
0 W 0
S ij  t  = ----------- (5-492)
E ij
The formulation given here uses the material description and is called Total Lagrange formulation in Marc. The stress in
this formulation is the 2nd Piola Kirchhoff stress and the strain is the Green Lagrange strain. The same results, however,
can be obtained using the spatial description when appropriate transformations of stress and strain are taken into account
k
and this formulation is called Updated Lagrange formulation in Marc. The variables Q i j in the Helmholtz free energy
function are internal variables that describe the relaxation process, and they are governed after Simo by the following
evolution equations:
· k 1 k gk 0  k  = 1 2  N

Q i j + ----- Q ij = --------- S i j  t 
k k
(5-493)
k
Qi j  0  = 0
The  k and g  k  are the relaxation time and the Prony factor of the kth Maxwell element. The superimposed dot over
k
terms in above expression denotes differentiation with respect to time. The potential function   Q i j  appearing in the
formulation is not explicitly needed as long as we are not interested in the viscoelastic strains. From the Clausius-Duhem
inequality applied to isothermal processes, the total stress at any time t is found as (Simo and Hughes, [Ref. 43]):
N N N
0 k 0 0 k  vk
S ij  t  = S ij  t  –  Qi j = g    Si j  t  +   g  k  Si j  t  – Qi j  = Si j  t  +  Si j (5-494)
k = 1 k = 1 k = 1
vk
where the S ij are the stresses in the kth Maxwell element and they are given as
vk 0 k · k

S ij = g  k  S ij  t  – Q i j = k Qi j (5-495)
Similarly as for the model given in Figure 5-84, a stress relaxation at constant strain E a b can be given as:
N t-
 – ----


0
S ij  t  =  g    +  g  k  e k S i j  E a b  (5-496)
 
 k = 1 
For this case, we may, therefore, introduce a Prony series representation of the strain energy of the form:
N t-
 – ---- 
 k  0 
W  t E ab  =  g    +  
g ek  W E ab (5-497)
 
 k = 1 
CHAPTER 5 411
Material Library
The stress in a relaxation at constant strain can then be given as:

W  t E ab 
S ij  t  = ---------------------------
- (5-498)
E ij
k
The solution for the internal variables Q i j  t  is given as:
– -
– t----------
k gk t k 0
Qi j  t  = ---------  e S i j    d (5-499)
k 0
The stress at time t can, therefore, be given as:
N – -
– t----------
0 0 gk t k 0
Si j  t  = g    S ij  t  +  g  k  Si j  t  – ---------  e S i j    d (5-500)
k 0
k = 1
If we assume small perturbations around a static equilibrium state (quantities related to this equilibrium state are
superscripted with an asterisk “*”), we can write the instantaneous stress as:
0 0* 0
S ij  t  = S ij + S ij  t  (5-501)
Substituting this in equation (5-500) and noting that any long term time effects of the equilibrium state die out, the stress
becomes:
N – -
– t----------
*0 0 0 gk t k 0
S ij  t  = g    S ij + g    S i j  t  +  g  k  S i j  t  – ---------  e S i j    d (5-502)
k 0
k = 1
The first term in the right-hand side of equation (5-502) is the long term equilibrium stress, so the perturbation stress
response at time t can be given as:
N – -
– t----------
0 0 gk t k 0
S ij  t  = g    S ij  t  +  g  k  S i j  t  – ---------  e S i j    d (5-503)
k 0
k = 1
By employing partial integration and noting again that any long term time effects of initial instantaneous stress increments
die out, we can also write this as:
N – -
– t---------- 0
0 t  k d  S i j    
S ij  t  = g    S ij  t  +  gk e --------------------------- d (5-504)
0 d
k = 1
We now assume periodic excitations with an angular frequency  or a vibration period T = 2   , which means:
0 0
S i j  t + T  = S ij  t  (5-505)
After some time, when all initial effects have died out, we also obtain a stationary periodic response satisfying:
S i j  t + T  = S ij  t  (5-506)
With equation (5-504), we can write for the stress at time t + T
N +T–
– t--------------------- 0
0 t+T k d  S i j    
S i j  t + T  = g    S i j  t + T +  gk e --------------------------- d (5-507)
0 d
k = 1
From this equation, together with the periodicity conditions equations (5-505) and (5-506), we find:
N T-
– ---- – -
– t---------- 0 N T-
– ---- 0
  k t  k d  S i j      d  S i j    
 gk 1 – e  e --------------------------
- d =  
g ek  k --------------------------
- d (5-508)
  0 d d
k = 1 k = 1
The g  k  are independent factors, so from equation (5-508) we can conclude that for periodic solutions we have for each k
– -
– t---------- 0 ----- 0
t  k d  S ij     1 T  d  S i j  t +   
0 e --------------------------- d = -----------------  e k ------------------------------------ d (5-509)
d T-
---- 0 d
k
e –1
where in the right hand side of equation (5-509), we have used a transformation replacing  – t by s and then replacing s
by  . The final periodic perturbation stress response can, therefore, be given as:
N 
----- 0
0 g  k  T  d  S i j  t +   
S i j  t  = g    S ij  t  +  -----------------  e k -----------------------------------
- d (5-510)
T-
---- 0 d
k = 1 k
e –1
When we apply equation (5-510) to the periodic excitation:
0
S i j  t  = S ij sin  t  , (5-511)
we obtain:
N N
g  k   2  k2 g  k   k
S ij  t  = g +  ----------------------- S i j sin  t  +  ----------------------- S i j cos  t  (5-512)
1 +  2  k2 1 +  2  k2
k = 1 k = 1
CHAPTER 5 413
Material Library
From the last expression, we recognize the storage factor:

N k 2 2
g  k
g'    = g +  ----------------------
- (5-513)
1 +  2  k2
k = 1
and the loss factor:
N k 2
g  k
g''    =  ----------------------
1 +  2  k2
- (5-514)
k = 1
These are the same factors as those present in equations (5-425) and (5-426), where they were derived from the Fourier
transforms given in equations (5-418) and (5-419).
0
Furthermore, we make the assumption that the perturbations S i j  t  are small enough to justify a linearization around
the static equilibrium state as:
0  2 W 0  E ab * 
S ij  t  = ----------------------------- E k l  t  (5-515)
E ij E kl
and we may write:
 2 W 0  E ab * 
S ij = ----------------------------- E k l (5-516)
E ij E kl
where E kl is the amplitude of the periodic strain increment of the form:
E kl  t  = E kl sin  t  (5-517)
A Model of Viscoelasticity including Thixotropic Effects after Lion

We now consider the ideas of Lion ([Ref. 42] and references herein) who introduces a thixotropic behavior of the rubber by
introducing a process dependent viscosity (i.e., a process dependent relaxation time) for each Maxwell element. The
k
internal variables Q ij are now governed by evolution equations of the following form:
· z· k 0
·  k  zk  k 
Q i j + ----- Q ij = g  k  ----- S i j  t  k = 1 2  N
k k
(5-518)
k
Qi j  0  = 0
We see that the constant relaxation time  k in equation (5-493) has been replaced by a time dependent relaxation time
 k  z· k and  k is now assumed to represent the initial relaxation time. The relation between the amplitude dependence
and the time dependent relaxation effects is modelled by the dimensionless function z· k which is assumed to have the
following form
z· k  t  = 1 + d k q k  t  (5-519)
In above expression, d k is a material time parameter and q k  t  is an internal variable whose evolution is driven by the
deformation process. Note that q k has dimension one over time (i.e. Hz) and that equation (5-518) reduces to equation
(5-493) when d k = 0 . The internal variable q k of each Maxwell element is assumed to be governed after Lion by the
following evolution equation:
1
q· k = -----  d – q k 
k
(5-520)
qk  0  = 0
In above expression, d is a norm of the rate of deformation tensor which provides a measure of amplitude when the
problem is considered in the frequency domain:
2
d = --- d ij d ij (5-521)
3
The constants  k are material parameters that can be interpreted as relaxation times of the microstructure (Lion, [Ref. 42]).
The function z k  t  can thus be interpreted as defining an intrinsic process time of Maxwell element k , which depends on
the material parameter d k and the internal variable q k  t  . The function z k  t  must be a monotonically increasing
function, which is guaranteed when the material constants  k and d k are positive. The solution of equation (5-520) can
be given as:
– -
– t----------
1 t k
q k  t  = -----  e d    d (5-522)
k 0
Since the integrand in equation (5-522) is always positive it can be concluded that q k  t  is positive for all t when  k is
positive. This implies z· k  t  is positive for all t when d k is positive, which in turn guarantees that z k  t  is a monotonically
increasing function.
The general solution to equation (5-518) can be given as:
zk  t  – zk   
– --------------------------------
-
k gk t k 0
Q i j  t  = ---------  e z· k   S i j    d (5-523)
k 0
CHAPTER 5 415
Material Library
This can easily be verified by substituting it in equation (5-518) while applying the Leibniz differentiation rule. Therefore,
the stress is given as:
zk  t  – zk   
N – --------------------------------
-
0 k 0 gk t k 0
S i j  t  = g    S ij  t  +  g Si j  t  – ---------  e z· k   S i j    d (5-524)
k 0
k = 1
By assuming an instantaneous stress perturbation around a static equilibrium state as in equation (5-501), employing partial
integration and noting that any effects of nonzero initial conditions die out over time we obtain the solution for the
perturbation stress response similar to Equation (5-504) as:
zk  t  – zk   
N – --------------------------------
- 0
 0 k t k d  S i j    
S i j  t  = g S ij  t  +  g 0 e --------------------------
- d (5-525)
d
k = 1
When we apply periodic excitations with a time period T (i.e., excitations satisfying equation (5-505)), we expect a
stationary periodic response with the same time period (i.e., a response satisfying equation (5-506)). We also expect that the
internal variables q k related to the response of the microstructure become periodic with the same time period, since the
driving strain increment variations are periodic with this time period. The rate of deformation tensor in equation (5-520)
when considering perturbations around a static equilibrium state can be approximated with the help of the rate of the Green
Lagrange strain tensor as:
d ij = F ki ·
* –1 F * –1 E (5-526)
lj kl
is the inverse of the deformation gradient in the static equilibrium state and E· kl is the time derivative of the
–1
Here F ij*
Green Lagrange strain tensor. With equation (5-517), we obtain for the latter rate:
E· kl  t  = E kl cos  t  (5-527)
and equation (5-520) becomes:

1 
q· k + ----- q k = -----  cos  t 
k k
(5-528)
qk  0  = 0
with  being the norm of the true strain increment computed from the rate of deformation in equation (5-526) as:
–1
* F * E –1 –T –1 –1
 ij = F ki lj kl   = F * EF *   = C * E 
(5-529)
We assume the periodic solution of q k can be approximated by a Fourier series of the form:

0 n n
qk  t  = qk +   q k cos  nt  + r k sin  nt   (5-530)
n = 1
Note that equation (5-530) does not have to satisfy the initial condition of equation (5-528), because we have not included
t-
– -----
k
the homogeneous solution q k homog.  = A k e , where A k is an additional integration constant which allows us to
satisfy the initial condition. For large t , the homogeneous solution dies out.
We also assume that differentiation and integration of the series provide well defined functions, so we can compute the
derivative of q k as:

n n
q· k  t  =   – nq k sin  nt  + nr k cos  nt   (5-531)
n = 1
We can solve the coefficients of the Fourier series by substituting equations (5-530) and (5-531) in equation (5-528) and
performing an integrated weighting with the independent basis functions of the Fourier series (Galerkin weighting)
T 1   T
 q· + ----
- q cos  nt  dt = ----- 
0  k  k k k 0  cos  t  cos  nt   dt
T T
(5-532)
1 
 q· + ---- 
0 - q sin  nt  dt = -----  0  cos  t  sin  nt   dt
 k  k k k
We find m =  1 2 3  
0 2
qk = ---  

4
– ---  – 1  m
 2m  
qk = ------------------------------------  
1 +  2m k  2
(5-533)
4 m
– ---  – 1  2m k
 2m  
rk = -----------------------------------------  
2
1 +  2m k 
 2m – 1   2m – 1 
qk = rk = 0
We see that only the even terms in the expansion of equation (5-530) contribute to the solution, which is not surprising,
since cos  t  is a periodic function with time period T  2 . Finally, we can write the solution for the intrinsic time z k
as:
CHAPTER 5 417
Material Library
 2m
0 dk qk d k r k 2 m 
z k  t  =  1 + d k q k t +  -------------------
- sin  2mt  – ------------------
- cos  2mt  (5-534)
2m 2m
m = 1
We can write this as:
0
z k  t  =  1 + d k q k t +  k  t  (5-535)
where  k  t  represents the periodic terms under the summation in equation (5-534) and it has a time period of T  2 .
Substituting this solution in equation (5-525), we can write the perturbation stress response as:
– -  k  t  –  k   
– t----------
N – ----------------------------------- 0
 0 k t * k d  S i j    
S i j  t  = g S ij  t  +  g 0 e k e --------------------------
- d (5-536)
d
k = 1
with:
k k
 k* = -------------------------
 0- = ------------------------------------
- (5-537)
1 + dk qk 2
1 + d k ---  

which is a modified frequency dependent and amplitude dependent relaxation time for the periodic excitation with angular
frequency  or time period T = 2   . Note that in the first exponential factor of the integrand in equation (5-536),
the modified relaxation time  k* is used and in the second exponential factor the initial relaxation time  k is used. Also,
note that the second exponential factor is periodic, since  k  t  is periodic. By the same procedure as in the previous
section, we can now find the periodic stress response as
- k  t +   – k  t 
N ----- -------------------------------------------- 0
 0 gk T * k d  S i j  t +   
S ij  t  = g S ij  t  +  ------------------  e k e ------------------------------------ d (5-538)
T-
----- 0 d
k = 1 *
k e –1
Lion makes the assumption that the relaxation times of the microstructure  k are much larger than the time period of the
excitation (i.e.,  k >>T or  k >>2 , so the  k  t  in equation (5-535) can be considered as a small oscillation around a
linearly increasing intrinsic time. We can, therefore, expand the second exponential factor in equation (5-538) in a Taylor
series and truncate the expansion after the constant term. With this assumption, the periodic response can be approximated
as:
-
-----
N 0
 0 gk T * d  S i j  t +   
S ij  t   g S ij  t  +  ------------------  e k ------------------------------------ d (5-539)
T-
----- 0 d
k = 1 *
k e –1
which is of exactly the same form as equation (5-510), but it uses the modified relaxation times  k* as given by equation
(5-537). We may, therefore, conclude that the storage and loss factors can be given by equations (5-513) and (5-514) when we
replace the relaxation times in these expressions by modified relaxation times given by equation (5-537). For the storage
factor, we find:
N
g  k   2  k2
g     = g +  -----------------------------------------------------------------
- (5-540)
2
 1 + d --2-    +  2  2
k = 1  k  k
For the loss factor, we find
2
g  k   1 + d k ---     k
N
  
g     =  ------------------------------------------------------------------ (5-541)
2 2

k = 1 1 + d k ---  
 + 2 2
   k
The storage and loss factors are now not only frequency dependent, but also amplitude dependent. We also note that the
relaxation times of the microstructure  k do not appear in equation (5-539); they only appear if we include higher order
harmonic terms in the stress response through the function  k in equation (5-538)). For the first order harmonic
approximation, we, therefore, only need the material constants d k in addition to the usual Prony constants to include the
amplitude dependence.
The material parameters that need to be entered by the user are the triples g  k  being the Prony factor,  k being the initial
relaxation time, and d k being a time factor of the intrinsic time scale for each Maxwell element present in the model. The
material parameters g  k  ,  k , and d k must all be positive. The parameters g  k  and  k are entered on the VISCELMOON or
similar model definition options, depending on the elasticity model that is being used to describe the static behavior of the
material. The parameters d k are entered on the PAYNE model definition option while choosing the thixotropic model
(model 1). The elastic material parameters that need to be entered for the static response are the short term material
parameters at zero (i.e., very small) amplitude. Temperature effects on the parameters  k and d k can be included by
assuming thermo-rheologically simple (TRS) material behavior using the SHIFT FUNCTION model definition option.
The Triboelastic Model

This type of model has been used by e.g. Ahmadi e.a., [Ref. 45] and Coveney e.a., [Ref. 46] and is based on a generalized
Prandtl model as shown in Figure 5-85, where a number of Prandtl elements (spring and friction/slip elements in series) are
combined in parallel to model the dissipative processes that take place during deformation.
CHAPTER 5 419
Material Library
G
G1 1
G2 2
GN N
Figure 5-85 Generalized Prandtl Model
Such a model is known from plasticity theory to model nonlinear hardening phenomena and can, for example, be used in
cyclic plasticity to model the Bauschinger effect. In plasticity theory, the model is also known as the fraction model or
subvolume model (Besseling, [Ref. 47]), where each Prandtl element represents elastic ideally plastic behavior. The
generalized Prandtl model has also been extended to contain an infinite number of Prandtl elements in parallel; in which
case, it is known as the Ishlinsky model (Palmov, [Ref. 48]), where the yield stresses for a finite number of different fractions
get replaced by a yield stress probability distribution. The most important feature of these models is that they can accurately
model a hysteresis loop of a cyclic plasticity test like the one shown in Figure 5-86. The hysteresis loop described by all these
models satisfies the Masing principle (Lemaitre and Chaboche, [Ref. 49]), which implies that the first hysteresis loop already
describes the stabilized hysteresis loop that is expected after repeated cyclic unloading and reloading with a certain double
strain amplitude ( 2 in Figure 5-86). Furthermore, the shape of the hysteresis loop only depends on the size 2 of the
double strain amplitude, but not on where it has been initiated in the loading or unloading process (Palmov, [Ref. 48]). The
area of the hysteresis loop determines the energy being dissipated in one load cycle.




III
I
–
 
II
–

2
Figure 5-86 Hysteresis Loop in Cyclic Plasticity
The material is initially loaded along path I (see Figure 5-86). At some level of strain  with stress  the material is unloaded
along path II until we reach strain level –  with stress –  after which the material is reloaded along path III back to strain
level  . The shape of unloading path II and reloading path III is found from the initial loading path I by scaling it
uniformly by a factor 2 in both coordinate directions. The stress-strain relations along these loading paths can thus be
represented as follows:
I :  =  First loading


II :  =  – 2  ------ Unloading
2 (5-542)

III :  = –  + 2  ------ Reloading
 2
The unloading reloading cycles are applied dynamically with angular frequency  or time period T = 2   by
applying the strain as:
  t  =  cos    (5-543)
CHAPTER 5 421
Material Library
The  along unloading path II and reloading path III can be given as:
II :   t  =  –   t  =   1 – cos  t  
(5-544)
III :   t  =  +   t  =   1 + cos  t  
The stationary stress response is expected to be periodic so we can expand it into a Fourier series as:

t a
--------- = ----0- +   a n cos  nt  + b n sin  nt  
 2
n = 1
The coefficients in this Fourier series can be computed as:

2 T
a n = ------    t  cos nt dt n = 0 1 2  N
T 0
2 T (5-545)
b n = ------    t  sin nt dt n = 1 2  N
T 0
The constant term in the Fourier series can be shown to vanish
a0 = 0 (5-546)
The first order terms of the Fourier series can be associated with the storage and loss modulus in the vibration. If we
normalize     with the initial stiffness G 0 of the material (i.e., the tangent at  = 0 of the first loading path I in
Figure 5-86), we obtain the storage and loss factors. For the storage factor, we find:
a 4  
g = ------1- = – -------------    ---  1 – cos  s   cos  s  ds (5-547)
G 0 G  0 0  2 
For the loss factor, we find:
b1 4  
g = – ------- = ------------- 0   ---  1 – cos  s   sin  s  ds –  (5-548)
G 0 G 0  2 

If we apply partial integration to equation (5-547) and make the transformation ---  1 – cos  s   =  , we can write the
2
storage factor as:
4  d    2 2
g = -------------  -------------- 1 –  1 – ----- d (5-549)
G  0 0 d  
If we make the same transformation in equation (5-548), we can write the loss factor as:
4 
g = ---------------- 0  2    –   d (5-550)
G 0  2
It is observed that the storage and loss factors in equations (5-547) and (5-548) or equations (5-549) and (5-550) do not depend
on the frequency, they only depend on the shape of the initial loading curve     . They, therefore, only model an
amplitude dependence of the stiffness and damping properties of the material. From equation (5-550), we can see that the
loss factor is proportional to the area of the hysteresis loop, as expected.
If we, for example, apply equations (5-549) and (5-550) to the stress strain curve of elastic, ideally plastic behavior as shown
in Figure 5-87, we obtain:

1   -------
G0
g = (5-551)
1 2   
2- -----------   
--- acos  1 – ---------
- – 2  1 – --------- - 1 – ---------
-   -------
  G  0   G  G0 
0  G 0  G0
and

0   -------
G0
g = (5-552)
4    
--- ----------  1 – ---------
-   -------
 G0  G 0  G0
as was reported by Lion in [Ref. 49]. In computing the storage factor, we have used the fact that:
4 2 2 2
---  1 –  1 – 2 2
----- d = acos  1 – ----- –  1 – 2
----- 1 –  1 – ----- = F     (5-553)
         
CHAPTER 5 423
Material Library


- 
 = ------ 
G0
Figure 5-87 Elastic, Ideally Plastic Material Behavior
If we approximate the     curve between  = 0 and the maximum amplitude of the vibration  =  m ax by a number
of piecewise linear segments as in Figure 5-88, we find that with the help of equation (5-553) the storage factor can be given as
n
1 n + 1 m – m – 1
g    = ---------- -------------  F     – F   n    +  --------------------------
-  F   m    – F   m – 1    (5-554)
G 0  n + 1 m – m – 1
m = 1
The loss factor for this curve can be given as

n
4
g    = ----------------   +  n    –  n  +   m + m – 1   m – m – 1  (5-555)
G 0  2
m = 1
In equations (5-554) and (5-555), we set  0 = 0 and  0 = 0 . When    m ax , we assume the curve continues with its
last slope. The stress-strain curve must be an increasing function with decreasing slope, so we also require that  n   n – 1
and  n   n   n – 1   n – 1 . We observe that, for amplitudes below  1 , the storage factor equals one and the loss
factor equals zero.

N

n
1
1 n   N =  max 
Figure 5-88 Plasticity Stress-strain Curve Approximated by Piecewise Linear Segments
The material parameters that need to be entered by the user are the points  0 0  and   n  n  ,  n = 1N  of the
plasticity stress-strain curve. The function values  n need to be normailized by the initial static stiffness G 0 of the material.

This initial stiffness is assumed to be given as G 0 = ----1- , so point 1 when normalized is simply   1  1  . These points are
1
entered on the PAYNE model definition option while choosing the triboelastic model (model 2). There is no need for
VISCELMOON or similar input depending on the elasticity model that is being used to describe the static behavior of the
material, so this model does not directly account for a frequency dependence. However, a frequency dependence can be
introduced by measuring the storage and loss moduli over a frequency range, and for each frequency, evaluate the
parameters of the triboelastic model (i.e., the points of a plasticity stress-strain curve) based on the measured storage and
loss moduli at that frequency as a function of amplitude. The material parameters  n can then be associated with tables
defined in the TABLE model definition to account for the frequency dependence. Note that all  n values of the different
plasticity stress-strain curves must be normalized with the same G 0 value that follows from the static elasticity input and
which represents the initial stiffness at zero frequency. Since no VISCELMOON or similar data is required for this model, it
cannot describe a temperature dependence through thermo-rheological simplicity (TRS), but temperature effects for the
 n parameters can still be included through tables if required. Note that with the definition of amplitude as given in the
Definition of Amplitude in the Payne Effect Formulations section the stresses  n represent normalized equivalent tensile stresses and
the amplitudes  n represent equivalent tensile strains.
CHAPTER 5 425
Material Library
Combining the Thixotropic and Triboelastic Models

In this section, two models are considered that combine the thixotropic and triboelastic models. The first is an additive
combination; the second is a multiplicative combination. An approach of this kind was followed by Rabkin, [Ref. 50], who
combined viscoelastic effects without thixotropy and triboelastic effects based on the Ishlinsky model. Since the thixotropic
contributions to these combined models are thermo-rheologically simple (TRS) and the triboelastic contributions are
frequency independent, it follows that the combined models are also thermorheologically simple as long as the material
constants in the triboelastic model have no explicit temperature dependence.
Additive Combination of the Thixotropic and Triboelastic Models

It is relatively easy to combine sophisticated nonlinear rheological models in parallel, since they are all subject to the same
strain. The relation between the stationary stress and the stationary strain for each element can be written as
̂e it = Ĝ  ̂ ̂e it (5-556)
where ̂ is the complex stress amplitude of the vibration, ̂ the complex strain amplitude and Ĝ the complex modulus.
The latter can be written as:
Ĝ = G' + iG'' = G 0  g' + ig''  (5-557)

In a parallel combination the total stress is the sum of the stresses in the individual elements, so the total complex modulus
is the sum of the complex moduli of the individual elements. Figure 5-89 shows the parallel combination of a thixotropic
element (V) and a triboelastic element (P).
̂
V P ̂
̂
Figure 5-89 Parallel Combination of a Thixotropic Element (V) and a Triboelastic Element (P)
Both elements undergo the same complex strain ̂ and the total complex stress is the sum of the complex stress in the
thixotropic element and the complex stress in the triboelastic element:
̂ = ̂ V + ̂ P
̂ V = Ĝ V ̂  ̂ =  Ĝ V + Ĝ P ̂ = G 0  ĝ V + ĝ P ̂ (5-558)
̂ P = Ĝ P ̂
The storage and loss modulus of the combined model can now be given as:
G = GV + GP
(5-559)
G = GV + G
P
The normalized moduli involved in this combination can directly be obtained from equations (5-540), (5-541), (5-554) and
1
(5-555). The long term modulus at zero amplitude is given in this model as G 0 g     V  + ------------ . The short term modulus
G 0 1
N
1
at zero amplitude is given as G0 gV +  g  k  + -----------
- , so the term inside the brackets must equal one
G 0 1
k = 1
N
1
gV +  g  k  + -----------
- = 1 (5-560)
G 0 1
k = 1
Multiplicative Combination of the Thixotropic and Triboelastic Models

An alternative combination can be obtained if we make a multiplicative combination of the viscous and the friction
behavior  V P  in the frequency domain and use this in parallel with an elastic  E  element as illustrated in Figure 5-90.
̂
V,P E ̂
̂
Figure 5-90 Multiplicative Combination of Thixotropic and Triboelastic Behavior  V P  in Parallel with Elastic Behavior
E
The combined behavior is assumed to be of the form:
̂ = G 0  g    +  g + ig   g + ig  ̂ (5-561)

V V P P
The viscous storage contribution gV is given by the summation term in equation (5-540) and g is given by equation
V
(5-541).
CHAPTER 5 427
Material Library
N k 2 2
g  k
g     =  -------------------------------------------------------
-
V  2 2 2 2
k = 1 1 + d k ---  +  k
  
(5-562)
k 2
N g 1 + d k ---   k
  
g     =  --------------------------------------------------------
V  2 2 2 2
k = 1 1 + d k ---  +  k
  
The plastic contributions g P and g P are given by a similar terms as in equations (5-554) and (5-555)
n
1 n + 1 – n m – m – 1
g    = ---------- -------------------------  F     – F   n    +  --------------------------
-  F   m    – F   m – 1   
P G 0  n + 1 –  n m – m – 1
m = 1 (5-563)
n
4
g    = -------------------   –  n    –  n + 1  –  +   m – m – 1   m – m – 1 
P G 0  g2
m = 1
The long term behavior at any amplitude is thus given as G 0 g    . The short term behavior at zero amplitude is given as
 N  
+   g  k  ------------ so the term inside the brackets must equal one
1
G0 g
  G0
k = 1  1
 N  
+   g  k  ------------ = 1
1
g (5-564)
  G0
k = 1  1
The material parameters that need to be entered by the user are the triples g  k  being the Prony factor,  k being the initial
relaxation time and d k being the time factor of the intrinsic time scale for each Maxwell element present in the model
together with the points  0 0  and   n  n  ,  n = 1N  of the plasticity stress-strain curve. The function values need
to be normailized by the initial short term static stiffness of the material. The material parameters g  k  ,  k , d k ,  n , and
 n must all be positive. The plasticity stress-strain curve must be an increasing function with decreasing slope, so it is also
required that  n   n – 1 and  n   n   n – 1   n – 1 . The parameters g  k  and  k are entered on the VISCELMOON
or similar model definition options, depending on the elasticity model that is being used to describe the static behavior of
the material. The parameters d k and the points of the plasticity stress-strain curve are entered on the PAYNE model
definition option while choosing the combined model (model 3). In addition, it needs to be specified if the combination
is additive (default) or multiplicative. This specification is also made on the PAYNE option. The elastic material parameters
that need to be entered for the static response are the short term material parameters at zero (i.e., very small) amplitude.
Temperature effects for the parameters  k and d k can be included by assuming thermo-rheologically simple (TRS)
material behavior using the SHIFT FUNCTION model definition option. In this model, the parameters  n of the plasticity
stress-strain curve cannot be associated with tables, since a frequency dependence in this model comes through the
thixotropic part and a temperature dependence through the TRS assumption.
Kraus/Ulmer Model
Kraus [Ref. 52] presented a model for the amplitude dependent storage and loss modulus of filled rubbers based on
considerations of breakage and recovery of bonds at the micro level. Kraus derived following relations for the storage and
loss modulus as a function of amplitude
G0 – G
G    = G     + ---------------------------------- (5-565)
1 +    c  2 m
and
2  G  m   – G         c  m
G    = G     + -----------------------------------------------------------------
- (5-566)
1 +    c  2 m
where  is the vibration amplitude and G  0  , G     , G     , G  m   ,  c , and m are material constants that have to be
evaluated from storage and loss moduli curves as shown for example in Figure 5-83. In these expressions, the superscripts do
not refer to short and long term time behavior, but to the size of the amplitude:
G  0  storage modulus at very low amplitude

G storage modulus at very high amplitude
G loss modulus at very high amplitude
G  m  maximum loss modulus
c amplitude where the loss modulus reaches its maximum
m exponent, which in practice lies around 0.5 for many materials
Ulmer [Ref. 53] modified the Kraus model to improve the correlation for the loss modulus with experimental data by adding
an exponentially decaying term:
2  G  m   – G         c  m –  2
- +  G  2   – G     e
G    = G    + ----------------------------------------------------------------- (5-567)
1 +    c  2 m
where G  2  and  2 are additional material constants. The correlation for the storage modulus was found to be good, so
this is still given by equation (5-565).
The eight material parameters G  0   , G     , G     , G  m   ,  c , m , G  2   , and  2 of this model are entered on the
PAYNE model definition option while choosing the Kraus/Ulmer model (model 4). The storage and loss terms need to be
normalized with the static initial stiffness of the material, which should coincide with the storage modulus at very low
amplitude G  0  . There is no need for VISCELMOON or similar input depending on the elasticity model that is being used
CHAPTER 5 429
Material Library
to describe the static behavior of the material and this model does not directly account for a frequency dependence.
However, storage and loss modulus can be measured as a function of amplitude at different frequencies and a set of material
parameters can be evaluated for each frequency. A frequency dependence can then be introduced by associating the material
parameters with tables defined in the TABLE model definition option. In that case the normalization of the storage and loss
terms is done with the storage modulus at very low amplitude and zero frequency. Since no VISCELMOON or similar data is
required for this model, it cannot describe a temperature dependence through thermo-rheological simplicity (TRS), but
temperature effects for its parameters can still be included through tables if required.
Table Model
It is possible to enter the data characterizing the Payne effect directly through tables defined in the TABLE model definition
option. These tables define a complex Payne factor that is multiplicative to the complex modulus that results from the
VISCELMOON or similar model definition options. One table is required for the real part of this complex Payne factor and
one table is required for its imaginary part. The independent variable used in the tables can be the amplitude or the
logarithmic amplitude (base 10). The complex modulus including effects of frequency and amplitude can thus be
represented as
Ĝ     = G 0   g  PA YNE     + ig  PA YN E      ĝ  V I S     (5-568)
where G 0 is the stiffness resulting from the static elasticity input and ĝ  V I S     is the normalized complex modulus
resulting from VISCELMOON or similar input accounting for the effects of frequency. The tables should satisfy
g  P AYN E   0  = 1 and g  PA YNE   0  = 0 ; i.e., at zero amplitude the Payne factor equals one. A static
pre-deformation dependence can optionally be defined in the VISCELCORR model definition option, which is then also
treated in a multiplicative sense with regard to the complex factor above.
The material parameters that need to be entered by the user on the PAYNE model definition option are the two table IDs
referencing the tables for the real and imaginary part of the factor, while choosing the table model (model 5). Also see the
Using Tables to Define Storage and Loss Moduli section which, in addition to the multiplicative table method of this section,
discusses a more general multidimensional table method.
User Subroutine Model

A user-defined model for the Payne effect can be entered with the UPAYNE user subroutine. The user needs to compute the
real and imaginary parts of a normalized complex modulus, which apart from depending on the amplitude can be made to
depend on the frequency, the static pre-deformation, the temperature and other state variables. The complex modulus can
thus be represented as:
Ĝ     I * T  = G 0  g  UP AY NE  + ig  UP AYN E   (5-569)
where G 0 is the stiffness resulting from the static elasticity input,  is the frequency,  is the amplitude, I * represents
the data related to the static pre-deformation, and T represents the temperature and other state variables. The factors
g  UP AY NE  and g  UP AYN E  need to be defined in the subroutine where they can incorporate the aforementioned
dependencies.
This model is entered when choosing the user subroutine model on the PAYNE model definition option (model 6).
Optionally, the user can enter a number of integer and a number of real values on this option that get passed into the user
subroutine. This provides a means to enter material specific properties. A frequency dependence can optionally be defined
in the VISCELMOON or similar model definition options, which is then treated in a multiplicative sense with regard to the
complex factor computed in UPAYNE. A static pre-deformation dependence can optionally be defined in the VISCELCORR
model definition option, which is then also treated in a multiplicative sense with regard to the complex factor computed
in UPAYNE. Thermo-rheologically simple (TRS) data can optionally be defined in the SHIFT FUNCTION model definition
option. The reduced frequency as well as the actual frequency are both passed into the user subroutine.
Thermo-Rheologically Simple Material Behavior in the Frequency Domain

The relaxation behavior of most viscoelastic materials changes with temperature. For many materials it is observed that
when displaying the relaxation curves for different temperatures on a logarithmic time scale, the shape of the curves doesn’t
change, but they shift along the logarithmic time axis. Such material behavior is called Thermo-Rheologically Simple
(TRS) material behavior. It basically means that all relaxation times that define a relaxation curve have the same
temperature dependence. When we know the relaxation behavior at a reference temperature T re f , and we know how the
relaxation curve shifts along the log-time scale, we can predict the relaxation behavior at other temperature levels as well.
In general, it is expected that when the temperature is higher than the reference temperature, the relaxation is faster so the
curve makes a left-shift and when the temperature is lower than the reference temperature, the relaxation is slower so the
curve makes a right-shift. This shifting along the log-time scale is represented by the so-called shift function, denoted as
a  T  , and the relation between the relaxation curve at temperature T and the reference temperature can be expressed as:
g  t T  = g  a  T t T r e f  (5-570)
The relaxation taking place on time scale t at other temperatures than the reference temperature is the same as the
relaxation taking place on a reduced time scale t r with respect to the reference temperature,
t r = a  T t (5-571)
When observing the behavior in the frequency domain, we conclude that vibration time periods T̃ at temperature T must
have the same stiffness and damping characteristics as vibration time periods T̃ r = a  T T̃ at the reference temperature
meaning that vibrations with frequencies related as:
2 2 
 r = ------ = --------------- = ----------- (5-572)
T̃ r a  T T̃ aT
will also have the same stiffness and damping characteristics and we may conclude that the product:
 r t r = t = invariant (5-573)
for each temperature. This basically means that when the relaxation curve shifts to the left on the log-time scale, the
accompanying storage and loss moduli curves shift to the right over the same amount on the log-frequency scale and vice
versa. In the frequency domain the relations expressing the shifting property for the storage and loss factors similar to the
time domain relation of equation (5-567) can be formulated as:
CHAPTER 5 431
Material Library

g   T  = g   r T r e f  = g  ----------- T r e f
 aT 
(5-574)

g   T  = g   r T r e f  = g  ----------- T r e f
 aT 
The TRS property means that all relaxation times have the same temperature dependence, which can be expressed as:
 k  T ref 
-------------------- = a  T  (5-575)
k  T 
This also means that the product  k  T  =  r  k  T r e f  is invariant, as can be observed from equations (5-513) and
(5-514).
In the thixotropic model, the TRS property is retained if we make the same assumption for the time factors d k of the
intrinsic time scales
d k  T ref 
-------------------- = a  T  (5-576)
dk  T 
which in turn means that the product d k  T  =  r d k  T re f  is invariant, as can be observed from equations ((5-540))
and (5-541).
The shift function parameters associated with Thermo-Rheologically Simple behavior are entered through the SHIFT
FUNCTION model definition option. A commonly used shift function is the Williams-Landel-Ferry (WLF) function. As an
alternative to the WLF function, Marc allows the use of a series expansion, or the use of an Arrhenius function, or the
specification via the TRSFAC user subroutine. See also the Thermo-Rheologically Simple Behavior section of this chapter.
Deformation Dependent Relaxation in the Frequency Domain

A dependency of the relaxation function in the frequency domain on the static deformation can be introduced by means
of a complex correction factor [Ref. 37]. The complex relaxation function of equation (5-424) gets multiplied by this complex
correction factor which depends on the static deformation of the equilibrium state and we obtain:
* 
ĝ    = g̃   c̃  E ab (5-577)
where,
c̃ is the complex correction factor

g̃    is the relaxation function in the frequency domain of the undeformed material
ĝ    is the modified relaxation function in the frequency domain
The complex correction factor may be represented with its modulus c and its argument  as:
c̃ = ce i (5-578)
In the absence of any deformation, the complex correction factor approaches unity, so c = 1 and  = 0 for E a*b = 0 .
With above representation of the correction factor it is easy to verify that:
Re  ĝ     = c  Re  g̃     cos  – Im  g̃     sin  
Im  ĝ     = c  Re  g̃     sin  + Im  g̃     cos   (5-579)
The constants c and  both depend on the deformation of the static equilibrium state. The constant c is assumed to be
a quadratic function of a suitable deformation parameter I d ef (for zero static deformation I d ef = 0 ) and the constant 
a linear function of this parameter, i.e.
2
c  I de f  = 1 +  c I def +  c I d e f
(5-580)
  I d ef  =   I def
where  c ,  c , and   are material parameters.
In case of uncoupled deviatoric and volumetric response, the deformation parameter for the deviatoric behavior can be
defined in three ways. The first is a normalized value of the instantaneous deviatoric strain energy function:
W 0 D  E ab *  W 0  I 1* I 2* 
D
I d ef = ----------------------------
- = -------------------------- (5-581)
G0 G0
where the normalization factor G 0 is the instantaneous shear modulus of material in the undeformed configuration.
The second is the so-called generalized octahedral shear strain defined as:
D 1
I d ef = ---  2I 1*2 – 6I 2*  (5-582)
6
In equations (5-581) and (5-582), I 1* and I 2* are the modified first and second invariant of deformation in the static
equilibrium state; i.e., they are based on the modified deformation gradient F i j that is present in equation (5-454).
The third is a user-defined deviatoric invariant quantity using the UINVAR user subroutine. If the volumetric response is also
viscoelastic, the deformation parameter for the volumetric behavior can also be defined in three ways. The first is again a
normalized value of the instantaneous volumetric strain energy function
V W 0 V  J * 
I d ef = ------------------------ (5-583)
K0
where the normalization factor K 0 is the instantaneous bulk modulus of the material in the undeformed configuration and
J * is the determinant of the deformation gradient in the static equilibrium state.
The second is based on the volumetric deformation
V
I d ef =  J* – 1 2 (5-584)
CHAPTER 5 433
Material Library
The third is a user-defined volumetric invariant quantity using the UINVAR user subroutine. In case of coupled
deviatoric/volumetric response, the deformation parameter is defined in the same way as for the deviatoric response in the
uncoupled case, but now based on either a normalized value of the total strain energy function with the initial shear
modulus G 0 taking the role of the normalization factor or based on the generalized octahedral shear strain defined in terms
of the unmodified invariants I 1 and I 2 or a generalized user defined invariant quantity using the UINVAR user subroutine.
If the strain energy function is being defined with the help of the UENERG or UELASTOMER user subroutine, the normalization
factors G 0 and K 0 are take as unity.
The type of deformation parameter I d ef and the values of the constants  c ,  c , and   are entered on the VISCELCORR
model definition option. In addition, it is possible to define the modulus c and the argument  of the correction factor
in the UCORRFAC user subroutine. By default no correction is used; i.e., c = 1 and  = 0 .
Alternatively, the complex correction factor can be defined entirely through tables by defining a table for the real part of
this factor, and a table for the imaginary part of this factor using the TABLE model definition option. The independent
variable of the table used for I d ef can be one of the three discussed earlier:
1. Octahedral shear strain or volumetric deformation

2. Strain energy function value
3. User defined invariant quantity
Note that these variables take context dependent values, depending if the table is evaluated for the deviatoric or the
volumetric or the coupled response. The use of these variables is exclusive, meaning that not more than one of them can
be present as independent variable in a table.
When the Payne effect is present in the model, the complex correction is multiplicative to the relaxation function in the
frequency domain of the undeformed material that includes the effects of the vibration amplitude and g̃    in equation
(5-574) can be viewed as g̃     .
The VISCELCORR option also allows input of objective stress increment factors for selected materials that are different from
the default value as specified by the VISCELHARM parameter.
Some caution needs to be exercised when this complex correction factor is used in conjunction with very low or very high
excitation frequencies. In both cases, the loss modulus tends to zero and no contribution to the damping matrix is expected.
However, with a complex correction factor the modified relaxation function in the frequency domain ĝ    will also
introduce damping in the low or high frequency range. Therefore, the application of this deformation dependent relaxation
behavior is primarily recommended for intermediate frequency ranges, but not for the very high and the very low frequency
ranges.
The complex correction discussed in this section can only be used for isotropic materials (isotropic linear elastic or isotropic
hyperelastic), but not for orthotropic materials.
Using Tables to Define Storage and Loss Moduli

It was mentioned earlier that tables that are used to enter the storage and loss behavior of the material should be normalized
with the instantaneous (short term) stiffness (i.e., the stiffness when the frequency approaches infinity). However, this short
term stiffness is not always available from the measured data, since the frequencies that were used in the experiment were
not high enough. Therefore, this condition may be relaxed, as is illustrated in Figure 5-91, which shows the curve of a storage
modulus on a logarithmic frequency scale with its typical sigmoidal shape.
G G     
G   ref 
G    0 
log   ref  log   

Figure 5-91 Normalization of Storage Modulus Table
In this figure, the lower asymptote represents the long term static equilibrium stiffness, which is reached as the frequency
approaches zero. The upper asymptote represents the instantaneous (short term) stiffness, which is reached as the frequency
approaches infinity. If this latter asymptote is not yet apparent from the measured data, we may also normalize the table
with the value G   r ef  at some reference frequency  r ef . The static elasticity data entered on the MOONEY or similar
options should then result in this reference stiffness, since the table value at this reference frequency evaluates to 1.
Furthermore, we should make sure that the table value evaluated for zero frequency when multiplied with the initial
stiffness that follows from the static elasticity input, results in the proper long term equilibrium stiffness of the material;
i.e., G    = G    0  = ĝ   = 0 G   r e f  . We may, in fact, choose this long term equilibrium stiffness as the
reference stiffness, in which case the storage table value at zero frequency should evaluate to 1. In that case, the long term
equilibrium stiffness data can be entered directly as the static elasticity data. The loss modulus curve should always be
normalized with the same stiffness value as the one used for the storage modulus curve.
The advantage of tabular data is that no phenomenological models are needed for which material parameters have to be
evaluated by curve fitting procedures. In general the most important independent variables for the tables are:
1. The frequency
2. A measure for the vibration amplitude
3. A measure for the static pre-deformation
4. The temperature.
The table input to combine these effects allows two input modes:
1. A multidimensional table mode for the storage and loss factors:
A storage factor g = g     I * T 
A loss factor g = g     I * T 
CHAPTER 5 435
Material Library
The complex factor is then given as: ĝ     I * T  = g + ig
The tables are entered in the TABLE model definition option. The use of this type of table is entered in the
VISCELPROP, VISCELMOON, VISCELOGDEN, or VISCELFOAM option. When this type of table is used for a material, any
information entered for this material on the VISCELCORR and PAYNE options is ignored.
2. A multiplicative table mode for the storage and loss factors:
One pair for the frequency dependence
g1 = g1   T 
g1 = g
1
  T 
One pair for the static pre-deformation dependence
g2 = g2 I * T 

g2 = g2  I * T 
One pair for the vibration amplitude dependence
g3 = g3   T 
g3 = g3    T 
The complex factor is then given as: ĝ     I * T  =  g + ig   g + g   g + g 

1 1 2 2 3 3
The tables are entered in the TABLE model definition option. The use of the first pair for the frequency dependence
is entered in the VISCELPROP, VISCELMOON, VISCELOGDEN, or VISCELFOAM option. The use of the second pair for the
static pre-deformation dependence is entered in the VISCELCORR option. The use of the third pair for the vibration
amplitude dependence is entered in the PAYNE option.
The independent variables that can be used in these tables are
1. The frequency (in vibrations per time; i.e., Hz) or logarithmic frequency (base 10)
2. The invariants characterizing the static pre-deformation
a. octahedral shear strain or volumetric deformation
b. strain energy function value
c. user-defined invariant quantity through the UINVAR user subroutine
3. The vibration amplitude or logarithmic vibration amplitude (base 10)
4. The temperature and other state variables
The storage and loss factors are normalized factors that usually get normalized with the short term modulus at zero
amplitude and zero pre-deformation at some reference temperature. As outlined earlier this condition may be relaxed and
the storage modulus at some other reference frequency may also be used for the normalization.
The complex stress-strain law generally takes the form:
 2 W 0  E a*b 
Ŝ i j = ĝ     I * T  ----------------------------- Ê k l (5-585)
E i j E k l
where Ê kl is the complex Green Lagrange strain increment and Ŝ i j is the complex 2nd Piola Kirchhoff stress increment
of the vibration and W 0 is the strain energy function that follows from the static elasticity input. At zero frequency, zero
amplitude and zero static pre-deformation, the stiffness tensor in equation (5-585) should evaluate to the long term static
initial stiffness of the material at some reference temperature
  2 W 0  E a*b = 0 
G i j kl = ĝ   = 0  = 0 I * = 0  T r e f  --------------------------------------- (5-586)
E i j E kl
The complex factor in the right hand side of equation (5-586) can only have a real part, at zero frequency its imaginary part
should be zero. Expressions for some important initial static stiffness quantities that follow from the strain energy functions
available in Marc are summarized in Table 5-12.
Table 5-12 Initial Shear and Bulk Modulus in Terms of the Entered Material Data
Initial Shear Modulus G Initial Bulk Modulus K
MOONEY 2  C 10 + C 01  K
E
GENT --- K
3
3 99 513 42039
ARRUDBOYCE nk  1 + ------- + ---------------- + ---------------- + ---------------------- K
 5N 175N 2 875N 3 67375N 4
W
MARLOW 2 --------  I 1 = 3  , see text N preceding Equation (5-228) K
I 1 1
OGDEN ---  N n n N K
21 1
FOAM --- n  = 1 n n ---   n   n – 3 n 
2 E 3 E
ISOTROPIC n--------------------
= 1 - n = 1------------------------
21 + v 3  1 – 2v 
Harmonic Equations of Motion

We employ the principle of virtual work to obtain the finite element formulation of small amplitude vibrations around
nonlinear equilibrium states. First, we state the conditions for static equilibrium, and then we consider small dynamic
deviations from this static equilibrium state. When using the undeformed configuration as the reference configuration, the
static equilibrium requirement can be formulated as follows:
 S ij* E ij* dV 0 =  b i* u i dV 0 +  p i* u i dA 0 (5-587)

V0 V0 A0
The left-hand side represents the virtual work of the internal stresses and evaluates to the internal force vector, the right
hand side represents the virtual work of the external loads and evaluates to the external force vector. S i*j is the second Piola-
CHAPTER 5 437
Material Library
Kirchhoff stress in the static equilibrium state and E i*j is the first variation of the Green-Lagrange strain in this state. The
Green-Lagrange strain is defined as:
1  u
* u * u k* u k*
E i*j = ---  --------i- + --------j- + --------- --------- (5-588)
2  X j X i X i X j 
Its first variation becomes:
1 * u k u
E ij* = ---  F ik ------------ + F j*k -----------k- (5-589)
2 X j X i 
The coordinates of the material points in the undeformed configuration are denoted by X i and u i* are their displacements.
The loads are given as distributed volume loads b i* and distributed surface loads p i* .
When we have small dynamic deviations from the static equilibrium state we may write:
u i = u i* + u i  t 
b i = b i* + b i  t  –  0 u··t  t 
p i = p i* + p i  t 
S ij = S ij* + S ij  t  (5-590)
E ij = E ij* +   E i j 
1 u k  t  u k u k  t  u k

  E ij  = --- -------------------- ------------ + -------------------- ------------
2 X i X j X j X i
where we have added the inertia terms as d’Alembert forces to the distributed volume loads. The linearized form of equation
(5-587) is:
u k  t  u k
  0 u··i  t u i dV 0 +  S i j  t E i*j dV 0 +  S i*j -------------------
X i X j
- ------------ dV 0 =
V0 V0 V0
(5-591)
 b i  t u i dV 0 +  p i  t u i dA 0
V0 A0
When we study small harmonic oscillations we have:
u i  t  = û i e it
u··i  t  =  2 û i e it
(5-592)
b i  t  = b̂ i e it
p i  t  = p̂ i e it
and equation (5-591) becomes:

û k u k
 –  0  2 û i u i dV 0 +  K i j k l   Ê k l E i*j dV 0 +  S i*j ------------
X i X j
------------ dV 0 =
V0 V0 V0
(5-593)
 b̂ i u i dV 0 +  p̂ i u i dA 0
V0 A0
Note that the stresses S ij* in equation (5-593) are the static (i.e., long term) equilibrium stresses, so they can also be written
as S i*j = g    S ij0 * . If we choose appropriate interpolations for the unknown fields and perform the integrations over a
finite element model, we obtain a complex matrix equation of the following form:
 –  2 M + iC los s    + K s t or a ge    + K G û = F̂ (5-594)

CHAPTER 5 439
Material Library
Here M is the mass matrix of the system, C l o ss is the damping matrix of the system that follows from the imaginary part
(i.e., the loss part) of the tensor K i j k l    , and K s t or a ge + K G is the stiffness matrix of the system where K st o r ag e
follows from the real part (i.e., the storage part) of the tensor K i j kl    and K G is the geometric stiffness matrix as a result
of the static equilibrium pre-stress S i*j . Both K s t or a ge and C l o ss are frequency dependent, but K G is not. The vector
û is the unknown complex displacement vector of the oscillation and F̂ is the complex load vector as a result of all the
harmonic loads.
When including the Payne effect, we get a similar set of equations, but now the matrices arising from the material behavior
also include an amplitude dependence. A matrix equation, similar to equation (5-594), can now be written as
 –  2 M + iC los s     + K s t or a ge     + K G û = F̂ (5-595)
The amplitude dependence arises, because at each material point the strain increment tensor  depends on the unknown
displacement increments û , and, thus, making the set of equations nonlinear.
This set of equations can be solved by a direct substitution procedure as summarized in Table 5-13.
Table 5-13 Direct Substitution Solution Procedure for the Nonlinear Harmonic Equations
i = 0
Make initial estimate for   i 
REPEAT
 –  2 M + iC l o ss     i   + K s t or a ge     i   + K G û  i + 1  = F̂
i = i+1
Update   i 
UNTIL converged
It should be noted that the material matrices can also include effects of the static pre-deformation and of the temperature,
but these do not make the equations nonlinear and are, therefore, not shown explicitly. Here only the dependency on
frequency (not a nonlinear effect) and the dependency on amplitude (the nonlinear effect) have been shown explicitly.
Convergence controls for solving this set of nonlinear equations can be entered in the CONTROL history definition option in
a harmonic loadcase. Relative and absolute convergence testing can be done on residual forces/moments and/or
displacements/rotations, but there is no strain energy criterion. Since the vector components involved in the convergence
testing are complex numbers of the form x + iy , the testing is done on the magnitude of the components x 2 + y 2 . By
default relative residual checking is used with a convergence tolerance of 0.1.
In most applications, a harmonic loadcase involves a frequency sweep from a lowest frequency to a highest frequency
including a number of intermediate frequencies. With the Payne effect each harmonic sub-increment is a nonlinear sub-
increment. By default, the amplitude of a previous sub-increment is used as the first amplitude estimate in the solution of
the current sub-increment. This default behavior may be switched off on the CONTROL option, in which case each new sub-
increment starts with zero amplitude as its first estimate.
It may sometimes be desirable to repeat the frequency sweep of a harmonic loadcase, while applying slightly higher or
slightly lower loads. For this case the same harmonic loadcase can be repeated a number of times, where in each instance
of the HARMONIC history definition option a different harmonic load factor (which defaults to 1) gets defined. The effect of
the harmonic load factor is that it scales all the harmonic loads listed in the LOADCASE history definition option with this
factor (i.e., it scales the right-hand side of equation (5-595) with this factor). The harmonic loads in general arise from
applied harmonic forces (point loads and distributed loads) and prescribed nonzero harmonic displacements. Since no
static loadcase is applied between two harmonic loadcases, the increment counter does not change, but the sub-increment
counter continues from the last value of the preceding harmonic loadcase. So in a way, the harmonic loadcases are being
chained. Although not required, it is recommended to apply the same excitation frequencies (in magnitude and number)
in all harmonic loadcases that are chained in this way. It is also recommended not to change loads and boundary conditions
between two chained harmonic loadcases.
The above equations of motion have been formulated using the undeformed configuration as the reference configuration,
and the problem has, therefore, been formulated within the total Lagrange framework. If the deformed static equilibrium
state serves as the reference configuration, the problem gets formulated within the updated Lagrange framework. Within
this framework, we would, of course, arrive at the same dynamic equations of motion in the frequency domain as given by
equation (5-594), so the choice of the analysis framework is primarily governed by considerations about how to analyze the
static loadcases preceding the harmonic loadcases.
Performing Viscoelastic Analysis in the Frequency Domain

In frequency domain viscoelasticity, we study the stationary response of a viscoelastic material subject to harmonic loads to
capture the dynamic stiffness and damping effects of the material. In general, these effects are frequency dependent; but in
some materials, they may strongly depend on the vibration amplitude as well, which is known as the Payne effect.
Sometimes the amount of static pre-deformation plays a role and under non-isothermal conditions the material properties
may be temperature dependent. Frequency domain viscoelasticity is always based on a linearization around a static
equilibrium state, which can be the undeformed stress-free state or some other deformed equilibrium state, and is,
therefore, limited to small amplitude vibrations around the static equilibrium state, as illustrated in Figure 5-92. Here, a
displacement excitation u  t  with harmonic amplitude û and static pre-deformation u * is applied and the corresponding
reaction force response R  t  is measured. This reaction force response consists of a static part R * and a harmonic part
with amplitude R̂ being out of phase with the applied harmonic displacement. This phase shift is characterized by  and
reflects the fact that the material has internal damping.
û sin  t 
u  t  = u * + û sin  t 
R  t  = R * + R̂ sin  t +  
u*
Figure 5-92 Illustration of a Model Subject to a Large Static Pre-deformation and a Small Harmonic Excitation
However, within this amplitude range, the size of the amplitude may still have a substantial effect on the dynamic response,
turning the frequency response problem into a nonlinear problem. The main ingredients to perform this type of analysis
are:
CHAPTER 5 441
Material Library
1. The elastic material properties specifying the instantaneous (short term) stiffness properties; although, when tables
are used to define storage and loss moduli, this condition can be relaxed. These properties are entered through the
ISOTROPIC, MOONEY, GENT, ARRUDBOYCE, MARLOW, OGDEN, FOAM, or ORTHOTROPIC model definition options.
2. The viscoelastic material data. These are entered through the VISCELPROP, VISCELMOON, VISCELOGDEN, VISCELFOAM, or
VISCELORTH model definition options depending on the type of elastic material being used. In addition amplitude
effects may be included through the PAYNE model definition option, temperature effects may be included through
the SHIFT FUNCTION model definition option specifying TRS material behavior and effects of the static pre-
deformation may be included through the VISCELCORR model definition option. The PAYNE and VISCELCORR options
are supported for isotropic materials only. For some models of the PAYNE option, no VISCELMOON or similar data is
required (see also Table 5-14).
3. The HARMONIC parameter. The complex damping flag on this parameter is required, since for any nonzero frequency
there are always damping contributions. It is also recommended to include the inertia effects in the calculation of
the reaction forces. When the Payne effect is present this is done automatically.
4. The static loadcases applying the pre-stress. Note that these loadcases use the long term elastic material properties,
since in an equilibrium state it is assumed that the material has fully relaxed. In static loadases, the viscoelastic
material data is, therefore, only used to compute the long term elastic properties of the material, but, in these
loadcases, there is no creep or relaxation.
5. The harmonic loadcases applying the harmonic excitations. In the sub-increments of these loadcases, the actual
viscoelastic analysis is performed for a specified frequency range. Multiple harmonic loadcases can be chained
together where each harmonic loadcase defines a different harmonic load factor on the HARMONIC history definition
option, but uses the same loads and boundary conditions and defines the same frequency sweep.
6. If the Payne effect is present in the analysis, the harmonic sub-increments become nonlinear and criteria for
convergence testing need to be entered via the CONTROL option.
In Table 5-14, an overview is given of the different supported combinations of the options mentioned in point 2.
Table 5-14 Possible Ways of Defining Frequency Dependent Material Behavior

Combination Type
I II III IV V VI VII VIII IX X
VISCELMOON,
VISCELOGDEN, Prony series x x x o
VISCELFOAM,
VISCELPROP
Tables x x o
VISCELORTH Prony series x
Tables x
Table 5-14 Possible Ways of Defining Frequency Dependent Material Behavior (continued)
Combination Type
I II III IV V VI VII VIII IX X
PAYNE 1: Thixotropic x
2: Triboelastic x
3: Combined x
4: Kraus/Ulmer x
5: Tables x x
6: UPAYNE x
VISCELCORR Any model x x x x x x o
SHIFT FUNCTION Any shift function x x x x x o x x
The “x” indicates which material models chosen on a specific model definition option can be combined together in each
combination type, it does not imply they are all required simultaneously. The “o” indicates that the options can optionally
be used in conjunction with the UPAYNE user subroutine.
When material properties are temperature dependent, it is assumed that the temperature distribution in the model
corresponds to a steady state temperature distribution; although, in general, this cannot be enforced or checked by Marc,
so the user must take care of this. The temperature distribution used in the harmonic sub-increments is the temperature
distribution of the last static increment preceding them and these temperatures cannot vary over the harmonic sub-
increments.
If CONTACT is used in a viscoelastic analysis in the frequency domain, it should be noted that the contact conditions of the
last static increment before a harmonic loadcase are “frozen” in the harmonic sub-increments of the harmonic loadcase.
This means that no new contact or separations can occur in the harmonic sub-increments of a harmonic loadcase.
The stiffness and damping aspects of viscoelastic materials, when observed in the frequency domain, are characterized by
the storage and loss moduli, which are generally frequency dependent. The viscoelastic data can be entered as Prony series
from which Marc can compute the storage and loss moduli itself as a function of frequency. In many instances this data is
taken from frequency response measurements where amplitude responses and phase changes have been measured. From
these responses, the storage and loss modulus can be evaluated directly, and it is not necessary to evaluate Prony parameters
through a curve fitting procedure to obtain the representation of the viscoelastic material data. Therefore, as an alternative
to the Prony series, the storage and loss moduli can each be entered directly through a TABLE. These tables have to be
normalized consistently with a storage modulus value at some reference frequency as outlined in the section Using Tables to
Define Storage and Loss Moduli. The storage factor value evaluated from the storage table at zero frequency multiplied with the
stiffness entered as the static elasticity stiffness should result in the long term equilibrium stiffness of the material. The static
elastic stiffness should correspond to the storage modulus at the reference frequency. If, e.g., the instantaneous (short term)
stiffness is used to normalize the tables, all table function values are between zero and one, but if the long term equilibrium
stiffness is used to normalize the tables, the storage table function value at zero frequency is one and all other storage table
function values are larger than one.
To illustrate this normalization, we may consider the following example of a rubber cube being loaded by a simple shear as
shown in Figure 5-93. The loading is displacement controlled by prescribing the displacement of the top surface of the cube;
CHAPTER 5 443
Material Library
i.e., the static preload is applied by the prescribed static displacement u* (here we make the convention that all quantities
associated with the static equilibrium state are superscripted with an asterisk, *) and the harmonic load is applied by the
prescribed harmonic displacement û sin  t  , which is assumed to be superposed to the prescribed static displacement.
The material is assumed to be incompressible, so the applied deformation field does not cause out-of-plane displacements.
u û sin  t 
u*

L0
x
L0
Figure 5-93 Rubber Cube Under Simple Shear Loading
u
The main deformation parameter for this model is  = tan    = ------ . If the instantaneous material response is
L0
represented by a strain energy function W 0  I 1 ,I 2  or W 0   1 , 2 , 3  where I 1 ,I 2 are the first two invariants of strain and
2 2 2
 1 , 2 , 3 are the principal stretch ratios (note that because of incompressibility the third invariant I 3 =  1  2  3 = 1
and the third principal stretch ratio  3 = 1 , so they are both independent of the prescribed displacement), then this
strain energy function can be completely expressed in terms of  and the associated instantaneous shear stress can be given
as:
W 0
 0 = -----------

Because of the viscoelastic material behavior, the general relation between the shear stress and the shear deformation based
on the Simo model for finite strain viscoelasticity can be given by the following hereditary integral

dg  s 
  t  = g 0 t  +  -------------   0  t – s  –  0  t   ds
ds
0
where t is the time parameter, s the history parameter, and g the normalized relaxation function. When studying small
dynamic oscillations around a static equilibrium state with shear stress *  *  , we can write the total shear deformation
as   t  = * +   t  and the total shear stress as   t  = * +   t  . The shear stress increment   t  can now be
given by following hereditary integral:

dg  s 
  t  = g   0  t  +  -------------   0  t – s  –  0  t   ds
ds
0
We may linearize the relation between the instantaneous shear stress increment and the shear deformation increment as:
 2 W 0  * 
 0  t  = -------------------------   t  = G 0  *   t 
 2
where G 0  *  can be interpreted as the deformation dependent instantaneous shear modulus. When ignoring the mass
(i.e. inertia) of the rubber material and with prescribed harmonic deformations of the form   t  = ̂e i  t , we obtain
harmonic shear stresses of the form   t  = ̂e i  t , where ̂ and ̂ are the complex amplitudes of the vibrations and i
is the imaginary unit  i 2 = – 1  , and the dynamic equation of motion in the frequency domain is given as
 g stor age    + ig los s    G 0  * ̂ = ̂
For high frequencies, we know that g s t or a ge  1 and g l o ss  0 ; so in this case, the equation reflects the instantaneous
static response. When applying a prescribed ̂ (for which we assume a zero phase angle; i.e., its imaginary part is zero) we
measure a stress ̂ in phase with the prescribed deformation and the ratio ̂  ̂ provides the instantaneous shear modulus.
For lower frequencies, we measure a stress out of phase with the deformation, where the phase angle  of the stress can be
Im  ̂ 
represented as tan    = --------------- and its amplitude as ̂ = Re  ̂  2 + Im  ̂  2 . The amplitude of stress and the phase
Re  ̂ 
angle are the actual results of our measurement. In this case, we obtain:
Re  ̂  = ̂ cos    = g s t or a ge   G 0  * ̂ = G   ̂

and
Im  ̂  = ̂ sin    = g l o ss   G 0  * ̂ = G   ̂
We observe that G    and G    are the storage and loss modulus of the material which appear in the complex stress-
strain relation:
 G    + iG    ̂ = ̂
This means, when we drive the deformation with a strain input of the form ̂ sin  t  , we obtain a stress output of the
form ̂  G    sin  t  + G    cos  t   , which can be written as , ̂ Ĝ    sin  t +   with
G = G    2 + G    2 . If, on the other hand, we drive the deformation with a stress input of the form
̂
̂ sin  t  , we obtain a strain output of the form ----------------- sin  t –   .
Ĝ   
CHAPTER 5 445
Material Library
As was indicated earlier, the material constants entered for Hooke’s law or for any strain energy function of a hyperelastic
material are the instantaneous (short term) elasticity constants; although, when tables are being used, this condition may
be relaxed. However, in any static loadcase preceding a harmonic loadcase, Marc uses the long term elasticity constants,
since in a static equilibrium state the material must have fully relaxed to its long term state. If the viscoelastic properties are
N t-
– ----

entered through the Prony series g  t  = g  +  gk e k , Marc can evaluate g  from the input as:
k = 1
N
g = 1 –  g k , so the long term material constants are easily established. If the storage and loss information is entered
k = 1
through tables, Marc needs to evaluate the long term storage factor from the table at zero frequency (for   0 , we obtain
the long term material behavior). Therefore, the table input must be such that the proper value at zero frequency can be
evaluated either by including a table input for zero frequency or by making sure that the method that Marc uses to evaluate
tables for independent variable values outside the range defined in the input can return the right value. By default, Marc
uses the endpoint value of the table, so this would then be the value at the lowest frequency in the table. However, if the
extrapolation method is used, the user has to make sure that the endpoint table value and the tangent at the endpoint that
is used in the extrapolation result in the right long term value.
The main results obtained from a viscoelastic analysis in the frequency domain are summarized in Table 5-15. The complete
list can be found in the list of the POST model definition option.
Table 5-15 Main Results from a Viscoelastic Analysis in the Frequency Domain
Nodal Results Post Codes
û Real and imaginary parts of the harmonic displacements By default on post file
R̂ Real and imaginary parts of the harmonic residual and reaction By default on post file
forces
F̂ Real and imaginary parts of the external forces By default on post file
Element Results Post Codes

̂ Real and imaginary parts of the harmonic stress tensor 351 (real part)
361 (imaginary part)
̂ Real and harmonic parts of the harmonic strain tensor 621 (real part)
631 (imaginary part)
W d iss The dissipated energy per unit time and unit volume (specific 620
dissipated power)
Element Results Post Codes

g The storage factor 861 (deviatoric or coupled response)
863 (volumetric response)
g The loss factor 862 (deviatoric or coupled response)
̂ The vibration amplitude 698 (deviatoric or coupled response)
Temperature Effects and Coefficient of Thermal Expansion

Experimental results indicate that a large number of material properties vary with temperatures. In Marc, almost all
material parameters may be a function of temperatures when using the table driven input procedure. A subset of which are
shown in Table 5-16 for different types of analysis.
Please note that T is temperature in the above expressions. With the exception of heat transfer, Joule heating, or coupled
thermal-mechanical analysis, the temperature is a state variable.
Table 5-16 Temperature-Dependent Material Properties

Analysis Type Material Properties
Stress Analysis Modulus of elasticity (Young’s Modulus) ET
Poisson’s Ratio T
Yield Stress y  T 
Workhardening Slope hT
Coefficient of Thermal Expansion T
Mooney Constants C 01  T  C 10  T 
C 11  T  C 20  T 
C 30  T 
Heat Transfer Analysis Thermal Conductivity KT
Specific Heat CT
Emissivity T
Couples Thermo-Electrical Electric Resistivity T
(Joule Heating Analysis)
Hydrodynamic Heating Viscosity T
CHAPTER 5 447
Material Library
Piecewise Linear Representation

In Marc, the temperature variation of a material constant F  T  may be entered as a piecewise linear function of
temperature as shown in Figure 5-94 or as an equation. The base value or reference value is entered in the ISOTROPIC,
ORTHOTROPIC, ANISOTROPIC, MOONEY, OGDEN, etc. model definition options. Using the nontable input format, these should
be at the lowest temperatures. The variation with temperature is entered on the TEMPERATURE EFFECTS and ORTHO TEMP or
TABLE model definition options.
Temperature
Dependent
Property
F(T) F4
F3
S3
F2
S2
F1 S1
Base
Value
F1
T4
T1 T2 T3 Temperature (T)
(1) Slope-Break Point Data

Slope Break Point
S1 = (F2 - F1)/(T2 - T1) T1
S2 = (F3 - F2)/(T3 - T2) T2
S3 = (F4 - F3)/(T4 - T3) T3
(2) Function-Variable Data
Function Variable
F1 T1
F2 T2
F3 T3
F4 T4
Figure 5-94 Piecewise Linear Representation of Temperature-Dependent Material Properties

Temperature-Dependent Creep
In Marc, the temperature dependency of creep strain can be entered in two ways. The creep strain rate may be entered as
·
a piecewise linear function. If the creep strain  c can be expressed in the form of a power law:
·c
 = AT m
(5-596)
where A and m are two experimental constants, input the experimental constants through the CREEP model definition
option.
For other temperature dependency, you may use an equation or use the CRPLAW user subroutine for explicit creep and
UCRPLW user subroutine for implicit creep to input the variation of creep strain with temperature.
Coefficient of Thermal Expansion

Marc always uses an instantaneous thermal expansion coefficient definition
th
d i j =  ij dT in general (5-597)
or,
th
d i j =  dT  ij for the isotropic case (5-598)
In many cases, the thermal expansion data is given with respect to a reference temperature
 th =   T – T 0  (5-599)
where a is a function of temperature:
 = T (5-600)
Clearly, in this case
d
d th =  + -------  T – T 0  dT (5-601)
dT
so the necessary conversion procedure is:
1. Compute and plot Equation (5-599) in the form Equation (5-601)
d
 =  + -------  T – T 0  (5-602)
dT
as a function of temperature.
2. Model Equation (5-602) in the ANEXP user subroutine, or with piecewise linear slopes and breakpoints in the
TEMPERATURE EFFECTS or TABLE option.
The anisotropic coefficient of thermal expansion can be input through either the ORTHOTROPIC model definition
option or the ANEXP user subroutine.
CHAPTER 5 449
Material Library
Phase Transformations
Materials undergo phase transformations when the material is subjected to a thermal load. For H2O, we commonly think
of this as ice to water to vapor. These changes of state have influences on the material properties, including mechanical,
thermal, electrical, and optical. Additionally, the change in the phase also results in a change in entropy in the system which
results in latent heat. Metals undergo solid-state transformation that is driven by the differences in the Gibbs free energy
between the initial state and the final state. A special example of metal phase transformations is Shape Memory Alloys,
which is discussed elsewhere. Another example of phase transformations of nonmetallic materials is modeled in Marc as
Curing. Carbon steel exhibits a change in mechanical or thermal properties when quenched or air cooled from a sufficiently
high temperature. At any stage during the cooling process, these properties are dependent on both the current temperature
and the previous thermal history. The properties are influenced by the internal microstructure of the material, which in
turn depends on the rate at which the temperature changes. Only in instances where the temperature is changed very
gradually does the material respond in equilibrium, where properties are simply a function of the current temperature. The
transformations can be accompanied by changes in volume. Simulating the phase transformations of steel is important in
manufacturing processes such as heat treatment, rolling, forging, and welding to determine the material properties,
manufacturability, distortion, and durability of the part.
Because, thermal processes are never uniform in the material, the resultant properties are never truly homogeneous. So,
while we often consider the initial material to be homogeneous, the final product is not. The objective of heat treatment is
often to achieve desired heterogeneous behavior such as surface hardening.
Marc has two procedures for modeling phase transformations in metals. This section highlights the use of the
MICROSTRUCTURE model definition option, which the following section discusses the older TIME-TEMP model definition
option.
The following discussion is focused on steel, though the procedure is viable for other metals as well. The process is driven
by the nucleation and growth of different phases in the material. These transformations can be driven processes.
1. Long term behavior associated with diffusion-controlled transformations.
2. Interface-controlled transformations.
In Marc, the solid-state transformations are based upon the Johnson-Mehl-Avrami-Kolmogorav (JMAK) kinetics theory
[Ref. 56] . The method was popularized by Avrami, and is often called the Avrami equation [Refs. 57, 58, and 59]. The theory
is based upon the nucleation, growth and impingement of randomly distributed nuclei. This can be expressed as an
exponential process, let  be the volume fraction of a phase, then:
  T t  = 1 – exp  –  e  T t  
During the phase transformation, two thermodynamic processes occur:
1. The production of the new phase releases Gibbs free energy.
2. The creation of the interface and generation of strain raises the Gibbs free energy.
The creation of a new phase must have a critical size, else it increases the energy and is forced to shrink. The critical radius
for nucleation of a spherical particle may be derived as:
R c = 2 I e   G v 
where,
Ie is the interfacial energy per unit interface.

G v is the free Gibbs energy difference due to the volume.
3 2
The Gibbs free energy of nucleation G * = 16  I e  3 G v which is approximately A  T 2 so one can see that the
nucleation is dependent on the change in temperature.
The nucleation rate per unit volume I  T  , that is, the rate of formation of particles with a critical radius is determined by
the atomic mobility that can be express in an Arrhenius form.
I  T  = I 0 exp  – H n  T 
where:
I0 is approximately v* exp  S    .
H n is the temperature dependent activation enthalpy for nucleation.
 is the Stefan-Boltzman constant.
v is the effective attempt frequency.
S is the entropy.
For the special case, where the nucleation and growth rates are constant with respect to time, the volume of the phase can
then be expressed as
  T t  = 1 – exp  –  k  T  * t n  
It is also, written as
  t  = 1 – exp  –  t    n 
K  T  is composed of geometric factors, I 0 given earlier, U  T  is the temperature dependent growth rate, and n is the
JMAK exponent. Note n is a function of both the temperature and the time.
f d I 0  T U d / m  T 
KT = -----------------------------------------
-
n d
---- + 1
m
where,
m is the growth mode factor

m = 1 for interface controlled growth
m = 2 for diffusion controlled growth
n = d + m  m for continuous nucleation
Expression for the JMAK exponent
CHAPTER 5 451
Material Library
Table 5-17 Isothermal Transformations for JMAK Model

Continuous Nucleation Site Saturation
d d
JMAK exponent  n  ---- + 1 ----
m m
 n – 1 H g + H N H g
Activation enthalpy  H  ------------------------------------------------
-
n
Rate constant  K  f d I 0  T U n – 1  T  n f d I 0  T U n  T 
n -------------------------------------------
n
Table 5-18 Approximate Values of JMAK Exponent

Interface Controlled Diffusion Controlled
Dimensionality Nucleation Model Transformation  m  = 1 Transformation  m  = 2
Three-dimensional Site Saturation 3 1.5
d = 3 Continuous Nucleation 4 2.5
Decreasing Nucleation 3-4 1.5 - 2.5
Two-dimensional Site Saturation 2 1
d = 2 Continuous Nucleation 3 2
Decreasing Nucleation 2-3 1-2
One-dimensional Site Saturation 1 0.5
d = 1 Continuous Nucleation 2 1.5
Decreasing Nucleation 1-2 0.5 - 1.5
In Marc, the parameters of the equation can be determined for each temperature by the points for start (1%) and end (99%)
of formation. Considering an isothermal behavior, this would be based upon the difference between P i , P s , and P f or
B s and B f in Figure 5-96.
A consequence of the JMAK equation is that transformations begin slowly, then accelerate and then slow down when the
transformation is complete. This is apparent in the typical S shaped figure shown as follows.
The diffusionless formation of Martensite is governed by Koistinen-Marburger equation.

Ms – T
------------------
 = 1– e KM
Parameters for the equation are determined by the values of Martensite start temperature and Martensite end temperature
as shown in Figure 5-96.
When temperature drops below Martensite start temperature M s all other transformations are stopped.
The material properties are based upon a linear mixture rule.
p ef f =  i pi
where p eff is the effective property and p i and  i are the property of phase i and  i is the phase fraction. It should be
noted that the phase transformation influence on thermal and electrical properties is significantly less that the influence on
the mechanical properties.
Associated with the change in phase the latent heat is added (subtracted) from the thermodynamic equilibrium equation.
The Latent heat is obtained as
L i
c L =  --------
T i
- 
where L i is the latent heat for phase i .
The phase transformation also results in volumetric changes in the material. These are represented as
 v =   vi  i
where  vi is volume change strain of phase i . This volumetric phase change is added to the thermal strain.
These are incorporated with the thermal strains. If the volumetric strains are large this may result in convergence issues.
There is an additional Transformation Induced Plasticity strain (TRIP) which is represented as:
 T RIP = 3.0  C TR I Pi  1 –  i  i s
CHAPTER 5 453
Material Library
where C TRIP is the TRIP coefficient of phase i ; s is the deviatoric stress.

i
Phases of Steel
There are many grades of steel and for a nice review see [Ref. 61] and [Ref. 62]. Steel is primarily iron, which is alloyed with
carbon, and other atoms including magnesium, chromium, nickel, tungsten, molybdium, cobalt, etc. Iron is an allotropic
element, that is it exists in multiple phases. When pure iron is slowly cooled from its liquid state to room temperature, it
undergoes transformations from liquid to  phase at 1534°C, and at 1390°C from  (Ferrite)to  (Austenite) at atC
from  to  phase. These phases have different crystal structures  and  phase are Body Centered Cubic (BCC) while
 is Face Centered Cubic (FCC).
These transformations can have a significant impact on the mechanical properties, for example the tensile strength of 100%
ferrite is 40 KSI, while that of 100% pearlite is 120 KSI.
The rate of decomposition of Austenite in steel depends on how fast the carbon atoms diffuse, which is a thermally driven
process. Hence, there is a significant different composition when the steel is cooled rapidly versus slowly.
The temperatures at which phase transformations occur are also dependent upon all the alloying elements. As can be seen
in Figure 5-95, the carbon content in the steel has a significant impact on the material behavior. This is typical represented
by the phase diagram for steel which appears as:
Figure 5-95 Phase Transformation Diagram
The carbon content is classifed as:

 Eutectoid – 0.77% carbon content
 Hypo-eutectoid – less than 0.8% carbon content
 Hyper-eutectoid – greater than 0.8% to 2% carbon content
Austenite (  FCC – generally only found at temperatures above 723°C. It can exist at room temperature if alloyed with
nickel or magnesium. It has maximum of 2% carbon content.
Ferrite (  and/or  ) – BCC – Alpha ferrite forms by slow cooling of austenite, so the diffusion of carbon occurs.  ferrite
contains very little carbon – 0.02% at Eutectoid temperature, and less at room temperature. Delta ferrite is the high
temperature form, when cooling low carbon content alloys, from the liquid state before transforming to austenite.
Cementite – Fe 3 °C is a very hard intermetallic compound with as much as 6.7% carbon content. Slower cooling results
in pearlite. Faster cooling results in harder and tougher ferrite and cementite compound. It is often considered to be Hyper-
eutectoid.
Pearlite – is a multi-phase state composed of alternating layers of ferrite and cementite. It is not a real phase, but a combined
microstructure. Formed during slow cooling of iron alloys that is diffusion controlled, and can begin at 1150°C.
CHAPTER 5 455
Material Library
Compound is softer with reduced toughness. Pure pearlite has less than 0.8% carbon content. It is often considered to be
Eutectoid or Hypo-eutectoid.
Bainite – may form if the (Austenitic) steel is cooled rapidly to between 200-400°C and held there. It has a finer structure
and has a higher strength than Pearlite. Like Pearlite, it is not a pure phase, but rather a combined microstructure.
Martensite – may form if the (Austenitic) steel is quenched to a lower temperature. It is extremely hard and brittle. As the
carbon does not diffuse, it is locked in the lattice distorts to a Body Centered Tetragonal (BCT). Material is extremely hard,
but brittle.
The very fast direct transformation from Austenite to Martensite is diffusionless and is modeled by the Koistinen-
Marburger model.
Transformation Temperatures
The critical temperatures A i symbolize arrests in heating or cooling, the letter A from the French word arret. The subscripts
r and c originates from the French words refroidissement (meaning cooling) and chauffage (heating), respectively. The faster
the rate of change, the larger the gap between A r and A c and the equilibrium point A.
The following principal temperatures can be distinguished for steels [Ref. 63]:
Ae1: Equilibrium temperature defining the lower limit of existence of Austenite.

Ae3: Equilibrium temperature defining the upper limit of existence of Ferrite
Aem: Equilibrium temperature defining the upper limit of existence of Cementite in a hypereutectoid steel.
AC1: temperature at which Austenite begins to form during heating.
AC3: Temperature at which Ferrite completes its transformation to Austenite during heating.
ACm: Temperature at which the Cementite in a hypereutectoid steel dissolves completely during heating.
Ar1: Temperature at which Austenite completes its transformation to Ferrite and Cementite during cooling.
Ar3: Temperature at which Ferrite begins to form during cooling.
Arm: Temperature at which Cementite begins to form in a hypereutectiod steel during cooling.
Ms: Temperature at which the Austenite begins to transform to Martensite during cooling.
Mf: Temperature at which the Austenite has completely transformed to Martensite during cooling.
Mx: Temperature at which x% of the Austenite has transformed to Martensite during cooling.
When carbon is present, transformation temperatures are lowered, and the transformations occur over a range of
temperatures, not a precise temperature.
In addition to the conventional material properties, one can enter either the TTT data the CCT data or both.
A Time-Temperature Transformation Curve (TTT) is a set of curves drawn in a semi-logarithmic coordinate, with
logarithmic time and temperature data providing the start and end of the transformation from the Austenite phase. A
typical T-T-T curve for a eutectic steel is shown as follows. The area of interest is often in the nose of this curve.
Figure 5-96 Steel TTT Diagram
In Marc, this data is entered as the Temperature versus Time for 1% and 99% volume fraction of the generated phase (see
Figure 5-99).
The CCT data for a typical steel is shown as follows:
Figure 5-97 Steel CCT Diagram

CHAPTER 5 457
Material Library
Heat Treatment
Cold working of metals is known to have been performed by ancient mankind, the first evidence of annealing occurred
around 5000 BCE, and carburization and quenching occurred as early as 1000 BCE. A scientific understanding of the
process did not begin until the mid-1800’s. The objective of heat treatment is to apply thermal processes such the material
takes on the desired characteristics from a ductility (or malleability), hardness, strength, durability or impact resistance
perspective. One can influence the resultant product by controlling the thermal process which changes the crystal structure
as discussed below.
Spheroidizing – the process where the high carbon steel is heated to greater than 700°C, and held for a prolonged time,
typically, between 8 and 30 hours. This is a diffusion controlled resulting in a fine structure of rods or spheres of cementite
within ferrite or pearlite, depending on the carbon content. If the fine structure appears as spheres, it is commonly called
spheroidite.
Annealing – carbon steel is heated to approximately 40°C above the Ac 3 or A C M for one hour, which results in recovery
and recrystallization, such that all the Ferrite transforms to Austenite. The steel is then cooled slowly (usually in the
furnace) resulting in a coarse pearlite structure. The annealed steel is softer and more ductile. The density of the dislocations
is decreased; hence, the effectively the equivalent plastic strain is removed and the internal stresses. The result is that the
yield stress becomes the initial yield stress. In addition, any anisotropy (or texture) induced by cold working is also
eliminated.
Process annealing – similar to annealing, for low carbon steels, but the heating temperature is usually lower, typically 80-
170°C below the A 1 temperature.
Normalizing – the carbon steel is heated to approximately 55°C above Ac 3 or A C M for one hour, such that the steel
completely converts to austenite, it is then air cooled. Result is a fine pearlite structure.
Quenching - used for steels with greater than 0.4 % carbon content. The carbon steel is heated to approximately 55 C
above Ac 3 or A CM for one hour and then rapidly cooled in water, brine or oil. During fast cooling, Austenite cannot
transform to Ferrite or Pearlite by diffusion. The result is a Martensite structure. This can result in surface cracks.
Martempering – an isothermal process where the temperature is right above the Martensite temperature in a brine, which
relieves the stresses associated with quenching.
Quench and tempering – reheating quenched (Martensite) steel to a temperature above the eutectoid temperature followed
by cooling. The heating allows carbon to diffuse out of the distorted martensite lattice structure, relieving internal stresses.
Some spheroidite form, which restores ductility. The combined effect of quenching followed by tempering, toughens the
material.
Case Hardening – hardens only the exterior surface creating a hard, wear resistant skin, but a tough ductile interior. This
may be done by infusing carbon into the surface through a controlled diffusion process. It may also involve very controlled
heating of the surface using induction heating.
Carburizing – a type of case hardening where the hypo-eutectoid steel surface absorbs carbon. Surface depth is as much as
6-7 mm.
The consequences of Annealing and Normalizing on the mechanical behavior is shown in Figure 5-98 [Ref. 62].
Figure 5-98 Influence of Annealing on Mechanical Properties
Material Input
All the required material data to perform a thermal/structural analysis involving phase transformations is assumed to be
defined in a material data base file with the extension .umt. Compared to a standard material, the most important
differences are:
 Material properties are defined separately for the individual phases (Austenite, Ferrite, Pearlite, Bainite, and
Martensite) and combined into a single material. Since the calculation will be based on effective properties, the
properties of the individual phases must be of the same type (one can, for example, not mix von Mises and Hill
plasticity).
 The transformation from one phase to another is driven by a TTT (Time Temperature Transformation) diagram
(see Figure 5-99a). If available, extra data can be provided using CCT (Continuous Cooling Transformation)
diagram. Note that the individual curves of the TTT diagram are provided as tables, where the temperature is the
independent variable(see Figure 5-99b).
CHAPTER 5 459
Material Library
(a) TTT Diagram (b) User-defined Temperature Time Curve

or 1% Transformation
Figure 5-99 Mentat Definition of Transformation Curves

 Phase transformations are accompanied by, for example, volume changes and heat effects, which are provided
using tables as a function of the temperature.
Once the various tables required to characterize the individual phases and the transformation behavior are known, creating
a .umt file is relatively easy. An example may be found in the Marc User’s Guide, Example 117.
Below the A1 temperature, Ferrite Phase formation is possible.

Below the A3 temperature, the Pearlite phase formation is possible.
0.2% Proof Stress (as a function of Cooling Rate). This is often considered
to be the yield stress for steels.
Tensile Strength (as a function of Cooling Rate). This is only used for
postprocessing.
Hardness (as a function of Cooling Rate). This is only used for
postprocessing.
Continuous Cooling Volumetric fraction of phase during the
transformation (as a function of Cooling Rate)
One needs to enter the transformation data from the first phase (usually
Austenite) to the other phases as a function of temperature at 1% and 99%
volume fraction.
The latent heat associated with the transformation. Note in many thermal
processes, such as laser welding, this value is considered to be small and can
be ignored.
Volume change – which is the volumetric strain induced by the phase
transformation (as a function of temperature.
Transformation Induced Plasticity.
Figure 5-100 Menu to Enter the Data for a .umt File
Time-Temperature-Transformation
The occurrence of phase change is also dependent on the rate of cooling of the material. This relationship is shown in a
typical cooling diagram (see Figure 5-101). The curves A, B, and C in Figure 5-101 represent the temperature history of a
structure that has been subjected to a different cooling rate. It is obvious that the structural material experiences phase
changes at different times and temperatures, depending on the rate of cooling. Under cooling rate A, the material changes
from phase 1 to phase 4 directly. The material undergoes three phase changes (phase 1 to phase 2 to phase 3 to phase 4)
for both cooling rates B and C. However, the phase changes take place at different times and temperatures.
The Time-Temperature-Transformation (TIME-TEMP) option allows you to account for the time-temperature-
transformation interrelationships of certain materials during quenching or casting analyses.
Use the T-T-T parameter to invoke the time-temperature-transformation. Input all the numerical data required for this
option through the TIME-TEMP model definition option.
In a transient heat transfer analysis, the thermal properties which can be defined as a function of time and temperature are
the thermal conductivity and the specific heat per unit reference mass. Here, the effects of latent heat or phase
transformation can be included through the definition of the specific heat.
In a thermal stress analysis, the mechanical properties which can be defined as a function of time and temperature are the
Young’s modulus, Poisson’s ratio, yield stress, workhardening slope, and coefficient of thermal expansion. The effects of
CHAPTER 5 461
Material Library
volumetric change due to phase transformation can be included through the definition of the coefficient of thermal
expansion.
A T1 T1
B C
T4 Line of Phase Change –

Change in Volume
Temperature (T)
T5
T3 T3
T2 T2
t1 t2 Time (t)
Figure 5-101 Simplified Cooling Transformation Diagram
Test data must be available in a tabular form for each property of each material group. For a given cooling rate, the value
of a property must be known at discrete points over a range of temperatures. There can be several sets of these discrete
points corresponding to measurements at several different cooling rates. The cooling tests must be of a specific type known
as Newton Cooling; that is, the temperature change in the material is controlled such that
T  t  = A exp  – at  + B (5-603)
In addition, a minimum and a maximum temperature that bracket the range over which the TIME-TEMP option is meant to
apply must also be given.
For the simulation of the cooling rate effect in finite element analysis, material properties of a structure can be assumed as
a function of two variables: time and temperature. Two-dimensional interpolation schemes are used for the interpolation
of properties.
Interpolation is based on making the time variable discrete. Stress analysis is carried out incrementally at discrete time
stations and material properties are assumed to vary piecewise linearly with temperature at any given time. These
temperature-dependent material properties are updated at each increment in the analysis. For illustration, at time t 1 , the
material is characterized by the phase 1 and phase 4 behaviors at temperature ranges T 1 to T 3 , and T 3 to T 2 , respectively
(see Figure 5-102). Similarly, at time t 2 , the material behavior must be characterized by all four phases, each in a different
temperature range (that is, phase 1, T 1 to T 4 ; phase 2, T 4 to T 5 ; phase 3, T 5 to T 3 ; phase 4, T 3 to T 2 ). The selection
of an interpolation scheme is generally dependent on the form of the experimental data. A linear interpolation procedure
can be effectively used where the properties are expressed as a tabulated function of time and temperature.
During time-temperature-transformation, the change in volume in a stress analysis is assumed to take place in a
temperature range T . The change in volume is also assumed to be uniform in space, such that the effect of the volume
change can be represented by a modification of the coefficient of thermal expansion. For a triangular distribution of   T 
in the temperature range T , the value of the modified coefficient of thermal expansion is:
2
 m = -------  3 1 +  – 1  (5-604)
T
where  is the change in volume. A schematic of the modified   T  is shown in Figure 5-102.
T
Coefficient of thermal Expansion ()
Temperature (T)
m
Figure 5-102 Modified Coefficient of Thermal Expansion for Short-Time Change in Volume
Low Tension Material

Marc can handle concrete and other low tension material. The CRACK DATA option assists in predicting crack initiation and
in simulating tension softening, plastic yielding and crushing. This option can be used for the following:
 Elements with a one-dimensional stress-strain relation (beam and truss elements)
 Elements with a two-dimensional stress-strain relation (plane stress, plane strain, axisymmetric, and
shell elements)
 Three-dimensional elements (bricks)
Analytical procedures that accurately determine stress and deformation states in concrete structures are complicated by
several factors. Two such factors are the following:
 The low strength of concrete in tension that results in progressive cracking under increasing loads
 The nonlinear load-deformation response of concrete under multiaxial compression
Because concrete is mostly used in conjunction with steel reinforcement, an accurate analysis requires consideration of the
components forming the composite structure. Steel reinforcement bars are introduced as rebar elements.
Each rebar element must be input with a separate element number. The REBAR model definition option or REBAR user
subroutine is used to define the orientation of the reinforcement rods.
CHAPTER 5 463
Material Library
Uniaxial Cracking Data

The cracking option is accessed through the ISOTROPIC option. Uniaxial cracking data can be specified using the CRACK DATA
option or the UCRACK user subroutine. When the CRACK DATA option is used to specify uniaxial cracking data, the following
must be specified: the critical cracking stress, the modulus of the linear strain softening behavior, and the strain at which
crushing occurs. Material properties, such as Young’s modulus and Poisson’s ratio, are entered using the ISOTROPIC option
and the WORK HARD option. This model is for a material which is initially isotropic; if the model is initially orthotropic, see
FAIL DATA for an alternative cracking model. A typical uniaxial stress-strain diagram is shown in Figure 5-103.
Low Tension Cracking

A crack develops in a material perpendicular to the direction of the maximum principal stress if the maximum principal
stress in the material exceeds a certain value (see Figure 5-104). After an initial crack has formed at a material point, a second
crack can form perpendicular to the first. Likewise, a third crack can form perpendicular to the first two. The material loses
all load-carrying capacity across the crack unless tension softening is included.
Tension Softening
If tension softening is included, the stress in the direction of maximum stress does not go immediately to zero; instead the
material softens until there is no stress across the crack. At this point, no load-carrying capacity exists in tension (see
Figure 5-104). The softening behavior is characterized by a descending branch in the tensile stress-strain diagram, and it may
be dependent upon the element size.
cr
Es
crush

E Young’s Modulus
Es Tension-Softening
E Modulus
y Yield Stress
cr Critical Cracking Stress
y
crush Crushing Strain
Workhardening
Figure 5-103 Uniaxial Stress-Strain Diagram

2
y
1

1
2
Figure 5-104 Crack Development
Crack Closure
After a crack forms, the loading can be reversed; therefore, the opening distance of a crack must be considered. In this case,
the crack can close again, and partial mending occurs. When mending occurs, it is assumed that the crack has full
compressive stress-carrying capability and that shear stresses are transmitted over the crack surface, but with a reduced shear
modulus.
Crushing
As the compressive stress level increases, the material eventually loses its integrity, and all load-carrying capability is lost;
this is referred to as crushing. Crushing behavior is best described in a multiaxial stress state by a crushing surface having
the same shape as the yield surface. The failure criterion can be used for a two-dimensional stress state with reasonable
accuracy. For many materials, experiments indicate that the crushing surface is roughly three times larger than the initial
yield surface.
Analysis
The evolution of cracks in a structure results in the reduction of the load carrying capacity. The internal stresses need to be
redistributed through regions that have not failed. This is a highly nonlinear problem and can result in the ultimate failure
of the structure. The AUTO INCREMENT or AUTO STEP option should be used to control the applied load on the structure.
Soil Model
Soil material modeling is considerably more difficult than conventional metals, because of the nonhomogeneous
characteristics of soil materials. Soil material usually consists of a large amount of random particles. Soils show unique
properties when tested. The bulk modulus of soil increases upon pressing. Also, when the preconsolidation stress is
exceeded, the stiffness reduces dramatically while the stiffness increases upon unloading. At failure, there is no resistance to
shear, and stiff clays or dense sands are dilatant. Over the years, many formulations have been used, including linear elastic,
nonlinear elastic, Drucker-Prager or Mohr-Coulomb, and Cam-Clay and variations thereof. In Marc, the material models
available for soil modeling are linear elasticity, nonlinear elasticity, the Cam-Clay model, von Mises, linear Mohr-Coulomb,
parabolic Mohr-Coulomb, orthotropic elastic, and exponential cap model.
CHAPTER 5 465
Material Library
Elastic Models
Linear elasticity is defined in the conventional manner, defining the Young’s moduli and the Poisson’s ratio in the SOIL
option. The nonlinear elasticity model is implemented through the NLELAST option or the HYPELA2 user subroutine. Some
of the simplest models include the bilinear elasticity, where a different moduli is used during the loading and unloading
path, or to represent total failure when a critical stress is obtained.
A more sophisticated elastic law is the hyperbolic model, where six constants are used. In this model, the tangent moduli are
R f  1 – sin     1 –  3  2  n
E = 1 – ---------------------------------------------------------  -----3- (5-605)
2c cos  + 2 3 sin   pa
Two of the elastic models are the E- and K-G variable elastic models. In the E- model, Poisson’s ratio is considered
constant and
E = E0 + E p + E  (5-606)
while, in the K-G model

K = K0 + K p (5-607)
G = G0 + G p + G  (5-608)
The key difficulty with the elastic models is that dilatancy cannot be represented.
Cam-Clay Model
The Cam-Clay model was originally developed by Roscoe, and then evolved into the modified Cam-Clay model of Roscoe
and Burland. This model, which is also called the critical state model, is implemented in Marc.
The yield surface is an ellipse in the p, plane as shown in Figure 5-105, and is defined by
2
F = ---------- – 2pp c + p 2 = 0 (5-609)
M cs 2
where p c is the preconsolidation pressure, and M c s is the slope of the critical state line.
The Cam-Clay model has the following no properties. At the intersection of the critical state line and the ellipse, the normal
to the ellipse is vertical. Because an associated flow rule is used, all plastic strain at failure is distortional; the soil deforms
at constant volume (Figure 5-105). The strain hardening and softening behavior are shown in Figure 5-106. Also, if the
preconsolidation pressure is large, the soil remains elastic for large stresses. The evolution of the preconsolidation pressure is
·
p· c = – p c tr   pl  (5-610)
where,
1+e
 = ------------- (5-611)
–
where,
e is the void ratio

 is the virgin compression index (see Figure 5-107)
 is the recompression index (see Figure 5-107)

(q)
Critical State Line Strain Hardening
Ceases
Strain Softening
Ceases 2
1
3
Region
originally
4 elastic
P   v p
CO
Figure 5-105 Modified Cam-Clay Yield Surface
q q
1 3
q2
4
q4
q q
Figure 5-106 Strain Hardening and Softening Behavior
1  1

ln  – p 
Figure 5-107 Response of Idealized Soil to Hydrostatic Pressure

CHAPTER 5 467
Material Library
The void ratio and the porosity are related by the expression:
e
 = ------------ (5-612)
1+e
In the modified Cam-Clay model, it is also assumed that the behavior is nonlinear elastic with a constant Poisson’s ratio,
the bulk modulus behave as:
1+e
K = – ------------ p (5-613)

Note that this implies that at zero hydrostatic stress, the bulk modulus is also zero. To avoid computational difficulties, a
cutoff pressure of one percent of the preconsolidation pressure is used.
This constitutive law is implemented in Marc using a radial return procedure. It is available for either small or large strain
analysis.When large displacements are anticipated, you should use the LARGE STRAIN parameter.
The necessary parameters for the Cam-Clay soil model can be obtained by the following experiments:
1. Hydrostatic test: determines volumetric elastic bulk modulus, yield stress, and the virgin and recompression ratio
of soils.
2. Shear-box test: determines the slope of critical state line and shear modulus of the soil. However, the tests have to
be calibrated with numerical simulations to get the necessary constants.
3. Triaxial shear test: the most comprehensive experimental information and obviates the need for the first two tests
and obtains all the necessary constants listed above.
Evaluation of Soil Parameters for the Critical State Soil Model

To illustrate how to extract soil parameters, namely M ,  , and  , a hypothetical data set for a normally consolidated clay
in presented in Figures 5-108 to 5-114. Now we shall use these data to show the procedure of determination of parameters
for the critical state model. The data presented here pertain to conventional triaxial conditions.
Figures 5-108 and 5-109 show the stress-strain relations for constant pressure tests under different initial conditions p0=10
(69), 20 (139), and 30 (207) psi (kPa), respectively. Here the mean effective pressure of the soil sample is kept constant
throughout the test. The last data point shown on the deviatoric stress (q) versus axial strain (1) plot for any test is
considered as the ultimate condition for that test. The void ratio values at the beginning and the end of each test are given
in Table 5-19.
Figures 5-111 to 5-113 show stress-strain relation plots for three fully drained tests performed at initial pressures 10, 20, and
30 psi. Here, there is no pore pressure development, and the effective mean pressure increases during the test. The void
ratio values at the beginning and end of each test are given in Table 5-19.
Figures 5-114 to 5-116 show stress-strain behavior under undrained conditions. Here there is no change in volume of the
sample. However, fluid pore pressure are developed during the test, thereby reducing the effective stresses until the ultimate
(failure) state is reached.
Figure 5-117 shows the variation of void ratio with mean pressure which is obtained from a hydrostatic compression (HC)
test with three load reversals (that is, unloading-reloading cycles). In this figure, void ratio is plotted to a linear scale while
the hydrostatic stress is plotted to a logarithmic (base 10) scale. In fact, this is the usual way of presenting one-dimensional
consolidation or hydrostatic test results in geotechnical engineering practices.
Determination of parameter, M: The parameter M is the slope of the critical state line on a p-q plot. To determine its value,
the values of p and q at ultimate conditions for each test are plotted as shown in Figure 5-118; the ultimate condition for
each test is assumed to be the last point plotted in Figures 5-108 to 5-116. At the ultimate conditions, the sample undergoes
excessive deformations under constant deviatoric stress, and hence it could be taken as the asymptotic stress to the curve.
The slope of the critical state line (Figure 5-118) is calculated as 1.0 for the soil above. That is, M = 1.0 .
Table 5-19 Test Values for Cam Clay Model

Initial Final
Pressure, p0 Initial Pressure, pf Final
(effective) void Ratio, (effective) Void Ratio,
Test (psi) e0 (psi) e1
p-constant 10 1.080 10 0.980
p-constant 20 0.959 20 0.860
p-constant 30 0.889 30 0.787
Drained 10 1.080 15 0.908
Drained 20 0.959 30 0.787
Drained 30 0.889 45 0.716
Undrained 10 1.080 05.55 1.080
Undrained 20 0.959 11.09 0.959
Undrained 30 0.889 16.64 0.889
CHAPTER 5 469
Material Library
(psi)
3

– 10
1

= 8
q
Deviatoric stress, 6
0 0.10 0.20
Axial strain, 1
0.10 0.20
0.01
0.02
Volumetric strain,
0.03
0.04
0.05
Figure 5-108 Constant Pressure Test: p0 = 10 psi

(psi)
3
 25
–
1
 20
=
q 15
Deviatoric stress,
10
0 0.10 0.20
Axial strain, 1
0.10 0.20
0.01
Volumetric strain, v
0.02
0.03
0.04
0.05

(psi)
3
30

– 25
1

= 20
q
15
Deviatoric stress, 10
0.10 0.20
Axial strain, 1
0.10 0.20
0.01

Volumetric strain,
0.02
0.03
0.04
0.05

(psi)
16
3
 14
–
112

= 10
q 8
Deviatoric stress,
6
4
2
0.10 0.20
Axial strain, 1
0.10 0.20

 0.02
Volumetric strain,
0.04
0.06
0.08
0.10
Figure 5-111 Drained Test: p0 = 10 psi

CHAPTER 5 471
Material Library
(psi)
3

– 30
1
25
=
q 20
Deviatoric stress,
15
10
5
0.10 0.20
Axial strain, 1
0.10 0.20

 0.02
Volumetric strain,
0.04
0.06
0.08
0.10

(psi)
3

– 50
1

= 40
q
30
Deviatoric stress,
20
10
0.10 0.20
Axial strain, 1
0.10 0.20
0.02

Volumetric strain,
0.04
0.06
0.08
0.10

(psi)
3
 6
–
1
 5
= 4
q
3
Deviatoric stress,
2
0.01 0.02 0.03 0.04

Axial strain, 1
6
Pore pressure, u (psi)
0.01 0.02 0.03 0.04

Axial strain, 1
Figure 5-114 Undrained Test: p0 = 10 psi

(psi)
3
 12
–
1 10

= 8
q
6
Deviatoric stress,
4
2
0.01 0.02 0.03 0.04

Axial strain, 1
10
8
6
4
2
0.01 0.02 0.03 0.04

Axial strain, 1

CHAPTER 5 473
Material Library
(psi)
16
3
 14
–
1
 12
= 10
q
8
Deviatoric stress,
6
4
2
0
0.01 0.02 0.03 0.04
Axial strain, 1
16
14
12
10
8
6
4
2
0.01 0.02 0.03 0.04

Axial strain, 1
1.2
1.1 A
Void ratio, e
1.0 Cc
0.9
C B
0.8
Cs
0.7
10 20 30 40 50
Pressure, p (psi) (logarithmic scale with base 10)
Figure 5-117 Hydrostatic Compression Test: Cc = 0.04, Cs = 0.06

45
40
Deviatoric stress, q (psi)

30
Drained
Constant pressure
Undrained
20
10 Slope = M = 1.0
0 M
10 20 30 40 50
Mean Pressure, p (psi)
Figure 5-118 Critical State Line in q-p (psi)
Determination of parameters  and : The values of gamma and kappa can be related to the commonly known quantities
such as compression index ( C c ) and swelling index ( C s ). The compression index, C c , is defined as the slope of virgin
loading line on e-log10p plot while the swelling index, C s , is defined as the unloading-reloading curves on the same plot.
Usually, the compression index and swelling index are defined with respect to a one-dimensional consolidation test.
However, it can be shown that the e-ln(p) curve for any constant stress ratio test, that is, for constant q/p ratio, is parallel
to that obtained from a hydrostatic test, (Figure 5-119).
1.2
e0 vs. p0
ef vs. pf
1.1
1.0
Void ratio, e
Hydrostatic loading
0.9
0.8 At critical state
0.7
20 30 40 50 60 70 80 90
Pressure (psi) (logarithmic scale with base 10)
Figure 5-119 Isotropic Consolidation (Data from 5-8)

CHAPTER 5 475
Material Library
In fact, one-dimensional hydrostatic test is parallel to that obtained under critical state conditions. The values of gamma
and kappa can be related to C c and C s as follows. The virgin compression line can be expressed as:
p
e – e 0 = C c log 10  ----- (5-614)
 p 0
or
p
e – e 0 =  ln  ----- (5-615)
 p 0
and the swelling line (unloading-reloading) can be expressed as:

p
e – e 0 = C s log 10  ----- (5-616)
 p 0
or
p
e – e 0 =  ln  ----- (5-617)
 p 0
Therefore, comparing Equations (5-614) and (5-615), we have:

Cc Cc
 = ------------
- = ------------
- (5-618)
ln 10 2.303
and comparing Equations (5-616) and (5-617) yields:
Cs Cs
 = ------------
- = ------------
- (5-619)
ln 10 2.303
The value of C c can be computed by considering two points, A and B, in Figure 5-117 as:
eA – eB
C c = -------------------------------------
-
log p B – log p A
(5-620)
1.08 – 0.08
= -------------------------------------
log 50 – log 10
= 0.40
Here the subscript denotes the value at that point. The swelling index, C s , can be computed by considering points B and
C of the same figures
eC – eB
C s = -------------------------------------
-
log p B – log p C
0.842 – 0.80 (5-621)
= -------------------------------------
log 50 – log 10
= 0.06
Hence, the values of gamma and kappa can be computed from Equations (5-618) and (5-619) as:
0.40
 = ------------- = 0.174 (5-622)
2.303
Damage Models
In many structural applications, the finite element method is used to predict failure. This is often performed by comparing
the calculated solution to some failure criteria, or by using classical fracture mechanics. Previously, we discussed two models
where the actual material model changed due to some failure, see Progressive Composite Failure and the previous section on
Low Tension Material. In this section, the damage models appropriate for ductile metals and elastomeric materials will be
discussed.
Ductile Metals
In ductile materials given the appropriate loading conditions, voids will form in the material, grow, then coalesce, leading
to crack formation and potentially, failure. Experimental studies have shown that these processes are strongly influenced by
hydrostatic stress. Gurson studied microscopic voids in materials and derived a set of modified constitutive equations for
elastic-plastic materials. Tvergaard and Needleman modified the model with respect to the behavior for small void volume
fractions and for void coalescence.
In the modified Gurson model, the amount of damage is indicated with a scalar parameter called the void volume fraction
f. The yield criterion for the macroscopic assembly of voids and matrix material is given by:
 2  q 2  k k
F =  ------ + 2q 1 f cosh  --------------- –  1 +  q 1 f  2  = 0 (5-623)
  y  2 y 
as seen in Figure 5-120.
CHAPTER 5 477
Material Library
e  M
1.0
·
f* = 0
·
0.5 f*  f u* = 0.01
0.1
0.3
0.6
0.9
0
0 1 2 3 4  k k  3 M
Figure 5-120 Plot of Yield Surfaces in Gurson Model
The parameter q 1 was introduced by Tvergaard to improve the Gurson model at small values of the void volume fraction.
For solids with periodically spaced voids, numerical studies [Ref. 10] showed that the values of q 1 = 1.5 and q 2 = 1
were quite accurate.
The evolution of damage as measured by the void volume fraction is due to void nucleation and growth. Void nucleation
occurs by debonding of second phase particles. The strain for nucleation depends on the particle sizes. Assuming a normal
distribution of particle sizes, the nucleation of voids is itself modeled as a normal distribution in the strains, if nucleation
is strain controlled. If void nucleation is assumed to be stress controlled in the matrix, a normal distribution is assumed in
the stresses. The original Gurson model predicts that ultimate failure occurs when the void volume fraction f, reaches unity.
This is too high a value and, hence, the void volume fraction f is replaced by the modified void volume fraction f in the
yield function.
The parameter f is introduced to model the rapid decrease in load carrying capacity if void coalescence occurs.
f = f if f  fc
 f u* – f c (5-624)
f = f c +  ----------------  f – f c  if f > fc
 f F – f c
where f c is the critical void volume fraction, and f F is the void volume at failure, and f u* = 1  q 1 . A safe choice for f F
would be a value greater than  1  q 1  namely, f F = 1.1  q 1 . Hence, you can control the void volume fraction, f F ,
at which the solid loses all stress carrying capability.
Numerical studies show that plasticity starts to localize between voids at void volume fractions as low as 0.1 to 0.2. You can
control the void volume fraction f c , beyond which void-void interaction is modeled by Marc. Based on the classical studies,
a value of f c = 0.2 can be chosen.
The existing value of the void volume fraction changes due to the growth of existing voids and due to the nucleation of
new voids.
· · ·
f = f gro wth + f nuclea t i on (5-625)
The growth of voids can be determined based upon compressibility of the matrix material surrounding the void.
· ·p
f gr owth =  1 – f  kk (5-626)
As mentioned earlier, the nucleation of new voids can be defined as either strain or stress controlled. Both follow a normal
distribution about a mean value.
In the case of strain controlled nucleation, this is given by:
fN p 2
· 1 m – n ·p
f nu cle ation = -------------- exp – ---  -------------------  m (5-627)
S 2 2 S 
where f N is the volume fraction of void forming particles,  n the mean strain for void nucleation and S the standard
deviation.
In the case of stress controlled nucleation, the rate of nucleation is given by:
· fN 1  – n 2
f nu cle ation = --------------  exp – ---  ---------------- (5-628)
S 2 2 S 
1
where  =  + ---  kk .
3
If the second phase particle sizes in the solid are widely varied in size, the standard deviation would be larger than in the
case when the particle sizes are more uniform. The Marc user can also input the volume fraction of the nucleating second
phase void nucleating particles in the input deck, as the variable f N .
A typical set of values for an engineering alloy is given by Tvergaard for strain controlled nucleation as:
 n = 0.30 ; f N = 0.04 ; S = 0.01 . (5-629)
It must be remarked that the determination of the three above constants from experiments is extremely difficult. The
modeling of the debonding process must itself be studied including the effect of differing particle sizes in a matrix. It is safe
to say that such an experimental study is not possible. The above three constants must necessarily be obtained by intuition
keeping in mind the meaning of the terms.
When the material reaches 90 percent of f F , the material is considered to be failed. At this point, the stiffness and the stress
at this element are reduced to zero.
CHAPTER 5 479
Material Library
Elastomers
Under repeated application of loads, elastomers undergo damage by mechanisms involving chain breakage, multi-chain
damage, micro-void formation, and micro-structural degradation due to detachment of filler particles from the network
entanglement. In Marc, the methods to model damage in elastomers basically follow the continuum damage approach, as
developed by Kachanov to model fracture under creep. Here the free energy W is written as:
0
W = KW (5-630)
0
where W is the nominal strain energy function and K is a factor accounting for damage. The first method to be discussed
is that of Miehe, the second is that of Ogden and Roxburgh.
Miehe method
Two types of phenomenological models, namely discontinuous and continuous damage, exist to simulate the phenomenon
of damage using the Miehe method.
1. Discontinuous Damage:
The discontinuous damage model simulates the so-called Mullins effect as shown in Figure 5-121.
Figure 5-121 Discontinuous Damage

This involves a loss of stiffness below the previously attained maximum strain. The higher the maximum attained
strain, the larger is the loss of stiffness. Upon reloading, the uniaxial stress-strain curve remains insensitive to prior
behavior at strains above the previously attained maximum in a cyclic test. Hence, there is a progressive stiffness
loss with increasing maximum strain amplitude. Also, most of the stiffness loss takes place in the few earliest cycles
provided the maximum strain level is not increased. This phenomenon is found in both filled as well as natural
rubber although the higher levels of carbon black particles increase the hysteresis and the loss of stiffness. The
damage factor is a function of:
0
 = max  W  (5-631)
0
which determines the evolution of the discontinuous damage, with max  W  the evolving maximum deviatoric
strain energy density for a material point in its deformation history.
The reduced form of Clausius-Duhem dissipation inequality yields the stress as (the parameter  is used for
continuous damage and will be discussed below):
0
W
S = 2K     ----------- (5-632)
C
where C is the deformation tensor.

Mathematically, the discontinuous damage model has a structure very similar to that of strain space plasticity.
Hence, if a damage surface is defined as:
 = W–0 (5-633)
The loading condition for damage can be expressed in terms of the Kuhn-Tucker conditions:
· · (5-634)
0 0  = 0
The consistent tangent can be derived as:
2 0 0 0
 W K W W
D = 4 K --------------- + ----------- -----------  ----------- (5-635)
CC W 0 C C
2. Continuous Damage:
The continuous damage model can simulate the damage accumulation for strain cycles for which the values of
effective energy is below the maximum attained value of the past history as shown in Figure 5-122. This model can
be used to simulate fatigue behavior. More realistic modeling of fatigue would require a departure from the
phenomenological approach to damage to a more detailed micro-mechanical damage approach. The evolution of
the continuous damage parameter is governed by the arc length, s , of the effective strain energy as:
t
 0
 =  ----- W  s  ds (5-636)
s
0
Hence,  accumulates continuously within the deformation process.

CHAPTER 5 481
Material Library
Figure 5-122 Continuous Damage
The damage factor K     is implemented in Marc through both an additive as well as a multiplicative decomposition
of these two effects as:
2 2
    
K     = d +  dn exp  – ------ +  d n exp  – ----- (5-637)
  n   n
n = 1 n = 1
2
  + n 
K     = d +  d n exp  – -------------------- (5-638)
 n 
n = 1
   
The user specifies the phenomenological parameters d n  d n   n  n d n  n and d . If d is not defined, it is
automatically determined such that, at zero values of  and  , the factor K = 1 . If, according to Equation (5-637) or
Equation (5-638) the value of K exceeds 1, K is set back to 1.
Ogden-Roxburgh method
Like Miehe’s method for discontinuous damage, the Ogden-Roxburgh (O-R) damage model can be used to simulate the
Mullins effect in incompressible filled elastomers. While producing sufficiently accurate results, the main strength of the
O-R damage model is that the primary loading and unloading/reloading are uncoupled processes as far as material data
fitting is concerned, making it easier to get a good experimental data fit.
In the O-R damage model (originally proposed by Ogden and Roxburgh, see [Ref. 74], and modified by Bose et.al., see
[Ref. 75]), the damage factor is defined by an internal state variable,  (which is similar to K discussed above), tracking the
damage state in the material:
0 0
1  max  W  – W 
 = 1 – --- erf  --------------------------------------- +     (5-639)
r  m + max  W 0 
where r , m and  are material parameters, erf  x  is the error function, and     is a damage function. The material
parameters should fulfill the requirements r  1 , m  0 and   0 . As the damage function     does not affect the
stresses, it will not be further discussed. Note that, as opposed to the Miehe method for discontinuous damage, the damage
variable  is activated and accumulating during unloading. The O-R model does not support continuous damage.
The above damage models are available for deviatoric behavior and are flagged by means of the OGDEN, MOONEY,
ARRUDBOYCE, GENT, MARLOW, ANISO HYPE and DAMAGE model definition options. If, in addition, viscoelastic behavior is
desired, the VISCELOGDEN, VISCELMOON or BERGBOYCE option can be included. Finally, the UELDAM user subroutine can be used
to define damage functions different from equations (5-637) to (5-639).
If damage in elastomers modeled, it is recommended to use the Updated Lagrange formulation, as this supports all the
above mentioned options. In the Total Lagrange formulation, only the Ogden strain energy function is supported.
The damage value is available for postprocessing using element post code 80, Damage. For the Miehe method, this post
0
1  max  W  
code is given by 1 – K , for the O-R method, this post code is given by --- erf  --------------------------------------- .
r  m + max  W 0 
Damage in elastomers can be combined with nonlinear visco-elastic and nonlinear elastic-plastic behavior using the Parallel
Rheological Framework (PRF) in Marc (see Parallel Rheological Framework for details).
Using experimental data, the parameters required for the damage and the PRF models can be obtained using the data fitting
capability in Mentat.
Fatigue Analysis
Fatigue analysis aims at assessing the design life and/or the level of damage of a structure undergoing cyclic loading. In this
context, the main role of the finite element method in a classical metal fatigue analysis [79.] is to provide stresses or strains
at specific locations of the structure with respect to specific material properties, constraints and loading conditions. The
computed stresses and strains during the finite element analysis are then used to calculate the fatigue life and the level of
damage of the structure based on the material endurance. Although the nonlinear behavior of elastomers calls for different
fatigue rules than those used for metals, some of the concepts of metal fatigue can also be used for elastomers. A brief
CHAPTER 5 483
Material Library
overview of the main assumptions adopted for a fatigue analysis of rubber components as implemented in Marc is provided
below, with emphasis on carbon filled rubbers, as used in an example engine mounts.
Fatigue life assessment of rubber components is usually performed experimentally by means of cyclic uniaxial tension or
torsion tests, returning the fatigue life or number of cycles to failure (Nf) for a set of applied displacements or rotations.
This fatigue life can often be expressed by a power law
n
Nf = A  k 
in which A and n are material parameters and k is the damage parameter. An easy but relatively accurate model can be
obtained by using the peak maximum principal logarithmic strain or the peak maximum principal Green-Lagrange strain
as the damage parameter [80.,81.]. The above mentioned fatigue life expression is sometimes referred to a as the material
Wöhler curve (see Figure 5-123).
Figure 5-123 Typical representation of fatigue experimental results and fitted power law equation for rubbers
The above equation is available in Marc for the finite element-based fatigue analysis of rubber components. As an
alternative to the power law, the user can provide a table defining the damage parameter versus the number of cycles to
failure. According to either the power law or the table, the fatigue life is estimated knowing the strain field and history in
the rubber component from nonlinear finite element calculations. This strain history is typically the result of an oscillating
load on the structural component.
The easiest case is represented by the constant amplitude load where, as shown in Figure 5-124, the strain oscillates between
a maximum (or peak) Smax and a minimum (or valley) Smin. These two quantities define the strain amplitude:
S –S
S max min
= ----------------------------------
a 2
and the strain mean value

S +S
S max min-
= ----------------------------------
m 2
Figure 5-124 Constant Amplitude Load
Where in metal fatigue corrections are applied for the mean strain value, it is assumed here that the fatigue behavior is
determined by the peak value. For this constant amplitude load, the number of cycles of failure follows directly from the
peak strain value and the chosen fatigue material model.
In most cases, the strain amplitude is not always constant. The simplest extension to the constant amplitude load is the
block loading sequence (see Figure 5-125a). In this case, the load consists of a set of amplitudes Sai, and average values Smi,
each of them repeated ni times. Like with metal fatigue, the number of blocks to failure is calculated using the Palmgren-
Miner rule. Deriving the number of cycles to failure Ni for each strain amplitude Sai, the level of damage reached during
one loading sequence is
n n n
D = -----1- + -----2- + -----3- +   1,
N1 N2 N3
The reciprocal of D is the number of block loading sequences to failure.

A more complex load case is represented by the variable amplitude load (see Figure 5-125b). In this case, a rain flow cycle
counting algorithm is employed to reduce the spectrum of varying strains into an equivalent set of simple strain reversals,
and consequently a set of strain amplitudes [79.]. For each reversal, the corresponding number of occurrences is calculated,
and the cumulative damage reached during the full history load is derived using the Palmgren-Miner rule.
Figure 5-125 a) Block Loading Sequence and b) Variable Amplitude Loading

CHAPTER 5 485
Material Library
Cohesive Zone Modeling

Marc has a library of so-called interface elements (186, 187, 188, 189, 190, 191, 192,and 193), which can be used to simulate
the onset and progress of delamination. The constitutive behavior of these elements is expressed in terms of tractions versus
relative displacements between the top and bottom edge/surface of the elements (see Figure 5-126).
8
top face
4
v1 7
˜ v3
˜ 3
5
v2 bottom face
1 ˜
6
2
Figure 5-126 3-D Linear Interface Element
Considering a 3-D interface element, the relative displacement components are given by one normal and two shear
components, expressed with respect to the local element system (see Marc Volume B: Element Library for the definition of the
local element systems):
top bo t t om
vn = u1 – u1
top b o t t om (5-640)
vs = u2 – u2
top b ot t o m
vt = u3 – u3
Based on the relative displacement components, the effective opening displacement is defined as:
2 2 2
v = vn + vs + vt (5-641)
Later on, some modifications of this definition will be discussed.

The effective traction t is introduced as a function of the effective opening displacement and is characterized by an initial
reversible response followed by an irreversible response as soon as a critical effective opening displacement v c has been
reached. The irreversible part is characterized by increasing damage ranging from 0 (onset of delamination) to 1 (full
delamination). This damage value is available for postprocessing using element post code 80, Damage.
Three standard functions are currently available; namely, a bilinear, an exponential, and a linear-exponential function (see
Figure 5-127):
2G c v if 0  v  v c
t = ---------- -----
vm vc
2G c  v m – v  if v  v  v
t = ----------  ------------------ c m Bilinear (5-642)
v m  v m – v c
t = 0 if v  v m
v –v  vc
t = G c ----- e Exponential (5-643)
2
vc
2qG c v if 0  v  v c
t = ----------------------- -----
vc  q + 2  vc
Linear-exponential (5-644)
2qG c q  1 – v  v  if v  v
t = ----------------------- e c c
vc  q + 2 
in which G c is the energy release rate (cohesive energy), v m is the maximum effective opening displacement (which is only
used by the bilinear model), and q is the exponential decay factor (which is only used by the linear-exponential model).
t t t
v v v
vc vm vc vc
Figure 5-127 Bilinear (left), Exponential (middle), and Linear-exponential (right)
Cohesive Material Model
It can easily be verified that the maximum effective traction t c , corresponding to the critical effective opening displacement
v c is given by:
CHAPTER 5 487
Material Library
2G
t c = ---------c- Bilinear (5-645)
vm
G
t c = -------c- Exponential (5-646)
ev c
2qG c
t c = ----------------------- Linear-exponential (5-647)
vc  q + 2 
So if the maximum effective traction is known, the critical or maximum effective opening displacement can be determined
by:
2G
v m = ---------c- Bilinear (5-648)
tc
G
v c = ------c- Exponential (5-649)
et c
2qG c
v c = ---------------------- Linear-exponential (5-650)
tc  q + 2 
Note that for the bilinear model, the critical effective opening displacement does neither affect the cohesive energy nor the
maximum effective traction.
Until now, the behavior in the normal and shear direction is treated similarly. However, sometimes the behavior of an
interface material may be different in tension and shear. The first method to include such differences is incorporated by
the shear-normal stress ratio  1 , which defines the ratio of the maximum stress in shear and the maximum stress in tension
[Ref. 26]. This ratio is used to redefine the effective opening displacement according to:
2 2 2 2 2
v = vn + 1 vs + 1 vt
The effect of  1 = 0.5 is depicted in Figure 5-128 for the bilinear model.
t t
tension shear
only only
Gc Gc
vc vm v vc vm v
Figure 5-128 Response in Tension and Shear for a Shear-normal Stress Ratio  1 = 0.5 (Bilinear Model)
Although the use of the shear-normal stress ratio offers some flexibility, it assumes that the cohesive energy in tension and
shear is the same. If one wants to define a different value of the cohesive energy in shear than in tension, the shear-normal
energy ratio  2 can be used. In a general state of deformation, when  2  1 , the curve defining the effective traction versus
the effective opening displacement is defined as a linear combination of the response in pure tension and pure shear. Using
 1 = 0.5 and  2 = 0.75 , Figure 5-129 shows the response in tension and shear for the bilinear model.
t t
tension shear
only only
Gc 0.75G c
vc vm v vc vm v
Figure 5-129 Response in Tension and Shear for a Shear-normal Stress Ratio  1 = 0.5 and a Shear-normal Energy Ratio
 2 = 0.75 (Bilinear Model)
In order to avoid convergence problems in a finite element simulation of delamination, one may activate so-called viscous
energy dissipation. The basic idea of the dissipation model is that when delamination starts, the rate of deformation may
suddenly increase. This increase is used to augment the constitutive behavior with a viscous contribution being equivalent
to this rate of deformation:
t c v·
t vis = ---------
- (5-651)
v· r
in which  is the viscous energy factor, v· is effective opening displacement rate and v· 0 is the reference value of the effective
opening displacement rate. This reference value can either be user-defined or calculated by the program. In the latter case,
the reference value is given by the maximum effective opening displacement rate in any interface element, as long as the
response in all the interface elements is reversible. The viscous energy dissipation model does not directly have a physical
background, but is basically numerical in nature.
In the equations discussed above, no distinction has been made between tensile and compressive loading in the normal
direction. Assuming that in compression the behavior will remain reversible, Equation (5-641) will be adapted as:
2 2 2
v =  max  v n 0   + v s + v t (5-652)
Since, irrespective of the damage level, the interface elements should be able to sustain ongoing loading in compression (so
that inter-penetration is prohibited), it is possible to make the stiffness in compression a function of the corresponding
(negative) opening displacement. By default, the stiffness in compression is constant and given by the slope of the traction
CHAPTER 5 489
Material Library
versus opening displacement curve at the origin. If a non-default value of the stiffening factor in compression F is given,
the stiffness at v = – v c is given by:
t t n
-------n- = F --------
v n –vc v n 0
So far, the constitutive behavior has been discussed in terms of an effective traction versus an effective opening
displacement. The traction components follow from the effective traction according to:
v v
t n =  1 t -------- ; t s ,t =  2 t ----------- (5-653)
v n v s ,t
If  2 = 1 , then  1 =  2 = 1 . If  2  1 , then  1 and  2 are used to scale the traction components to be

proportional to their corresponding cohesive energies at the current state of deformation. This is illustrated below, where
the cohesive energy in shear is assumed to be significantly larger than in tension (  1 =  2 = 5 ). In Figure 5-130 (a) and
(b), the responses in pure tension and pure shear are given. For mixed mode loading with v n = v s , the responses are
shown in In Figure 5-130 (c) and (d). With  1 =  2 = 1 , the responses given by the dashed lines would be obtained,
showing that the maximum traction in the normal direction would largely exceed that of pure tension. When setting
FEATURE,12602, the program always uses  1 =  2 = 1 .
tn ts tn ts
shear only
mixed mode
tension only mixed mode
(a) vn (b) vs (c) vn (d) vs
Figure 5-130 Traction components in pure tension, pure shear and mixed mode loading ( v n = v s ) for a material with
1 = 2 = 5 .
As an alternative to the above mentioned standard models, the UCOHESIVE user subroutine can be used to enter a
user-defined material behavior.
There is a corresponding thermal cohesive model to what was discussed above which uses heat transfer cohesive elements.
In this model, the thermal conductivity can be a function of the displacement opening. The thermal gradient through the
thickness is based upon the temperatures along the top and bottom surfaces. The UCOHESIVET user subroutine may be used
to provide a more complex thermal conductivity.
Lemaitre Model
The Lemaitre damage model is a phenomenological approach to ductile damage in ferrous materials that are subject to large
plastic deformations as they occur in the manufacturing processes. The model is based on the thermodynamic dissipation
potential of the material where ductile damage is considered as a specific energy that is released when macroscopic fracture
occurs. Without going further into detail the mathematical derivation can be found in: Lemaitre, J.: A Course on Damage
Mechanics, 2nd Ed., Springer Verlag, Berlin, 1996. In the following paragraphs, some basic formulas are given that serve
for the calculation and the interpretation of the damage values. The material parameters can be derived by uniaxial tensile
tests for the assumed forming conditions (strain rate, temperature). Some standard values are given in this context, more
information can be found in the mentioned literature.
The Lemaitre damage model calculates three damage values which have different meanings. Macroscopic damage is
characterized by plastic deformation that leads to pore growth, pore coalescence and final rupture of the material matrix.
The damage growth begins approximately after an equivalent plastic strain threshold,  d . This is the first material
parameter to be defined in an experiment. For mild steels, it is assumed to be between 0.1 and 0.2. The so-called absolute
damage D represents the ductile damage growth in the material. The incremental damage law is given as follows:
2
f  
dD = ------------------------------------------- d p , 0D1
2E  S   1 – D  2
where the triaxiality function f    contains information about the state of stresses and is defined as follows:
2 m
f    = ---  1 + v  + 3  1 – 2v  2  = ------
-
3 
 is the von Mises stress, E the Young’s modulus, v the Poisson’s ratio,  m the mean normal stress, D the current
integrated value of the absolute damage at that material point and d p the effective plastic strain increment. If the
NONLOCAL parameter is used, then d p is the effective nonlocal plastic strain increment. For most steels one can assume a
maximum value of D from 0.15 to 0.4 at fracture. Copper might even reach D = 0.9 . The more ductile the material,
the higher D becomes. S is called the damage resistance, a material parameter to be determined from tensile tests. This
parameter influences mostly the growth of D and can also be determined by data correlation (for example, simulate the
material test then derive the correct material parameter).
The critical damage D c is used to compare the ductile damage D with the “state” of the material; such as, whether the
actual conditions (stresses, strains, state of stresses, already reached damage, etc.) might be critical for macroscopic failure:
2
U 2
D c = D 1c --------------------------
-1 – D 1  Dc  0
2
 f
D 1c is the critical damage in the uniaxial loadcase, the third material parameter to be determined by tensile tests. Most
steels show a D 1c from 0.15 to 0.4.  v is the ultimate stress during the tensile test (before necking begins). The lower
D c , the more likely a material damage is. Note that D c behaves contrary to D : D c = 1 for the “safe” material and low
for possible damaged regions.
CHAPTER 5 491
Material Library
In fact the comparison between D and D c is necessary to identify critical forming zones: as long as D is (much) lower
than D c , the forming operation is safe. When D reaches D c , the damage probability tends to be 100%. The comparison
is done by the relative damage value D r e l (reflected in the postprocessing):
D
D re l = ------ 0  Dr e l  1
Dc
When D r el approaches 1 in a specific region, fracture is highly possible whereas small values indicate a “safe” region.
Simplified Model
For elastic, elastic-plastic, or rigid-plastic materials, there is the option for you to define a simplified damage model. You
define the damage factor (df ) in the UDAMAG user subroutine.
If model 9 is used, then:
p ·p
 y =  y      T *  1.0 – df 
If model 10 is used, then:

p ·p
 y =  y      T *  1.0 – df  and E = E  T *  1.0 – df 
The normal data for a specific material are defined with the ISOTROPIC, WORK HARD, and OGDEN options. Cross-reference to
this material is made with the material number.
Cockroft-Latham Damage Indicator

Cockroft-Latham damage indicator does not affect the yield stress. It is a postprocessing value to indicate a possible damage
area. It can also be used to initiate crack by removing elements in the area.
 max ·
 ------------

 dt  C
·
where  m ax is the maximum principal stress,  is the effective von Mises stress and  is the effective plastic strain rate.
C is material constant threshold for damage.
Principal-tension Damage Indicator

This damage indicator does not affect the yield stress.
 max
 ------------

dt  C
where  m ax is the maximum principal stress and  is the effective von Mises stress. C is material constant threshold for
damage.
Figure 5-131 shows the application of the Cockroft-Latham damage criterion used to predict cracks known as Chevron
cracks in the drawing process.
Figure 5-131 Chevron Crack Prediction with Cockroft-Latham Damage Criterion
Oyane Damage Indicator

m ·
  ------

- + B  dt  C

Similar to Cockraft-Latham, Oyane damage indicates a possible damage area where  m is the mean or average stress, B
and C are both material constants.
The UDAMAGE_INDICATOR user subroutine can be used to define unsupported damage criteria. The UACTIVE user subroutine
can be used for element removal.
Figure 5-132 show the application of the Oyane damage criterion used to predict cracks known as Chevron cracks in the
extrusion process.
CHAPTER 5 493
Material Library
Figure 5-132 Chevron Crack Prediction with Oyane Damage Criterion
Bonora Damage Model

Bonora damage model, like Lemaitre, uses continuum damage mechanics (CDM) as the basis to derive the evolution of
damage in the material. It is also known as nonlinear CDM model. It takes the three processes of damage evolution in
ductile metal (pore initiation, growth and coalescence) into account in a more explicit way. For more detail description of
the Bonora model, see [Ref. 32].
The Bonora damage model utilizes the current state of the stress to determine the Young’s modulus changes with damage.
In one dimension system, the damage is assumed to be completely closed under compression which means that the material
temporarily recovers its virgin elasticity.
The incremental damage is given as follows:
1
D cr –1 1
dD = --------------------  D cr – D  ------------- f    --------- d p +
f   p+
ln  --------
  th 
where  p+ is the active equivalent plastic strain. The “active” equivalent plastic strain is the accumulated plastic strain
during the tensile loading. f    is the triaxiality function as shown in the Lemaitre Damage Model. It contains
information about the state of stresses. This model has shown good correlation with experiments under cyclic compression-
tension loading [Ref. 33].
There are four materials parameters to be identified using experiments. The damage starts to accumulate when the active
equivalent plastic strain is greater than the threshold,  t h .  f and D cr are the strain and damage at failure, respectively.
Please note D cr is not necessarily 1.  is a factor to control how the damage evolution progresses. The procedure to
identify these values is given in [Ref. 32]. It also provides data for a number of materials.
The coupling of the damage variable to the elastic properties of the material, by using the effective Young’s modulus, is
derived using the strain equivalence hypothesis, [Ref. 34]: the strain associated with a damage state under the applied stress
is equivalent to the strain associated with the undamaged state under the effective stress. The effective Young’s modulus is
given as follows:
E = E0  1 – D 
where E 0 is the original Young’s modulus.
The damage variable is not coupled directly to the plasticity potential as normally done (based on the strain equivalence
hypothesis) in the CDM model. This is based on the proposition that the stress-strain plasticity data already contains
implicitly the damage effect [Ref. 33]. This means that Bonora damage model does not generate additional material
softening behavior.
For user’s convenience, the UNLCDM user subroutine is introduced for users who want to have their own damage model
within the frame work of the Bonora model.
Johnson-Cook Damage Model

Damage strain is defined as:
D * ·
 d =  D 1 + D 2 * e 3  *  1 + D 4 * ln *  *  1 + D 5 * T * 
 
where D 1 to D 5 are constants depending on materials.  * equals to hydrostatic pressure divided by the effective stress, as
shown in the following,
m
 * = ------
-

The damage indicator is measured by the following equation:
 p
D =  --------

-
d
The material is considered damage if D > 1.0.
Nonstructural Materials
In addition to stress analysis, Marc can be used for heat transfer, coupled thermo-electrical heating (Joule heating), coupled
electrical-thermal-mechanical analysis (Joule-mechanical), hydrodynamic bearing, fluid/solid interaction, electrostatic,
magnetostatic, magnetodynamic, piezoelectric, acoustic, and fluid problems. Material properties associated with these
analyses and Marc options that control these analyses are described below.
Heat Transfer Analysis

In heat transfer analysis, use the ISOTROPIC, ORTHOTROPIC, or ANISOTROPIC model definition options to input values of thermal
conductivity, specific heat, and mass density. If the latent heat effect is to be included in the analysis, the value of latent
heat and associated solidus and liquidus temperatures must be entered through the LATENT HEAT or TEMPERATURE EFFECTS
model definition option. Both the thermal conductivity and specific heat can be dependent on temperatures. The mass
density must be constant throughout conventional heat transfer analysis. In addition, the ANKOND user subroutine can be
used for the input of anisotropic thermal conductivity. For radiation analysis, you must enter the Stefan-Boltzmann
constant in the RADIATION parameter and the emissivity through the EMISSIVITY or ISOTROPIC option. The emissivity can be
temperature dependent.
For high temperature analysis where pyrolysis occurs, the THERMO-PORE option is used to define the advanced material
behavior.
Piezoelectric Analysis
In a piezoelectric analysis, the material properties contain a mechanical part, an electrostatic part, and a part connecting
these two. The material properties for the mechanical part are the same as for an elastic stress analysis. The material
properties for the electrostatic part, the permittivity, and the material properties for the piezoelectric coupling can be
CHAPTER 5 495
Material Library
defined through the PIEZOELECTRIC option. The coefficients for the piezoelectric coupling can be either stress based or strain
based.
Thermo-Electrical Analysis
In addition to thermal conductivity, specific heat and mass density, the value of electric resistivity must be entered using
the ISOTROPIC and ORTHOTROPIC model definition options in a coupled thermo-electrical analysis. Input the variation of
electric resistivity with temperature through the TEMPERATURE EFFECTS option.
Coupled Electrical-Thermal-Mechanical Analysis

Material properties for coupled electrical-thermal-mechanical analysis are the same for stress analysis in addition to those
of Joule heating analysis (thermal conductivity, specific heat and mass density and electric resistivity).
Hydrodynamic Bearing Analysis

In a hydrodynamic bearing analysis, the ISOTROPIC and ORTHOTROPIC model definition options can be used for entering both
the viscosity and specific mass. Define the temperature dependency of viscosity through the TEMPERATURE EFFECTS option.
Fluid/Solid Interaction Analysis – Added Mass Approach

In a fluid/solid analysis, the density of a fluid is given in the first field of the ISOTROPIC option. The fluid is assumed to be
nonviscous and incompressible.
Electrostatic Analysis
In an electrostatic analysis, the permittivity can be defined through either the ISOTROPIC or ORTHOTROPIC options. In
addition, the UEPS user subroutine can be used for the input of anisotropic permittivity.
Magnetostatic Analysis
In a magnetostatic analysis, the permeability can be defined through either the ISOTROPIC or ORTHOTROPIC options. A
nonlinear relationship can be entered via the B-H RELATION option. The UMU user subroutine can be used for the input of
isotropic permeability.
Magnetodynamic Analysis
In an magnetodynamic analysis, the permittivity, permeability and conductivity can be defined using either the ISOTROPIC
or ORTHOTROPIC options. An assumption made is that the permittivity is a constant (does not vary with time) in the analysis.
A nonlinear permeability can be entered via the B-H RELATION option. The UEPS, UMU, and USIGMA user subroutines can be
used for the input of anisotropic permittivity, permeability, and conductivity, respectively. Both the dependent and
harmonic analysis can be performed.
Coupled Electrostatic-Structural
Material properties for a coupled electrostatic-structural are the same as for a stress analysis in addition to those for an
electrostatic analysis (permittivity).
Acoustic Analysis
In an acoustic analysis of a cavity with rigid boundaries, the bulk modulus and the relative density of the medium can be
entered through the ISOTROPIC option.
The ACOUSTIC parameter is used to indicate that a coupled acoustic-structural analysis is performed. In addition to the
CONTACT option, the ACOUSTIC and REGION model definition options are used to define the material properties of the acoustic
medium and to set which elements correspond to the solid and the fluid region.
Fluid Analysis
In the fluid analysis, viscosity and density can be defined using the ISOTROPIC option. In addition, conductivity and specific
heat are defined for coupled fluid-thermal analysis. Non-Newtonian fluid behavior can be defined using the STRAIN RATE
option. The TEMPERATURE EFFECTS option can be used to define the temperature dependence of the properties above.
References
1. Auricchio, F. and Taylor, R.L., “Shape-memory alloy: modeling and numerical simulations of the finite-strain
superelastic behavior”, Comput. Methods Appl. Mech. Engrg., Vol. 143, pp.175-194 (1997).
2. Auricchio, F., “A robust integration-algorithm for a finite-strain shape-memory-alloy superelastic model”, Int. J.
Plasticity, Vol.17, pp.971-990 (2001).
3. Arruda, E. M. and Boyce, M. C. “A three-dimensional constitutive model for the large stretch behaviour of rubber
elastic materials”, J. Mech. Phys. Solids, Vol.41, No. 2, 1993.
4. Buyukozturk, O., “Nonlinear Analysis of Reinforced Concrete Structures”, Computers and Structures, 7, 149-156
(1977).
5. Gent, A. N., “A new constitutive relation for rubber”, Rubber Chem. Tech., 69, 1996.
6. Barlat, F., Lege, D.J. and Brem, J.C., “A six-component yield function for anisotropic metals”, Int. J. Plasticity, 7,
693-712 (1991).
7. Chung, K. and Shah, K., “Finite element simulation of sheet metal forming for planar anisotropic metals”, Int. J.
Plasticity, 8, 453-476 (1992).
8. Yoon, J.W., Yang, D.Y. and Chung, K. and Barlat. F., “A general elasto-plastic finite element formulation based
on incremental deformation theory for planar anisotropy and its application to sheet metal forming”, Int. J.
Plasticity, 15, 35-67 (1999).
9. Yoon, J.W., Barlat, F., Chung, K., Pourboghrat, F. and Yang, D.Y., “Earing predictions based on asymmetric
nonquadratic yield function”, Int. J. Plasticity, 16, 1075-1104 (2000).
10. “Current Recommended Constitutive Equations for Inelastic Design Analysis of FFTF Components.” ORLN-
TM-360Z, October 1971.
CHAPTER 5 497
Material Library
11. “Finite Element Calculation of Stresses in Glass Parts Undergoing Viscous Relaxation”, J.Am.Ceram.Soc. Vol.
70[2], pp. 90-95, 1987.
12. Mooney, M. J. Appl Phys., Vol. II, p. 582, 1940.
13. Naghdi, P. M. “Stress-Strain Relations in Plasticity and Thermoplasticity.” In Plasticity, Proceedings of the Second
Symposium on Naval Structural Mechanics, edited by E. H. Lee and P. S. Symonds. Pergamon Press, 1960.
14. Narayanaswamy, O.S., “A Model of Structural Relaxation in Glass”, J.Am.Ceram.Soc., Vol. 54[10], pp. 491-498,
1971.
15. Prager, W. Introduction to Mechanics to Continua. New York: Dover Press, 1961.
16. “Proposed Modification to RDT Standard F9-5T Inelastic Analysis Guidelines.” ORNL, October 1978.
17. Rivlin, R. S. Phil Trans Roy Soc (A), Vol. 240, 459, 1948.
18. Rivlin, R. S. Phil Trans Roy Soc (A), Vol. 241, 3-79, 1948.
19. Timoshenko, S. P., and J. N. Goodier. Theory of Elasticity. Third Ed. New York: McGraw-Hill, 1970.
20. Simo, J. C. “On a Fully Three-dimensional Finite Strain Viscoelastic Damage Model: Formulation and
Computational Aspects,” Computer Methods in Applied Mechanics and Engineering, Vol. 60, 1987, pp. 153-173.
21. Tvergaard, V., “Influence of Voids on Shear Band Instabilities under Plane Strain Conditions”, Int. J. Fracture, Vol.
17, pp. 389-407, 1981.
22. Tvergaard, V., “Material Failure by Void Coalescence in Localized Shear Bands”, in Int. J. Solids Struct., Vol. 18,
No. 8, pp. 652-672, 1982.
23. Chaboche, J. L., “Constitutive Equations for Cyclic Plasticity and Cyclic Viscoplasticity”, International Journal of
Plasticity, Vol. 5, pp. 247-302, 1989
24. van den Boogard, A.H., “Implicit integration of the Perzyna viscoplastic material model”, TNO Building and
Construction Research, The Netherlands, TNO-report, 95-NM-R711, 1995
25. Puck, A. and Schürmann, H., “Failure Analysis of FRP Laminates by Means of Physically Based Phenomenological
Models”; Composite Science and Technology, 58, pp. 1045-1067, 1998.
26. Camacho, G.T. and Ortiz, M., “Computational modelling of impact damage in brittle materials”, Int. J. Solids
Struct., Vol. 33, pp 2899-2938, 1996.
27. Greve, L. and Picket, A. K., “Modelling Damage and Failure in Carbon/Epoxy Non-crimp Fabric Composites
Including Effects of Fabric Pre-shear”, Composites: Part A, Vol. 37, pp. 1983-2001, 2006.
28. Knops, M. and Bögle, C. “Gradual Failure in Fibre/Polymer Laminates”, Composites Science and Technology, Vol.
66, pp. 616-625, 2006.
29. Buchanan, David L., Gosse, Jonathan H., Wollschlager, Jeffrey A., Ritchey, Andrew, and Pipes, R. Byron,
“Micromechanical enhancement of the macroscopic strain state for advanced composite materials”, Composites
Science and Technology, pp. 1974-1978 (2009).
30. Gosse, Jon H., Boeing Company, “Strain Invariant Failure Criteria for Fiber Reinforced Polymeric Composite
Materials”.
31. Brown, S. B., Kim, K. H., and Anand, L., 1989, “An Internal Variable Constitutive Model for Hot Working of
Metals,” Int. J. Plast., 5, pp. 95–130.
32. Bonora, N., “A nonlinear CDM model for ductile failure, Engineering Fracture Mechanics, 1997, 58(1/2), 11-28.
33. Pirondi, A., Bonora, N. “Modeling ductile damage under fully reversed cycling”, Computational Materials Science,
2003, 26, 129-141.
34. Lemaitre, J., “Local approach to fracture”, Engineering Fracture Mechanics, 1986, 25(5/6), 523-537.
35. Morman, K. N. (Jr.) and Nagtegaal, J. C., “Finite element analysis of sinusoidal small-amplitude vibrations in
deformed viscoelastic solids. Part I: Theoretical Development”, International Journal for Numerical Methods in
Engineering, 19, 1079-1103, 1983.
36. Coleman, B. D. and Noll, W., “Foundations of Linear Viscoelasticity”, Reviews of Modern Physics, 33/2, 239-249,
1961.
37. Kim, B. K. and Youn, S. K., “A viscoelastic model of rubber under small oscillatory load superimposed on large
static deformation”, Archive of Applied Mechanics, 71/11, 748-763, 2001.
38. G.Y. Qiu and T.J. Pence, “Loss of ellipticity in plane deformation of a simple directional reinforced incompressible
nonlinearly elastic solid”, Journal of Elasticity, 1997 (49), 31-63.
39. L.W. Brown and L.M. Smith, “A simple transversely isotropic hyperelastic constitutive model suitable for finite
element analysis of fiber reinforced elastomers”, Journal of Engineering Materials and Technology (ASME),
2011(133), 021021:1-13
40. T. C. Gasser, R.W. Ogden and G.A. Holzapfel, “Hyperelastic modeling of arterial layers with distributed collagen
fiber orientations”, J. R. Soc. Interface 2006 (3), 15-35.
41. Anand, L., “Constitutive Equations for the Rate-Dependent Deformation of Metals at Elevated Temperatures”,
ASME Journal of Engineering Materials and Technology, Vol. 104, (1982) pp. 12-17.
42. Rendek M. and Lion A., [2010], “Amplitude dependence of filler-reinforced rubber: Experiments, constitutive
modelling and FEM-Implementation”, International Journal of Solids and Structures, 47/21, 2918-2936.
43. Simo J.C. and Hughes T.J.R., [1998], Computational Inelasticity, Springer, New York.
44. Rendek M. and Lion A., [2010], “Amplitude dependence of filler-reinforced rubber: Experiments, constitutive
modelling and FEM-Implementation”, International Journal of Solids and Structures, 47/21, 2918-2936.
45. Ahmadi H., Dalrymple T., Kingston J. and Muhr A., [2008], “Dynamic properties of filled rubber – Part III: Use
of nonlinear kinematic hardening plasticity material models”, Paper – American Chemical Society of Rubber
Chemistry, 4, 1724-1741. Paper presented at the Fall 174th Technical Meeting of the Rubber Division of the
American Chemical Society, Inc., Louisville, KY, ISSN: 1547-1977.
46. Coveney V.A., Johnson D.E. and Turner D.M., [1995], “A Triboelastic Model for the Cyclic Mechanical Behavior
of Filled Vulcanizates”, Rubber Chemistry and Technology, 68/4, 660-670.
47. Besseling J.F., [1953], “A Theory of Plastic Flow for Anisotropic Hardening in Plastic Deformation of an Initially
Isotropic Material”, Nat. Aero. Res. Inst. A’dam (NLL) Report S. 410.
48. Palmov V., [1998], Vibrations of Elasto-Plastic Bodies, Spring-Verlag Berlin Heidelberg New York, ISBN 3-540-
63724-9.
49. Lemaitre J. and Chaboche J.-L. [1998], Mechanics of solid materials, Cambridge University Press, ISBN 0-521-
47758-1.
50. Lion A., [2005], “Phenomenological Modelling of Strain-Induced Structural Changes in Filler-Reinforced
Elastomers”, KGK Kautschuk Gummi Kunststoffe, 58/4, 157-162.
CHAPTER 5 499
Material Library
51. Rabkin M., [2008], “Simulation of the Fletcher-Gent effect by using of the subroutine UPHI MSC/Marc©”,
Constitutive Models for Rubber V – Boukamel, Laiarinandrasana, Méo & Verron (eds), Taylor & Francis Group,
London, ISBN 978-0-415-45442-1, 263-268.
52. Kraus G., [1984], “Mechanical Losses in Carbon-Black-Filled Rubbers”, Journal of Applied Polymer Science: Applied
Polymer Symposium, 39, 75-92.
53. Ulmer J.D., [1996], “Strain dependence of dynamic mechanical properties of carbon black-filled rubber
compounds”, Rubber Chemistry and Technology, 69/1, 15-47.
54. F. Barlat, H. Aretz,. J.W. Yoon, M.E. Karabin, J.C. Brem, and R.E. Dick, “Linear transformation-based anisotopric
yield function”, International Journal of Plasticity, 21, (2005) 1009-1039.
55. F. Barlat, J.C. Brem, J.W. Yong, K. Chung, R.E. Dick, D.J. Lege, F. Pourboghrat, S.H. Choi, and E. Chu, “Plast
stress yield function for aluminum alloy sheets -- part 1: theory”, International Journal of Plasticity, 19 (2003) 1279-
1319.
56. A. M. Johnson, R. F. Mehl, “Reaction kinetics in processes of nucleation and growth”, Trans. Am. Inst. Min.
Engrs. 135 (1939) 416.
57. M. Avrami, Kinetics of phase change I”, J. Chem. Phys. 7 (1939) 1103.
58. M. Avrami, Kinetics of phase change II”, J. Chem. Phys. 8 (1940) 212.
59. M. Avrami, Kinetics of phase change III”, J. Chem. Phys. 9 (1941) 117.
60. ASM – Practical Heat Treating, 2nd Edition, ASM International , 2006.
61. https://en.wikipedia.org/wiki/Steel.
62. https://www.twi-global.com/terms.
63. http://www.twi-global.com/technical-knowledge/faqs/material-faqs/faq-what-are-the-microstructural-
constituents-austenite-martensite-bainite-pearlite-and-ferrite/
64. Armstrong, P. J. and Frederick, C. O. (1966): A mathematical representation of the multiaxial Bauschinger effect,
CEGB Report RD/B/N 731 (or in Materials at High Temperature, 24, 1-26 (2007)).
65. Hashiguchi, K. (1980): Constitutive equations of elastoplastic materials with elastic-plastic transition, J. Appl.
Mech. (ASME), 47, 266-272.
66. Hashiguchi, K. (1989): Subloading surface model in unconventional plasticity, Int. J. Solids Structures, 25, 917-
945.
67. Hashiguchi, K. (2016): Exact formulation of subloading surface model: unified constitutive law for irreversible
mechanical phenomena in solids, Arch. Compt. Meth. Eng., 23, 417-447.
68. Hashiguchi, K. (2017): Foundation of Elastoplasticity: Subloading Surface Model, Springer.
69. Hashiguchi, K. and Ueno, M. (1977): Elastoplastic constitutive laws of granular materials, Constitutive Equations
of Soils (Proc. 9th Int. Conf. Soil Mech. Found. Eng., Spec. Ses. 9), Tokyo, JSSMFE, pp. 73-82.
70. Hashiguchi, K. and Ueno, M. (2017): Elastoplastic constitutive equation of metals under cyclic loading, Int. J. Eng.
Sci., 111, 86-112.
71. Hashiguchi, K. (2000): Fundamentals in constitutive equation: continuity and smoothness conditions and loading
criterion, Soils and Foundations, 40(3), 155-161.
72. Kachanov, L. M. (1958): Time of the rupture process under creep conditions, Tzv. Akad. Nauk. SSR. Otd. Tekh.
8, 26-31.
73. Miehe, C., (1995): Discontinuous and continuous damage evolution in Ogden-type large-strain elastic materials.
European Journal of Mechanics - A/Solids 14, 697-720.
74. Ogden, R., Roxburgh, D. (1999): A pseudo-elastic model for the Mullins effect in filled rubber. Proceedings of the
Royal Society of London. Series A: Mathematical, Physical and Engineering Sciences 455, 2861–2877.
75. Bose, K., Hurtado, J.A., Snyman, M.F., Mars, W.V. and Chen, J.Q. (2003): Modeling of Stress Softening in Filled
Elastomers. Constitutive Models for Rubber III Proceedings of the 3rd European Conference on Constitutive Models for
Rubber (ECCMR), pp. 223-230. Edited by J. C. Busfield and A. H. Muhr. Taylor and Francis, London.
76. H. Vegter and A. H. van den Boogaard, A plane stress yield function for anisotropic sheet material by interpolation
of biaxial stress states, International Journal of Plasticity, 22:557-580, 2006
77. M. Abspoel, M. E. Scholting, M. Lansbergen, Y. An and H. Vegter, A new method for predicting advanced yield
criteria input parameters from mechanical properties, Journal of Materials Processing Tech., 248:161-177, 2017
78. M. Abspoel, M. E. Scholting, M. Lansbergen, B. N. Neelis, Accurate anisotropic material modelling using only
tensile tests for hot and cold forming, Journal of Physics: Conference Series
79. N.W.M. Bishop and F. Sherratt, Finite element based fatigue calculations, NAFEMS Ltd, 2000.
80. W. D. Kim, H. J. Lee, J. Y. Kima and S.-K. Koh, Fatigue life estimation of an engine rubber mount, International
Journal of Fatigue, 553-560:26, 2004 .
81. Chang S. Woo, Hyun S. Park, and Wan D. Kim, The Effect of Maximum Strain on Fatigue Life Prediction for
Natural Rubber Material, World Academy of Science, Engineering and Technology, Vol:7 2013-04-26.
Chapter 6: Contact
6 Contact
 Introduction 502

Definition of Contact Bodies 502
 Contact Tables 506
 Motion of Bodies 506
 Node-to-Segment Contact 510
 Segment-to-Segment Contact 563
 Interference Fit 583

Symmetry Conditions 586
 References 587
Introduction
The simulation of many physical problems requires the ability to model the contact phenomena. This includes analysis of
interference fits, rubber seals, tires, crash, and manufacturing processes among others. The analysis of contact behavior is
complex because of the requirement to accurately track the motion of multiple geometric bodies, and the motion due to
the interaction of these bodies after contact occurs. This includes representing the friction between surfaces and heat
transfer between the bodies if required. The numerical objective is to detect the motion of the bodies, apply a constraint
to avoid penetration, and apply appropriate boundary conditions to simulate the frictional behavior and heat transfer.
Several procedures have been developed to treat these problems including the use of Perturbed or Augmented Lagrangian
methods, penalty methods, and direct constraints. Furthermore, contact simulation has often required the use of special
contact or gap elements. Marc allows contact analyses to be performed automatically without the use of special contact
elements. Two numerical procedures to simulate these complex physical problems have been implemented in Marc, namely
a node-to-segment and a segment-to-segment procedure. Since for both procedures the model set up is very similar,
switching from one procedure to the other only involves minor effort.
The node-to-segment procedure in Marc was implemented before the segment-to-segment procedure, and remains the
default contact procedure. As will be discussed later on in more detail, important reasons to switch from the node-to-
segment to the segment-to-segment procedure are:
 Reduction of the dependency of the solution on the contact body numbering (in case of contact between
deformable bodies).
 Improvement of the stress distribution in the area of contact (in case of contact between deformable bodies).
 Double-sided shell contact.
 Contact involving beam elements.
Although the majority of contact options is available for both procedures, there are still a few (like deactivation of glued
contact and anisotropic friction) which are only supported for the node-to-segment procedure. On the other hand, the
beam contact options within the segment-to-segment procedure are superior to those within the node-to-segment
procedure.
In this chapter, some common aspects of the contact procedures will be discussed first. This includes the various types of
contact bodies, their numbering and the ways to define their motion. Then separate information will be provided on
various aspects of the node-to-segment and the segment-to-segment procedures, like contact detection, friction, separation
and a number of parameters the user can enter to control the analysis.
The discussion will mainly focus on contact in a structural analysis. Analyses involving other physics types, like heat transfer
or magnetostatics, will be separately mentioned.
Definition of Contact Bodies

There are two types of contact bodies in Marc – deformable and rigid. More general, a distinction can be made between
meshed bodies and geometric bodies, where meshed bodies consist of finite elements and geometric bodies consist of curves
(2-D) or surfaces (3-D). In a structural analysis, a meshed body is commonly called a deformable body and a meshed body
a rigid body.
In terms of data input, a deformable body is simply a collection of finite elements, shown in Figure 6-1.
CHAPTER 6 503
Contact

Figure 6-1 Deformable Contact Bodies defined by Finite Elements
Based on the set of elements defining the body, Marc automatically determines which nodes and element edges (2-D) or
faces (3-D) are on the outer surface of the body. Depending on the contact procedure used, they are used to determine if
contact will occur, either with the body itself or with another contact body.
Note that a body can be multiply connected (have holes in itself ). It is also possible for a body to be composed of both
triangular and quadrilateral elements in 2-D or tetrahedral, pentahedral and brick elements in 3-D (provided that
connected elements share faces with the same number of nodes). Beam elements and shells are also available for contact;
the algorithms for beam contact will be discussed separately. Both lower-order (linear) and higher-order (quadratic)
elements may be used. One should not mix continuum elements, shells, and/or beams in the same contact body.
Each node and element should be in, at most, one body. The elements in a body are defined using the CONTACT option. It
is not necessary to identify the nodes on the exterior surfaces as this is done automatically. The algorithm used is based on
the fact that nodes on the boundary are on element edges or faces that belong to only one element. As all nodes on free
surfaces are considered contact nodes (node-to-segment procedure) or all edges/faces defining the free surfaces are
considered contact segments (segment-to-segment procedure), if there is an error in the mesh generation such that internal
holes or slits exist, undesirable results can occur.
Note: These problems can be visualized by using Mentat to create an outline plot and fixed by using the sweep
command in Mentat.
In the node-to-segment procedure, each node on the exterior surface is treated as a potential contact node. In many
problems, it is known that certain nodes never come into contact; in such cases, the CONTACT NODE option can be used to
identify the relevant nodes. The potential segments composed of edges or faces are treated in potentially two ways. The
default is that they are considered as piece-wise linear (PWL). As an alternative, a cubic spline (2-D) or a Coons surface (3-
D) can be placed through them. The SPLINE option is used to activate this procedure. This improves the accuracy of the
representation of the true geometry of the contact body if this is curved.
Rigid bodies are composed of curves (2-D) or surfaces (3-D) or meshes with only thermal elements in coupled problems.
The most significant aspect of rigid bodies is that they do not deform. Deformable bodies can contact rigid bodies, but
contact between rigid bodies is generally not considered. There is one exception to this rule. It is possible to assign nodes
to rigid bodies using the RIGID NODE option. Nodes assigned to a rigid body (which typically should not be used in the
connectivity of a finite element) follow the motion of this rigid body and they may come into contact with another rigid
body, provided that at least one of these rigid bodies is a load-controlled body (see Detection of Contact for a further discussion
on the rigid body types). This feature can be conveniently used to e.g. model complex tool sets in a forming process.
Rigid bodies can be created either in CAD systems and transferred through Mentat or Patran into Marc, created within
Mentat or Patran, or created directly through the Marc input. There are several different types of curves and surfaces that
can be entered including:
Table 6-1 Geometrical Entities Used in

Modeling Contact
2-D 3-D
line 4-node patch
circular arc ruled surface
spline surface of revolution
NURBS Bezier
ploycurve poly-surface
cylinder
sphere
NURBS
trimmed NURBS
Within the Mentat GUI, all curves or surfaces are mathematically treated as NURBS curves or surfaces or faceted
curves/surfaces. Within the analysis, these rigid curves or surfaces can be treated in the following three ways:
1. Discrete piecewise linear lines (2-D) or patches (3-D)
When the discrete approach is used, all geometric primitives are subdivided into the straight line segments or
bilinear patches. You have the control over the density of these subdivisions to approximate a curved line or a
surface within a desired degree of accuracy. This subdivision is also relevant when the corner conditions (see
Detection of Contact) are determined. Note that too few subdivisions lead to a poor surface description whereas too
many subdivisions lead to a more expensive analysis due to more complex contact checking.
2. Analytical NURBS curves or surfaces
When using the analytical description, the number of subdivisions given is less relevant. It should be large enough
to roughly visualize the shape of the NURBS. In 3-D, a too large number is automatically reduced to ensure a
reasonable analysis cost. The treatment of the rigid bodies as NURBS surfaces is advantageous as it leads to a greater
accuracy in the representation of the geometry and a more accurate calculation of the surface normal. Additionally,
the variation of the surface normal is continuous over the body leading to a better calculation of the friction
behavior and usually results in better convergence.
CHAPTER 6 505
Contact
3. Polycurves/ polysurfaces
In this case the geometrical body is represented by a series of straight lines (polycurves) or a series of conforming
triangles/quads (polysurface). The user has the possibility to enrich this faceted description via the SPLINE option.
In that case the straight lines or triangles/quads will be defined through cubic splines or coons surfaces using the
information of the neighboring facets. Straight edges (edges with a discontinuous normal) can be automatically
detected or specified by the user.
Either the analytical of the polysurves/polysurface description is recommended.
To create a rigid body, you can either read in the curve and surface geometry created from a CAD system or create the
geometry in Mentat or directly enter it in Marc. You then use the CONTACT option to select which geometric entities are to
be a part of the rigid body. An important consideration for a rigid body is the definition of the interior side and the exterior
side. For two-dimensional analysis, the interior side is formed by the right-hand rule when moving along the body.
This is easily visualized with Mentat by activating Identify Contact Bodies.

!

Figure 6-2 Orientation of Rigid Body Segments
For three-dimensional analysis, the interior side is formed by the right-hand rule along a patch. The interior side is
visualized in Mentat as the front surface (pink); whereas, the exterior side is visualized in Mentat as the back surface (gold).
It is recommended that rigid bodies in 2-D defined with curves and rigid bodies in 3-D defined with surfaces have no
branches. If branches exist, sliding contact in the area of the branches is unpredictable and may lead to wrong behavior.
It is not necessary for rigid bodies to define the complete body. Only the bounding surface needs to be specified. You should
take care, however, that the deforming body cannot slide out of the boundary curve in 2-D (Figure 6-3). This means that it
must always be possible to decompose the displacement increment into a component normal and a component tangential
to the rigid surface.
Incorrect Correct
Figure 6-3 Deformable Surface Sliding Out of Rigid Surface
Contact Tables
Although the contact procedures in Marc use efficient searching algorithms, it can be advantageous to reduce the number
of possible contact body pairs, especially if there are many contact bodies present in the model to be analyzed. Where by
default all deformable contact bodies can potentially touch all other deformable and rigid bodies, by defining a CONTACT
TABLE, the number of potential contact pairs can be set by the user. Via a contact table, not only the possible contact pairs
are defined, but also the parameters used for each contact pair can be defined. A typical example of the latter is the
coefficient of friction, but later on in this chapter several other parameters will be discussed as well, which can have either
a physical or numerical significance.
In Marc versions prior to version 2013, both the contact pairs and the corresponding parameters were entered on the
CONTACT TABLE option itself. Since in many cases the same set of parameters is used for multiple contact pairs, it could be
laborious to enter this data or modify it in an existing model. As of version 2013, the user can enter all possible contact
parameters for a contact pair separately on the CON INTERA option and assign data to a contact pair using the CON TABLE
option. In order to actually apply the data entered on these two options, the CONTACT TABLE version 4 or higher has to be
used. Since the data is referred to by ID’s, the format of the CONTACT TABLE is very compact and entering or changing data
needs to happen only on the CON INTERA or CON TABLE options. These options allow the user to enter many sets of contact
parameters and possible contact pairs. They can be referred to by repeated use of the CONTACT TABLE option in the history
definition part of the Marc input file.
It should be noted that in this chapter no distinction will be made between entering data directly on a CONTACT TABLE or
on the CON INTERA and CON TABLE options. Both ways of entering data are equivalent.
Motion of Bodies
The motion of deformable bodies is prescribed using the conventional methods of applying displacements, forces, or
distributed loads to the bodies. It is advantageous to not apply displacements to nodes which might come into contact with
other bodies, as this may cause conflicts with the constraints from the contact procedure. If a prescribed displacement is to
be imposed, it is better to introduce another rigid body and apply the motion to the rigid body. Symmetry surfaces are
treated as a special type of bodies which have the property of being frictionless and where the nodes are not allowed to
separate. If a distributed load is applied to an edge or face, it is possible for this load to be deactivated if all nodes on the
edge or face are in contact with another body (node-to-segment procedure) or if the edge or face are in partial contact with
another body (segment-to-segment contact).
CHAPTER 6 507
Contact
There are four ways to prescribe the motion of rigid surfaces:

 Prescribed velocity.
 Prescribed position.
 Prescribed load.
 Prescribed scaling.
Associated with the rigid body is a point labeled the centroid. When the first two methods are chosen, you define the
translational motion of this point, and the angular motion about an axis through this point. The direction of the axis can
be defined for three-dimensional problems. For two-dimensional problems, it is a line normal to the plane. For complex
time-dependent behavior, the MOTION user subroutine can be used to prescribe the motion as an alternative to the input.
The motion during a time increment is considered to be linear. The position is determined by an explicit, forward
integration of the velocities based upon the current time step. A time increment must always be defined even if a static,
rate-independent analysis is performed.
When load controlled rigid bodies are used, two additional nodes, called the control nodes, are associated with each rigid
body. In 2-D problems, the first node has two translational degrees of freedom (corresponding to the global x- and y-
direction) and the second node has one rotational degree of freedom (corresponding to the global z-direction). In 3-D
problems, the first node has three translational degrees of freedom (corresponding to global x-, y-, and z-direction) and the
second node has three rotational degrees of freedom (corresponding to the global x-, y-, and z-direction). In this way, both
forces and moments can be applied to a body by using the POINT LOAD option for the control nodes. Alternatively, one may
prescribe one or more degrees of freedom of the control nodes by using the FIXED DISP or DISP CHANGE options. Generally
speaking, load-controlled bodies can be considered as rigid bodies with three (in 2-D) or six (in 3-D) degrees of freedom.
These degrees of freedom may be in a user-defined system of either the TRANSFORMATION or COORD SYSTEM option is used.
The prescribed position and prescribed velocity methods (see Figure 6-4) have less computational costs than the prescribed
load method (see Figure 6-5).
2
Centroid
1 3
V 
2
1
Figure 6-4 Velocity Controlled Rigid Surface
Fy
Mz
Fx
Extra Node
Figure 6-5 Load Controlled Rigid Surface
If the second control node is not specified, the rotation of the body is prescribed to be zero. If the second control node is
identical to the first control node, the rotation of the body is defined by the third degree of freedom in a 2-D analysis, or
by the fourth, fifth and sixth degrees of freedom in a 3-D analysis. In case of two identical control nodes, the RBE parameter
option is required if the model does not contain shell or beam elements.
It should be noted that the nodal coordinates of the control nodes are defined by the center of rotation of the body. So
coordinates entered by the user will be replaced by the coordinates of the center of rotation.
It is also possible to either expand or contract the size of the rigid body through the use of specified scale factors multiplied
by a table to make the scaling time dependent. Alternately, the UGROWRIGID user subroutine may be used to provide scale
factors. The scale factors must be initialized to 1.0 but can change afterwards. Abrupt or discontinuous changes in the scale
factors should be avoided while the rigid surface is in contact with other bodies. If no rotations are applied, the scale factors
can be different in the x-, y-, and z-directions.
Initial Conditions
At the beginning of the analysis, bodies should either be separated from one another or in contact. Bodies should not
penetrate one another unless the objective is to perform an interference fit calculation. Rigid body profiles are often
complex, making it difficult for you to determine exactly where the first contact is located. Before the analysis begins
(increment zero), if a rigid body has a nonzero motion associated with it, the initialization procedure brings it into first
contact with a deformable body. No motion or distortion occurs in the deformable bodies during this process. In a coupled
thermal mechanical analysis, no heat transfer occurs during this process. If more than one rigid body exists in the analysis,
each one with a nonzero initial velocity is moved until it comes into contact. Because increment zero is used to bring the
rigid bodies into contact only, you should not prescribe any loads (distributed or point) or prescribed displacements
initially. For a multi-stage contact analysis (often needed to simulate manufacturing processes), the APPROACH and
SYNCHRONIZED options in conjunction with the CONTACT TABLE and MOTION CHANGE options allow you to model contact
bodies so that they just come into contact with a deformable body (often called the blank or the work piece). A similar
procedure to establish initial contact can be applied to deformable bodies.
In assembly analysis, it is possible that multiple bodies are initially in contact. Because of mesh discretization, it is possible
that the contact is not perfect, which would result in inducing stresses due to overclosure. The CONTACT TABLE option may
be used to specify that the initial contact should be stress free. For node-to-segment contact, stress-free projection is handled
CHAPTER 6 509
Contact
by relocating the coordinates of the touching nodes onto the touched surface. For segment-to-segment contact, the stress-
free projection is handled in two steps:
1. A nodal stress-free projection is done at the start of the analysis (increment 0) for each body pair for which stress-
free projection is flagged. Between the two bodies of the pair, the body that has more touching nodes within the
distance tolerance is projected onto the other body – i.e., the coordinates of the nodes are physically adjusted to be
put onto the other body.
Note: The nodal projection is assessed by default for Version 15 and later input files. If not desired, it can be turned
off via FEATURE,17702. For pre-version 15 files, nodal projection is not done by default in order to maintain
backward compatibility. If nodal projection is desired for pre-version 15 files, it can be turned on via
FEATURE,17703.
2. After the one-time nodal adjustment, any additional offsets for the patches are computed at the polygon points of
the patch. This offset is applied on the fly during the analysis at the individual point level. Note that the associated
patch positions are not physically adjusted to reflect the polygon point offsets. Also note that when there is large
sliding and new polygon points are generated, there is no provision to compute and apply the stress-free projection
at the new points. This limitation is mitigated by applying the nodal projection as a first step.
A stress-free projection tolerance can be independently defined for the initial contact to define the zone in which the
relocation of nodes / offset takes place. In some cases, it is useful to prescribe a larger stress-free projection tolerance for the
initial contact so that more entities are adjusted at the start. Note that the bias factor is applied to the stress-free tolerance
as well. If left undefined, the stress-free projection tolerance is taken to be the same as the regular contact tolerance. For
both node-to-segment and segment-to-segment contact, the nodal stress-free projection vector can be visualized on the post
file. This vector is available by default on the post file for Version 15 and later input files. For pre-version 15 files, it can be
put on the post file by choosing nodal vector 110 (Stress Free Projection Vector).
Initial Gap
This option can be used to define the initial gap or overlap between contacting bodies.
A uniform gap between the contacting bodies can be specified in Mentat. To define the spatially varying gap, the user
should use the UINITGAP user subroutine. Optionally, the user can specify the search direction at a touching body node using
the same user subroutine. The initial gap option can only be used at the beginning of analysis; hence, it has to be specified
in the initial contact table in the Jobs menu. Unlike the stress free projection option, the contacting node coordinates are
not modified when using the initial gap option.
Following are the inputs required to define the initial gap for a contact pair:
1. Gap/overlap magnitude (Positive value indicates gap between contact bodies, and negative value indicates
overclosure between contact bodies)
2. Search tolerance
3. Contact body to be adjusted (required for double sided contact)
4. Search direction (through user subroutine)
With the initial gap option, the nodes of the touching body are projected onto the nearest segment of the touched body to
get the projection vectors at the touching nodes. In case of double-sided contact, the user must specify the contact body,
the nodes of which will be projected to obtain the projection vector. Search for the contact is limited to the search tolerance
zone. The projection vector at a touching body node is then modified to accommodate the user-specified gap or
overclosure. During the analysis, the projection vector is continuously updated based on the displacement and the rotation
of the associated node. In case of a gap between the contacting bodies, the nodes are not in contact at the beginning of the
analysis. As the bodies moves towards each other, contact is detected. In case of a overlap between the contacting bodies,
the interference is resolved in the first increment of the loadcase.
Following points should be noted to avoid unwanted contact detection:
1. To specify the selective contact nodes involved in initial gap adjustment, user can use CONTACT NODE option for node
to segment contact. Same can be achieved using EXCLUDE option in case of segment to segment contact. In this case
nodes not belonging to excluded segment will be the candidate for initial gap adjustment.
2. Comparatively finer meshed body should be selected for the initial gap adjustment.
3. Use of analytical boundary description is recommended for the faceted touched segments.
4. In case of penetrating shell bodies, a top or bottom surface should be defined using boundary redefinition option
available in Contact Table menu.
An alternative way to define initial gap is to specify project stress free option available in Contact Interaction menu at
increment zero. And in subsequent loadcases, the user can define allowable interference magnitude using the contact
normal option. In this case, the table referred for the interference magnitude is of constant magnitude throughout the
analysis.
Node-to-Segment Contact
Numbering of Contact Bodies

When defining contact bodies for a deformable-to-deformable analysis, it is important to define them in the proper order.
As a general rule, a body with a finer mesh should be defined before a body with a coarser mesh.
Note: For problems involving local adaptivity or global remeshing, care must be taken to satisfy this rule before as well
as after the mesh change.
If one has defined a body numbering which violates the general rule, or if the rule is violated upon remeshing, then a
CONTACT TABLE model or history definition option can be used to modify the order in which contact is established. This order
can be directly user-defined or decided by the program. In the latter case, the order is based on the rule that if two
deformable bodies might come into contact, searching is done such that the contacting nodes are on the body having the
smallest element edge length and the contacted segments are on the body having the coarser mesh. It should be noted that
this implies single-sided contact for this body combination, as opposed to the default double-sided contact.
For deformable-deformable contact, the CONTACT option offers the following three global contact search controls:
 SINGLE-SIDED
 DOUBLE_SIDED
 OPTIMIZE CONTACT CONSTRAINT EQUATIONS
If flexible region A comes into contact with flexible region B, the following rules will be satisfied by the program to
determine the constraint equations:
CHAPTER 6 511
Contact
1. Region A and region B belong to the same contact body:

a. If OPTIMIZE CONTACT CONSTRAINT EQUATIONS is active, the finer meshed region contacts the
coarser meshed region.
b. If OPTIMIZE CONTACT CONSTRAINT EQUATIONS is not active, the region containing the lower
node number coming into contact contacts the other region.
2. Region A and region B do not belong to the same contact body:
a. If SINGLE-SIDED contact is active, the region belonging to the lower contact body number contacts the
region belonging to the higher contact body number.
b. If DOUBLE_SIDED contact is active, the region containing the lower node number coming into contact
contacts the other region. Because of the search order for contact is from lower body numbers to higher body
numbers, this often leads to the same solution as mentioned above.
c. If OPTIMIZE CONTACT CONSTRAINT EQUATIONS is active, the region with the softer material
contacts the region with the harder material, or the region with the finer mesh contacts the region with the
coarser mesh (soft-hard criterion has priority over mesh-density criterion).
d. All the above settings can be overruled by the CONTACT TABLE model or history definition option where you
can specify that a certain body AA should contact body BB and not vice-verse, or that the body with the smallest
element edge should contact the other body.
As an example, two identical bodies shown in Figure 6-6, but created using two different approaches so that the
node numbering order is different. Each U-Section is one contact body. In particular, for the model on the left,
the left leg of the U-section has lower node ids that the right leg; while for the model on the right, the opposite
is true.
Figure 6-6 Node Numbers on Models
A uniform distributed load is placed on the right leg, such that self-contact occurs. Without using optimized
contact, the behavior is shown in Figure 6-7.
Figure 6-7 Contact Based upon Original Model.
The dark squares indicate nodes that are in contact based upon the Contact Status. Observe that the model on
the left gives good behavior, while that on the right does not. This is because the lower node numbers on the
left leg detect contact with the right leg before the corner node on the right leg (which has a higher id) detects
contact with the left leg. Then, activate the Optimized Contact Constraint option, and the results are shown
in Figure 6-8. The node number IDs are no longer controlling the solution.
Figure 6-8 Contact using Optimized Contact Constraints
Each model is then subdivided into two contact bodies as shown in Figure 6-9, the contact table option is not
used and the optimized procedure is not used.
CHAPTER 6 513
Contact
Figure 6-9 Contact Based upon Contact Body Numbering
Observe now that in the left model, the left leg is a lower body number, so the nodes on this body try to contact
the segments on the right leg. In the model on the right then, the right leg has the lower body number, and so
the nodes on this leg contact (successfully) with the segments on the left leg.
When either the Optimized Contact Constraint Procedure or a well-selected Contact Table utilizing Single-
Sided Contact, the consistent results may be obtained for either model as shown in Figure 6-10
Figure 6-10 Contact Using either Optimized Contact Constraint or Contact Table.
Detection of Contact
During the incremental procedure, each potential contact node is first checked to see whether it is near a contact segment.
The contact segments are either edges of other 2-D deformable bodies, faces of 3-D deformable bodies, or segments from
rigid bodies. By default, each node could contact any other segment including segments on the body that it belongs to.
This allows a body to contact itself. To simplify the computation, it is possible to use the CONTACT TABLE option to indicate
that a particular body will or will not contact another body. This is often used to indicate that a body will not contact itself.
During the iteration process, the motion of the node is checked to see whether it has penetrated a surface by determining
whether it has crossed a segment.
Because there can be a large number of nodes and segments, efficient algorithms have been developed to expedite this
process. A bounding box algorithm is used so that it is quickly determined whether a node is near a segment. If the node
falls within the bounding box, more sophisticated techniques are used to determine the exact status of the node.
During the contact process, it is unlikely that a node exactly contacts the surface. For this reason, a contact tolerance
(Figure 6-11) is associated with each surface.
nce
2 x Tolera
Figure 6-11 Contact Tolerance
If a node is within the contact tolerance, it is considered to be in contact with the segment. The default contact tolerance
is calculated by the program (see Tolerance Values in this chapter). It is also possible to define the contact tolerance through
the input, either globally via the CONTACT option or per pair of contact bodies via the CONTACT TABLE option.
t  t+ t   t+  t 
During an increment, if node A moves from A to A , where A is beyond the contact tolerance, the node
is considered to have penetrated (Figure 6-12). In such a case, a special procedure is invoked to avoid this penetration. More
details are discussed in Numerical Aspects of the Direct Constraint Procedure in this chapter.
A(t)
A(t + t)
Figure 6-12 Trial Displacement with Penetration
The size of the contact tolerance has a significant impact on the computational costs and the accuracy of the solution. If
the contact tolerance is too small, detection of contact and penetration is difficult which leads to higher costs. Penetration
of a node happens in a shorter time period leading to more recycles due to iterative penetration checking or to more
increment splitting and increases the computational costs. If the contact tolerance is too large, nodes are considered in
contact prematurely, resulting in a loss of accuracy or more recycling due to separation. Furthermore, the accepted solution
might have nodes that “penetrate” the surface less than the error tolerance, but more than desired by the user. The default
error tolerance is recommended.
CHAPTER 6 515
Contact
Many times, areas exist in the model where nodes are almost touching a surface (for example, in rolling analysis close to
the entry and exit of the rolls). In such cases, the use of a biased tolerance area with a smaller distance on the outside and
a larger distance on the inside is advised. This avoids the close nodes from coming into contact and separating again and
is accomplished by entering a bias factor. The bias factor should be given a value between 0.0 and 0.99. The default is 0.0
or no bias. Also, in analyses involving frictional contact, a bias factor for the contact tolerance is recommended. The outside
contact area is (1. - bias) times the contact tolerance on the inside contact area (1. + bias) times the contact tolerance
(Figure 6-13). The bias factor recommended value is 0.95.
In some instances, you might wish to influence the decision regarding the deformable segment a node contacts (or does
not contact). This can be done using the EXCLUDE option.
(1 - Bias) tolerance
(1 + Bias) tolerance
Figure 6-13 Biased Contact Tolerance
Shell Contact
A node on a shell makes contact when the position of the node plus or minus half the thickness projected with the normal
comes into contact with another segment (Figure 6-14). In 2-D, this can be shown as:
x1 = A + n t  2
x2 = A – n t  2
Shell
Midsurface
2
x A t
rance
2x tole
x
Figure 6-14 Default Shell Contact
If point x or y falls within the contact tolerance distance of segment S, node A is considered in contact with the segment
S. Here x 1 and x 2 are the position vectors of a point on the surfaces 1 and 2 on the shell, A is the position vector of a
point (node in a discretized model) on the midsurface of the shell, n is the normal to the midsurface, and t is the shell
thickness.
As the shell has finite thickness, the node (depending on the direction of motion) can physically contact either the top
surface, bottom surface, or mathematically contact can be based upon the midsurface. You can control whether detection
occurs with either both surfaces, the top surface, the bottom surface, or the middle surface. In such cases, either two or one
segment will be created at the appropriate physical location. Note that these segments will be dependent, not only on the
motion of the shell, but also the current shell thickness (Figure 6-15).
Note: Shell elements should be oriented consistently. Also, a shell segment cannot be in contact simultaneously on
both the Top and Bottom sides. If simultaneous contact with rigid bodies is required, use the solid shell
element or switch to segment-to-segment contact.
S1
n n S1
2
S2
2
1
1
Include Both Segments Top Segment Only
2
n
S1
S2 2
1
1
Bottom Segments Only Ignore Shell Thickness
Figure 6-15 Selective Shell Contact
S 1 S 2 are segments associated with shell consisting of node 1 and 2.
Neighbor Relations
When a node is in contact with a rigid surface, it may slide from one segment to another. In 2-D, the segments are always
assumed to be continuous and so are the segment numbers. Hence, a node in contact with segment n slides to segment
n – 1 or to segment n + 1 (Figure 6-16). This simplifies the implementation of contact.
n-1
n
n+1
Figure 6-16 Neighbor Relationship (2-D)

CHAPTER 6 517
Contact
In 3-D, the segments are often discontinuous (Figure 6-17). This can be due to the subdivision of matching surfaces or, more
likely, the CAD definition of the under lying surface geometry.
Hole Physically
Discontinuous
Nonmatching
Segments
Continuous Surface Segments Discontinuous Surface Geometry
Figure 6-17 Neighbor Relationship (3-D)
Continuous surface geometry is highly advantageous as a node can slide from one segment to the next with no interference
(assuming the corner conditions are satisfied). Discontinuous surface geometry results in additional operations when a
node slides off a patch and cannot find an adjacent segment. Hence, it is advantageous to use geometry clean-up tools to
eliminate small sliver surfaces and make the surfaces both physically continuous and topologically contiguous.
Implementation of Constraints
For contact between a deformable body and a rigid surface, the constraint associated with no penetration is implemented
by transforming the degrees of freedom of the contact node and applying a boundary condition to the normal
displacement. This can be considered solving the problem:
K ââ K âb  u â   f â 
  =  
K b â K bb  u b   fb 
where â represents the nodes in contact which have a local transformation, and b represents the nodes not in contact and,
hence, not transformed. Of the nodes transformed, the displacement in the normal direction is then constrained such that
u ân is equal to the incremental normal displacement of the rigid body at the contact point.
t
P
n
Figure 6-18 Transformed System (2-D)
As a rigid body can be represented as either a piecewise linear or as an analytical (NURBS) surface, two procedures are used.
For piecewise linear representations, the normal is constant until node P comes to the corner of two segments as shown in
Figure 6-19. During the iteration process, one of three circumstances occur. If the angle  is small
 –  smooth     smooth  , the node P slides to the next segment. In such a case, the normal is updated based upon the
new segment. If the angle  is large (    smooth or   –  smooth ) the node separates from the surface if it is a convex
corner, or sticks if it is a concave corner. The value of  smooth is important in controlling the computational costs. A larger
value of  smooth reduces the computational costs, but might lead to inaccuracies. The default values are 8.625° for 2-D
and 20° for 3-D. These can be reset using the PARAMETERS option in the model definition or history definition section.



Convex Corner Concave Corner

Figure 6-19 Corner Conditions (2-D)
In 3-D, these corner conditions are more complex. A node (P) on patch A slides freely until it reaches the intersection
between the segments. If it is concave, the node first tries to slide along the line of intersection before moving to segment
B. This is the natural (lower energy state) of motion.
These corner conditions also exist for deformable-to-deformable contact analysis. Because the bodies are continuously
changing in shape, the corner conditions (sharp convex, smooth or sharp concave) are continuously being re-evaluated.
When a rigid body is represented as an analytical surface, the normal is recalculated at each iteration based upon the current
position. This leads to a more accurate solution, but can be more costly because of the NURBS evaluation.
CHAPTER 6 519
Contact
B
P
Figure 6-20 Corner Conditions (3-D)
When a node of a deformable body contacts a deformable body, a multipoint constraint (called tying) is automatically
imposed. Recalling that the exterior edges (2-D) or faces (3-D) of the other deformable bodies are known, a constraint
expression is formed. For 2-D analysis using lower-order elements, the number of retained nodes is three – two from the
edge and the contacting node itself. For 3-D analysis, the number of retained nodes is five – four from the patch and the
contacting node itself. When using higher-order elements and true quadratic contact, the number of retained nodes for 2-
D becomes four, for 3-D, (hexahedrals) nine, and for 3-D (tetrahedrals) seven. The constraint equation is such that the
contacting node should be able to slide on the contacted segment, subject to the current friction conditions. This leads to
a nonhomogeneous, nonlinear constraint equation. In this way, a contacting node is forced to be on the contacted segment.
This might introduce undesired stress changes, since a small gap or overlap between the node and the contacted segment
will be closed (note that using the Single Step Houbolt or generalized alpha/HHT operator in dynamics, forcing a node to
be on the contacted segment is switched off by default, but can be activated via the PARAMETERS option). During initial
detection of contact (increment 0), the stress-free projection option avoids those stress changes for deformable contact by
adapting the coordinates of the contacting nodes such that they are positioned on the contacted segment. This stress-free
projection can be activated using CONTACT TABLE. A similar option exists for glued contact; however, in this case, overlap
will not be removed.
During the iteration procedure, a node can slide from one segment to another, changing the retained nodes associated with
the constraint. A recalculation of the bandwidth is automatically made. Because the bandwidth can radically change, the
bandwidth optimization is also automatically performed.
A node is considered sliding off a contacted segment if is passes the end of the segment over a distance more than the contact
tolerance. As mentioned earlier, the node separates from the contacted body if this happens at a convex corner. For
deformable contact, this tangential tolerance at convex corners can be enlarged by using the delayed sliding off option
activated via CONTACT TABLE.
Friction Modeling
Friction is a complex physical phenomenon that involves the characteristics of the surface such as surface roughness,
temperature, normal stress, and relative velocity. An example of this complexity is that quite often in contact problems
neutral lines develop. This means that along a contact surface, the material flows in one direction in part of the surface and
in the opposite direction in another part of the surface. Such neutral lines are, in general, not known a priori.
The actual physics of friction and its numerical representation continue to be topics of research. Currently, in Marc the
modeling of friction has basically been simplified to two idealistic models. The most popular friction model is the Coulomb
friction model. This model is used for most applications with the exception of bulk forming as encountered in e.g. forging
processes. For such applications the shear friction model is more appropriate.
Coulomb Friction
The Coulomb model can be characterized by:
 t   n (stick) and  t = –   n  t (slip)
where
t is the tangential (friction) stress

 n is the normal stress
 is the friction coefficient
t is the tangential vector in the direction of the relative velocity:
vr
t = ----------
- , in which v r is the relative sliding velocity.
vr
Similarly, the Coulomb model can also be written in terms of nodal forces instead of stresses:
f t  f n (stick) and f t = –  f n  t (slip)
where
ft is the tangential (friction) force

fn is the normal force
When Coulomb friction is used with the stress-based model, the integration point stresses are first extrapolated to the nodal
points and then transformed, so that a direct component is normal to the contacted surface. Given this normal stress and
the relative sliding velocity, the tangential stress is then evaluated and a consistent nodal force is calculated. For shell
elements, the nodal force based Coulomb model must be used, since a reduced stress state is assumed with  n = 0 . For
continuum elements, both the stress-based and force-based Coulomb model can be used.
For a given normal stress or normal force, the friction stress or force has a step function behavior based upon the value of
the relative sliding velocity v r or the tangential relative incremental displacement u t , as outlined in Figure 6-21 for a 2-D
case, where the relative velocity and incremental displacement are scalar values.
CHAPTER 6 521
Contact
 t or f t
Stick
v r or u
t
Slip
Figure 6-21 Coulomb Friction Model
Since this discontinuity in the friction value may easily cause numerical difficulties, different approximations of the step
function have been implemented. They are graphically represented in Figure 6-22 and they will be successively discussed.
 t or f t ft ft
vr u t u t
arctangent model stick-slip (modified step bilinear model

function) model
Figure 6-22 Different Approximations for the Coulomb Friction Model
Arctangent Model
The first procedure is based on a continuously differentiable function in terms of the relative sliding velocity:
2 vr
 t = –   n --- arctan  -------------------------  t
  RVCNST
for the friction stress, and:
2 vr
f t = –  f n --- arctan  -------------------------  t
  RVCNST
for the friction force.
Physically, the value of RVCNST can be seen as the value of the relative velocity below which sticking occurs. The value
of RVCNST is important in determining how closely the mathematical model represents the step function, as shown in
Figure 6-23. A very large value of RVCNST results in a reduced value of the effective friction. A very small value may result
in poor convergence. It is recommended that the value of RVCNST be 1% to 10% of a typical relative sliding velocity,
vr .
The friction implementation not only affects the external force vector, but also the stiffness matrix of the final set of
equations to be solved on a global level. This stiffness matrix contribution follows from:
f t v r
k
K ij = – ----------i -------------
v r u t
k j
This later contribution, if fully implemented, would lead to a non-symmetric system. Because of the additional
computational costs, both in terms of memory and CPU costs, the contribution to the stiffness matrix is symmetrized.
ft
f n
vr
– 10 10
RVCNST = 10
RVCNST = 1
– fn RVCNST = 0.1
Figure 6-23 Step Function Approximation for Different Values of RVCNST (-10<vr<10)
i
During iteration  i  of the Newton-Raphson process, the change of the friction force f t is related to the change of the
i
relative sliding velocity v r . Since the latter can be written as:
i i – 1 i
i u t – u t u t
v r = ----------------------------------------
- = -----------
-
t t
the resulting equation can still be expressed in displacement degrees of freedom. During the very first iteration of an
increment, it turns out that the convergence can be significantly improved by using:
1  previous increment  1
1 u t u t u t  previous increment 
v r = -------------
- – ------------------------------------------------ = ------------
- – vr
t t
 previous increment  t
CHAPTER 6 523
Contact
 previous increment 
where v r results in a contribution to the right hand side of the global set of equations.
As a result of the smoothing procedure, a node in contact always has some slipping. Besides the numerical reasons, this ‘ever
slipping node’ model has a physical basis. Oden and Pires pointed out that for metals, there is an elasto-plastic deformation
of asperities at the microscopic level (termed as ‘cold weld’) which leads to a nonlocal and nonlinear frictional contact
behavior. The arctangent representation of the friction model is a mathematical idealization of this nonlinear friction
behavior.
A complicating aspect of the arctangent model is that it may be difficult to estimate a priori the typical relative sliding
velocity. This is particularly true in quasi-static deformable contact analyses (where the velocity is not an input quantity)
and in cases where the sliding velocity varies strongly during the analysis.
Stick-slip (Modified Step Function) Model

The second procedure is based on a slightly modified step function and can be used to simulate true stick-slip behavior. In
this procedure, each node in contact gets a friction status, being either stick or slip. Depending on this status, different
constraints are applied and after each iteration in the iterative solution process, the correctness of the friction status is
checked and if necessary adapted.
The typical parameters used (see also Figure 6-24) are the friction coefficient multiplier  , the slip to stick transition region
 and the friction force tolerance e .
The multiplier  (with a default value of 1.05 ) can be seen as an overshoot parameter used to simulate the difference
between static and dynamic friction.
–6
The slip-to-stick transition region  (with a default value of 1 10 ) can be seen as a tolerance on the friction solution.
If a node is in slipping mode and moves in the direction of the friction force (which is physically incorrect), but the
corresponding relative displacement magnitude is within the slip-to-stick transition region, then this will not cause the
increment to be restarted with modified friction conditions.
–6
Related to the parameter  , a reduction factor  = 1.0 10 is used. This fixed factor causes the product of  and 
to be very small, so that it can be used as a criterion to estimate if a node is initially in slipping or sticking mode. This initial
assumption serves as the starting point of the above mentioned process of checking and adapting the friction status, which,
for a 2-D analysis, is given in Figure 6-25 (note that u t  0 implies u t   ).
ft  : friction coefficient multiplier
 : slip to stick transition region

–6
f n   : reduction factor ( 1 10 )
f n
 u t
Figure 6-24 Stick-slip or Modified Step Function Friction Parameters
Initial Contact
No Yes
u t  0
Assume Slipping Assume Sticking

Mode Mode
Determine Solution
of Next Iteration
Remain in Slipping Mode if: Remain in Sticking Mode if:

f t  u t  0 and u t   f t  f n
Change to Sticking Mode if:

f t  u t  0 and u t   Change to Slipping Mode if:
f t  f n
or if u t  0
Figure 6-25 Flow Diagram for the Stick-slip or Modified Step Function Friction Model
CHAPTER 6 525
Contact
When a node is in slipping mode, the contribution to the stiffness matrix is:
f t
i
K i j = – -------------
u t
j
while in sticking mode constraints are applied which exactly enforce a zero relative tangential displacement. It should be
noted that in slipping mode the current friction force is not only a function of the current normal force and relative
displacement, but also of the friction force during the previous iteration. This generally improves the stability of the
nonlinear solution process.
As opposed to the procedure based on the arctangent model, this procedure requires additional friction-based testing.
Changes in the friction status as well as too large changes in the friction forces compared to the previous iteration may cause
the program to perform an additional iteration. In 2-D, the latter implies that the friction force should fulfil1:
ft
1 – e  ----p-  1 + e
ft
where f p is the friction force of the previous iteration and e is the above mentioned friction force tolerance, with a default
t
value of 0.05. In 3-D, in addition to this requirement on the magnitude of the friction force, a check is performed on the
change in direction of the friction force vector:
1 – e  cos   f   1 + e
in which  f is the angle between the friction vector in the current and the previous iteration.
In order to limit the amount of friction-related iterations, the above mentioned requirements will not cause extra iterations
if the number of nodes not fulfilling the friction requirements compared to the total number of nodes in contact is small.
This can be expressed in the friction force tolerance as:
number of nodes which do not fulfil friction tolerances
-----------------------------------------------------------------------------------------------------------------------------------  e
total number of nodes in contact
and must be true for all contact bodies in the model.
If the maximum number of iterations as defined on the CONTROL model definition option has been reached, additional
friction iterations will not be enforced and the analysis will continue with the next increment.
The stick-slip model is always based on nodal forces. This model is not recommended for quadratic elements in 3-D,
because of the varying magnitude and sign of nodal forces of such elements.
When using the PRINT parameter to get extra contact-related information in the output file, the program will echo the
contact status, the ratio of the magnitude of the current and previous friction force vector (and in 3-D also the cosine of
the angle between the current and previous friction force vector), as well as the total number of nodes with a friction
contribution and the number of nodes with a friction contribution which do not fulfill the friction tolerance criteria.
Bilinear Model
The third procedure is called the bilinear model. Similar to the modified step function model, it is based on relative
tangential displacements. Instead of defining special constraints to enforce sticking conditions, the bilinear model assumes
that the stick and slip conditions correspond to reversible (elastic) and permanent (plastic) relative displacements,
respectively. The clear resemblance with the theory of elasto-plasticity will be used to derive the governing equations.
First, Coulomb’s law for friction is expressed by a slip surface  :
 = f t – f n
The stick or elastic domain is given by   0 , while   0 is physically impossible.

Next, the rate of the relative tangential displacement vector is split into an elastic (stick) and a plastic (slip) contribution
according to:
e p
u· t = u· t + u· t (6-1)
and the rate of change of friction force vector is related to the elastic tangential displacement by:
f·t = Du· t
e
(6-2)
in which matrix D is given by:
f
-------n- 0
D = 
f n
0 --------

with  the slip threshold or relative sliding displacement below which sticking is simulated (see Figure 6-26). The value of
 is by default determined by the program as 0.0025 times the average edge length of the finite elements defining the
deformable contact bodies.
Now attention is paid to the case that, given a tangential displacement vector, the evolution of f t would result in a
physically impossible situation, so   0 . This implies that the plastic or slip contribution must be determined.
Equations (6-1) and (6-2) yield:
· · ·p
ft = D  ut – ut 
CHAPTER 6 527
Contact
ft
f n
 u t
 slip threshold
Figure 6-26 Bilinear Model
It is assumed that the direction of the slip displacement rate is given by the normal to the slip flow potential  , given by:
 = ft
·
So that, by indicating the slip displacement rate magnitude as :
·p · 
u t =  -------
f t
with the slip surface,  , different from the slip flow potential,  an analogy to nonassociative plasticity can be observed.
Since a ‘force point’ can never be outside the slip surface, it is required that:
·  T ·
 =  ------ f t = 0
 f t
In this way, the magnitude of the slip rate can be determined. To this end, the equations above can be combined to:
 T  · 
- D u t – · ------- = 0
 -----
 f t  f t
or:
- T ·
 ----- Du t
·  f t
 = ----------------------------
 T 
 ----- - D -------
 f t f t
Utilizing this result, the final set of rate equations reads:
   T 
 D -------  ------ D
· f t  f t  · * ·
f t =  D – --------------------------------
- u t =  D – D u t
  T  
  ----- - D ------- 
  f t f t 
*
Similar to non-associative plasticity, matrix D will generally be non-symmetric. In Marc, a special procedure has been
used which results in a symmetric matrix, while maintaining sufficient numerical stability and rate of convergence.
The bilinear model also uses an additional check on the convergence of the friction forces, which has been achieved if the
following equation is fulfilled:
p
Ft – Ft
--------------------------------
-e
Ft
p
where F t is the current total friction force vector (the collection of all nodal contributions), F t is the total friction force
vector of the previous iteration and e is the friction force tolerance, which has a default value of 0.05.
In the output file one can find, per deformable contact body, information about the current total friction force, the current
total normal force and the total friction force change, compared to the previous iteration.
If a node comes into contact and the structure is still stress-free, then the friction stiffness matrix according to the derivation
above will still be zero. This could result in an ill-conditioned system during the next solution of the global set of equations.
To avoid this problem, the initial friction stiffness will be based on the average contact body stiffness (following from the
trace of the matrix defining the material behavior), which is determined during increment 0 of the analysis. Alternatively,
the user can define this initial friction stiffness via the PARAMETERS model definition or history definition option.
Limitations of the Coulomb Model

When the normal force or stress becomes large, the Coulomb friction model may not correlate well with experimental
observations. This is caused by the fact that the Coulomb model predicts that the frictional shear stresses increase to a level
that can exceed the flow stress or the failure stress of the material. As this is not physically possible, a different friction
calculation should be applied. The choices are either to have a nonlinear coefficient of friction, or to introduce the friction
stress limit in the bilinear model or to use the shear based friction model.
CHAPTER 6 529
Contact
t
Linear Coulomb Model

Observed Behavior
n
Figure 6-27 Linear Coulomb Model versus Observed Behavior
A nonlinear friction coefficient can be defined via UFRIC user subroutine or via table driven input.
limit
The friction stress limit for the bilinear model (  t ) can be entered via the CONTACT TABLE option and is used to bound
the maximum friction stress, based on the assumption that the extrapolated and averaged shear (friction) stress in a node
is proportional to the applied shear (friction) force (see Figure 6-28). If the shear stress reaches the limit value, then the
limit
applied friction force is reduced, so that the maximum shear stress is given by min   n  t .
Finally, instead of using an adapted Coulomb friction model, one may use the shear friction model, which is the second
idealistic friction model in Marc.
t
Friction Stress Limit
n
Figure 6-28 Friction Stress Limit used by Bilinear Model
Shear Friction
The shear based model states that the frictional stress is a fraction of the equivalent stress  in the material:
 
 t  m ------- (stick) and  t = – m -------  t (slip)
3 3
where m is the friction factor.

Given the similarity with the Coulomb model, this shear model is implemented using two approximations of the
theoretical step function. First, the arctangent function to smooth out the step function is applied:
 2 vr
 t = – m ------- --- arctan  -------------------------  t
3  RVCNST
This model is available for all elements using the distributed load approach.
Second, the bilinear model has been adapted based on the assumption that the shear (friction) stress in a node is
proportional to the applied shear (friction) force. Similar to the friction stress limit, the shear stress due to friction is limited
by:

 t = min  m n m -------
 3
Friction Coefficient
When a node contacts a rigid body, the coefficient of friction associated with the rigid body is used. When a node contacts
a deformable body, the average of the coefficients for the two bodies are used. The CONTACT TABLE option can be used if
complex situations occur. In general, entering the friction coefficient for a contact body pair via a CONTACT TABLE is strongly
recommended.
Recalling that friction is a complex physical phenomenon, due to variations in surface conditions, lubricant distribution,
and lubricant behavior, relative sliding, temperature, geometry, and so on, it was decided to implement the above two
friction models (Coulomb and shear), and to allow you to extend them, if necessary, by means of the UFRIC user subroutine
or by referencing a table. In user subroutine UFRIC, you provide the friction coefficient or the friction factor as:
 =   x f n T v r  y 
or
m = m  x f n T v r  y 
where,
x is the position of the point at which friction is being calculated

fn is the normal force or stress at the point at which friction is being calculated (depending on the
friction model, see UFRIC for more details)
T is the temperature at the point at which friction is being calculated
vr is the relative sliding velocity between point at which friction is being calculated and surface
y is the equivalent stress at the point at which friction is being calculated
Note that when defining the friction coefficient using a table, the equivalent stress is not available as an independent
variable; instead, the normal stress can be used as an independent variable.
Anisotropic Friction Model

The anisotropic friction model implemented in Marc allows for different friction coefficients in two orthogonal directions
on the contact surface. These directions are referred to as slip directions. This model is an extension of the previously
discussed bilinear friction model. The anisotropic friction model can be expressed in mathematical form as:
CHAPTER 6 531
Contact
1---
f t1 2 f t2 2 2
 -----
- +  --------
-  fn (6-3)
  1  2 
and can be graphically represented as shown in Figure 6-29.
Figure 6-29 Slip Surface for Anisotropic Friction Model in Terms of Contact Force Components
For anisotropic friction, the user needs to define two friction coefficients (  1 and  2 ) and the direction associated with
the first friction coefficient (  1 ). This direction is henceforth referred to as the first slip direction.
For any contact pair, the anisotropic friction can be activated using the CONTACT TABLE option. It should be noted that with
the anisotropic friction model, the friction coefficients specified through the CONTACT TABLE option are the only ones used
in the simulation; friction coefficients defined on a per body basis are neglected. The friction coefficients defined on the
contact table can be associated with a table ID or can be modified by using the UFRIC user subroutine.
The first slip direction definition is specified for an individual body through the CONTACT option. For the anisotropic
friction behavior, at least one of the contact bodies should have orientations defined. If the orientations are defined on both
the contact bodies, then the touching body orientations are used. To facilitate the definition of the slip direction, five
orientation options are provided in Marc:
1. COORDINATE SYSTEM:
In this option, the user can define a coordinate system and select the axis which defines the first slip direction.
2. CURVE:
In this option, the user can assign a list of NURBS curves to the contact body on which slip directions are to be
computed. To compute the slip direction at a node, the closest point on any of the given curves is automatically
detected and the first slip direction is given by the direction of the curve at this point.
3. 1-D ELEMENT DIRECTION:
For a deformable touching body made up of 1-D elements like beams or cables, the first slip direction is computed
as the vector from node one to node two. The slip direction is updated based on the displacements of the two nodes.
4. SHARP EDGE DIRECTION:
This orientation option is for special applications wherein a higher sliding resistance is expected at the sharp edges
of a structure due to its penetration in underlying soft material. For example, a underwater cable resting on the sea
floor. In this option, the set of element edges with discontinuous outward normals define the first slip direction.
The second slip direction is computed by the cross product of the first slip direction with the contact normal
direction. As shown in fig. 8-30 the first slip direction is selected such that the second slip direction is aligned with
the average normal of faces attached to the sharp edge. With this option, the second coefficient of friction is used
only for the sharp edge moving outwards (i.e., in positive second direction).
(a) First Slip Direction: Along Sharp Edge (b) Second Slip Direction: Perpendicular to Sharp Edge
Figure 6-30 First and Second Slip Direction
5. USER SUBROUTINE (UCONORT):

In this option the user can write a subroutine to return the unit vector in the first slip direction. It allows the user
to define the first slip direction at a touching node based on the updated nodal positions. All the required inputs
are passed through the argument.
Updating Slip Directions

The slip direction defined using the first two options are mapped on the contact segments at the start of the analysis. As
the associated contact segment deforms or rotates, the slip direction is automatically updated based on the current
orientation of the segment. For orientations defined on rigid contact bodies, the orientations are updated based on the
current position of the rigid body. For option 3, the slip direction is updated based on the displacement of the two nodes.
For option 4, the slip directions are updated based on the displacement of the nodes belonging to the sharp edges. For
option 5, it is the user’s responsibility to specify any updates to the slip direction.
CHAPTER 6 533
Contact
Additional Output
In addition to standard contact output, additional post quantities like relative slip, frictional forces and frictional stresses
along the first and second slip directions are made available using nodal post codes 82, 83 and 84 respectively. A vector plot
of the slip directions can be visualized using nodal post codes 84 and 85.
Limitations
 The anisotropic friction model can be used with node-to-segment contact algorithm only.
 Anisotropic friction is only available with the bilinear friction model.
 With local adaptivity (see the ADAPTIVE option) or global remeshing (see the ADAPT GLOBAL option), the update in
slip direction is not possible. In that case, only a user subroutine option can be used.
Separation
After a node comes into contact with a surface, it is possible for it to separate in a subsequent iteration or increment.
Mathematically, a node should separate when the reaction force between the node and surface becomes tensile or positive.
Physically, you could consider that a node should separate when the tensile force or normal stress exceeds the surface
tension. Rather than use an exact mathematical definition, you can enter the force or stress required to cause separation.
Separation can be based upon either the nodal forces or the nodal stresses. The use of the nodal stress method is
recommended as the influence of element size is eliminated.
When using higher-order elements and true quadratic contact, the separation criteria must be based upon the stresses as
the equivalent nodal forces oscillate between the corner and midside nodes.
In many analysis, contact occurs but the contact forces are small; for example, laying a piece of paper on a desk. Because
of the finite element procedure, this could result in numerical chattering. Marc has some additional contact control
parameters that can be used to minimize this problem. As separation results in additional iterations (which leads to higher
costs), the appropriate choice of parameters can be very beneficial.
When contact occurs, a reaction force associated with the node in contact balances the internal stress of the elements
adjacent to this node. When separation occurs, this reaction force behaves as a residual force (as the force on a free node
should be zero). This requires that the internal stresses in the deformable body be redistributed. Depending on the
magnitude of the force, this might require several iterations.
There are two schemes available for the nodal separation check in node-to-segment contact:
1. Default check: Nodes are checked for separation only after satisfying all specified convergence criteria. For example,
if the user has flagged Residual Checking with Local Node Equilibrium and Displacement Checking, then nodal
separation is checked only after residuals converge at both the global level (maximum residual force in model /
maximum reaction force in model < tol1) and at the local level (maximum residual force at node / maximum
element force at node < tol2) and displacements have converged (iterative change in displacement in model /
incremental change in displacement in model < tol3).
2. Accelerated check: Nodes are checked for separation as soon as the maximum residual force in the model drops to
less than 10% of the maximum reaction force in the model. If there is separation, the iteration number is reset to
1 and the increment is continued with new contact constraints. If there is no separation, iterations are continued
till all specified convergence criteria are satisfied. This accelerated check can reduce the number of iterations
significantly for many separation dominated problems.
You should note that in static analysis, if a deformable body is constrained only by other bodies (no explicit boundary
conditions) and the body subsequently separates from all other bodies, it would then have rigid body motion. For static
analysis, this would result in a singular or nonpositive definite system. This problem can be avoided by appropriate
boundary conditions. When the direct solvers are used, the AUTOSPC parameter may also be used to suppress singular modes.
Release
A special case of separation is the intentional release of all nodes from a rigid body. This is often used in manufacturing
analysis to simulate the removal of the workpiece from the tools. After the release occurs in such an analysis, there might
be a large redistribution of the loads. It is possible to gradually reduce the residual force to zero, which improves the stability,
and reduces the number of iterations required. The RELEASE history definition option allows the release (separation) of all
the nodes in contact with a particular surface at the beginning of the increment. The rigid body should be moved away
using the MOTION CHANGE option or deactivated using the CONTACT TABLE option to ensure that the nodes do not
inadvertently recontact the surface they were released from.
Glue Contact Conditions

In a contact analysis, a distinction is made between touching and glue conditions. In a structural analysis, a touching
condition triggers the local application of a nonpenetration constraint still allowing relative sliding of the bodies in the
contact interface. The nonpenetration constraint is applied through a tying or boundary condition on the displacement
components normal to the contact surfaces. A glue condition suppresses all relative motions between bodies through tyings
or boundary conditions applying them to all displacement degrees of freedom of the nodes in contact. For elements with
nodes that also have rotational degrees of freedom, the rotations may be additionally constrained to provide a moment
carrying glue capability. For the case that a connection is made to the face of a solid element, the rotations of the touching
node are connected to the translations of the nodes of the contacted patch by a constraint relation based upon the large
rotation RBE3 theory. The activation of the glue conditions and their type is controlled by appropriate settings via the
CONTACT TABLE input. Unless a flag has been set, nodes in contact via the glue option are not allowed to separate.
In most nonstructural analysis types, such as thermal or mass diffusion, the touching condition by default does not trigger
the application of tyings, but allows interaction in the contacting interface through the use of a contact interface coefficient.
For small or zero values of the interface coefficient the bodies practically behave as being insulated. For high values the
interface coefficient acts as a penalty constraint. Alternatively the user may request glue conditions meaning the degrees of
freedom in the contact interface are being coupled through tyings as in the structural case. This has the advantage that the
user does not have to choose a high interface coefficient if ideal transport conditions are desired over the contact interface
(e.g., a high heat transfer coefficient for ideal heat transport in case of a heat transfer analysis or a high mass transfer
coefficient for ideal mass transport in case of a mass diffusion analysis) and it becomes relatively easy to connect dissimilar
meshes. The activation of the glue conditions is controlled by appropriate settings on the CONTACT TABLE input.
The electrostatic analysis type does not support contact interface coefficients and, by default tyings, are applied to the
electric potential degrees of freedom of the nodes in the contacting interfaces. For this case, there is no difference between
touching or glue conditions. It is possible, however, to disconnect bodies electrically by making appropriate settings on the
CONTACT TABLE.
In a three-dimensional magnetostatic analysis, the touching condition means that the tangential components of the
magnetic vector potential in the contact interface are coupled through tyings, but the normal components are not. The
glue condition means that all components of the magnetic vector potential are coupled through tyings. In a 2-D analysis,
CHAPTER 6 535
Contact
the one nonzero component of the magnetic potential is always coupled through a tying constraint and there is no
difference between touching or glue conditions.
In a weakly coupled multi-physics analysis where the equations are being solved in a staggered way, it may be desirable to
have touching conditions in one physics pass, but glue conditions in the other. For example, in a coupled thermo-
mechanical analysis, it may be desirable to have touching conditions in the structural pass allowing relative sliding between
the contact bodies, but glue conditions in the thermal pass simulating ideal heat transfer over the contact interface. In such
cases, the touching or glue conditions may be requested on the CONTACT TABLE for each physics pass separately.
In the strongly coupled analysis types that are supported in Marc (i.e., piezoelectric analysis and soil/pore pressure analysis),
the glue condition simultaneously couples all degrees of freedom of both physics types through tyings and it is not possible
to treat the different physics types separately.
Table 6-2 summarizes the meaning of touching and glue conditions for each possible physics type that is supported in Marc.
The exceptions are fluid and electromagnetic analysis where the cells are blank indicating that these analysis types do not
support contact.
Table 6-2 Contact Condition Summary per Physics Type

Physics Type Pass # Touching Condition Glue Condition
Structural 1 Tyings on normal displacements. Tyings on all displacement degrees of
freedom.
Tangential displacements are free, Moment carrying if desired and
unless there is friction. applicable.
Thermal 2 No tyings. Heat transfer may be Tyings on the temperatures.
controlled through a contact heat
transfer coefficient.
Fluid 3
Electrical in Joule 4 No tyings. Electric current may be Tyings on the voltages.
heating controlled through a contact electrical
coefficient.
Mass diffusion 5 No tyings. Mass transfer may be Tyings on the pressures.
controlled through a contact mass
transfer coefficient.
Electrostatic 6 Tyings on the electric potential degrees Tyings on the electric potential degrees
of freedom. There is no contact of freedom.
interface coefficient.
Table 6-2 Contact Condition Summary per Physics Type (continued)

Physics Type Pass # Touching Condition Glue Condition
Magnetostatic 7 2-D: Tyings on the non-zero magnetic 2-D: Tyings on the nonzero magnetic
potential degrees of freedom. There is potential degrees of freedom.
no contact interface coefficient.
3-D: Tyings on the tangential magnetic 3-D: Tyings on all magnetic vector
vector potential degrees of freedom. potential degrees of freedom.
Normal components are free.
There is no contact interface
coefficient.
Magnetodynamic 8 Tyings on all degrees of freedom for the Tyings on all degrees of freedom for the
magnetic vector potential and tyings on magnetic vector potential and tyings on
the electric potential degrees of the electric potential degrees of
freedom. There is no contact interface freedom.
coefficient.
Piezoelectric 1 Tyings on normal displacements. Tyings on all displacement degrees of
Tangential displacements are free. freedom.
Tyings on the electric potential degrees Tyings on the electric potential degrees
of freedom. of freedom.
Soil/pore pressure 1 Tyings on normal displacements. Tyings on all displacement degrees of
freedom.
Tangential displacements and pressure Tyings on the pressure degrees of
are free. freedom.
Deact Glue
In a structural analysis, an important option related to glued contact is the DEACT GLUE option. Suppose two bodies are glued
together and a part of the interface should have standard contact (in order to model, for example, a crack). The nodes in
the region that should have standard contact are then identified with DEACT GLUE. Without this option, one would have to
split up the contact bodies in order to model this case.
Breaking Glue
There are two capabilities in the CON INTERA option to break up the glued connection. The first one uses a stress criterion
and the second one uses a temperature criterion.
Breaking glue due to stress

When the following criterion is fulfilled at a node or at a polygon point, the glued contact is released:
 m  n
 -----n- +  -----t  1 (6-4)
 S n  S t
CHAPTER 6 537
Contact
Here  n is the contact normal stress (used in tension only),  t the contact tangential stress, and S n , S t , m, and n are
user-defined parameters. A user-defined criterion can be given with the UBREAKGLUE user subroutine.
When a node is released due to breaking glued, its status is changed from being glued to standard contact, permitting
separation and friction.
The contact stresses are calculated using extrapolated stresses for solid elements and as contact force divided by equivalent
area for shell elements.
Nodal post codes are available for postprocessing the separate terms in Equation (6-4) as well as the sum. A nodal post code
for the deact glue status is also available.
Breaking glue due to temperature
The breaking glue due to temperature capability is based on the mean temperature between a touching body node or
polygon point and its projection to the touched body. Two criteria have been defined. The breaking capability can be
triggered either when the local mean temperature exceeds a temperature threshold or when the local mean temperature
decreases below a temperature threshold.
Based on the criterion, the glued contact is broken when the following criterion is fulfilled at a node or at a polygon point:
Breaking above the temperature threshold:
n+1 n+1
0.5   T i + T i pro j   T t h r e sh o l d
Breaking below the temperature threshold:

n+1 n+1
0.5   T i + T i pro j   T t h r e sh o l d
n+1 n+1
where, T i is the temperature at the touching body node or polygon point at the end of the increment and T i p r o j is
the temperature at the projection of the touching body node or polygon point to the touched body.
A thermal pass is required in order to use the current capability. Coupled analyses involving the structural pass, the diffusion
pass and joule heating pass are supported.
The temperature criterion is checked at:
 First contact. When the first contact is established the temperature criterion is checked, before any pass is
performed. If the criterion is fulfilled the breaking is immediately applied for all passes.
 The end of the thermal pass, after convergence. The criterion is checked based on the temperature at the end of
the current increment. If the breaking takes place the new touching contact is applied in the stress pass at the
current increment but will be applied at the next increment for the other passes.
When a node is broken, the structural contact is of touching (standard) type and permits separation and friction.
To know if the temperature criterion has been fulfilled, use the contact status post code that has the value of 1 or 1.5 in
case of standard touching contact or glued contact, respectively.
Bonding
There is a capability on the CON INTERA option to bond two bodies to each other. This turns a touching contact into a glued
contact. This capability is based on a temperature criterion. The mean temperature between a touching body node or
polygon point and its projection to the touched body is compared to a temperature threshold. The bonding capability can
be triggered either when the local mean temperature exceeds a temperature threshold or when the local mean temperature
decreases below a temperature threshold. An extra criterion can be chosen involving the peak temperature which is defined
as the highest temperature reached locally since the start of the analysis. When using this criterion the bonding takes place
only if the mean between the peak temperature at the touching body node or polygon point and the peak temperature at
its projection to the touched body exceeds a peak temperature threshold given by the user. The peak temperature criterion
is particularly useful when the contact is shortly lost in the heated phase and again established when the temperature has
decreased but the glued contact is still expected to take place.
Based on the criteria chosen by the user the bonding takes place at a node i or a polygon point i, when the following
criterion/a is/are fulfilled:
Bonding above the temperature threshold:
n+1 n+1
0.5   T i + T i pr oj   T t h r e sh o ld
Bonding above the temperature threshold combined with a peak temperature:

n+1 n+1
0.5   T i + T i pr oj   T t h r e sh o ld
n+1 n+1
0.5   max  T i  + max  T i p r oj    T p ea k t h r es ho l d
Bonding below the temperature threshold:

n+1 n+1
0.5   T i + T i pr oj   T t h r e sh o ld
Bonding below the temperature threshold combined with a peak temperature:
n+1 n+1
0.5   T i + T i pr oj   T t h r e sh o ld
n+1 n+1
0.5   max  T i  + max  T i p r oj    T p ea k t h r es ho l d
where,
n+1
Ti is the temperature at the node or polygon point at the end of the increment,
n+1
T i pr oj is the temperature of the projection of the touching body node or polygon point to the touched body,
n+1
max  T i  is the highest (peak) temperature reached at that node or polygon point since the beginning of the
analysis, and
n+1
max  T i pro j  is the peak temperature at the projection.
A thermal pass is required in order to use the current capability. Coupled analyses involving the structural pass, the diffusion
pass and joule heating pass are supported.
CHAPTER 6 539
Contact
The temperature criterion is checked at:

 First contact. When the first contact or when the first contact after separation in case of bonding option is
established the temperature criterion is checked, before any pass is performed. If the criterion is fulfilled the
bonding is immediately applied for all passes.
 The end of the thermal pass, after convergence. The criterion is checked by using the temperature at the end of the
current increment. If the bonding takes place the new glued contact is applied in the stress pass at the current
increment but will be applied at the next increment for the other passes.
When a node is not yet bonded the structural contact is of touching (standard) type and permits separation and friction.
To know if the temperature criterion has been fulfilled one can use the contact status post code that has respectively the
value of 1 or 2 in case of standard touching contact or glued contact.
Cohesive Contact
Many simulations involve assemblies of many parts into a single structure. The assembly process may be done through a
variety of processes such as rivets, bolts, spot weld, seam welds, or adhesives. It is often too costly to model each one of these
discrete connectors, and the glued contact capability is used to simplify and reduce the computation costs. While this
method is easy to use, it often results in too stiff of a structure because effectively the connection is rigid. To alleviate this,
a cohesive contact capability is available. Using this simulation process, the connections are not rigid but, in fact, results in
a finite stiffness between the parts. The implementation is done within the glued contact capability, and the requirement
is that the relative motion between the two bodies is small. It should be noted that when a finite contact stiffness is defined
with this method, slight penetration may occur between the two bodies. This is often referred to as soft contact. The contact
stiffness can be defined directly as the stiffness per displacement or the contact stress as a function of the displacement. This
can be provided for both the normal and tangential direction in an independent manner. The cohesive contact capability
is only available with segment-to-segment contact procedure. No augmentation is used.
When defining the contact (normal and tangential) stiffness, one can either define a constant value, or it can be defined as
a function of the relative displacement (independent variable 38) or temperature. It should be noted that contact stiffness
defined is mathematically the same as the penalty factor.
When defining the contact stress versus relative displacement, the gradient of the table is internally calculated and used for
the contact stiffness. To facilitate the gradient calculation, the contact stress needs to be specified as a function of relative
displacement (type 38). In addition, the contact stress can be optionally varied as a function of temperature.
The glues contact stiffness can further be modified with the help of the UGLUESTIF_STS user subroutine.
If displacement is used as an independent variable, then the relative displacement between contacting and contacted point
along contact normal direction will be used for interpolation in normal direction. Similarly for interpolating quantities in
tangential direction, a relative slippage in tangential direction will be used. When the table is a function of temperature,
then the average temperature at the touching and touched points will be used.
For the table referred for contact normal direction, a "-ve" displacement represents opening at contacting points and "+ve"
displacement represents penetration at contacting points. Hence, when using the contact stress method, the table should
be such that positive relative displacement gives positive stress.
For the table referred for contact tangential direction, absolute value of relative tangential displacement is used for
interpolation.
Table defining tangential stress versus relative displacement data should have zero stress at zero relative displacements.
It is recommended that TABLE used for stress table should have extrapolation option selected.
Coupled Analysis
A contact analysis has several consequences on performing a coupled analysis. When performing a coupled contact analysis
between multiple deformable bodies, each body deforms due to mechanical and thermal loads and undergoes heat
conduction. When the bodies come into contact, there is heat flux across the surfaces. You need to provide a coefficient of
heat transfer between the surfaces. This is often quite significant as a hot workpiece might come into contact with a cold
tool set. Additionally, if friction is present, heat is generated based upon:
q = f fr  v r  M eq
where M eq is the mechanical equivalent of heat. Half of the heat generated due to friction is contributed to each body. If
the model allows a body to be rigid, two options are available. In the first case, the rigid body is considered to have a
constant temperature (heat source or heat sink). Heat flux is exchanged between the rigid body and the deformable body.
In the second case, it is required to perform a heat transfer analysis in the rigid body. In this case, the rigid body must be
constructed of finite elements, but they are chosen as heat transfer types. Deformation does not occur in this body, but a
heat transfer analysis is performed. This is computationally more efficient than performing a coupled analyses where all
bodies are deformable. Symmetry bodies have different characteristics as they are frictionless (hence, no heat generated)
and because there is no heat flux across a symmetry plane.
Heat fluxes (FILMS) are automatically created on all the boundaries of the deformable-bodies. Three types of heat fluxes
can be distinguished, where each flux has its own physical characteristic. The type of flux Marc uses, depends on the
distance between the contacting bodies (see Figure 6-31). We define the distance, d co n t a ct , as the distance below which the
two bodies are touching each other. For a distance smaller than d co n t ac t , the first flux type is used. The second flux type
is used when the distance is between d c on t a c t and d n ea r , where d n ea r is called the near contact distance. This distance
represents a small gap between the contact bodies, and should not be chosen larger than the smallest element size. If this
distance is not set, this type of flux is disregarded. The third flux type is used when the distance is larger then d n ea r .
T1
d
T2
Figure 6-31 Contact Distance
If d  d contact , the bodies touch and conduction takes place between the two contacting bodies. The heat flux is written
as:
q = H TC  T 2 – T 1 
CHAPTER 6 541
Contact
where,
T1 is the surface temperature,

HT C is the film coefficient between the two surfaces,
T2 is the temperature of the same contact location, as obtained from interpolation of nodal
temperatures of the body being contacted.
If d c ontact  d  d near , a small gap between the bodies exists. The heat flux that exists through this gap can be
decomposed in the following physical processes: convection, natural convection, radiation, and a distance dependent term.
This flux is given by:
B NC B B 4 4
q = H CV  T 2 – T 1  + H NC  T 2 – T 1  + H NL  T 2 N L – T 1 N L  + f  T 2 – T 1 
 d d 
+  H TC  1 – ------------- + H BL -------------  T 2 – T 1 
 d  d
 near n e ar 
where
HC V is the convection coefficient for near field behavior.

H NC is the natural convection coefficient for near field behavior.
BC N is the exponent associated with natural convection.
H NL is the nonlinear convection coefficient.
BN C is the exponent associated with the natural convection coefficient for near field behavior.
, , and f are the coefficients for radiation, the Stefan-Boltzman coefficient, the emissivity, and the viewfactor,
respectively. The Stefan-Boltzman constant is given via the PARAMETERS option.
HB L is the separation distance dependent heat transfer coefficient.
T A2 T 2 + T ABS REF is the temperature in absolute scale. The offset between the user temperature and the
absolute temperature is given in the PARAMETERS option.
If d  d near , or d near has not been defined and d  d co n t a ct , the bodies are not touching and a heat flux to the
environment exists in the following form,
4 4
q = H CTV E  T SINK – T 1  + f  T A2 – T A1 
where,
H C T VE is the heat transfer coefficient to the environment

T SI NK is the environment sink temperature
When a deformable-body contacts a rigid-body, the coefficients associated with the rigid-body are used. When two
deformable bodies are in contact, the average value of the film coefficient specified for the bodies is used. This only works
for the coefficients discussed in the first and third heat flux type. A better way to prescribe the coefficients is to use the
CONTACT TABLE option. With this option, all the previously discussed coefficients can be set for each interface separately. For
the second heat flux type, this is the only way to specify the coefficients. As with all other coupled problems, heat generated
by plastic deformation can be calculated and applied as a volumetric flux. The heat generated by friction is also calculated
and applied as a surface flux.
Three user subroutines are available to facilitate the creation of more sophisticated boundary flux definitions (such as
radiation and convections with variations in space, temperature, pressure, etc.). UHTCON allows you to specify a film
coefficient while the surface is in contact ( d  d c o nt a c t ), UHTNRC allows you to specify a film coefficient while the surface
is almost in contact ( d c on t a c t  d  d n e ar ), and UHTCOE allows you to specify a film coefficient while the surface is
free ( d  d near ).
Thermal Contact
A thermal contact analysis is performed when no structural pass is present, this results in a computationally more efficient
analysis. Possible analyses types are Heat Transfer, Coupled Thermal-Electrical, Magnetostatics and Electrostatics. For Heat
Transfer and Coupled Thermal-Electrical the analysis is comparable to a coupled analysis. The two types of rigid bodies
that are described in the coupled analysis are allowed. So one is a rigid which is considered to have a constant temperature
(heat source or heat sink), and the other is a finite element body with thermal properties. The three types of heat fluxes as
described in Coupled Analysis can also be used here.
The coefficients are prescribed in the same way as in a Coupled Analysis where it is advised to use the CONTACT TABLE option.
The three user subroutines UHTCON, UHTNRC, and UHTCOE as previously discussed in Coupled Analysis can also be used in a Heat
Transfer and Coupled Thermal-Electrical analysis.
Joule Heating
In a Joule heating analysis similar conditions as thermal contact apply regarding the flow of current between contacting
bodies.
If d  d contact , the bodies touch and electrical conduction takes place between the two contacting bodies. The current is
written as:
i = H EC  V 2 – V 1  or H EC  V b od y – V 1 
where,
V1 is the surface voltage,

HE C is the electrical contact conductivity,
V2 is the voltage of the same contact location, as obtained from interpolation of nodal voltages of the body being
contacted.
V B od y is the voltage of the rigid body being touched.
CHAPTER 6 543
Contact
If d co n tact  d  d near , a small gap between the bodies exists. The current that exists through this gap can be decomposed
into:
 d d 
i = H NE  V 2 – V 1  +  H EC  1 – ------------- + H BL E -------------  V 2 – V 1 
 d  d
 n e ar n ea r 
where,
H NE is the coefficient for near field behavior.

HB L E is the separation distance dependent conductivity coefficient.
If d  d near , or d near has not been defined and d  d c on t a ct , the bodies are not touching and a current leakage to the
environment exists in the following form,
q = H VE  V SINK – V 1 
where,
H VE is the coefficient to the environment

V SI NK is the environment sink voltage
Electrical contact between bodies does contribute a term to the heat generation.
Note that near electrical contact for Joule heating does not represent electrical arcing between bodies, which is a much more
complex phenomena.
In a Joule heating analysis, the THERMAL CONTACT model definition option is also used to identify elements which are in a
conducting body to perform a resistance calculation using EMRESIS history definition option.
Three user subroutines are available to facilitate the creation of more sophisticated boundary definitions. UVTCON allows you
to specify a film coefficient while the surface is in contact ( d  d co n t a ct ), UVTNRC allows you to specify a film coefficient
while the surface is almost in contact ( d c o nt a c t  d  d n ea r ), and UVTCOE allows you to specify a film coefficient while the
surface is free ( d  d near ).
Contact in Electrostatic or Piezoelectric Analysis

In an electrostatic analysis, if two bodies are in contact, the scalar potential is tied between the bodies composed of elements.
For an electrostatic analysis the THERMAL CONTACT option is used. When a node is in contact with a symmetry surface, it will
be treated as a free surface, that is no current transmitted through the symmetry surface. The THERMAL CONTACT option is
also used to identify the elements which are in a conducting body to perform a capacitance calculation using EMCAPAC
history definition option. In such cases, the conducting bodies should not be in contact.
Contact in Magnetostatic Analysis

For 3-D magnetostatics, by default the tangential components of the magnetic vector potential are tied for nodes in contact.
When the glue flag is set on the CONTACT TABLE input all the components of the magnetic vector potential are tied. For 2-
D magnetostatics the one non-zero component of the magnetic vector potential gets tied in contact irrespective of the value
of the glue flag.When a node is in contact with a symmetry surface, a constraint is imposed such that only the normal
component of the potential is zero. For a magnetostatic analysis, the THERMAL CONTACT option is used.
Contact in Magnetodynamic Analysis

In magnetodynamic analysis, if two bodies are in contact, both the magnetic vector potential and the scalar electric
potential is tied between the bodies composed of elements. All components are tied. This is independent of the contact
condition; so, for both glue type and touching type conditions, all components are tied. For magnetodynamic analysis, the
THERMAL CONTACT option is used.
Contact in Soil Analysis

When two deformable bodies are in contact, the displacements are treated in the conventional manner. The pressure degree
of freedom associated with the fluid hydrostatic pressure is tied when the glue flag is set on the CONTACT TABLE input. If a
node contacts a rigid body, nothing is done with the pressure degree of freedom.
Contact in Acoustic Analysis

In an acoustic analysis with flexible and/or rigid walls composed of finite elements and/or rigid bodies, a coupling interface
between the acoustic medium and the walls is created to connect the fluid pressure to the wall movement taking into
account accelerations and reactive boundary conditions.
Element Considerations
Marc allows contact with almost all of the available elements, but the use of certain elements has a consequence on the
analysis procedure. Contact analysis can be performed with all of the structural continuum elements, either lower order or
higher order, including those of the Herrmann (incompressible) formulation, except axisymmetric elements with twist.
Friction modeling is available in all of these elements except the semi-infinite elements. Higher-order isoparametric
elements use shape functions which, when the elements are loaded by a (for example) uniform pressure, lead to equivalent
nodal loads that oscillate between the corner and midside nodes. This has a detrimental effect on determining contact
separation and two procedures have been implemented to eliminate this problem:
1. Linearized Contact: The midside nodes on the exterior surface are automatically tied to the corner nodes in this case.
This effectively results in a linear variation of both the geometry and the displacement on the exterior element edges.
Hence, the elements with edges on the exterior do not behave as full bi-quadratic (2-D) or tri-quadratic (3-D)
elements.
All elements in the interior of the body behave in the conventional higher-order manner, but the constraints on the
exterior easily cause the behavior of the complete structure to be too stiff; while in the area of contact, the stress
distribution might be irregular.
CHAPTER 6 545
Contact
2. True Quadratic Contact: No special constraints introduced on the exterior surface other than coming from contact
in this case. Both the midside and the corner nodes may come into contact and when contact is established with
another deformable body consisting of quadratic elements, a constraint equation corresponding to the complete
quadratic shape function is automatically incorporated. Since the above mentioned oscillating nodal loads cannot
be used for separation, the decision whether or not a node should separate is based on the contact normal stress
rather than the contact normal force.
In many manufacturing and rubber analyses, the lower-order elements behave better than the higher-order elements
because of their ability to represent the large distortion; hence, these lower-order elements are recommended.
The constraints imposed on the nodal degrees of freedom are dependent upon the type of element.
1. When a node of a continuum element comes into contact, the translational degrees of freedom are constrained.
2. When a node of a shell element comes into contact, the translational degrees of freedom are constrained and no
constraint is places on the rotational degrees of freedom. The exception to this is when a shell contacts a symmetry
surface. In this case, the rotation about the element edge is also constrained.
Additionally, beams and shells contact is governed by the rules outlined below.
The node-to-segment procedure has limited capabilities to simulate beam contact. Contact of beams with non-beams is
established at the beam nodes without any allowance made for the beam dimensions. For simulating beam-beam contact,
an equivalent beam radius has to be defined by the user, and multi-point constraints across the beam segments are
established based on a proximity check that includes the beam radii.
2-D Beams
All nodes on beams are potential contact nodes. Beam elements can be used in contact in two modes.
1. The two-dimensional beams can come into contact with rigid bodies composed of curves in the same x-y plane.
The normal is based upon the normal of the rigid surface.
2. The two-dimensional beams can come into contact with deformable bodies either of continuum elements or other
beam elements. As the beams are in two dimensions, they do not intersect one another.
3-D Beams
Three-dimensional beam elements can be used in contact in three methods.
1. The nodes of the beams can come into contact with rigid bodies composed of surfaces. The normal is based upon
the normal of the rigid surface.
2. The nodes of the beam and truss elements can also come into contact with the faces of three-dimensional
continuum elements or shell elements. The normal is based upon the normal of the element face.
3. The three-dimensional beam and truss elements can also come into contact with other beam or truss elements
(beam-to-beam contact).
The first two methods are activated by default if contact bodies consisting of beam or truss elements are defined using the
CONTACT option. The third method must be activated explicitly by additionally switching on the beam-to-beam contact flag
on the CONTACT option.
In the beam-to-beam contact model, a beam or truss element is viewed as a conical surface with a circular cross-section.
The radius of the cross-section can vary linearly between the start and end node of a beam element. For each beam or truss
element of a contact body, a contact radius must be entered via the GEOMETRY option. The contact radius at a node follows
from the average contact radius of the elements sharing that node. Hence, the start and end node of an element may have
different contact radii.
Contact is detected between two beam or truss elements if the associated conical surfaces touch each other; that is, if the
distance d between the closest points on the conical surfaces is smaller than the distance below which bodies are considered
touching each other. This is outlined in Figure 6-32, where beam elements and their contact body representation are given.
It should be emphasized that the contacting points are points on the conical surfaces and not nodes of the finite element
model.
X Z
Figure 6-32 Beam-to-Beam Contact: Finite Element Model (top) and Contact Body Representation (bottom)
If two beam or truss elements are in contact, a multipoint constraint equation (tying) is automatically imposed to ensure
that the conical surfaces will not interpenetrate. This constraint equation involves the displacements of the begin and end
nodes of both elements. The tied node in that equation is automatically selected, taking into account the location of the
contacting points with respect to the elements, any boundary conditions applied to the nodes and any contact between the
nodes and rigid surfaces or faces of continuum or shell elements.
During the iteration procedure, the contacting points of two beam or truss elements can change if the elements slide with
respect to each other. In addition, the points in contact can move from one element to another. In that case, the nodes
involved in the multipoint constraint equations are automatically updated. During sliding, friction may be taken into
account. Since for beam or truss elements the normal stress in the contact points is not available, only Coulomb friction
based on nodal forces (either the arctangent or the bilinear model) is supported for beam-to-beam contact. The glue model,
which imposes that there is no relative tangential motion, is also available.
A limitation of the beam-to-beam contact model is that a contact body cannot contain branches, that is, every element in
the contact body must have a unique successor and predecessor.
CHAPTER 6 547
Contact
Shell Elements
All nodes on shell elements are potential contact nodes. As the midside nodes of shell elements are automatically tied, the
high-order shell element (type 22) has no benefit. Shell elements can contact either rigid bodies, continuum elements, or
other shell elements. Shell-shell contact involves a more complex analysis because it is necessary to determine which side
of the shell contact occurs.
By default, the top and bottom face of a shell element (which follow from the mid face by using the thickness offset vector)
are separately taken into account. This implies that a node may be found to be in contact based on the top or the bottom
face of an element. However, due to the nature of the shell formulation, it is not possible to simultaneously use contact
conditions following from the top and the bottom face. Touching a shell element may also occur at its top or bottom face.
The user has the option to set per pair of contact bodies on the CONTACT TABLE option how the geometry of the shell
elements should be handled. This means that both for the contacting and the contacted body it can be defined that the
contact description will be based on one of the following options: both top and bottom (default), top only (with or without
the thickness offset vector) or bottom only (with or without the thickness update vector). For glued contact it is possible
to use the top and bottom face, but without the thickness update vector.
The beam-to-beam contact model discussed above can also be used to model shell edge-to-edge contact. This requires
switching on the beam-to-beam contact flag on the CONTACT option and, for a shell contact body combination, the edge-
to-edge contact flag on the CONTACT TABLE option. For edge-to-edge contact, half of the thickness of the shell is used to set
the contact radius.
Dynamic Impact
The capability is available for both implicit time integration via the Newmark-beta or Houbolt operators, and the explicit
time integration via the (fast) central difference operator.
The Newmark-beta, Single Step Houbolt and Generalized Alpha procedure have the capability to allow variable time steps
and, when using the user-defined fixed time step procedure, the time step is split by the algorithm to satisfy the contact
conditions. Since the central difference operator is constrained to small time steps by the stability requirement, time
splitting is not used by the contact algorithm.
For most dynamic impact problems, the Single Step Houbolt method is recommended, as this procedure possesses high-
frequency dissipation. This is often necessary to avoid numerical problems by contact-induced high-frequency oscillations.
If the other dynamic operators are used, it is recommended that numerical damping is used during the analysis, or, when
using the Generalized Alpha method, that a small spectral radius is defined. Note that for a zero spectral radius, the
1
Generalized Alpha method is identical to the Single Step Houbolt method with default parameters (  = 1.5 ,
 = – 0.5 ).
In dynamic analysis, the requirement of energy conservation is supplemented with the requirement of momentum
conservation. In addition to the constraints placed upon the displacements, additional constraints are placed on the velocity
and acceleration of the nodal points in contact, except for the Single Step Houbolt and Generalized Alpha method. As
mentioned in the paragraph Implementation of Constraints, for these time integration methods, small gaps or overlaps in the
contact area are not instantaneously closed, as this may introduce undesired inertia effects. In particular for contact
problems involving large sliding along curved contact surfaces, not correcting at all for such gaps or overlaps can result in
growing gaps or overlaps. For this reason, the surface projection factor (defined via the PARAMETERS option) has been
introduced, representing the fraction of the gap or overlap to be corrected per iteration. If its value is 0, no correction is
applied. If its value is 1, the full correction is applied, i.e. the gap or overlap is closed instantaneously.
When a node contacts a rigid surface, it is given the velocity and acceleration of the rigid surface in the normal direction.
The rigid surfaces are treated as if they have infinite mass, hence, infinite momentum.
Global Remeshing and Rezoning

In manufacturing simulations the objective is to deform the workpiece from some initial (simple) shape to a final, often
complex shape. This deformation of the material, results in mesh distortion. For this reason, it is often required to perform
a rezoning/remeshing step. At this point, a new mesh is created; the current state of deformation, strains, stresses, etc. is
transferred to the new mesh, and the analysis is continued. This has several consequences for contact analysis.
1. When manual rezoning is used, it is necessary to redefine the elements associated with the deformable body that is
being rezoned. This is done through the CONTACT CHANGE option.
2. If a contact tolerance was not defined by you, a new contact tolerance is calculated based upon the new mesh. This
distance can be smaller than the previously calculated tolerance, leading to more iterations.
3. After the new mesh is created, the program redetermines the potential contact nodes. The memory is automatically
allocated to accommodate the contact data in the new mesh.
4. At the first increment after remeshing, Marc determines which nodes are in contact. If the new mesh does not reflect
the exterior surface of the old model, it is possible that a region that was previously in contact is no longer in contact.
This is usually not so serious; as most likely, the node subsequently comes into contact. A more significant problem
is if a node in the new mesh is created which, in fact, has penetrated another body. If the newly created node is
beyond another body’s surface by more than the contact tolerance, it is possible that a poor contact analysis results.
With automatic global remeshing, steps are taken to made sure contact conditions are preserved and incorrect
penetrations are removed in the new mesh. This is achieved by correcting the nodal position after a new mesh is
created.
If rezoning occurs, rigid surfaces must be kept the same and cannot be changed.
Marc auto-rezoning feature, activated by the parameter REZONING,1 for 2-D problems and REZONING,2 for 3-D problems,
automatically creates new meshes and performs rezoning to resolve mesh distortion problems. In such cases, points (1) and
(4) are treated automatically.
Local Adaptivity
Contact between a deformable-body and a rigid body is insured such that the nodes do not penetrate the rigid surface. It
is possible that an edge of an element penetrates a rigid surface, especially where high curvature is present because of the
finite element discretization. The use of the adaptive mesh generation procedure can be used to reduce these problems.
In addition to the traditional error criteria such as Zienkiewicz-Zhu or maximum stress, etc., you can request that the mesh
be adaptively refined due to contact. In such a case, when a node comes into contact, the elements associated with that
node are refined. This results in a greater number of elements and nodes on the exterior region where contact occurs. This
can lead to a substantial improvement in the accuracy of the solution.
CHAPTER 6 549
Contact
Penetration of Element Edge due to Mesh Discretization Result of Adaptive Meshing

Figure 6-33 Contact Closure Condition
The adaptive meshing procedure has consequences to the contact procedure similar to the rezoning procedure.
If a contact tolerance was not defined by you, a new contact tolerance is calculated based upon the new mesh. This distance
can be smaller than the previously calculated tolerance, leading to more iterations.
Results Evaluation
For Node-to-Segment contact, the Marc post file contains contact results for both deformable bodies and rigid bodies.
There are three types of result categories possible for a contact analysis:
Contact Status and Touched Body

For deformable bodies, the status is available as a nodal output. The convention is as follows:
 A value of 0.0 means that the node is not in contact.
 A value of 0.5 means the node is in near thermal contact.
 A value of 1.0 means that the node is in regular contact.
 A value of 1.5 means that the node is in glued contact.
 A value of 2.0 means that the node is on a cyclic symmetry boundary.
Note that contact status is only seen at the nodes of the touching body.
The Contact Touched Body is also available as a nodal output. For Node-to-Segment contact, up to 3 touched bodies can
be output (2 in 2D) and this output is only available for the touching nodes.
Contact Forces and Stresses

Three types of response output are possible:
a. Normal Force / Friction Force / Normal Stress / Friction Stress
These vectorial quantities are available as nodal quantities for touching and touched segments of deformable
bodies. Contact normal force and friction force are included as Default nodal output – contact normal stress
and friction stress need to be selected via Custom output if desired. Note that the contact normal and friction
stresses are computed by two different methods depending on the element type and the method used for
computing separation stress:
Extrapolation: Stresses are extrapolated from the integration points of the surface elements to the associated
nodes. They are then transformed into normal and tangential components based on the average normals at the
nodes. This method is used by default for contact involving quadratic continuum elements. For contact
involving linear continuum elements, this method is used only when the Separation criterion is changed from
Force to Stress and the associated Derivation method for computing the Stress is set to Extrapolation. Note that
this method is not available for shells. For accurate stress results, this is the recommended method where
possible.
Force / Area: The contact normal force or friction force at a contact node is divided by the area assigned to the
node to compute the corresponding contact normal stress or friction stress. This method is used for linear
continuum elements and shells by default.
b. Resultant Body Forces and Moments
The resultant contact force and moment vectors are available for each deformable body and rigid body. The
time history of these resultant quantities are of significant interest in many engineering analyses. The moments
are computed about the user-defined centroid for rigid bodies (default = 0,0,0) and about the origin for
deformable bodies.
c. Resultant Body Force and Moments per Body Pair
The interaction forces and moments for each body pair can also be output. The time history of the forces and
moments can be plotted. In addition, by associating a node with a specific body pair, the total normal and
tangential interaction force / moment quantities can be visualized as vectors.
For the resultant quantities in categories (b) and (c), it should be noted that the contact interaction output do not include
self-equilibrating forces between the bodies. The computed interaction quantities only refer to the net contact force and
moment vectors that satisfy free-body diagrams for each body. Note that if there are no resultant forces found for a
deformable body or a rigid body, it implies that either the body is not in contact with any other body or all forces on the
body are self-equilibrating. For example, a cylinder press-fit that introduces uniform radial forces around the circumference
would indicate a net interaction force of zero since all the contact forces are self-equilibrating.
CHAPTER 6 551
Contact
Miscellaneous Contact Response Quantities

These includes the following:
a. Contact Stress-Free Projection Vector
This nodal vector is available by default for Version 15 style input if stress-free projection is activated for at least
one body pair via the Contact Interaction option.
b. Near Contact Distance
This nodal quantity is only available for Node-to-Segment contact. The actual distance of a node that is within
the near distance tolerance to a touched segment is output as a scalar quantity. This value is typically available
for analyses involving a heat transfer pass with Near Contact activated via the Contact Interaction option.
c. Contact Interaction Heat Flux
For thermal analyses, the heat flux transmitted between the bodies can be output as a nodal scalar quantity. By
associating a node with a specific body pair through the Contact Table, the interaction heat flux can be
visualized on the post file.
Contact Additional Print Output

Additional print about a contact analysis can be obtained in the output file through the PRINT parameter. Information
on when a node comes into contact, what is the touched segment, when separation occurs, when a node slides from one
segment to another, etc. can be printed out through the PRINT,5 option. Incremental displacements in the local coordinate
system can be printed out through the PRINT,8 option. Note that, for large problems, this can result in a significant
amount of output.
Tolerance Values
On the third data block for contact, five tolerances can be set for determination of the contact behavior. Not entering any
values here means that Marc calculates values based on the problem specification.
Relative Sliding Velocity Between Surfaces Below Which Friction Forces Drop
As discussed in the Friction Modeling section, the equations of friction are smoothed internally in the program to avoid
numerical instabilities. The equations are inequalities whenever two contacting surfaces stick to each other and
equalities whenever the surfaces slide (or slip). Thus, the character of contact constraints change depending on whether
there is sticking or slipping. The smoothing procedure consist of modifying it in such a way, that there is always slip; the
amount is a function of the relative velocity and a constant RVCNST. The value of this constant must be specified here. It
actually means, that if we specify a small value in comparison to the relative velocity, the jump behavior is better
approximated, but numerical instabilities can be expected. A large value means, that we need a large relative velocity before
we get the force at which the slip occurs.
It is suggested to use values between 0.1 and 0.01 times a typical surface velocity.
Distance Below Which a Node is Considered Touching a Surface

In each step, it is checked whether a (new) node is in contact with other surfaces. This is determined by the distance
between the nodes and the surfaces. Since the distance is a calculated number, there are always roundoff errors involved.
Therefore, a contact tolerance is provided such that if the distance calculated is below this tolerance, a node is considered
in contact. A too large value means that a high number of body nodes are considered to be in contact with the surface and
are consequently all moved to the surface, which can be unrealistic in some applications. A too small value of this number
means that the applied deformation increment is split into a high number of increments, thus increasing the cost
of computation.
The tolerance must be provided by the analyst or can be calculated by Marc. The analyst can provide a global contact
distance tolerance (see CONTACT (2-D) option) or an individual value for each body pair (see CON INTERA option). In general,
the contact tolerance should be a small number compared to the geometrical features of the configuration being analysed.
At the global level, if the user has not specified the contact distance tolerance, Marc calculates it by using the dimensions
of the contact bodies. A couple of options are provided for the user to control this computation:
 Global tolerance across all contact bodies: This is the default. The value calculated by Marc is determined as
1/20th of the smallest element size of all elements in any contact body OR 1/4th of the smallest thickness of shell
or beam elements in any contact body, whichever is the smallest. Note that for a continuum element, the element
size is defined here as the smallest edge of the surrounding rectangle (2D) or box (3D) set up in the global
coordinate system, while for a shell element, it is the smallest element edge and for a beam element, it is the
element length.
 Tolerance for each body pair: The same computations are done as mentioned above (1/20th of the smallest
continuum edge, 1/4th of the smallest shell thickness), but it is separately done for each body pair. This option is
useful when there is a wide variation in mesh densities or thicknesses across contact bodies.
In addition, the user can optionally provide a scale factor for the Marc computed default tolerance. The scale factor defaults
to 1.0 and any value above 0.0 can be specified by the user. The scale factor is uniformly applied to the common tolerance
computed for all contact bodies OR to the tolerance computed for each body pair. Note that no scale factor is applied when
the user specifies the contact distance tolerance.
For global remeshing problems, the distance tolerance assessed by Marc is automatically recomputed due to the mesh size
change. Note that this recomputation is done for either a common tolerance across all bodies or for an individual tolerance
for each body pair. If the user has prescribed a contact tolerance, there are two choices:
 Allow Marc to scale the user-provided value based on the new mesh size (default behaviour).
 Indicate that the original user-provided value should be fixed (i.e., even after remeshing, the user value is retained).
Tolerance on Nodal Reaction Force or Nodal Stress Before Separation Occurs

If a tensile force occurs at a node which is in contact with a surface, the node should separate from the surface. Rather than
using any positive value, a threshold value can be specified. This number should theoretically be zero. However, because a
small positive reaction might be due only to errors in equilibrium, this threshold value avoids unnecessary separations. A
too small value of this force results in alternating separation and contact between the node and the surface. A too large
value, of course, results in unrealistic contact behavior.
Marc calculates this value as the maximum residual force in the structure. Note that the maximum residual force can be
specified in the CONTROL option. Default for this value, a 10 percent of the maximum reaction force is used. Consequently,
if locally high reaction forces at a particular point are present, the separation force is large as well. In most cases, however,
the default value is a good measure.
If you indicate that separation is to be based upon stresses, a value of the separation stress should be entered. The default
value is the maximum residual force at a node divided by the contact area of node n.
CHAPTER 6 553
Contact
Numerical Aspects of the Direct Constraint Procedure

Marc divides contact problems into two domains; the first is when a deformable body makes contact with a rigid surface
and the second is when a deformable body makes contact with another deformable body or itself.
Two-dimensional or three-dimensional bodies are treated conceptually in the same manner.
Deformable-Rigid Contact
In such a problem, a target node on the deformable body has no constraint while contact does not occur. Once contact is
detected, the degrees of freedom are transformed to a local system and a constraint is imposed such that:
u no r mal = v  n
where v is the prescribed velocity of the rigid surface. This local transformation is continuously updated to reflect sliding
of point A along the rigid surface.
If the glue option is activated, an additional displacement constraint is formed as:
u tangential = v  t
The determination of when contact occurs and the calculation of the normal vector are critical to the numerical simulation.
Historically in Marc, three procedures are used for detecting penetration and imposing contact, but now two of them are
preferred and are discussed here. The first (and most general one) is the Iterative Penetration Checking Procedure. The
second one (Time Step Reduction Procedure) is recommended only for a transient dynamic analysis with automatic time
stepping.
In Figure 6-35, node A is not in contact with the rigid body at the beginning of an iteration. After time step t , node A
penetrates rigid surface if no constraint is imposed during the iteration which is not an acceptable solution.
B B
A
A
t y
x
n
Before Contact After Contact
z
Figure 6-34 Contact Coordinate System
A A
Start of Iteration, No Contact End of Iteration, If No Constraint

Figure 6-35 Contact Constraint Requirement
The Iterative Penetration Checking Procedure checks for every iteration in the Newton-Raphson process if the
displacement solution found would cause nodes to penetrate. If this is the case, the iterative displacement solution is scaled
such that at the start of the subsequent iteration, new contact will be established. Some mathematical details can be found
at the end of this chapter.
When in a transient dynamic analysis, the adaptive time stepping procedure AUTO STEP is used, Marc reduces the time step
of the current increment such that nodes, which would penetrate based on the original time step, will just come into
contact. In the current increment, no contact constraints are applied to these nodes. In the next increment, the nodes are
given the appropriate contact constraints and a new time step is determined based upon the convergence criteria.
Analytical Contact
Utilizing the analytical representation of rigid surfaces influences the numerical procedure in several ways:
1. The rigid surface is represented by a nonuniform rational B-spline surface (NURBS) which provides a precise
mathematical form capable of representing the common analytical shapes – circle, conic curve, free-form curves,
surfaces of revolution, and sculptured surfaces.
The surface can be expressed in simple mathematical form to model complex multiple surfaces with advantage of
continuity in first and second derivative. Those analytical forms are expressed in four-dimensional homogeneous
coordinate space by:
n+1 m+1
  B i j h i j N i k  u M j l  v 
i = 1 j = 1
P  u v  = -----------------------------------------------------------------------------------------
n+1 m+1
  h i j N i k  u M j l  v 
i = 1 j = 1
CHAPTER 6 555
Contact
For curves, it simplifies to:

n+1
 B i h i N i k  u 
i = 1
P  u  = -----------------------------------------
-
n+1
 h i N i k  u 
i = 1
where the B are 4-D homogeneous defining polygon vertices, N i k and M j l are nonrational B-spline basis
functions, hi,j is homogeneous coordinates. For given parameters u and v in local system, the location (x, y, z) in
three-dimensional space, first derivative and second derivative (if required), are calculated. Given a point with x, y
and z Cartesian coordinates, there is no explicit mathematical form to find out the parameters u and v in the local
space, so they are obtained using an iterative procedure.
2. The location of the closest point on the contact surface corresponding to node A needs to be determined. This
process, determined analytically when using flat patches, requires an iterative approach for rigid bodies modeled
with NURBS.
A
Figure 6-36 Closest Point Projection Algorithm
3. Once the point P is known, the normal is calculated based upon the NURBS. This normal is used in a manner
consistent with the PWL procedure.
4. Because the normal can change from iteration to iteration, the imposition of the kinematic boundary conditions is
continuously changing.
In the PWL approach:
0
 u n  = vn
i
 u n  = 0
In the analytical approach:

0
 u n  = v  n0
i
 u n  = v  n 1 – v  n 0
where the subscripts represent the iteration number.

5. In the PWL approach, when a node slides from one segment to another, the corner condition logic is activated.
The advantages of analytical surfaces is that this logic is not necessary until you come to the end of the spline. This
reduces the amount of iterations required.
6. The friction calculation is dependent upon the surface normal and tangent. When using the analytical approach,
the friction calculation includes the effect of changes in the direction of the normal vector from iteration to
iteration. This improves the accuracy and convergence behavior.
7. Because the normal is continuous, the calculation of nodal forces is more accurate. This allows for the
determination of nodal separation to occur beginning with the second iteration as opposed to when equilibrium
convergence has been achieved using the PWL approach.
Solution Strategy for Rigid Contact

Step 1: At the start of an increment, all boundary nodes are checked for contact with surfaces and flagged if so. If the
Iterative Penetration Checking Procedure has detected potential penetration, this is repeated at the start of an
iteration.
Step 2: Next, transformations and imposed displacements are determined for each touching node. With the knowledge
of the time increment and surface velocity, the configuration of the surface at the end of the increment is found.
Then, the distance from the node starting position to the current surface segment is determined and is the
probable normal displacement to be imposed. If there is a previous solution for displacement increments, these
are used to estimate the tangential displacement increment (Figure 6-37). If this estimate still puts the node at the
end of the increment in the same surface segment, then this one is used to determine the transformation matrix
between local and global coordinates, and the normal displacement is accepted. Otherwise, the procedure is
repeated for the adjacent segment. If a concave corner is hit, the boundary node is fixed to such corner. If the
node goes out of a sharp convex corner, it is immediately released from the surface.
SEGB
P
u
SEGB- Die segment at beginning of increment

un SEGE- Die segment at end of increment
u - Displacement increment of previous increment
SEGE un - Imposed normal displacement
utest
utest- Estimate of tangential displacement

Figure 6-37 Node P Already In Contact (2-D)
Step 3: One increment of the problem is solved iteratively. During each iteration, the Iterative Penetration Checking
Procedure checks if the current iterative displacement solution must be scaled to avoid penetration (Figure 6-38).
If needed, Step 1 is repeated to establish new contact constraints. Step 2 is repeated at each iteration for possible
changes in surface segment.
CHAPTER 6 557
Contact
SEGB
A SEGE
B SEGB- Die segment at beginning of iteration
SEGE- Die segment at end of iteration
u - Iterative Displacement calculated
u
PB
P -------- - Factor to scale iteration
PA
Figure 6-38 New Node in Contact P
Step 4: Once convergence is obtained, nodal forces are checked. Whenever a positive normal force (or stress) is detected
which is larger than the tolerance, a node is released from the surface segment at which such force (or stress)
corresponds and Step 3 is repeated. This tolerance can be modified using the SEPFOR or SEPSTR user subroutine.
Step 5: Go to next increment.
PATB
u
un
utest
PATE
3
1
2
PATB- Patch at beginning of increment

PATE- Patch at end of increment
u - Previous displacement increment of Node P
un - Imposed normal displacement increment (local direction 3)
Figure 6-39 Node P Already in Contact (3-D)
Deformable-Deformable Contact
When a node contacts a deformable body, a tying relation is formed between the contacting nodes and the nodes of the
contacted segment on the other body. This constraint relationship uses information regarding the normal vector to the
segment and the closest point projection of the contacting node on the contacted segment. For lower-order elements, the
normal vector and the projection are calculated based upon the piecewise linear representation of the element. This has the
consequence that the constraint relation may not be accurate because the normal vector is constant over the complete
segment, whereas the actual structure may be curved with a varying normal vector. When a node slides from one segment
to another, there may be a discontinuity of the normal vector, which leads to potential numerical difficulties or an
inaccurate solution. This is especially important for contact analyses where the deformations are small.
When higher-order elements are used, the normal vector and projection are created based upon the true quadratic element
description. Provided that the midside nodes are defined to be on the actual curved geometry, this improves the accuracy,
but does not necessarily result in continuity of the normal vector at nodes shared by multiple segments.
Both for lower- and higher-order elements, you can fit a smooth curve (2-D) or surface (3-D) through the finite element
segments. For 2-D, a cubic spline is used (see Figure 6-40). This spline is calculated based upon the tangent and position
vectors of the nodes on the segment. For lower-order elements, each segment is internally replaced by one cubic spline;
where for higher-order elements, each segment is replaced by two cubic splines. This gives a more accurate representation
of the actual physical geometry and a more accurate calculation of the normal vector. You must identify where actual
discontinuities exist. This is done by defining nodes where the normal vector is discontinuous or by activating the option
that the program will determine such nodes based on the angle spanned by the normal vectors to adjacent segments. You
can use the SPLINE option to define deformable bodies in 2-D or 3-D where the smoothness is required.
Actual Geometry Finite Element Approximation Cubic Spline Representation
: Nodes with a normal vector discontinuity
Figure 6-40 2-D Deformable Contact; Piece wise Linear FE Description and Cubic Spline Representation
For 3-D problems, the concept of a cubic spline can be expanded to generate a Coons surface to improve the geometric
representation of a segment (see Figure 6-42). When calculating the closest point projection, it is advantageous to replace
the Coons surface description by a cubic serendipity surface. For 2-D problems, you have to define nodes where the actual
normal vector has a discontinuity; for 3-D problems, you have to specify edges where the normal vector is discontinuous.
The SPLINE option allows you to enter a list of nodes (where each edge is defined by its start and end node) or to activate
the option that the program will determine such edges based on the angle spanned by the normal vectors to adjacent
segments. At edges with a discontinuous normal vector, the smooth surface description of adjacent segments can be set up
independently of each other, which usually results in C0-discontinuity. Alternatively, the smooth surface description of
adjacent segments can be forced to include information from the interface they have in common, which results in C0-
continuity. You can force C0-continuity in 3-D also on the SPLINE option.
If the SPLINE option is used for curved contact bodies defined by higher-order elements, the coordinates of the midside
nodes can be recalculated based on the cubic splines defined by the position and tangent vectors of the corner nodes. This
can be advantageous if the structure to be modeled is available only based on lower-order elements, so that a change from
CHAPTER 6 559
Contact
lower- to higher-order elements yields midside nodes which are not on the curved boundary, but on the straight edges
between the corner nodes.
C0-discontinuous Coons
Surface Representation
Actual Geometry Finite Element Approximation
C0-continuous Coons
Surface Representation
Figure 6-41 3-D Deformable Contact; Piece wise Linear FE Description and Coons Surface Representations
Solution Strategy for Deformable Contact

1. Upon initiation, the program determines the boundary nodes on each body. For 2-D, these nodes are ordered in
sequence in order to describe the boundary in a counterclockwise manner. Linear geometrical entities are then
created out of subsequent pairs of nodes, to define a surface profile as in a rigid surface. If higher-order elements
are used in full quadratic mode, quadratic entities are created from the three nodes. For 3-D using lower-order
elements, 4-node patches are created. For 3-D higher-order hexahedral elements, a quadratic surface is created.
2. Once contact between a node and a deformable surface is detected, a tie is activated. The tying matrix is such that
the contacting node follows the shape of the surface (Figures 6-42 and 6-43); it can slide along it or be stuck
according to the general contact conditions.
3. Contact occurs between all deformable bodies unless the CONTACT TABLE option is used. In deformable contact, there
is no master or slave body; each body is checked against every other body, except single-sided contact or the CONTACT
TABLE option is used.
4. During the iteration process of finding an incremental solution, a node can slide from one segment to another,
changing the retained nodes of its tie. A recalculation of the bandwidth is, therefore, made every iteration. From
increment (or subincrement) to increment, the number of nodes in contact can change. In such cases, an
optimization of the bandwidth is automatically available to cope with the possible drastic changes in bandwidth
that a new tie produces.
When the glue option is used between two deformable bodies, a simpler tying relationship is formed, such that no
relative motion occurs.
CHAPTER 6 561
Contact
Tied NodeA
A
Retained NodesA B C
segt
B
UB UA UC
segt+t
Figure 6-42 Tyings in Deformable Contact (2-D) using Lower-order Elements
R2
R3 T/R1
R5
t Tied NodeT
Retained NodesT, R2, R3,R4,R5
R4
t+ t
Figure 6-43 Tying in Deformable Contact (3-D) using Lower-order Elements
Iterative Penetration Checking

Except for a transient dynamic analysis with automatic time stepping where the Time Step Reduction Procedure is used,
the commonly used method to prevent penetration in a contact analysis is the Iterative Penetration Checking Procedure.
Using this procedure, the iteration process is done simultaneously to satisfy both the contact constraints and global
equilibrium using the Newton-Raphson procedure. This procedure is accurate and stable. This procedure is automatically
activated when the AUTO STEP procedure is used in static analysis or when beam-beam contact is present.
In a conventional iteration process, the finite element system calculates for each iteration:
K T u i = R i – 1
where K T is the tangent stiffness matrix and R i – 1 are the residuals based on displacements from the previous iteration.
Using the iterative penetration method K T is now based upon the contact status at this iteration, both from a constraint
and friction perspective. Furthermore, after the solution for u i is obtained, the contact procedure is used to determine if
new penetration will occur. If, at least, one node penetrates a contact surface, a scale factor is applied to the change in
displacements such that the penetrating nodes are moved back to the contact surface. This procedure can be considered a
type of line search. Given that s is the fraction of u i such that no new penetration occurs, the displacement increment
then becomes
U i = U i – 1 + su i
and the total displacement is
U n = U n – 1 + U i
The strains, stresses, and residuals are based upon these quantities. Separation is based upon these values. When global
equilibrium is achieved, based upon the user’s criteria, the solution proceeds to the next increment. Because the procedure
can reduce the change in displacements, it may require more iterations to complete an increment. It is important to
ensure that the maximum allowable number of iterations to complete an increment is set to a sufficiently large value.
Node to Segment Linear Contact

The Linear Contact feature is specifically designed and implemented for analysis involving very small relative displacements
and rotations. This includes engine block analysis, small strain inertia relief analysis or bolted assembly analysis with limited
relative motion between the contact bodies.
Contact is established only at the start of an analysis and no new nodes can come into contact later. Contact constrains are
always established in the initial configuration. Both the position of the node and of the touched patch are not updated
through the analysis and shell normal are based on the initial geometry. The touching nodes can separate and re-establish
contact anytime during the analysis. As the displacements are small and the parts are initially in contact, nodes are not
checked for sliding across segments or for penetration.
Linear contact is available with parallel processing and multi-threading capabilities of Marc. The various contact features
like friction modeling, glued contact are supported with linear contact.
Limitations: Any option that requires regeneration of the contact boundary is not supported for linear contact – this
includes Global Remeshing, Local Adaptive Meshing, Element Deactivation and Mesh Splitting.
Instabilities
In some analyses, because of instabilities such as buckling or a loss of contact, large displacement increments may occur. In
such cases, it is possible to specify a maximum iterative displacement component allowed u al l o w ed . If the magnitude of
any iterative displacement component turns out to be larger than this maximum value, the iterative displacement solution
is scaled using a factor s , such that:
s u i  u allow ed
for any node in the model. If u a l l ow e d is not specified, the value used defaults to the average of the maximum dimension
of the global rectangular boxes enveloping the contact segments. In case of fine meshes, this default value can be small and
scaling of the solution vector can be reduced by specifying a more physically meaningful length scale.
CHAPTER 6 563
Contact
Finally, if the solution is still not able to converge and the cutback feature is activated, the time step is reduced in magnitude
and the increment is repeated.
Segment-to-Segment Contact
The contact algorithm discussed above is based on nodes being in contact with a segment (a curve, surface, element edge, or
element face), and can thus be called a node-to-segment algorithm. This contact algorithm has matured and has been
successfully applied to a large variety of contact problems. However, there are a few weaknesses of this algorithm especially
for deformable contact, which call for the development of an alternative contact algorithm:
 When contact has been detected, the nonpenetration constraint is enforced on a nodal basis. Because of this point-
wise application of constraints, the node-to-segment algorithm does not generally maintain stress continuity across
the contact interface of deformable contact bodies. This can be easily illustrated using the contact patch test,
where a uniform stress distribution should be obtained in two deformable contact bodies with dissimilar meshes
(see [Ref. 7]).
 Since the nonpenetration constraints are enforced using multi-point constraint equations, there is a potential
dependency of the solution on the selection of the master and the slave nodes. In other words, the solution
depends on which nodes are touching and which nodes correspond to a touched contact segment. Although there
are various options to optimize the multi-point constraint equations, it is not always possible to achieve the best
results everywhere in the model and to completely eliminate the dependency on the contact body numbering.
 If there would be contact detected at the top and the bottom face of a shell element, it is not possible to only use
the nodes of the shell elements to apply both multi-point constraint equations. This implies that there are limited
modeling options for so-called double sided shell contact. A similar problem occurs when the edge of a shell
element is in contact with the face of another element. If this happens, then the contact constraints are based on
either the shell top or the shell bottom face. Consequently, the “footprint” of the shell edge is not directly related
to the shell thickness.
In the segment-to-segment algorithm the nonpenetration constraints are enforced using augmented Lagrangians. In this
section, the basic aspects of this algorithm will be discussed. It should be noted that currently the segment-to-segment
algorithm does not support:
 Anisotropic friction
 Brake squeal
 Pore pressure and fluid-solid analyses
 Wear
Basic Theory for Normal and Frictional Contact

For a set of deformable contact bodies, the virtual work equation commonly used as the basis for the derivation of the
standard set of equations to be solved within a nonlinear structural finite element analysis can symbolically be written as:
G  u u  = 0 (6-5)
where u is the displacement field and u the kinematically admissible variation of this field. Contact conditions can be
incorporated by adjusting this virtual work equation by two surface integrals:
T
G  u u  +   n g n d +   t g t d = 0 (6-6)
 
Here  is that part of boundary of the bodies being in contact and the subscript n is used to indicate the normal direction
to the contact boundary, where the subscript t indicates the tangential direction. The function g n is called a gap function
and expresses the distance between a point and its closest point projection on the contact boundary. If g n  0 , a point is
outside the contact boundary, if g n = 0 , a point is on the contact boundary and if g n  0 , a point would be beyond the
contact boundary, which is physically inadmissible. The Lagrange multiplier  n represents the contact normal stress.
Similarly, g t is the tangential gap vector and  t is the tangential stress vector.
The tangential or frictional behavior is assumed to be governed by Coulomb’s friction law, which, as already stated before
for node-to-segment contact, can be expressed as:
 =  t –  n  0 (6-7)
where  is the friction coefficient;   0 corresponds to sticking and  = 0 corresponds to slipping.
Since the gap functions g n and g t depend on the geometry of the bodies (defining the normal and tangential vectors) and
the displacement field u , the contact part of Equation (6-6) can be expressed as:
g n T g t
G c  u u  =   n -------
u
- u d +   t ------- u d
u
(6-8)
 
Since Equation (6-6) has to be valid for all kinematically admissible variations u , the contact contribution to the
equilibrium equation is given by:
T g n T g t
Fc =   n -------
u
- d +   t ------- d
u
(6-9)
 
The overall solution procedure will be based on an augmented Lagrangian approach. It is assumed that the solution at step
n – 1 is known and that the solution at step n has to be determined in an iterative fashion, using a Newton-Raphson
process. Based on the iterative displacement solution obtained at iteration i , the following trial solutions for the Lagrange
multipliers are introduced (note that the superscript n to indicate the step number is omitted to simplify the notation):
t ri a l i–1 i trial i–1 i
n = pn + En gn ; t = tt + Et gt (6-10)
in which p n and t t are fixed estimates of  n and  t , E n and E t are penalty factors for the contact behavior in the normal
and tangential directions (actually E t is a diagonal matrix with the penalty factor E t on the diagonal) and g n and g t
follow from the global iterative displacement solution. The default values of the penalty factors are derived from the contact
body properties (this will be discussed later on), but can also be user defined.
CHAPTER 6 565
Contact
Looking first at contact in the normal direction, it can easily be seen that if a contact condition is exactly fulfilled, then the
i trial i–1
gap function g n = 0 and  n = pn . The idea of the augmentation procedure is to better approximate the contact
conditions by adjusting the contact normal stress. The evolution of the contact normal stress during the Newton-Raphson
0
process is illustrated in Figure 6-44. The first displacement solution is obtained using p n = p n (which initially can be zero,
but in an arbitrary increment is given by the value at the end of the previous increment). Given the displacement solution,
1 1 0 1
the gap function g n is calculated and the contact normal stress update is given by p n = p n + E n g n . This new contact
normal stress is used in a subsequent iteration. If the augmentation is not performed, then the method reduces to a penalty
method with a constant penalty factor E n . As for the contact conditions in the tangential direction, the augmentation
procedure can be applied also if there is sticking contact. Then the sticking conditions (no relative tangential displacements)
will be better approximated by adjusting the tangential stresses (see Figure 6-45).
pn
exact value of pn
2
pn
1
pn
atan  E n 
0
pn
2 1 gn
gn gn
Figure 6-44 Updating Contact Normal Stress according to the Augmented Lagrange Procedure
tt
 n
1
tt
atan  E t 
0
tt
2 1
gt gt gt
–  n
Figure 6-45 Updating Contact Friction Stress according to the Augmented Lagrange Procedure (2-D)
Within the Newton-Raphson solution procedure, not only the contact contributions to the global force vector (as given by
Equation (6-9)), but also the contributions to the global stiffness matrix are needed:

K c = ------ F c (6-11)
u
Further evaluation of Equation (6-9) and Equation (6-11) depends on whether there is sticking or slipping. At the very
beginning of an analysis, an assumption has to be made (typically sticking), but after having determined the trial solutions,
it is possible to check if there will be sticking or slipping:
t ri a l trial trial trial t r i al
 = t –  n ;  0 : sticking;   0 : slipping (6-12)
and previously made assumptions on the contact state can be adjusted accordingly. Now the states of sticking and slipping
will be discussed separately.
In case of sticking, the predicted solution for the Lagrange multipliers is:
n trial n trial
n = n ; t = t (6-13)
For the contact contributions to the global stiffness matrix, the second order derivatives of the gap functions with respect
to the displacements are neglected, so that the following terms are needed:
 n g n  g t
-------- = E n -------- ; -------t = E t ------- (6-14)
u u u u
In case of slipping, the tangential stresses need to be adjusted in order to agree with Coulomb’s friction law:
trial
n trial n t trial
n = n ; t = -------------------  n (6-15)
trial
t
and the contact contribution to the global stiffness matrix now involves also changes in contact in the normal direction:
trial trial t r i al
 n g n  t  n t r i a l   t 
-------- = E n -------- ; -------t = -------------------  ---------------- +  n ------  ------------------- =
u u u t
t r i a l u u   t r i al

t
 2 
 trial tr i al t r i al 

 1 – --------------------
t1
–  t1  t2 
- -----------------------------
-
trial trial  trial 2 t r i al 2 
t g n  n  t t g t
= ------------------- E n -------- + ------------------- E t  ------- (6-16)
t
trial u t
trial
 2 u 
trial trial t r i al
 –t 1 t 2 t 2 
 ------------------------------ 1 – -------------------- - 
 trial 2 t r i al 2 
  t  t 
The first term in the expression for the frictional behavior in Equation (6-16) results in a non-symmetric stiffness matrix. If
the assumption is made that the contact normal stress is constant during a Newton-Raphson iteration, then this term
CHAPTER 6 567
Contact
cancels and a symmetric matrix is retained. Using a non-symmetric matrix will significantly increase the memory
requirements. It should be noted that a non-symmetric matrix may only have impact on the convergence behavior; if
convergence will be obtained, the final solution won’t be affected.
In Force Vector and Stiffness Matrix Contribution below, it will be shown how the Equation (6-9) will be numerically integrated in
a 2-D or 3-D contact analysis.
Contact Body Definition

The definition of contact bodies for the segment-to-segment algorithm is identical to that of the node-to-segment
algorithm. This implies that one can define deformable contact bodies consisting of finite elements, and rigid bodies
consisting of curves (2-D) or surfaces (3-D). Rigid bodies can be load, velocity or displacement controlled; load controlled
rigid bodies have one (translation only) or two (translation and rotation) control nodes.
Based on the finite elements assigned to a contact body, the program will automatically set up the outer boundary of the
deformable bodies. This boundary is defined by element edges (2-D) or element faces (3-D). Figure 6-46 shows the basic
concept for 2-D continuum and 2-D shell/beam elements. For 2-D continuum elements, the contact segments coincide
with the element edges. For 2-D shell/beam elements, not only the top and the bottom are separately included in the
contact body description, but also the free ends of the finite element model are converted into contact segments. The same
concept is applied to 3-D continuum elements and 3-D shell elements, where for the shell elements the free edges of the
finite element model are converted into contact segments. The user can also decide to ignore the shell thickness, in which
case the top and the bottom segments have the same position, but an opposite orientation. The contact segments of the
free ends will always include the shell thickness.
: nodal points
: auxiliary points
Finite Elements Contact Segments Auxiliary Versus Nodal Points

Figure 6-46 2-D Contact Bodies: Continuum Elements (Upper); Shell/beam Elements (Lower)
The SPLINE option, as discussed before for node-to-segment contact, is available for segment-to-segment contact. When this
option is selected for a shell element or a 2-D beam element, then the transition from the top and bottom contact segments
to the segment at the free ends is automatically marked to have a normal vector discontinuity.
Beam Contact
Beam contact can be easily analyzed using the segment-to-segment contact procedure. For the element stiffness and stress
recovery, the beam elements are used as is, but for contact purposes, the beams are exactly represented by internally
expanding them to three-dimensional beam patches. The contact forces and constraints are automatically transferred from
the patches to the beam nodes.
For 3-D beam elements, the line representation of the beam contact body is automatically replaced by an expanded
representation of the beam contact body based off the section properties. Five types of beam sections are supported:
1. Hollow pipes via the GEOMETRY option (Figure 6-47a).
2. General thin-walled sections via Method A of the BEAM SECT parameter (Figure 6-47b).
3. Non-Integrated sections via the GEOMETRY option or Method B of the BEAM SECT parameter.
4. Solid template sections via Method C of the BEAM SECT parameter (Figure 6-47c).
5. General solid sections via Method D of the BEAM SECT parameter (Figure 6-47d).
(a) Hollow Pipe Section (b) General Thin-walled Section
(c) Solid Template Section (d) General Solid Section

Figure 6-47 Four Types of Supported Beam Sections
CHAPTER 6 569
Contact
A sample expanded representation of four beam contact bodies comprising of a solid circular section, hollow pipe section,
thin-walled channel section, and arbitrary thin-walled section is shown in Figure 6-48. This expanded form is automatically
created internally from the given section and orientation data.
Figure 6-48 Expanded Representation of Four Beam Contact Bodies with Different Beam Sections
The following general points should be kept in mind for the expanded beam bodies:
 The expanded beam representation is the only possible scheme available for 3-D beams in segment-segment
contact. The capability is implemented for element types 14, 25, 52, 76, 77, 78, 79, and 98. For the quadratic beams
(types 76, 77), the midside node is ignored in the contact representation, and the patches are created between the
end nodes. Truss elements (types 9, and 64) are not supported in segment-segment contact.
 Applications of beam contact in the expanded mode include tube-tube contact (internal contact between beams),
beam-beam contact (external contact between beams), beam contact with solids, beam contact with shells, and
beam contact with rigid bodies. This form of beam contact can be used for statics, dynamics (modal and transient)
as well as multi-physics (e.g., thermo-mechanical contact).
 The expanded beam representation takes into account the current position of the beam based on it’s nodal
coordinates, the beam cross-sectional properties, the beam orientation (EGEOM4, EGEOM5, and EGEOM6 of the
GEOMETRY option) and beam offsets at the nodes (if any). Geometrically, beam patches are only used for contact
detection, separation and penetration. Internal rigid links are automatically generated from each beam node to the
generated beam patch nodes. No additional degrees of freedom are introduced and the movement of the beam
patch nodes is completely based on the translations and rotations of the parent beam nodes. Offsets that are
specified at the parent beam nodes are applied to the beam patches also. The contact stiffnesses and forces are
calculated at the patches and transferred to the original beam nodes using the patch vectors.
 EGEOM7 (7th field of the GEOMETRY option) is used to provide additional control information for the expanded
beams in segment-segment contact. EGEOM7 is a packed entity consisting of three flags: NSEG - number of
segments used to represent certain cross-sections. This is only relevant for circular solids, elliptical solids and
hollow pipes. IPTCH - patch creation flag. This is only relevant for hollow pipes. IESCAP - end cap/connecting
cap creation flag. IESCAP is applicable to all sections and is used to control generation of both end caps and
connecting side caps. When flagged, end cap patches are generated at the free ends of the beam. The method
varies depending on the type of section. For connecting side patches, there are two available choices for
generation: (a) Creating discontinuous patches across the beam elements - the usual method used is to generate
separate end-cap like patches for each beam element. (b) Averaging the patches across the beam elements - in this
case, the main patches themselves are adjusted through averaging and no separate side patches are created. This is
only possible if the adjacent sections are identical but with differing dimensions.
 The distance tolerance calculation for the expanded beam sections incorporates two schemes: The normal
evaluation that is similar to node-to-segment contact for beam elements is done. In addition, a second evaluation
based on the detailed section properties is carried out. The minimum of the node-to-segment scheme and the
section property scheme is the final distance tolerance. The section property based distance tolerance evaluation is
based on 0.25 * min(patch edge length, section thickness). Note that the minimum patch edge
length is relevant for all sections but the minimum section thickness is only relevant for thin-walled sections
(general and hollow pipes) when both outer and inner patches are flagged to be created.
 The local direction cosine matrix at each beam node is used to establish the 3-D patch vector directions. The
direction cosine matrix for each beam element comprises of the local Z axis along the beam element and the local
X-Y axes in the cross-sectional plane of the element. At each beam node, the nodal direction cosine matrix is
computed as the average of the matrices of two adjacent beam elements meeting at the node. This average
direction cosine matrix is established only if the angle between the local Z axes of the two elements and the angle
between the local X axes of the two elements is less than a certain value. This value can be user specified (7th field
of the 6th data block of the PARAMETERS option). If this value is blank or specified as 0.0, the default value of 160°
is used. If this value is specified as a negative value, the averaging is skipped and each beam element is given a
separate set of patch vectors at the beam node based on the direction cosine matrix for that element. If this value is
> = 180°, the averaging is forced irrespective of the angle between the two beam elements.
 The SPLINE option is available for the beams in expanded mode. This can be specially useful for beams with curved
cross-sections. Only automatic discontinuities are supported. It is worthwhile checking the spline fit for beams in
expanded mode by using the zeroth increment spline mfd file for visual verification.
 Limitations of the expanded beam representation include the following:
• Beams that are used in the contact definition cannot be candidates for adaptive meshing, global rezoning, or
element activation/deactivation. In other words, any analysis which requires a recreation of the expanded beam
boundary at a later stage is not possible, and the analysis will error out at that stage. It should be noted that the
restriction only applies to the beams used as contact bodies and other contact bodies (non-beams) in the same
analysis can be used as candidates for adaptive meshing, global rezoning or element activation/deactivation.
• The beam cross-section does not change with finite strains. Note that this is a general restriction for all Marc
beam elements in a large strain analysis and is a limitation for the expanded beam representation also. This
means that while the beam patches can undergo finite translations and rotations, the cross section does not
change for large strains, thermal strains as well as contact interference.
CHAPTER 6 571
Contact
• Multiply connected beams are not possible. The analysis will continue with a warning but the neighbor
determination for each of the beams is based on the assumption of a singly connected beam with one neighbor
each at the start and end nodes of the beam element.
Additional information for each method available to create beam sections is as follows:
 Hollow Pipes (GEOMETRY option)
• The thickness (T) and radius (R) provided via EGEOM1 and EGEOM2 are used for creating the patches. By
default, outer patches are created at R+T/2 and inner patches are created at R-T/2.
• Additional controls are provided via EGEOM7 for controlling the number of segments (NSEG), the creation of
outer/inner patches (IPTCH) and the creation of end-caps/connecting side caps (IESCAP). NSEG defaults to 32
each for the outer and the inner patches. IPTCH allows one to specify outer patches only, inner patches only,
or both (default). For example, for a tube-tube contact problem, one can specify that only the outer patches of
the inner tube need to be created to contact the inner patches of the outer tube. The thickness can be included
(default) or excluded while creating outer patches only or inner patches only. If flagged via IESCAP, end cap
patches and discontinuous connecting side patches are generated from the inner patches to the outer patches of
the hollow pipe.
 Thin-walled sections (Method A of the BEAM SECT parameter)
• Additional controls are provided for excluding a branch. For included branches, the user can control the branch
sides that are created. Default is to include all branches and create all sides (outer/top, inner/bottom and
thickness sides). The definition of outer and inner is with respect to the centroid of the section. Outer and inner
patches can be created with the thickness offset or at the branch mid-line (i.e., thickness is ignored). If flagged
via IESCAP, end cap patches and discontinuous connecting side-cap patches are created from the inner patches
to the outer patches of the thin-walled section.
• A straight branch is represented by a single segment while a curved branch uses two times the number of stress
divisions for the branch.
• In order to get the correct orientation of normals to the patches, it is recommended that every thin-walled
section should follow a counter-clockwise direction. For example, for a L-section, the vertical leg should be first
defined in the -Y direction followed by the horizontal leg in the +X direction and not the other way about. For
continuous segments that are either perpendicular to each other or of equal thickness, the normals of the
meeting segments and thicknesses at the meeting segments are automatically averaged. If a section is defined
clockwise, the program will automatically flip the branches to get the correct orientation - however, the
averaging of the normal and thickness would not take place in this case.
 Non-Integrated Sections (GEOMETRY option or Method B of BEAM SECT parameter)
• It is not recommended to use these types of sections since the exact section shape to be used for the expanded
beam mode is not known.
• An attempt is made to fit a circular solid or a rectangular solid section based on the given section properties. If
this attempt is unsuccessful, the analysis errors out and the user is requested to provide the actual section
properties. If successful, the rules for solid templates are followed.
 Solid Template Sections (Method C of the BEAM SECT parameter)
• Eight solid templates are supported. These include: Elliptical, Circular, Square, Rectangular, Triangular,
Trapezoidal, Diamond, Hexagonal. Number of segments to be used for the elliptical and circular solids are
specified via NSEG of EGEOM7 field. Default is 32.
• If flagged via IESCAP of EGEOM7, end cap patches are generated by creating an additional center node and
placing triangular patches from this center node to the beam patch nodes. Also, there are two sub-options
available for generating discontinuous side-patches across the beam elements:
(i) For sections with identical shapes but differing dimensions, patches are created from the main patches
of the smaller section to the main patches of the larger section.
(ii) For sections with different shapes, separate end-cap type patches are generated for each beam element.
 General Solid Sections (Method D of the BEAM SECT parameter)
• Additional controls are provided for excluding a segment. For included segments, the user can control the sides
that are created. Default is to include all segments and create all sides (side 1-2, side 2-3, side 3-4, side 4-1).
• In order to get the correct orientation of normals to the patches, it is recommended that each segment should
follow the right-hand rule. The program will automatically flip incorrect segments to get the correct orientation.
• If flagged via IESCAP of EGEOM7, end cap patches and discontinuous side cap patches are generated by creating
an additional center node and placing triangular patches from the center node to the beam patch nodes.
Contact Detection
As also shown in Figure 6-46, each contact segment will get a number of auxiliary points. These auxiliary points are only used
during the contact detection phase of the analysis and they are defined at fixed positions on the contact segments. Since
the auxiliary points are defined on a per segment basis, there are multiple points with possibly different normal vectors at
adjacent contact segments. Especially at sharp corners, this allows for multi-body contact in such a way that each (part of
a) segment can touch only one other segment.
The actual contact detection phase consists of two passes. During the first pass, there is a check whether an auxiliary point
is sufficiently close to a contact segment and if the corresponding normal vectors are “almost opposite”. Hence there is a
distance check followed by a direction check (see also Figure 6-49). During the distance check, the distance between an
auxiliary point and its closest point projection on a contact segment is calculated and compared with a reference value. This
reference value is determined by the contact tolerance  and the bias factor B , where 0  B  1 . The default contact
tolerance is calculated by the program as the smaller of 5% of the smallest element side or 25% of the smallest (beam or
shell) element thickness, based on all the elements defined in any contact body. Alternatively, the user can define the contact
tolerance. The default bias factor is zero. A non-zero bias factor, entered by the user, implies that the outside zone becomes
smaller and the inside zone larger and usually avoids unnecessary iterations following from physically unrealistic contact
detection. If the position of an auxiliary point is x a and that of its closest point projection on a segment s of contact body
is x s , then the distance check is passed if x a – x s   1 – B  when the auxiliary point lies outside the contact body,
and if x a – x s   1 + B  , when the auxiliary point lies inside the contact body. If the distance check is passed, then
the direction check is performed. This direction check is passed if the angle between the normal vector at the auxiliary point
and the normal vector at the potentially contacted segment is larger than a threshold value  , so  n a n s    . The
default value of  is 120 , but it can be user defined using the PARAMETERS option.The angle between the normal vectors
CHAPTER 6 573
Contact
is defined according to Figure 6-50. Once both the distance and the direction check are passed, the corresponding segments
are marked as a pair of potentially contacting segments and for each of these pairs, the projection of one segment on the
other is determined. In 2-D, this projection defines a polyline, in 3-D it defines a polygon. In Figure 6-51, a 2-D and a 3-
D example is given. Each polyline/polygon has a number of polygon/polygon points, which will serve as the basis for the
numerical integration of Equation (6-9). It should be noted that the polyline/polygon points are not the same as the
auxiliary points introduced before. The auxiliary points only serve to find potential contact pairs, the polyline/polygon
points characterize the actual contact area between such contact pairs.
: closest point projection

na 1
 1 + B 
 1 – B 
n a2
ns ns
Distance Check And Direction Check Passed Distance Check Passed, Direction Check Failed
Figure 6-49 Contact Detection, Phase 1: Distance And Direction Check
na
 n a n s 
ns
Figure 6-50 Definition of the Angle Between Normal Vectors
: polyline points
Figure 6-51 Common Areas of Contact Segments: Polylines (Left) And Polygons (Right)
During the second pass of the contact detection phase, for each polyline/polygon point it is checked whether or not the
segments are locally in contact. This check is equivalent to the check which has been described above, so a distance check
followed by a direction check. At the end of this second pass, there is a set of polyline/polygon points being in contact and
representing the contact area.
Force Vector and Stiffness Matrix Contribution

The polyline/polygon points discussed above basically define a local connection between two contact segments. Indicating
such a point by a subscript i , the gap function g n i can be evaluated as the normal displacement difference between a point
and its closest point projection on the contacted segment. Similarly, by looking at the tangential displacements, the gap
function g ti can be evaluated. The displacement of a point on a contact segment is a function of the displacements and
(in case of shell elements) rotations of the nodes corresponding to that segment. By collecting all the nodal displacement
and rotation degrees of freedom in a vector U , the gap functions can be written as:
T T T T
g n i =  G ni – G' ni U ; g t i =  G t i – G' t i U (6-17)
Here, vector G ni and matrix G t i express the dependency of the normal and tangential displacements of a point on a
contact segment on the total set of degrees of freedom. In a similar way, G' n i and G' t i are used for the closest point
projection of this point on the contacted segment. It should be noted that, if applicable, G n i , G' n i , G t i and G' t i also
include the effect of the shell thickness and shell offset vectors. Care has been taken to define them in such a way that they
are able to represent rigid body modes of structures being in contact without introducing spurious energies.
Equation (6-17) can be used to numerically evaluate equation (6-9). First, the contribution to the force vector is given by:
N N
g T g T
 -------n-  d +  -------t  d =
  u  n   u  t   G n i – G' n i  n i  i +   G t i – G' t i  t i  i (6-18)
  i = 1 i = 1
in which N is the total number of polyline/polygon points,  n i and  t i are the estimated contact normal and tangential
stresses in polyline/polygon point i (see Equations (6-13) and (6-15)) and  i is the area corresponding to this point. Note
that a possible dependency of G n i , G' n i , G t i and G' t i on U has been neglected.
Next, for sticking contact, the contribution to the global stiffness matrix is:
N
g n T g n g T g T T
  -------
u 
- E n -------- d +   -------t E t -------t d =
u  u  u   G n i – G' n i E n i  G n i – G' n i  i
  i = 1
(6-19)
N
T T
+   Gt i – G' t i E t i  G t i – G' t i  i
i = 1
CHAPTER 6 575
Contact
Finally, for slipping contact the contribution is found to be:
2
t 1 –t 1 t 2
T
1 – ------------ - -------------------
2 2
g n T g n g n T  t g t g t T  n  t g t
 -------- E -------- d +  -------- E ---------- ------- d +  ------- ---------- E t ------- d =
  u  n u   u  n  t u   u   t t 2 u
   –t 1 t 2 t 2
------------------- -------------
2 2
t t
N N T
T T t i T T
=   G ni – G' n i E n i  G n i – G' n i  i +   Gn i – G' n i E n i ------------  G t i – G' t i  i + (6-20)
t i
i = 1 i = 1
2
t 1 i –t 1 i t 2 i
N - -----------------------
1 – --------------
2 2
 n t i t i T T
+   G t i – G' t i  ---------- E t  G t i – G' t i  i
t 2
i = 1 –t 1 i t 2 i t 2 i
----------------------- 1 – ---------------
2 2
t i t i
The right choice of the penalty factors is important to get accurate results in an acceptable number of iterations. If the
penalty factor for normal contact, E n , is too small, then there will be a relatively large amount of penetration, which has
to be removed by the augmentation process outlined above. However, if the penalty factor is small, then the first order
correction process illustrated in Figure 6-45 will converge slowly. On the other hand, if the penalty factor is too large, the
overall system of equations may be ill-conditioned and a non-smooth contact stress field may be found.
The default value of the penalty factor E n i depends on the body stiffness of the two contact bodies involved and a
characteristic length (note that the dimension of the penalty factor is force per cubic length). The body stiffness is either
defined by the average trace of the initial stress-strain law of the elements defining the two contact bodies or by the average
bulk modulus for (nearly) incompressible rubber materials, whichever of the two is the largest. For continuum elements,
the characteristic length is given by one half of the average length of all the edges being part of the contact boundary. For
shell elements, the characteristic length is given by half of the average thickness of all the shell elements being part of a
contact body. When there is contact between a solid and a shell element, then the characteristic length is defined by the
shell element. If polyline/polygon point i defines contact between the deformable bodies k and l , with body stuffiness
S k and S l , and the characteristic length of the model is L , then the default value of E n i is given by:
0.5  S k + S l 
E n i = ------------------------------
- (6-21)
L
In case of contact with a rigid body, the equations shown before can be applied with minor modifications. First, the
displacements G' ni U and G' t i U of the closest point projection directly follow from the motion of the rigid body. Second,
the issue of an ill-conditioned system does not easily occur and the default penalty factor can be chosen to be larger. Since
there is no body stiffness associated with a rigid body, the default value of E n i is related to the deformable body only and
given by:
1000S
E n i = -----------------k- (6-22)
L
For tangential contact, the default penalty factor E t i is related to the penalty factor E n i according to:
E t i = E ni (6-23)
–3 –6
where  = 1 10 for deformable contact and  = 1 10 for rigid contact.
As mentioned before, the user can define the penalty factors. This can be done either globally via the CONTACT (2-D) option
or per body combination via the CONTACT TABLE (with TABLE Input) option. The penalty factors are by default printed in the
output file. The program computed default penalty factors can be optionally scaled by the user. The default penalty factors
are 1.0 and any value above 0.0 is allowed. If excessive penetrations are seen in a problem, this may indicate that the default
penalty factor is too small and a scale factor in the order of 10.0 could provide for a stiffer contact solution. If excessive
iterations are seen in a problem, this may indicate that the default penalty factor is too large and a scale factor in the order
of 0.1 could provide for a softer and convergent contact solution.
Iterative Penetration Checking and Contact Stress Update

After each solution in the global Newton-Raphson process, there is a check if the iterative displacement field would cause
penetration of contact bodies. This check involves first the auxiliary points on the contact segments not yet being marked
as contact pairs and next the polyline/polygon points not yet being in contact. The maximum amount of penetration at
the auxiliary and polyline/polygon points is computed and if this penetration is larger than the inside contact tolerance
distance  1 + B  , then the entire iterative displacement solution is scaled back to avoid this penetration.
i i T T i
Once the iterative displacement solution U is known, the gap function can be updated with g n i =  G i – G' i U
i i–1 i
and g ni = g ni + g n i , where the superscript i indicates the iteration number. The update of the contact normal stress
is given by:
i i–1 i i
p n i = p ni + E ni g n i + E n i g n i (6-24)
Note that the last term in Equation (6-24) results from the augmentation procedure.
Since there are in general more polyline/polygon points on a contact segment than there are nodes on such a segment,
enforcing a zero gap function everywhere would in general result in an over-constrained system. To this end, a reduction
of the penetration field has to be applied. The following options are available:
 No augmentation. This results in a pure penalty method and is recommended for most analyses, since it gives
reasonably accurate results in a relative small number of iterations.
 Augmentation based on a constant penetration field. This is the recommended procedure for linear finite
elements.
CHAPTER 6 577
Contact
 Augmentation based on a (bi-)linear penetration field. This procedure should be used only for quadratic finite
elements.
It is also possible to let the program decide which of the augmentation methods should be applied (constant or linear).
Then the choice is made based on the combination of elements corresponding to the contacting segments. It is noted that
in case of contact with a rigid body, the augmentation procedure is always applied.
In case of augmentation based on a constant penetration field, the average gap function per contact segment is determined
and applied to each polyline/polygon point corresponding to that segment.
In case of augmentation based on a (bi-)linear penetration field, a least squares method is used to determine the values of
the gap function in the corner nodes of the segment. Based on the position of a polyline/polygon point on this segment, a
(bi-)linear interpolation between the corner values then gives the local value in the point.
Since an exact fulfillment of a zero gap function is difficult to achieve, a threshold value is needed to determine whether or
not an extra iteration due to the augmentation procedure is needed. This penetration distance beyond which an
augmentation will be applied can be calculated by the program or entered by the user. The default penetration distance
beyond which an augmentation will be applied is defined as:
 au g = 0.001L (6-25)
where the characteristic length L is the same as used in Equation (6-21). Similar to the penalty factor, the user can define
the augmentation distance either globally for the analysis on the CONTACT option or per body pair on the CONTACT TABLE
option.
If an augmentation procedure is selected and the maximum penetration is beyond  au g , then an extra Newton-Raphson
iteration will take place, as long as the maximum number of iterations has not been exceeded. If the maximum number of
iterations has been reached, then the program will continue with the next increment if the usual convergence criteria have
been fulfilled; so the analysis will not stop because of the augmentation procedure. In case of friction with augmentation,
the same logic is applied to the tangential behavior.
Separation
Since in each polyline/polygon point the contact normal stress is calculated, separation will always be based on stresses. By
default, if the contact normal stress is in tension, then the corresponding polyline/polygon point will separate, which
implies that this point will no longer contribute to the global stiffness matrix and force vector.
In order to speed up the separation process on a structural level, there is an extra check performed on a contact segment
basis. If there are points on a segment which separate, then the total contact normal force F n t due to points on this segment
being in tension (negative) is compared with the total contact normal force F n c due to points on this segment being in
compression (positive). These forces are obtained by integrating the contact normal pressure over the corresponding
contact area. If F nc  0.75 F n t , then all the points on this segment will separate. This extra check is only active if the
separation stress is default (zero). As soon as the user enters a non-zero separation stress, either globally for the analysis on
the CONTACT option or per body pair on the CONTACT TABLE option, this extra check will be skipped.
Finite Sliding for Deformable Contact

In the discussion above, the position of the polyline/polygon points with respect to the element edges/faces is based on the
geometry of the elements at the first time of touching. In case of contact between deformable contact bodies, this implies
that the equations are valid only for simulations where the relative displacements between touching deformable bodies are
small. Since in many contact applications the relative displacements are finite, the default option is to allow for large relative
displacements. If large relative displacements are allowed, the relative displacements of the polyline/polygon points are
monitored and once a threshold value d m a x is exceeded, new points will be created. The default threshold value equals
five times the contact tolerance  , but it can also be defined by the user via the CONTACT option.
For a 2-D configuration, the update process for large relative displacements is illustrated in Figure 6-52. The initial position
of the polyline points is shown on the left-hand side. The blue and green areas represent the common areas between the
upper and the left-hand and right-hand lower element. A possible relative displacement d is given in the middle of
Figure 6-52. As long as d  d m a x , the contact simulation is based on the initial position of the polyline points. If d  d m ax ,
new polyline points are created, resulting in a situation as shown on the right-hand side. Redefining polyline/polygon
points will be done during an analysis as often as needed. After having created new polyline/polygon points, important
contact data (like the contact stresses) is mapped from the old on the new polyline/polygon points and used as a starting
point to continue the analysis.
Figure 6-52 Large Relative Displacements: Initial Position (left), Sliding (middle) and New Position (Right)
The CONTACT option allows the user to switch on the flag for small relative displacements, in which case there is no update
of the position of the polyline/polygon points. Note that for this case of Small Sliding, the constraints are still evaluated on
the updated patch positions, shell normals are updated continuously, new segments can come into contact at any time and
polygon points are checked for sliding across segments and for penetration. The only limitation for Small Sliding is that
new polyline / polygon points are not generated based on a check of d versus dmax.
Results Evaluation
For Segment-to-Segment contact, the Marc post file contains contact results for both deformable bodies and rigid bodies.
There are three types of result categories possible for a contact analysis:
Contact Status and Touched Body

For deformable bodies, the status is available as a nodal output. Note that though contact is computed and enforced at the
polygon points, the status is extrapolated and plotted at the nodes. In this case, the nodes of both touching and touched
CHAPTER 6 579
Contact
segments of the contact bodies are assigned a non-zero contact status. Furthermore, note that though the touching and
touched segment nodes indicate contact, the contact footprint is limited to the actual intersection between the touching
and touched segments. The convention is as follows:
 A value of 0.0 means that none of the polygon points of the segments associated with the nodes are in contact.
 A value of 0.5 means at least one of the contacting polygon points of the segments associated with the nodes is in
near thermal contact and no other points are in regular contact / glued contact / cyclic symmetry contact.
 A value of 1.0 means that at least one of the contacting polygon points of the segments associated with the nodes
is in regular contact and no other points are in glued contact / cyclic symmetry contact.
is in glued contact and no other points are in cyclic symmetry contact.
is on a cyclic symmetry boundary.
The Contact Touched Body is also available as a nodal output. The Touched Body output is available for nodes of both
participating segments, i.e., the Touched Body for X is Y and vice versa.
Contact Forces and Stresses

Three types of response output are possible:
a. Normal Force / Friction Force / Normal Stress / Friction Stress:
These vectorial quantities are available as nodal quantities for touching and touched segments of deformable
bodies. Contact normal force and friction force are included as Default nodal output – contact normal stress
and friction stress need to be selected via Custom output if desired. Note that the quantities are computed as
follows:
• Contact Normal Force: This is a nodal vector directly following from the integration of the contact normal
stress in the polyline/polygon points
• Contact Friction Force: This is a nodal vector directly following from the integration of the contact friction
stress in the polyline/polygon points
• Contact Normal Stress: This is a nodal vector representing the normal stress in the polyline/polygon points.
This nodal representation is based on a constant field approximation per contact segment of the
polyline/polygon values.
• Contact Friction Stress: This is a nodal vector representing the friction stress in the polyline/polygon points.
This nodal representation is based on a constant field approximation per contact segment of the
polyline/polygon values.
Since there are usually many more polyline/polygon points than there are nodal points, in many cases, the
contact stress vectors can only approximately represent the actual stress field calculated by the contact
algorithm.
b. Resultant Body Forces and Moments

The resultant contact force and moment vectors are available for each deformable body and rigid body. The
time history of these resultant quantities are of significant interest in many engineering analyses. The moments
are computed about the user-defined centroid for rigid bodies (default = 0,0,0) and about the origin for
deformable bodies.
c. Resultant Body Force and Moments per Body Pair
The interaction forces and moments for each body pair can also be output. The time history of the forces and
moments can be plotted. In addition, by associating a node with a specific body pair, the total normal and
tangential interaction force / moment quantities can be visualized as vectors.
For the resultant quantities in categories (b) and (c), it should be noted that the contact interaction output do not include
self-equilibrating forces between the bodies. The computed interaction quantities only refer to the net contact force and
moment vectors that satisfy free-body diagrams for each body. Note that if there are no resultant forces found for a
deformable body or a rigid body, it implies that either the body is not in contact with any other body or all forces on the
body are self-equilibrating. For example, a cylinder press-fit that introduces uniform radial forces around the circumference
would indicate a net interaction force of zero since all the contact forces are self-equilibrating.
Miscellaneous Contact Response Quantities

These includes the following:
a. Contact Stress-Free Projection Vector
This nodal vector is available by default for Version 15 style input if stress-free projection is activated for at least
one body pair via the Contact Interaction option. Note that this vector only captures the initial adjustment
made to the nodal coordinates – it does not capture any additional adjustments made to the polyline/polygon
points.
b. Contact Interaction Heat Flux
For thermal analyses, the heat flux transmitted between the bodies can be output as a nodal scalar quantity. By
associating a node with a specific body pair through the Contact Table, the interaction heat flux can be
visualized on the post file.
Contact Additional Print Output

Additional print about a contact analysis can be obtained in the output file through the PRINT parameter. Information
on the segments that come into contact, the number of generated polygons and polygon points, which polyline/polygon
points come in contact, what is the touched segment, when does a polygon point separate, etc. can be printed out through
the PRINT,5 option. Incremental displacements in the local coordinate system can be printed out through the PRINT,8
option. Note that, for large problems, this can result in a significant amount of output.
Beam post file and combined model

When beam contact bodies are used in a segment-segment analysis, in addition to the job post file, two additional post files
are produced on user request. The beam post file allows the visual verification of the expanded beam representation. It has
the same type (binary or ascii) and frequency as the job post file. Note that the beam post file only supports post file versions
9 and higher. If the job post file version is less than 9, the post file version for the beam post file is set to 12. This post file
CHAPTER 6 581
Contact
only contains the expanded representations of the beam contact bodies and other non-beam contact bodies (deformable or
rigid) are skipped. The contact status of the beam patch nodes and the displacements of the beam patch nodes are available
for every post increment. The model and beam post file is a wrapper file that allows the simultaneous visualization of the
job post file and the beam post file. The combined visualization is supported for most cases including remeshing but is not
supported for the following situations involving the job post file : post file revisions < 9, iterative post file option used for
the job post file, continuous post option used for restart.
Multi-physics
The segment-to-segment algorithm supports thermal, coupled thermal-mechanical, Joule heating, electrostatic and
magnetostatic analyses. The basic equations for the thermal pass have already been discussed in the node-to-segment session
of this chapter. In case of glued contact, for segment-to-segment contact no multi-point constraint equations are applied.
Instead, a penalty factor is used to approximate the continuity of the primary variable (like temperature, voltage and electric
potential). The numerical integration over the contact surface is performed using the common area of contact segments,
as shown in Figure 6-51.
Trouble Shooting
Compared to the node-to-segment procedure, the segment-to-segment procedure uses more numerical parameters. An
effort has been made to define default values for these parameters which work for various kind of problems (small and finite
sliding, small and large deformations, with and without friction, with and without global remeshing, etc.), but in some
cases it may be necessary to modify the default settings. It is recommended to start an analysis using the default settings,
but if they do not result in a sufficiently accurate solution, the following can be used as a set of guide lines to improve the
situation:
 If the amount of penetration is too large, then the penalty factor can be increased (note that the default value of
the penalty factor as calculated by Marc can be found in the output file) or the augmentation procedure can be
activated. Typically, the penalty factor should be increased by a factor of 10. If the amount of penetration is
growing during sliding contact and the augmentation procedure isn’t used, then a modified calculation of the
contact normal pressure can be activated using FEATURE,8401; this switches off the mapping of the contact normal
stress as discussed in Finite Sliding for Deformable Contact, and directly determines the contact normal stress based on
the current penetration.
 If there is sliding contact, but the amount of friction is too small, then most likely the default value of the penalty
factor to simulate sticking is too small and one can increase this penalty factor or activate the augmentation
procedure for friction. Typically, this penalty factor should be increased by a factor of 10. Alternatively, the
penalty factor used to simulate sticking can be derived from the current contact normal pressure using
FEATURE,11701.
 If there is little to no penetration, but a very slow convergence behavior is observed, then usually the penalty factor
is too large and should be decreased. Typically, the penalty factor should be decreased by a factor of 10. If there is
contact between deformable bodies with a large difference in stiffness or if there are shell elements involved in the
calculation and they are mostly loaded in bending, then FEATURE,12001 can be activated. This causes the default
penalty factor to be based on the softest of two contacting bodies, without taking into account the shell thickness
when determining the characteristic length (see Force Vector and Stiffness Matrix Contribution).
 If the augmentation procedure is activated and the number of iterations is large for every increment, then the
distance beyond which an augmentation will be applied can be significantly increased. If this is done, the
augmentation procedure is still applied, but will not cause extra iterations if the convergence criteria are fulfilled.
Bullet points 2, 3, and 4 above can also be achieved by using modified defaults, activated by a 1 in the 3rd field of the 1st
data block of the CONTACT (2-D) option.
In the discussions above, the angle between two segments (see Figure 6-50) is used only to determine whether or not
segments will contact each other. Once there is contact, the actual value of this angle does not influence the value of the
applied penalty factor. Especially for cases where the angle  is close to the threshold value (which by default is 120 ),
this may result in inconsistent behavior: values just below the threshold value do not result in any contact constraint, where
values just above the threshold result in fully applying the contact constraints between the contacting segments. By entering
a 1 in the 11th field of the 3rd data block of the CONTACT (2-D) option, the contact stiffness between the segments is
multiplied by a factor which is 1.0 if the angle  is 180 (parallel segments) and it goes smoothly to 0.0 at the above
mentioned threshold value. This way, a small variation of the angle  around the threshold value only has a small impact
on the analysis.
As outlined above, when two segments come into contact, the contacting segment is projected on the contacted segment
and the contact pressure and stiffness are created in the normal direction of the contacted segment. For parallel segments,
it does hardly make a difference which segment is contacting and which segment is contacted, but for non-parallel
segments, this may not be true. By entering a 1 in the 11th field of the 3rd data block of the CONTACT (2-D) option, an
algorithm is activated to minimize the effect of the contact body numbering. To this end, some rules are applied based on
the following segment types:
1. Linear Shell Edge segments;
2. Quadratic Shell Edge segments;
3. Expanded Beam segments;
4. Linear Shell Top and Bottom segments;
5. Quadratic Shell Top and Bottom segments;
6. Linear Continuum Element segments;
7. Quadratic Continuum Element segments.
When two different segments contact each other, the lowest type will be the contacting segment and the highest will be the
contacted segment. When two identical segments of different contact bodies contact each other, the contacting segment
will be from the body with the largest average element edge length. If the segments belong to the same body, in general the
lowest element number will be the contacting segment. The rules on identical segment types will not be followed if the
contacting segment belongs to a larger smooth surface and the contacted segment has sharp edges: in this case, the segment
belonging to the larger smooth surface will be the contacted segment.
Segment to Segment Linear Contact

The Linear Contact feature is specifically designed and implemented for analysis involving very small relative displacements
and rotations. This includes engine block analysis, small strain inertia relief analysis and bolted assembly analysis with
limited relative motion between the contact bodies.
CHAPTER 6 583
Contact
Contact is established only at the start of an analysis and no new segments can come into contact later. The contact stiffness
is computed based on the initial configuration. Both the position of the touching segment and of the touched segment are
not updated through the analysis and shell normal are based on the initial geometry. The polygon/polyline points can
separate and re-establish contact anytime during the analysis. As the displacements are infinitesimal and the parts are
initially in contact, polygon/polyline points are not checked for sliding across segments or for penetration.
Linear contact is available with parallel processing and multi-threading capabilities of Marc. The various contact features
like friction modeling, glued contact are supported with linear contact.
Limitations: Any option that requires regeneration of the contact boundary is not supported for linear contact – this
includes Global Remeshing, Local Adaptive Meshing, Element Deactivation and Mesh Splitting.
Interference Fit
This feature can be used to simulate situations where bodies with overlaps or gaps in the geometry are required to fit
together. Some examples of interference fit usage include spur gears that are mounted tightly on a shaft, couplings, collar
rings, tire mounted on wheel rim, bearing bushings in hubs, valve seats, rubber seals, bushings made of synthetic resin
pressed material, etc.
The interference fit capability is treated in Marc using the contact algorithm. The penetration associated with interference
fit can be much larger compared to any contact analysis. Ideally, interference fit computations are performed in a separate
loadcase, which might take more than a one increment. The user can control the number of increments or time required
to accomplish the fit. The user is equipped with many methods which can be adopted depending on the geometry.
There are five methods available in Marc for the interference fit option. These are applicable to both Node-to-Segment
Contact (N2S) and Segment-to-Segment Contact (S2S). These methods are as follows:
1. Contact normal
2. Translation
3. Scaling
4. Automatic
5. User-subroutine
1. Contact Normal
This method is generally recommended for situations with small values of interference to be resolved along the
normal direction to the touched interface. The other methods are more suitable when the interference is large.
In this case, the nodes (N2S) or auxiliary points (S2S)of the touching body are projected in a direction normal to
the segments of the touched body. The individual touching entities are positioned at a certain distance (over-closure
or gap) from the touched surface along the normal of the touched surface. The user needs to provide this
interference distance which signifies the maximum overlap or gap between the bodies. This provided value is
specified as a negative value for an overlap and as a positive value for a gap. The detection algorithm works as
follows:
(Overlap/Gap + Interference) at a node (N2S) or auxiliary point (S2S) < D (1+B) on the inside or < D (1-B) on the
outside
Where D = regular distance tolerance and B = bias factor.

To control the number of increments or time during which the interference fit is achieved, the user can optionally
specify a table in which the magnitude of interference distance is ramped down from the possible maximum value
of interference to zero. Additionally, the user can specify the “project stress-free” option to correct the existing
geometry to make the interference uniform across all the touching nodes. Using this option, interference can be
gradually resolved as shown in Figure 6-53. Note that if a clearance or overlap needs to be maintained for the entire
job, then the interference value needs to be maintained as a constant value in all the loadcases of the job. This
method may not perform adequately when the interference distance is large – in this case, the touching entities may
find multiple touched segments within the given closures distance.
Figure 6-53 Gradual Resolution of Interference
Methods 2 to 5 are more general and allow larger amounts of interference (overlap or gap) between the touching and
touched bodies. The general scheme for these methods is to internally compute an initial shift vector between the nodes of
the touching body and the segments of the touched body. This initial vector is also referred to as a pseudo displacement vector.
The following points should be noted about the pseudo displacement vector:
 The magnitude of the pseudo movement is varied based on a table provided by the user, which is normally a ramp
down table from unity to zero over the loadcase. Note that the projection of the touching entity to a touched
segment is done for each iteration but the position of the touching entity is always based on a scaling of the pseudo
displacement vector computed at the start of the loadcase. Therefore, large rotations of the interfering bodies
should not be permitted during the interference loadcase.
CHAPTER 6 585
Contact
 The treatment of the interference across two successive loadcases should be noted. Note that the interference can
be specified for any loadcase, though normally, it is specified at the start of the analysis (zeroth increment and first
loadcase). Three situations are possible:
1. The interference is specified in the zeroth increment and also in the first loadcase – this is the recommended
option. During the zeroth increment, the contact between the interfering bodies is established based on the
given data but the interference itself is not resolved. The actual interference is resolved in the first loadcase.
2. The interference is not specified in the zeroth increment but specified in the first loadcase – this is not
recommended. In this case, it is possible that some contact is detected between the two bodies at increment 0
based on the regular contact algorithm and this may cause a conflict with the pseudo displacement contact in
loadcase 1.
3. The interference is specified in the zeroth increment but not specified in the first loadcase – this is also not
recommended. In this case, the contact between the interfering bodies is established at increment 0 and the
computed “pseudo displacement” is immediately reduced to 0.0 at the start of increment 1. This may cause
convergence difficulties.
Each interference method provides a different scheme for the computation of this pseudo vector.
4. Translation
In this method, for each interfering body pair in the contact table, the user specifies the following:
• Magnitude of the vector
• Direction cosines of the vector
• The coordinate system in which the direction cosines are specified (default is the global Cartesian system)
• The body to which to apply the interference vector and
• A table specifying the variation of the interference vector during the loadcase
The algorithm performs an initial pseudo movement of the interfering body along the direction specified by the
user such that the overclosure is removed. This is followed by a removal of the pseudo movement such that the
interference fit is achieved. This method is especially useful for large initial over-closure between contacting bodies
in a specific direction.
5. Scaling
• Position of the Centroid about which the scaling is to occur
• X, Y and Z Scaling Factors
• The coordinate system in which the scaling should occur (default is the global Cartesian system)
• A table specifying the variation of the interference resolution during the loadcase
The pseudo adjustment is done by scaling one of the bodies such that the overlap is removed. The rest of the
algorithm is similar to that of the Translation option.
6. Automatic
• A penetration tolerance
The overlap distance vectors are found between the nodes of the overlapping body (specified by the user) into the
other body. The penetration tolerance should be a little more than the highest penetration between contact body
pair. Nodes within this tolerance are considered for the interference calculation. The pseudo displacement vector
at each node is established by taking the perpendicular distance between the node and the closest touched segment.
The rest of the algorithm is similar to that of the Translation option. The following additional points should be
noted for the AUTOMATIC method:
• In single sided contact, touching body has to be selected as the interfering body.
• In general, use of analytical boundary description for the touched body is recommended while using this
method.
• For shell contact bodies, user needs to specify the top or bottom surface for detecting contact.
7. USHFTVEC User Subroutine
In this method, for each interfering body pair in the contact table, the user specifies the following while setting up
the job:
In the user-subroutine, user is provided with the co-ordinates and displacements of the boundary nodes of the body
and needs to return the initial pseudo movement (distance and direction) of the given node such that the body is
not interfering. This pseudo movement is then ramped down using the specified table. USHFTVEC is called at the
beginning of a loadcase.
Limitations of Interference Fit

 In Methods 2 to 5, the contact bodies cannot change due to global remeshing, local adaptivity, or element
activation/deactivation until the interference is fully resolved.
 Cannot be used for self contact.
 Cannot be used with beam-to-beam contact.
 In Methods 2 to 5, large rotations should not be present during the interference loadcase.
 Interference is not supported for non-mechanical analyses.
Symmetry Conditions
Symmetry conditions are supported via the contact option. It is defined as a special body which is defined by a curve in 2-
D and a surface in 3-D. This body extends to infinity independent of the size of the curve or surface. The behavior depends
on the physics pass in which the symmetry plane is used.
 For the structural pass, the nodes on this surface are not allowed to separate, but they can move frictionless.
 For Poisson type pass, the flux through this surface is zero.
 For the magnetostatic and magnetodynamic pass, a symmetry and antisymmetry condition is available.
• For a symmetry condition the magnetic vector potential is fixed normal to the symmetry plane, but free in the
symmetry plane.
CHAPTER 6 587
Contact
• For an antisymmetry condition the magnetic vector potential is fixed in the symmetry plane, but free normal
to the symmetry plane.
• For magnetodynamics the same type of symmetry condition is applied to the electric potential. For a symmetry
condition the flux through the surface is zero and for an antisymmetry condition the electric potential is fixed
at the surface. A typical symmetry condition will be applied to a plane in which a current flows. A typical
antisymmetry condition will be applied to a plane which is perpendicular to a flowing current. This is illustrated
in Figure 6-54, the right image shows the flow of a current. In the left image, the yellow plane is the symmetry
plane and the orange plane and the pink plane are the antisymmetry planes.
Figure 6-54 Symmetry and Antisymmetry planes for magnetostatics and magnetodynamics
Note that when in a coupled analysis an antisymmetry condition for the magnetostatic or magnetodynamic pass is defined,
it will be treated as a symmetry condition in the other physics passes.
References
1. Oden, J. T. and Pires, E. B. “Nonlocal and Nonlinear Friction Laws and Variational Principles for Contact
Problems in Elasticity,” J. of Applied Mechanics, V. 50, 1983.
2. Ju, J. W. and Taylor, R. L. “A perturbed Lagrangian formulation for the finite element solution of nonlinear
frictional contact problems,” J. De Mechanique Theorique et Appliquee, Special issue, Supplement, 7, 1988.
3. Simo, J. C. and Laursen, T. A. “An Augmented Lagranian treatment of contact problems involving friction,”
Computers and Structures, 42, 1992.
4. Peric, D. J. and Owen, D. R. J. “Computational Model for 3-D contact problems with friction based on the Penalty
Method,” Int. J. of Meth. Engg., V. 35, 1992.
5. Taylor, R. L., Carpenter, N. J., and Katona, M. G. “Lagrange constraints for transient finite element surface
contact,” Int. J. Num. Meth. Engg., 32, 1991.
6. Wertheimer, T. B., “Numerical Simulation Metal Sheet Forming Processes,” VDI BERICHTE, Zurich,
Switzerland, 1991.
7. Feng, Q. and Prinja, N.K., NAFEMS Benchmark Tests for Finite Element Modelling of Contact, Gapping and
Sliding, NAFEMS, 2001
Chapter 7: Introduction to Mesh Definition
7 Introduction to Mesh
Definition
 Overview 590

Direct Input 590

User Subroutine Input 597
 MESH2D 597
 Mentat 602

FXORD Option 602

Incremental Mesh Generators 607
 Bandwidth Optimization 608

Rezoning 608
 Substructure 609
 BEAM SECT Parameter 612

Error Analysis 619

Local Adaptivity 619
 Automatic Global Remeshing 625
Overview
This chapter describes the techniques for mesh definition available internally in Marc. Mesh definition is the process of
converting a physical problem into discrete geometric entities for the purpose of analysis. Before a body can undergo finite
element analysis, it must be modeled into discrete physical elements. An example of mesh definition is shown in Figure 7-1.
Figure 7-1 Finite Element Mesh of a Brake Rotor (click figure to activate 3D content)
Mesh definition encompasses the placement of geometric coordinates and the grouping of nodes into elements. For Marc
to have a valid mesh definition, the nodes must have geometric coordinates and must be connected to an element. It can
be done as follows:
1. Describe the element by entering the element number, the element type, and the node numbers that make up the
element.
2. Enter the physical coordinates of the nodal points.
Note: You need not enter element numbers and node numbers sequentially or consecutively.
Direct Input
You must enter two types of data into Marc for direct mesh definition as follows:
1. Connectivity data: This describes the nodal points for each element.
2. Coordinate data: This gives the spatial coordinates of each nodal point.
This section describes how to enter these data.
Element Connectivity Data

You can enter connectivity data from either the input option file (FORTRAN unit 5) or from an auxiliary file. Several
blocks of connectivity can be input. For example, the program can read one block from tape and subsequently read a block
CHAPTER 7 591
Introduction to Mesh Definition
from the input option file. Each block must begin with the word CONNECTIVITY. In the case of duplicate specification,
Marc always uses the data that was input last for a particular element.
Enter the nodal points of two-dimensional elements in a counterclockwise order. Figure 7-2 illustrates correct and incorrect
numbering of element connectivity data.
4 3 2 3
Y,R
1 2 1 4
X,Z
Correct Numbering Incorrect Numbering

Figure 7-2 Correct/Incorrect Numbering of Two-Dimensional Element Connectivity of 4-Node Elements
When there are eight nodal points on a two-dimensional element, number the corner nodes 1 through 4 in
counterclockwise order. The midside nodes 5 through 8 are subsequently numbered in counterclockwise order. Figure 7-3
illustrates the correct numbering of element connectivity of 8-node elements.
Y,R 4 7 3
X,Z 8 6
1 5 2
Figure 7-3 Numbering of Two-Dimensional Element Connectivity for 8-Node Quadrilateral Elements
Lower-order triangular elements are numbered using the counterclockwise rule.
X,R 3
X,Z
1 2
Figure 7-4 Numbering of 3-Node Triangular Element
Note that quadrilateral elements can be collapsed into triangular elements by repeating the last node.
The higher order triangular elements have six nodes, the corner nodes are numbered first in a counterclockwise direction.
The midside nodes 4 through 6 are subsequently numbered as shown in Figure 7-5.
6 5
1 4 2
Figure 7-5 Numbering of 6-Node Triangular Element
Number three-dimensional elements in the same order as two-dimensional elements for each plane. Enter nodes for an 8-
node brick in counterclockwise order as viewed from inside the element. First, enter nodes comprising the base; then enter
ceiling nodes as shown in Figure 7-6.
CHAPTER 7 593
8 7
5 6
Z
4 3
X 1 2
Figure 7-6 Numbering of Element Connectivity for 8-Node Brick
A 20-node brick contains two 8-node planes and four nodes at the midpoints between the two planes. Nodes 1 through 4
are the corner nodes of one face, given in counterclockwise order as viewed from within the element. Nodes 5 through 8
are on the opposing face; nodes 9 through 12 are midside nodes on the first face, while nodes 13 through 16 are their
opposing midside nodes. Finally, nodes 17 through 20 lie between the faces with node 17 between 1 and 5. Figure 7-7
illustrates the numbering of element connectivity for a 20-node brick.
8 15 7
16 14
13 20 19
5 6
Z
4 11 3
17 18
12
Y 10
1 9 2
X
Figure 7-7 Numbering of Three-Dimensional Element Connectivity for 20-Node Brick
The four node tetrahedral is shown in Figure 7-8.

1 2
Figure 7-8 Numbering of Four-Node Tetrahedral
The ten-node tetrahedral is shown in Figure 7-9. The corner nodes 1-4 are numbered first. The first three midside nodes
occur on the first face. Nodes 8, 9, and 10 are between nodes 1 and 4, 2 and 4, and 3 and 4, respectively.
10
8 9 3
7 6
1 5 2
Figure 7-9 Numbering of 10-Node Tetrahedral
The 6-node pentahedral element is shown in Figure 7-10.
5
3
Figure 7-10 6-Node Pentahedral
The 15-node pentahedral element is shown in Figure 7-11.

CHAPTER 7 595
3
15
6
9 8
12 11
1 7 2
13 14
4 10 5
Figure 7-11 15-Node Pentahedral
The 5-node pyramid element is shown in Figure 7-12.
Figure 7-12 5-Node Pyramid element

The 13-node pyramid element is shown in Figure 7-13
Figure 7-13 13-Node Pyramid element
Nodal Coordinate Data

You can enter nodal coordinates directly from the input option file (FORTRAN unit 5) or from an auxiliary file. You can
enter several blocks of nodal coordinate data in a file. In the case of duplicate specifications, the program uses data entered
last for a particular nodal point in the mesh definition.
Direct nodal input can be used to input local corrections to a previously generated set of coordinates. These options give
the modified nodal coordinates.
The CYLINDRICAL option can be used to transform coordinates given in a cylindrical system to a Cartesian system.
Note: Requires the final coordinate data in terms of a single Cartesian system. Refer to Marc Volume B: Element Library to
determine the required coordinate data for a particular element type.
Activate/Deactivate
You have the ability to turn on and off elements using this option, which is useful when modeling ablation or excavation.
When you enter the mesh connectivity, the program assumes that all elements are to be included in the analysis unless they
are deactivated. This effectively removes this material from the model. These elements can be reinstated later by using the
ACTIVATE option. If the element is activated, one can select if the level of stress is to be reinstated or set to zero. The use of
these options results in nonlinear behavior and have an effect upon convergence.
CHAPTER 7 597
User Subroutine Input

User subroutines can be used to generate or modify the data for mesh definition. User subroutine UFCONN generates or
modifies element connectivity data. The UFCONN model definition option activates this subroutine. The user subroutine is
called once for each element requested. Refer to Volume D: User Subroutines and Special Routines for a description of the
UFCONN user subroutine and instructions for its use.
The UFXORD user subroutine generates or modifies the nodal coordinates. The UFXORD model definition option activates this
subroutine. The user subroutine is called once for each node requested. Refer to Volume D: User Subroutines and Special
Routines for a description of the UFXORD user subroutine and instructions for its use.
MESH2D
MESH2D generates a mesh of quadrilateral or triangular elements for a two-dimensional body of any shape. The generated
mesh is written to a separate file and must be read with the CONNECTIVITY, COORDINATES, and FIXED DISP, etc., options.
Block Definition
In MESH2D, a physical object or domain is divided into quadrilateral and/or triangular parts, called blocks. Quadrilateral
blocks are created by Marc by mapping with polynomials of the third order from a unit square. These blocks can, therefore,
be used to approximate curved boundaries. The geometry of a quadrilateral block is defined by the coordinates on 12 nodes
shown in Figure 7-14. If the interior nodes on an edge of the block are equal to zero or are not specified, the edge of the
block is straight.
Triangular blocks have straight edges. The geometry of a triangular block is defined by the coordinates of the three vertices.
Merging of Nodes
Marc creates each block with a unique numbering scheme. The MERGE option fuses all nodes that lie within a small circle,
renumbers the nodes in sequence, and then removes all gaps in the numbering system. You can select which blocks are to
be merged together, or you can request that all blocks be merged. You must give the closeness distance for which nodes will
be merged.
 P3
P9
y P4 P10
4 10 9 3 P8
P11 P7
11 8 P12 P2
2  P6
P1
P5
12 7 x
1 5 6 2 12
x   ,  =  x  P i  i   , 
i = 1
2
12
y   ,  =  y  P i  i   , 
i = 1
Figure 7-14 Typical Quadrilateral Block Mapping
Block Types
MESH2D generates four types of quadrilateral blocks.
1. Block Type 1:
Is a quadrilateral block that is covered by a regular grid. The program obtains this grid by dividing the block edge
into M by N intervals. Figure 7-15 illustrates the division of block edges into intervals with M = 4, N = 3.
The P1 P4 face of the block is the 1-4 face of triangular elements and the P1 P2 face of the block becomes the 1-
2 face of quadrilateral elements.
CHAPTER 7 599


P4 P3
16 17 18 19 20
24 16 17 18 19 20
2/3
17 12 13 14
2/3 9 10 11 12
11 15
N = 3 2/3 16 11 12 13 14 15

9 7 8 9
N = 3 2/3 5 6 7 8 
6 10
2 4 5 8 6 7 8 9 10
2/3 P2
1 3 6 7 2/3 1 2 3 4
1 2 3 4 5 1 2 3 4 5
P1
1/2 1/2 1/2 1/2 1/2 1/2 1/2 1/2
M=4 M=4
Triangular Quadrilateral
Figure 7-15 Block Type 1
2. Block Type 2:
Is a quadrilateral block that allows the transition of a coarse mesh to a finer one. In one direction, the block is
divided into M, 2M, 4M ...; while in the other direction, the block is divided into N intervals. Figure 7-16
illustrates the division of Block Type 2 edges into intervals.
The P1 P2 face of the block becomes the 1-2 face of quadrilateral elements, and the P2 P3 face of the block is the
2-3 face of triangular elements.

P4 P3
19
18 34 
2/7
9 17 1/2 1/2 1/2 1/2
7 10 11 12 13 14 15 16
4/7
8 7 8 9 10 11
M=2 5 6 7
4 8
N=3 x 1 3 4 6
1 3 4 6
4 5 6 x
3/7 2
2 5 2 5
1 2 3
P1 1 2 3 P2
1 1
1 1

3. Block Type 3:
Is a triangular block. The program obtains the mesh for this block by dividing each side into N equal intervals.
Figure 7-17 illustrates Block Type 3 for triangular and quadrilateral elements.
P3 P3
15 10
13
16
10 6
14 8 9
P1 6 11
8
1 2 7 12 4 5
1 6 7
3 4 8 5
2
9 P1 1 2 3
3 P2
7
4 P2
1 2 3 4
N=4 5
4. Block Type 4:
Is a refine operation about a single node of a block (Figure 7-18). The values of N and M are not used.
If quadrilateral elements are used in a triangular block, the element near the P2 P3 face of the block is collapsed by
MESH2D in every row. The P1 P2 face of the block is the 1-2 face of the generated elements.
5 6 7
4 3
4
3
1 2
1 2
Symmetry, Weighting, and Constraints

MESH2D contains several features that facilitate the generation of a mesh: use of symmetries, generation of weighted
meshes, and constraints. These features are discussed below.
MESH2D can use symmetries in physical bodies during block generation. An axis of symmetry is defined by the
coordinates of one nodal point, and the component of a vector on the axis. One block can be reflected across many axes to
form the domain. Figure 7-19 illustrates the symmetry features of MESH2D.
CHAPTER 7 601
2
1 1
Original Block One Symmetry Axis
3 2 3 2
4 1 4 1
5 8
6 7
Two Symmetry Axes Three Symmetry Axes

Figure 7-19 Symmetry Option Example
A weighted mesh is generated by the program by spacing the two intermediate points along the length of a boundary. This
technique biases the mesh in a way that is similar to the weighting of the boundary points. This is performed according to
the third order isoparametric mapping function.
Note: If a weighted mesh is to be generated, be cautious not to move the interior boundary points excessively. If the
points are moved more than 1/6 of the block length from the 1/3 positions, the generated elements can turn
inside-out.
The CONSTRAINT option generates boundary condition restraints for a particular degree of freedom for all nodes on one side
of a block. The option then writes the constraints into the file after it writes the coordinate data. The FIXED DISP, etc., option
must be used to read the boundary conditions generated from the file.
Additional Options
Occasionally, you might want to position nodes at specific locations. The coordinates of these nodes are entered explicitly
and substituted for the coordinates calculated by the program. This is performed using the SPECIFIED NODES option. Some
additional options in MESH2D are as follows:
 MESH2D can be used several times within one input file.
 The START NUMBER option gives starting node and element numbers.
 The CONNECT option allows forced connections and/or disconnections with other blocks. This option is useful
when the final mesh has cracks, tying, or gaps between two parts.
 The MANY TYPES option specifies different element types.
Mentat
Mentat is an interactive program which facilitates mesh definition by generating element connectivity and nodal
coordinates. Some of the Mentat capabilities relevant to mesh generation are listed as follows:
 Prompts you for connectivity information and nodal coordinates. Accepts input from a keyboard or mouse.
 Accepts coordinates in several coordinate systems (Cartesian, cylindrical, or spherical).
 Translates and rotates (partial) meshes.
 Combines several pre-formulated meshes.
 Duplicates a mesh to a different physical location.
 Generates a mirror image of a mesh.
 Subdivides a mesh into a finer mesh.
 Automatic mesh generation in two- and three-dimensions.
 Imports geometric and finite element data from CAD systems.
 Smooths nodal point coordinates to form a regular mesh
 Converts geometric surfaces to meshes.
 Refines a mesh about a point or line.
 Expands line mesh into a surface mesh, or a surface mesh into a solid mesh.
 Calculates the intersection of meshes.
 Maps nodal point coordinates onto prescribed surfaces.
 Writes input data file for connectivity and coordinates in Marc format for use in future analyses.
 Apply boundary conditions to nodes and elements.
 Define material properties.
 Submit Marc jobs.
FXORD Option
The FXORD model definition option (Volume C: Program Input) generates doubly curved shell elements of element type 4,
8, or 24 for the geometries most frequently found in shell analysis. Since the mathematical form of the surface is well-
defined, the program can generate the 11 or 14 nodal coordinates needed by element type 8, 24, or 4 to fit a doubly curved
surface from a reduced set of coordinates. For example, you can generate an axisymmetric shell by entering only four
coordinates per node. The FXORD option automatically generates the complete set of coordinates required by the elements
in the program from the mathematical form of the surface.
A rotation and translation option is available for all components of the surface to give complete generality to the surface
generation. The input to FXORD consists of the reduced set of coordinates given in a local coordinate system and a set of
CHAPTER 7 603
coordinates which orient the local system with respect to the global system used in the analysis. The program uses these
two sets of coordinates to generate a structure made up of several shell components for analysis.
The FXORD option allows for the generation of several types of geometries. Because you may need to analyze shells with well-
defined surfaces not available in this option, you can use the UFXORD user subroutine to perform your own coordinate
generation (Volume D: User Subroutines and Special Routines). The FXORD option can also be used to convert cylindrical
coordinates or spherical coordinates to Cartesian coordinates for continuum elements.
Major Classes of the FXORD Option

The cases are considered are as follows:
 Shallow Shell (Type I)
 Axisymmetric Shell (Type 2)
 Cylindrical Shell Panel (Type 3)
 Circular Cylinder (Type 4)
 Plate (Type 5)
 Curved Circular Cylinder (Type 6)
 Convert Cylindrical to Cartesian (Type 7)
 Convert Spherical to Cartesian (Type 8)
Shallow Shell (Type I)

Type 1 is a shallow shell with:
 1 = x 1 2 = x2 (7-1)
The middle surface of Figure 7-20 (Type 1) is defined by an equation of the form:
x 3 = x 3  x 1 x 2  (7-2)
and the surface is determined when the following information is given at each node:
x 3 x 3  2 x 3
x 1 x 2 x 3 --------- --------- ------------------ (7-3)
x 1 x 2 x 1 x 2
The last coordinate is only necessary for Element Type 4.

X3
X3
R
f
X2
X2
q
X1
Type 1 Type 2
X3
X2
R
X2 
3
X1
X3
X1
Type 4
X3
Type 3
Figure 7-20 Classification of Shells
Axisymmetric Shell (Type 2)

The middle surface symmetric to the x 3 axis (Figure 7-20, Type 2) is defined as:
x 1 = R    cos  cos 
x 2 = R    cos  sin 
(7-4)
x 3 = R    sin 
where  and  are the angles shown in Figure 7-20. In this case, the surface is defined by:
dR
  R ------- (7-5)
d
The angles  and  are given in degrees.

CHAPTER 7 605
Cylindrical Shell Panel (Type 3)

The middle surface is the cylinder defined by Figure 7-20:
x1 = x1  s 
x2 = x2  s  (7-6)
x3 = x3
The nodal geometric data required is:

dx 1 dx 2
s x 3 x 1 x 2 --------- --------- (7-7)
ds ds
Circular Cylinder (Type 4)

This is the particular case of Type 3 where the curve,
x 1  s  x 2  s  (7-8)
is the circle given by Figure 7-20 (Type 4).
x 1 = R cos 
(7-9)
x 2 = R sin 
The only nodal information is now:

 x 3 R (7-10)
Note that  is given in degrees and, because R is constant, it needs to be given for the first nodal point only.
Plate (Type 5)
The shell is degenerated into the plate,
x3 = 0 (7-11)
The data is reduced to,

x 1 x 2 (7-12)
Curved Circular Cylinder (Type 6)

Figure 7-21 illustrates this type of geometry.
X3
Shell Middle Surface
q1
q2 R
f
X2
q
X1
Type 6
Figure 7-21 Curved Circular Cylinder
The middle surface of the shell is defined by the equations as:

x 1 = r cos 
x 2 = r sin  cos  + R  1 – cos   (7-13)
x 3 =  R + – r sin  r sin   sin 
The Gaussian coordinates on the surface are:

 1 = r
(7-14)
 2 = R
and form an orthonormal coordinate system. The nodal point information is:
  r R (7-15)
 and  in degrees. You need to specify the radii r and R only for the first nodal point.
Convert Cylindrical to Cartesian (Type 7)

Type 7 allows you to enter the coordinates for continuum elements in cylindrical coordinates, which are converted by Marc
to Cartesian coordinates. In this way, you can enter R ,  , Z and obtain x, y, z where  is given in degrees and,
x = R cos 
y = R sin  (7-16)
z = Z
CHAPTER 7 607
Convert Spherical to Cartesian (Type 8)

Type 8 allows you to enter the coordinates for continuum elements in spherical coordinates, which are converted by Marc
to Cartesian coordinates. In this way, you can enter R ,  ,  and obtain x, y, z where  and  are given in degrees and,
x = R cos  sin 
y = R sin  cos  (7-17)
z = R cos 
Recommendations on Use of the FXORD Option

When a continuous surface has a line of discontinuity, for example, a complete cylinder at  = 0 = 360 , you must
place two nodes at each nodal location on the line to allow the distinct coordinate to be input. You must use tying element
type 100 to join the degrees of freedom. Generally, when different surfaces come together, you must use the intersecting
shell tyings.
The FXORD option cannot precede the COORDINATES option, because it uses input from that option.
Incremental Mesh Generators

Incremental mesh generators are a collection of options available in Marc to assist you in generating the mesh. Incremental
mesh generators generate connectivity lists by repeating patterns and generate nodal coordinates by interpolation. Use these
options directly during the model definition phase of the input.
During the model definition phase, you can often divide the structure into regions, or blocks, for which a particular mesh
pattern can be easily generated. This mesh pattern is established for each region and is associated with a single element
connectivity list. Use the CONNECTIVITY option to input this element connectivity list. The incremental mesh generators then
generate the remainder of the connectivity lists.
Critical nodes define the outline of the regions to be analyzed. Use the COORDINATES option to enter the critical nodes. The
incremental mesh generators complete the rest and join the regions by merging nodes.
A special connectivity interpolator option generates midside nodes for elements where these nodes have not been specified
in the original connectivity. A separate mesh generation run is sometimes required to determine the position of these nodes.
This run can be followed by mesh display plotting.
The incremental mesh generators are listed below:
Element Connectivity Generator – The CONN GENER option repeats the pattern of the connectivity data for
previously defined master elements. One element can be removed for each series of elements, allowing the
program to generate a tapered mesh. Two elements can be removed for each series with triangular elements.
Element Connectivity Interpolator – The CONN FILL option completes the connectivity list by generating midside
nodes. You first generate the simpler quadrilateral or brick elements without the midside nodes. You can then fill
in the midside nodes with this option.
Coordinate Generator – The NODE GENER option creates a new set of nodes by copying the spacing of another
specified set of nodes.
Coordinate Interpolator – The NODE FILL generates intermediate nodes on a line defined by two end nodes. The
spaces between the nodes can be varied according to a geometric progression.
Coordinate Generation for Circular Arcs – The NODE CIRCLE option generates the coordinates for a series of
nodes which lie on a circular arc.
Bandwidth Optimization
Marc can minimize the nodal bandwidth of a structure in several ways. The amount of storage is directly related to the size
of the bandwidth, and the computation time increases in proportion to the square of the average bandwidth.
The OPTIMIZE option allows you to choose from several bandwidth optimization algorithms.
The four available OPTIMIZE options are listed in Table 7-1.
Note: This option creates an internal node numbering that is different from your node numbering. Use your node
numbering for all inputs. All output appears with your node numbering. The occurrence of gap or Herrmann
elements can change the internal node numbers. On occasion, this change can result in a non optimal node
numbering system, but this system is necessary for successful solutions.
Table 7-1 Bandwidth Optimization Options

Option Number Remarks Solver
2 Cuthill-McKee algorithm Profile (0)
9 Sloan (Recommended) Profile (0)
10 Minimum Degree Algorithm Pardiso (11), MUMPS (12)
11 METIS (Recommended)* Multifrontal (8), Direct-Iterative (10), Pardiso (11), MUMPS (12)
*Reference “A Fast and Highly Quality Multilevel Scheme for Partitioning Irregular Graphs”, George Karypis and Vipin
Kumar, SIAM Journal on Scientific Computing, Vol. 20, No. 1, pp. 359-392, 1999. METIS documentation may be
downloaded from http://glaros.dtc.umn.edu/gkhome/fetch/sw/metis/manual.pdf
The nodal correspondence obtained through this process can be saved and then used in subsequent analyses. This
eliminates the need to go through the optimization step in later analyses. The correspondence is used to relate the
user-defined node (external) numbers to the program-optimized (internal) node numbers and vice versa.
Rezoning
The REZONING parameter defines a new mesh and transfers the state of the old mesh to the new mesh. Elements or nodes
can be either added to or subtracted from the new mesh. This procedure requires a sub increment to perform the definition
of the new mesh. The rezoning capability can be used for two- and three-dimensional continuum elements and for shell
elements 22, 75, 138, 139, and 140. Rezoning is not available for Pyramid elements. See Figure 7-22 for an example of
rezoning.
CHAPTER 7 609
Rezoning or remeshing can be carried automatically using the mesh generators. This feature is described in detail in the
Automatic Global Remeshing section.
Before Rezoning After Rezoning

Figure 7-22 Mesh Rezoning
Substructure
Marc is capable of multilevel substructuring that includes:
 Generation of superelements
 Use of superelements in subsequent Marc analyses
 Recovery of solutions (displacements, stresses, and strains) in the individual substructures
The Marc multilevel procedure allows superelements to be used. One self-descriptive database stores all data needed during
the complete analysis. You only have to ensure this database is saved after every step of the analysis.
The advantages of substructuring are the following:
 Separates linear and nonlinear parts of the model
 Allows repetition of symmetrical or identical parts of the model for linear elastic analysis
 Separates large models into multiple, moderate-size models
 Separates fixed model parts from parts of the model that may undergo design changes
A disadvantage of substructuring is the large amount of data that must be stored on the database.
Three steps are involved in a substructuring run as follows:
1. The superelement generation step is done for every superelement at a certain level.
Note: The SUPERELEM model or history definition option may be used to create a reduced matrix. As opposed
to many linear programs, if the history definition is used, the reduced matrix includes the nonlinear
effects such as initial stress stiffening and material behavior. This file is written to a file
named .jid.dmigst_xxxx, where xxxx is the increment number where the reduction occurred.
But when these reduced matrices are used in a subsequent analysis, they are treated as linear or
constant. The interface or external nodes are defined through the SUPERELEM option. Note that
superelements can only be created using the multifrontal direct solver, solver type 8 or the Pardiso
solver, solver type 11.
2. The use of superelements in subsequent Marc runs is done at the highest level, or is incorporated into Step 1 for
the intermediate levels.
Note: To utilize a previously created superelement, use the DMIG option. One can have multiple DMIGs in the
model pointing to multiple DMIG files. A previously created superelement (DMIG) must be activated
using the K2GG option. In Marc, a superelement may under go rigid body rotation if a control node
(six degrees of freedom) or three control nodes forming a triad (three degrees of freedom) is given on
the K2GG option. In order to facilitate the recovery of displacements and stresses in the interior of the
superelement in a subsequent run, the interface DOF solution is stored by default in a file
JIDNAME.t31 where JIDNAME is the name of the current job. When used in conjunction with the
FEATURE,22901 parameter, the -sid option can be specified to store the interface DOF solution in
SIDNAME.t31.
3. Recovery of solutions within a certain superelement can be done by specifying the interface DOF database.
Note: To recover the displacement and stress solution, one should use the BACKTOSUBS option and identify
the database using the -sid command line parameter. Only one superelement stress recovery can be
done per job. It is possible to utilize a previously created superelement via K2GG while doing stress
recovery for another superelement via BACKTOSUBS. In this case, the interface DOF solution for
superelement recovery is read from SIDNAME.t31 in conjunction with -sid while the interface DOF
solution for superelement assembly is written by default to JIDNAME.t31.
CHAPTER 7 611
Substructuring in Marc is only possible for static analysis. Nonlinearities are not allowed with a superelement. You, as a
user, must ensure that nonlinearities are not present.
The maximum number of levels in a complete analysis is 26. The maximum number of substructures in the complete
analysis is 676.
Each step can be done in an individual run, or an unlimited number of steps can be combined into a single run.
During superelement generation in Marc, you can generate a complete new superelement or you can copy a previously
defined superelement with identical or newly defined external load conditions. Any number of superelements can be
formed in a generation run.
Marc offers flexibility in the use of superelements by allowing rotation or mirroring of a superelement. If a run is nonlinear,
superelements are treated as linear elastic parts. At every increment, you can perform a detailed analysis of certain
substructures by descending down to the desired superelements. Use the normal Marc control algorithm (AUTO INCREMENT,
AUTO LOAD, PROPORTIONAL INCREMENT) to control the load on the superelements.
The system of equations for a linear static structure is:
Ku = P (7-18)
When local degrees of freedom (subscripted l ) and external degrees of freedom (subscripted e ) are considered, this can
be rewritten as:
K l l K el  u l   P 
 l 
= (7-19)
K l e K ee  u e   P 
 e 
To obtain both the stiffness matrix and the load vector of the substructure, it is necessary to eliminate u 1 and rewrite the
above system with u e as the only unknown splitting the above equation.
K l l u l + K el u e = P l
and (7-20)
K l e u l + K ee u e = P e
The first equation can be written as:
u l = – K ll– 1  K el u e + K l–l1 P l (7-21)
Substituting this equation into the second:
– K l e K ll– 1 K el u e + K le K l–l1 P l + K e e u e = P e (7-22)

This can be rewritten as:
K e*e u e = P e* (7-23)
where,
–1
K e*e = K ee – K le K ll K e l (7-24)
and,
–1
P e* = P e – K le K ll P l (7-25)
K e*e and P e* are solved by the triangularization of K l l , the forward and backward substitution of K el and P l ,
respectively, and premultiplication with K l e , K e*e , and P e* are used in the next part of the analysis with other substructures
or with another element mesh. That analysis results in the calculation of ue.
You can now calculate the local degrees of freedom and/or the stresses using the following procedure:
K ll  u l = P l – K el u e (7-26)
which can be written as:

–1
u l = K ll  P l* (7-27)
where,
P l* = P e – K el u e (7-28)
The displacement of the substructure is, therefore, known, and stresses and strains can be calculated in the normal way.
Scaling Element Stiffness

Occasionally, it is desirable to perform a scalar multiplication of the stiffness, mass, and load matrix to represent a selective
duplication of the finite element mesh.
The STIFSCALE option can be used to enter the scaling factor for each element. In this case, the global stiffness, mass, and
load matrices are formed as follows:
K g = s i K iel M g = s i M ie l and F g = s i f iel (7-29)
Note that no transformation of the stiffness matrix occurs and that point loads are not scaled.
BEAM SECT Parameter

The BEAM SECT parameter inputs data to define the sectional properties for three-dimensional beam elements. Include this
option if you are using:
CHAPTER 7 613
 Element types 13, 77, or 79

 Element types 14, 25, 76, or 78 with a noncircular section
 Element types 52 or 98 and torsional and shear stiffness must be defined independently
 Element types 52 or 98 with numerical integrated solid cross-section
The convention adopted for the local (beam) coordinate system is: the first and second director (local X and Y) at a point
are normal to the beam axis; the third director (local Z) is tangent to the beam axis and is in the direction of increasing
distances along the beam. The director set forms a right-handed system.
Orientation of the Section in Space

The beam axis in an element is interpolated from the two nodes of the element.
dx dy dz
x y z ------ ------ ------ (7-30)
ds ds ds
where the last three coordinates are only used for element 13.
The beam section orientation in an element is defined by the direction of the first director (local X) at a point, and this
direction is specified via the coordinates of an additional node or through the GEOMETRY option (see Marc Volume C:
Program Input).
Definition of the Section

You can include any number of different beam sections in any problem. Data options following the BEAM SECT parameter
of the Marc input (see Marc Volume C: Program Input) define each section. The program numbers the sections in the order
they are entered. To use a particular section for a beam element, set EGEOM2 (GEOMETRY option, 3rd data block, columns
11-20) to the floating-point value of the section number, for example, 1, 2, or 3. The program uses the default circular
section for the closed section beam elements (14, 25, 76, 78) if EGEOM1 is nonzero. The program uses the default solid
rectangular cross section for elements 52 or 98 if EGEOM1 is nonzero. Figure 7-23 shows how the thin-walled section is
defined using the input data.
Section 1 Section 2
10 xl 8
xl
5 4 6 3 4
yl yl
16
20
3 1 2
2 1
8
Branch Divisions Thickness Branch Divisions Thickness

1-2 8 1.0 1-2 10 1.0
2-3 4 0.0 2-3 10 1.0
3-4 8 0.3 3-4 10 1.0
4-5 4 0.0
5-6 8 1.0
Section 3 Section 4
XI
1 6 8
R=15
10
R=5
2 1 3 4 R=5 7
60° 5 YI
R=5
17.85
2 6
3 4 Branch Divisions Thickness
xl 1-2 4 0.2
2-3 4 0.2
yl 3-4 4 0.0
Branch Divisions Thickness 4-5 4 0.2
5-3 4 0.2
1-2 8 0.5 3-4 4 0.0
2-3 4 0.5 4-6 4 0.2
3-4 8 0.5 6-7 4 0.2
4-5 4 0.5 7-8 6 0.2
5-6 8 0.5 8-1 6 0.2
Figure 7-23 Beam Section Definition Examples for Thin-walled Sections
CHAPTER 7 615
The rules and conventions for defining a thin-walled section are listed as follows:
1 1 1
1. An x – y coordinate system defines the section, with x the first director at a point of the beam. The origin of
1 1
the x – y system represents the location of the node with respect to the section.
2. Enter the section as a series of branches. Branches can have different geometries, but they must form a complete
traverse of the section in the input sequence so that the endpoint of one branch is the start of the next branch. It is
often necessary for the traverse of the section to double back on itself. To cause the traverse to do this, specify a
branch with zero thickness.
3. You must divide each branch into segments. The stress points of the section are the branch division points. The
stress points are the points used for numerical integration of a section’s stiffness and for output for stress results.
Branch endpoints are always stress points. There must always be an even number of divisions (nonzero) in any
branch. Not counting branches of zero thickness, you can use a maximum of 99 stress points (98 divisions) in a
complete section.
4. Branch thickness varies linearly between the values given for branch endpoint thickness. The thickness can be
discontinuous between branches. A branch is assumed to be of constant thickness equal to the thickness given at
the beginning of the branch if the thickness at the end of the branch is given as an exact zero.
1 1
5. The shape of a branch is interpolated based on the values of x and y and their directions, in relation to distance
1 1
along the branch. The data is input at the two ends of the branch. If both dx  ds and dy  ds are given as exact
zeros at both ends of the branch, the branch is assumed to be straight. The section can have a discontinuous slope
at the branch ends. The beginning point of one branch must coincide with the endpoint of the previous branch.
1 1
As a result, x and y for the beginning of a branch need to be given only for the first branch of a section.
6. Stress points are merged into one point if they are separated by a distance less than t  10 , where t is a thickness at
one of these points.
7. Arcs must be defined in a counterclockwise direction.
Figure 7-23 shows four example sections of a beam. Notice the use of zero thickness branches in the traverse of the I section.
Section 4 shows an example of multi-direction arcs; in which case, the zero thickness branches can be used to overcome the
restriction on clockwise arcs. The program provides the following data: the location of each stress point in the section, the
thickness at that point, the weight associated with each point (for numerical integration of the section stiffness), and the
warping function at each section.
Solid sections can be defined by using one of the standard sections and specifying its typical dimensions or by using
quadrilateral segments and specifying the coordinates of the four corner points of each quadrilateral segment in the section.
The local axes of a standard section are always the symmetry axes. For the standard sections, the symmetry axes are also the
principal axes. The principal axes for each section are shown in Figure 7-24. For the more general sections, it is not required
to enter the coordinates of the corner points of the quadrilateral segments with respect to the principal axes. The section is
automatically realigned and you can make specifications about how this realignment should be carried out. The details
about these orientation methods are described in the Cross-section Orientation and Location of the Local-section Axis sections in
Marc Volume B: Element Library .
Figure 7-24 shows the standard solid sections, the dimensions needed to specify them, and the orientation of their local axes.
If the dimension b is omitted for an elliptical section, the section will be circular. If the dimension b is omitted for a
rectangular section, the section will be square. If the dimension c is omitted for a trapezoidal or a hexagonal section, the
section will degenerate to a triangle or a diamond.
y y
x x
b b
a a
Elliptical Section Rectangular Section
Circular (set b = 0) Square (set b = 0)
y y
c c
[2(a+c)b]/[3(a+c)]
x
b b
x
a a
Trapezoidal Section Hexagonal Section
Figure 7-24 Standard Solid Section Types with Typical Dimensions
Figure 7-25 shows the default integration point locations and numbers for standard solid section types. For the elliptical
section, the center point is the first integration point. The second is located on the negative y-axis and the points are
numbered radially outward and then counterclockwise. For the rectangular, trapezoidal and hexagonal sections the first
integration point is nearest to the lower left corner and the last integration point is nearest to the upper right corner. They
are numbered from left to right and then from bottom to top. The default integration scheme for all standard solid sections
uses 25 integration points.
CHAPTER 7 617
Elliptical Section Rectangular Section
Trapezoidal Section
Hexagonal Section
Figure 7-25 Default Integration Point Locations and Numbers for Standard Solid Section Types
The rules and conventions for defining a general section solid are listed as follows:
1. Each quadrilateral segment is defined by entering the coordinates of the four corner points in the local xy-plane.
The corner points must be entered in a counterclockwise sense.
2. The quadrilateral segments must describe a simply connected region; i.e., there must be no holes in the section, and
the section cannot be composed of unconnected “islands” and should not be connected through a single point.
3. The quadrilateral segments do not have to match at the corners. Internally, the edges must, at least, partly match.
4. The segments must not overlap each other (i.e., the sum of the areas of each segment must equal the total area of
the segment).
5. Each section can define its own integration scheme and this scheme applies to all segments in the section.
6. A section cannot have more than 100 integration points in total. This limits each section to, at most, 100 segments
that, by necessity, will use single point integration. If less than 100 segments are present in a section, they may use
higher order integration schemes, as long as the total number of points in the section does not exceed 100.
7. Integration points on matching edges are not merged, should they coincide.
This method allows you to define an arbitrary solid section in a versatile way, but you must make sure that the conditions
of simply connectedness and nonoverlap are met. The assumptions underlying the solid sections are not accurate enough
to model sections that are thin-walled in nature. It is not recommended to use them as an alternative to thin-walled sections.
Solid sections do not account for any warping of the section and, therefore, overestimate the torsion stiffness as it is
predicted by the Saint- Venant theory of torsion.
The first integration point in each quadrilateral segment is nearest to its first corner point and the last integration point is
nearest to its third corner point. They are numbered going from first to second corner point and then from second to third
corner point. The segments are numbered consecutively in the order in which they have been entered and the first
integration point number in a new segment simply continues from the highest number in the previous segment. There is
no special order requirement for the quadrilateral segments. They must only meet the previously outlined geometric
requirements. For each section, the program provides the location of the integration points with respect to the principal
axes and the integration weight factors of each point. Furthermore, it computes the coordinates of the center of gravity and
the principal directions with respect to the input coordinate system. It also computes the area and the principal second
moments of area of the cross section. For each solid section, all of this information is written to the output file.
All solid sections can be used in a pre-integrated fashion. In that case, the area A and the principal moments of area Ixx and
Iyy are calculated by numerical integration. The torsional stiffness is always computed from the polar moment of inertia
J=Ixx+Iyy. The shear areas are set equal to the section area A. The stiffness behavior may be altered by employing any of the
stiffness factors. In a pre-integrated section, no layer information is computed and is, therefore, not available for output.
Pre-integrated sections can only use elastic material behavior and cannot account for any inelasticity (e.g. plasticity). For
pre-integrated sections there is no limit on the number of segments to define the section.
Figure 7-26 shows a general solid section built up from three quadrilateral segments using a 2 x 2 Gauss integration scheme.
It also shows the numbering conventions adopted in a quadrilateral segment and its mapping onto the parametric space.
For each segment in the example, the first corner point is the lower-left corner and the corner points have been entered in
a counterclockwise sense. The segment numbers and the resulting numbering of the integration points are shown in the
figure. Gauss integration schemes do not provide integration points in the corner points of the section. If this is desired,
you can choose other integration schemes like Simpson or Newton-Cotes schemes. In general, a 2 x 2 Gauss or a 3 x 3
Simpson scheme in each segment suffices to guarantee exact section integration of linear elastic behavior.
y (x3,y3)
K
11 12 (x4,y4) (-1,1) (1,1)
3
9 10
7 8 x 3 4
2 [
5 6
1 2
(x2,y2)
3 4
1
1 2 (-1,-1) (1,-1)
(x1,y1)
Figure 7-26 Beam Section Definition for Solid Section plus Numbering Conventions Adopted
CHAPTER 7 619
Error Analysis
You can determine the quality of the analysis by using the ERROR ESTIMATE option.
The ERROR ESTIMATE option can be used to determine the mesh quality (aspect ratio, and skewness), and how they change
with deformation. While all of the Marc elements satisfy the patch test, the accuracy of the solution often depends on
having regular elements. In analyses where the updated Lagrangian method is used, the mesh often becomes highly
distorted during the deformation process. This option tells you when it would be beneficial to perform a rezoning step.
This option can also be used to examine the stress discontinuity in the analysis. This is a measure of the meshes ability to
represent the stress gradients in the problem. Large stress discontinuities are an indication that the mesh is not of sufficient
quality. This can be resolved by increasing the number of elements or choosing a higher order element or using local
adaptive meshing.
Local Adaptivity
The adaptive mesh generation capability increases the number of elements and nodes to improve the accuracy of the
solution. The capability is applicable for both linear elastic analysis and for nonlinear analysis. The capability can be used
for lower-order elements, 3-node triangular solids and shells, 4-node quadrilateral solids and shells, 4-node tetrahedrals, 8-
node hexahedral elements, and the 8-node solid-shell elements. Local adaptive meshing cannot be used with the triangular
and tetrahedral Herrmann elements, type 155 - 157. Local adaptive meshing can also not be used with the improved
integration triangular and tetrahedral elements (type 239 - 241).
When used in conjunction with the ELASTIC parameter for linear analysis, a steady-state heat transfer, electrostatic, or
magnetostatic, the mesh is adapted and the analysis repeated until the adaptive criteria is satisfied. When used in a nonlinear
analysis, an increment is performed. If necessary, this increment is followed by a mesh adjustment which is followed by the
analysis of the next increment in time. While this can result in some error, as long as the mesh is not overly coarse, it should
be adequate.
Local adaptivity is available in a parallel analysis. In this case, the new elements remain in the same domain as the parent
element.
Number of Elements Created

The adaptive meshing procedure works by dividing an element and internally tying nodes to insure compatibility.
Figure 7-27 shows the process for a single quadrilateral element.
Original Element Level 1 Refinement Level 2 Refinement Level 3 Refinement

Figure 7-27 Single Quadrilateral Element Process
A similar process occurs for the triangles, tetrahedrons, and hexahedrons elements. You can observe that for quadrilaterals
the number of elements expands by four with each subdivision; similarly, the number of elements increases by eight for
hexahedrals. For the 8-node solid-shell element, the subdivision is similar to 4-node quadrilaterals; there is no refinement
 levelx 2 
through the thickness. If full refinement occurs, you observe that the number of elements is 2 for quadrilaterals
 levelx3 
and 2 for hexahedrons elements. The number of levels also limits the amount of subdivisions that may occur.
Number of Elements
Level Quadrilaterals Hexahedrals
0 1 1
1 4 8
2 16 64
3 64 512
4 256 4096
For this reason, it is felt that the number of levels should, in general, be limited to three.
When adaptive meshing occurs, you can observe that discontinuities are created in the mesh as shown in the following
image :
A
B C
To ensure compatibility, node B is effectively tied to nodes A and C, and node D is effectively tied to nodes C and E. All
of this occurs internally and does not conflict with other user-defined ties or contact. It should be noted that for shell
elements this automated tying does not apply to the out-of-plane displacement component; this is to avoid that a state of
pure bending cannot be represented if adjacent elements have a different refinement level.
Boundary Conditions
When mesh refinement occurs, boundary conditions are automatically adjusted to reflect the change in mesh. The rules
listed below are followed:
1. Fixed Displacement:
For both 2-D and 3-D, if both corner nodes on an edge have identical boundary conditions, the new node created
on that edge has the same boundary conditions. For 3-D, if all four nodes on a face have identical boundary
conditions, the new node created in the center of the face has the same boundary conditions. Note that identical
here means the same in the first degree of freedom, second degree of freedom, etc. independently of one another.
CHAPTER 7 621
2. Point Loads:
The point loads remain unchanged on the original node number.
3. Distributed Loads:
Distributed loads are automatically placed on the new elements. Caution should be used when using the FORCEM
user subroutine as the element numbers can be changed due to the new mesh process.
4. Contact:
The new nodes generated on the exterior of a body are automatically treated as potential contact nodes. The
elements in a deformable body are expanded to include the new elements created. After the new mesh is created,
the new nodes are checked to determine if they are in contact. It should be noted that new nodes on shells that are
completely tied to the corresponding edge nodes are not checked for contact.
Caution: None of the nodes of an element being subdivided should have a local coordinate system
defined through the TRANSFORMATION or COORD SYSTEM option.
Location of New Nodes

When an element is refined, the default is that the new node on an edge is midside to the two corner nodes. As an
alternative, the POINTS, CURVES, SURFACES, ATTACH EDGE and ATTACH FACE options or the UCOORD user subroutine can be used.
The CURVES and SURFACES options can be used to describe the mathematical form of the curve or surface. If the corner nodes
of an edge are attached to the surface, the new node is placed upon the actual surface. This option to attach new nodes is
available both in a linear as well as a nonlinear analysis.
This is illustrated in Figure 7-28 and Figure 7-29, where initially a single element is used to represent a circle. The circle is
defined with the CURVES option and the original four nodes are placed on it using the ATTACH EDGE option. Notice that the
new nodes are placed on the circle.
Figure 7-28 Original Mesh and Surface

Level 1 Refinement Level 2 Refinement
Level 2 and 3 Refinement

Figure 7-29 Levels of Refinement
Adaptive Criteria
The adaptive meshing subdivision occurs when a particular adaptive criterion is satisfied. Multiple adaptive criteria can be
selected using the ADAPTIVE model definition option. These include:
Mean Strain Energy Criterion

The element is refined if the strain energy of the element is greater than the average strain energy in a chosen set of elements
times a given factor, f 1 .
total strain energy
element strain energy  ----------------------------------------------- * f 1 (7-31)
number of elements
Zienkiewicz-Zhu (Z-Z) Criterion

The error norm is defined as either
2 2

2
=
   –   dV
------------------------------------------------------------- 
2
=
  E – E  dV
------------------------------------------------------------- (7-32)
2 2 2 2
  dV +    –   dV  E dV +   E – E  dV
CHAPTER 7 623
The stress error and strain energy errors are:

2 2
X =    –   dV and Y =   E – E  dV (7-33)
The stresses at the integration points are extrapolated to the nodes and averaged. This is considered the smoothed nodal
* .  * is then these stresses interpolated back to the integration points.
stresses  N
*
 * = N N
 is the calculated stress. Similarly, E is for energy.

An element is subdivided if,
  f 1 and (7-34)
X el  f 2 * X/NUMEL + f 3 * X * f 1    NUMEL (7-35)
or
  f 1 and (7-36)
Y el  f 4 * Y/NUMEL + f 5 * Y * f 1    NUMEL (7-37)
where NUMEL is the number of elements in the mesh. If f 2 , f 3 , f 4 , and f 5 are input as zero, then f 2 = 1.0 .
The Zienkiewicz-Zhu criterion may be placed on the post file using post codes 491-494.
Zienkiewicz – Zhu Plastic Strain Criterion

p p 2
The plastic strain error norm is defined as 
2
=
   * –   dV
--------------------------------------------------------------- .
p2 p p 2
 dV +    * –   dV
p* p 2
The plastic strain error is A =   –   dV . The allowable element plastic strain error is
AEPS = f 2 * A  NUMEL + f 3 * A * f 1    NUMEL . The element will be subdivided when   f 1 and
A el  AEPS . NUMEL is the number of elements in the mesh.
Zienkiewicz-Zhu Creep Strain Criterion

c c 2
Zienkiewicz-Zhu creep strain error norm is defined as 
2
=
   * –   dV
-------------------------------------------------------------- . The creep strain error is
c2 c c 2
 dV +    * –   dV
c* c 2
B =   –   dV . The allowable element creep strain error is:
AECS = f 2 * B  NUMEL + f 3 * B * f 1    NUMEL .
The element will be subdivided when   f 1 and B e l  AECS . NUMEL is the number of elements in the mesh.
Equivalent Values Criterion

This method is based upon either relative or absolute testing using either the equivalent von Mises stress, the equivalent
strain, equivalent plastic strain or equivalent creep strain. An element is subdivided if the current element value is a given
fraction of the maximum (relative) or above a given absolute value.
max
 vm  f 1  vm or  vm  f 2
max
 vm  f 3  vm or  vm  f 4
Node Within A Box, Cylinder, or Sphere Criterion

An element is subdivided if it falls within the specified box, cylinder, or sphere, respectively. If all of the nodes of the
subdivided elements move outside the box, the elements are optionally merged back together.
The location of this box can be repositioned based upon the motion of a rigid body, the displacement of a node, or by using
the UADAPBOX user subroutine.
The criteria may also be used in conjunction with the WELD FLUX model definition option. In such cases, the motion of the
weld source controls the location of the box.
Nodes In Contact Criterion

An element is subdivided if one of its nodes is associated with a new contact condition. In the case of a deformable-to-rigid
contact, this implies that the node has touched a rigid surface. For deformable-to-deformable contact, the node can be
either a tied or retained node. One can exclude symmetry bodies from those considered for local adaptive meshing. One
can also explicitly list which three bodies should be considered for nodes in contact criterion. Note that if chattering occurs,
there can be an excessive number of elements generated. Use the level option to reduce this problem. If none of the nodes
of the subdivided elements is involved in contact, the elements are optionally merged back together.
Contact Penetration Criterion

An element will be subdivided if a node in the element penetrates another body by a specified distance.
CHAPTER 7 625
Temperature Gradient Criterion

An element is subdivided if the temperature gradient in the element is greater than a given fraction of the maximum
gradient in the solution. This is the recommended method for heat transfer. It is also possible to specify that this only occurs
if the temperature is above or below a user-specified threshold. One can also indicate if the elements should be merged back
if the temperature is below a threshold.
Pressure Gradient Criterion

An element is subdivided if the pressure gradient in the element is greater than a given fraction of the maximum gradient.
This is the recommended method for diffusion analysis.
Electrical Potential Criterion

An element is subdivided if the electrical potential in the element is greater than a given fraction of the maximum gradient.
This is the recommended method for electrostatic analysis.
Magnetostatic Potential Criterion

An element is subdivided if the magnetic potential in the element is greater than a given fraction of the maximum gradient.
This is the recommended method for magnetostatic analysis.
User-defined Criterion
The UADAP user subroutine can be used to prescribe a user-defined adaptive criteria. The user subroutine UADAP2 can be
used to prescribe a criterion for merging elements back (unsubdivide).
Previously Refined Mesh Criterion

Use the refined mesh from a previous analysis as the starting point to this analysis. The information from the previous
adapted analysis is read in.
Angle Between Shell Elements

An element is refined if the change in angle between neighboring shell elements is larger than the given value. For each
node, a normal is calculated by averaging shell element normals for elements connected to the node. If an element is to be
subdivided, the angle used for checking is two times the angle between the averaged nodal normal and the element normal.
Only the change in the normals from the undeformed shape is used to avoid that elements originally connected at an angle
get subdivided immediately.
For equally oriented quadrilateral or hexahedral elements, an initial refinement level can be entered for criteria which
support merging back of subdivided elements. If this initial refinement level is non-zero, this merging of element can go
beyond the initial mesh density. This way, one can further control the total number of elements in the model.
Automatic Global Remeshing

The automatic global remeshing is a procedure where the mesh in a part of the model can be automatically replaced during
the analysis. This is very useful if the deformations are large and the mesh gets distorted during the analysis. Another
situation is when one wants to have a fine mesh in a certain region, and this region changes during the analysis.
In applications like metal forming or complex rubber seals, the materials may be deformed from some initial (maybe
simple) shape to a final, very often, complex shape. During the process, the deformation can be so large that the mesh used
to model the materials become highly distorted, and the analysis cannot go any further without using some special
techniques (see Figure 7-30). Remeshing is very useful for this type of analysis.
Figure 7-30 Mesh too Distorted and Simulation Stops (Left); Using Global Remeshing, the Simulation Completes
Another case when remeshing is useful is when we want to adapt the mesh to the solution in order to obtain better accuracy
at a lower cost. Figure 7-31 shows a shell example with a plate in tension. On the left-hand side, we have the distribution of
the von Mises stress; and on the right-hand side, the mesh obtained by using the stress distribution for controlling the mesh
density variation. This type of mesh density control is supported for all types of elements used in remeshing.
Figure 7-31 Solution Dependent Mesh Density Control

CHAPTER 7 627
Global remeshing is currently carried out on a contact body. The remeshing itself is independent of contact, but the
definition of what to remesh is done through a contact body. In an analysis where contact is not needed, you have to create
a dummy contact body for the part which is to be remeshed.
When performing an analysis with global remeshing, we need to specify the type of mesher to use, when to do the
remeshing, and define controls for mesh size and other settings for remeshing.
The basic steps in remeshing are as follows:

1. The outline or surface of the current, deformed shape of the remeshed body is extracted. For 2-D, this is the outline,
for shells the entire body; and for 3-D, solids the closed surface surrounding the body.
2. The outline is cleaned up and a new mesh is generated based upon this outline.
3. A data mapping is performed to transfer necessary data from the old, deformed mesh to the new mesh.
4. Contact parameters are refined on the new mesh with a new outline and new tolerances.
5. Boundary conditions, if any, are transferred to the new mesh.
6. Marc continues the computation based on the new mesh.
Global remeshing can be used for 2-D or 3-D bodies consisting of continuum elements or 3-D bodies consisting of shell
elements. A 2-D solid, a 3-D solid, and a 3-D shell example are shown in Figure 7-32.
CHAPTER 7 629
2-D Global Remeshing of Rubber Seal Insertion
3-D Global Remeshing of Connecting Rod Forging
Deep Drawing of a Box with Shell Remeshing

Figure 7-32 2-D and 3-D Automatic Remeshing and Rezoning Examples
The following are the supported features and limitations:
1. Analysis type:
Mechanical analysis, thermal-mechanical, thermal Joule mechanical, and electrostatic-structural coupled analysis
are supported.
2. Element types:
In 2-D, lower order continuum element types are supported. These include both quadrilateral and triangular
elements. In 3-D, lower and higher order tetrahedral elements, and lower order hexahedral element types are
supported. Global remeshing is not available for pentahedral and pyramid elements. For 3-D shell elements, the
lower order triangular and quadrilateral elements are supported.
3. Contact analysis:
The remeshing body is a meshed contact body. Contact information, including boundary conditions defined
through contact definitions, is re-determined based on the new mesh. Meshes that are not defined in contact bodies
are not supported.
• Support Rigid-Deformable contact.
• Support Deformable-Deformable contact including self-contact.
4. Boundary conditions:
Boundary conditions are allowed in a remeshing body. These boundary conditions are transferred to the new mesh
after remeshing. The new table style input format is required. These boundary conditions include:
• Distributed load, flux, and current.
• Point load, flux, and current.
• Fixed displacement, temperature, and potential.
The above boundary conditions can be defined on the boundary. For 2-D remeshing using the Mesh on Mesh
technique they can also be defined in the interior.
5. Local adaptivity is not supported together with global remeshing.
6. Mesh density can be based upon the results by using full mesh density control.
7. Domain Decomposition is only supported if the body to be remeshed is included in a single domain.
8. Other supported analysis options:
• Model sections and initial conditions from previous analysis (pre state).
• Cyclic symmetry.
• Flow lines and particle tracking.
Remeshing Region – Partial Remeshing

The main remeshing region is a deformable contact body. The contact body always has to be selected when defining the
region to be remeshed. In addition, it is possible to define a sub-region of the contact body so the entire body does not have
to be remeshed. This partial remeshing is currently supported for tetrahedral meshing and 2-D meshing using the Mesh
on Mesh technique. The mesh in the sub-region is replaced by the remeshing, and the part which was connected to the rest
of the body is preserved. After remeshing the new mesh will fit seamlessly with the other part of the contact body.
The following options are available for defining the sub-regions:
 A set of elements
This explicitly defines the sub-region. The set is automatically updated after remeshing to allow for repeated
remeshing operations on the same region.
 A geometric entity
CHAPTER 7 631
The sub-region is defined by all elements of the contact body that are completely within the geometric entity.
Supported entities are box, cylinder and sphere.
 Vicinity of a crack
The sub-region is defined as a volume around a crack. All elements of the contact body that are completely within
a distance to the crack will be part of the sub-region. The entity which defines what it takes the distance to
depends on the status of the crack. During crack initiation it is the entire crack initiator surface. For a stationary
crack it is the crack front. For a growing crack it is the surface defined by sweeping from the old crack front to the
new. This means that the remeshing region will follow a growing crack.
The sub-regions can freely overlap within the contact body. Care must be taken if multiple remeshing definitions are used
for the same contact body and the option with element sets is used. The element set will be invalidated if part or whole of
it is remeshed in another remeshing definition, and after this the element set can no longer be used as a remeshing region.
This could happen if the different remeshing definitions are used in different loadcases. Within a loadcase, only one
remeshing definition can use a given contact body.
It is possible to use multiple sub-regions in a contact body. However, each sub-region should have a reasonable size to allow
for a meaningful remeshing operation.
The meshing of the sub-region will be constrained by the fact that the faces that are common with the surrounding mesh
must be preserved. It is advisable to avoid a too drastic change in mesh density. An example is shown in the figure below.
The sub-region is defined by the selected elements in green. A target edge length of 1/10 of the original element edge length
is used in the remeshing. A correct mesh is obtained, but a quite abrupt mesh size change was needed.
Remeshing Criteria
The remeshing criteria determine when remeshing will occur. There are both automatic criteria and criteria that can be
specified by the user. Remeshing is performed at the start of an increment. In case of automatic remeshing during the
increment, the state at the start of the increment is restored, the remeshing is performed, and the current increment is
repeated with the new mesh.
Remeshing is automatically performed if,
 Elements go inside out during the solution;
 Crack initiation or growth is performed.
 The current loadcase is a TRIM loadcase.
In the following, we describe the user specified remeshing criteria.
Note: In general, frequent remeshing should be avoided for an effective and computationally efficient analysis. Also,
since each remeshing and subsequent rezoning step involves interpolation and extrapolation of element
variables, a possibility of error accumulation exists as the analysis progresses when remeshing occurs too
frequently.
Increment
Remeshing occurs at specified increment frequency.
Element Distortion
The identified body is remeshed when the distortion in the elements becomes large.
For 2-D analysis, the distortion check is based on corner angles. Remeshing is performed if the following conditions are
met:
2-D
 Any inner angle is greater than 175° or less than 5°
 Any inner angle change is greater than the user input data
For 3-D analysis, a volume ratio is measured to determine if remeshing is required. A volume ratio is calculated based on
each corner node and its connecting nodes. If v is the volume of a tetrahedron formed by nodes 1, 2, 3, and 4 and s is
the triangle area of nodes 2, 3, and 4, then the ratio:
h
r = ---
l
l
or if h  l , r = ---
h
Where h is the height and l is the equivalent length of the triangle. They can be calculated respectively by:
3v
h = ------
s
l = s
The default control ratio is 0.01. Any volume ratio of each corresponding corner node smaller than this value forces the
analysis to perform remeshing. Users can change this number to control the remeshing.
Contact Penetration
The identified body is remeshed when the curvature of the contact body is such that the current mesh cannot accurately
detect penetration.
For 2-D analysis, the penetration remeshing criteria is based upon examining the distance between the edge of an element
and the contacted body.
CHAPTER 7 633
For 3-D analysis, the penetration is measured from the center of each boundary element face to the contacting surface.
By default, remeshing is carried out if the penetration is greater than twice the contact
tolerance and less than the target element size, where the contact tolerance is 0.05 of the
smallest element length and the target element size is the element size for remeshing. This
check does not apply to the self-contact situation. Also, this penetration limit can be given
by user input to avoid too many remeshing operations.
Remeshing is activated when the penetration distance reaches or exceeds the given
penetration tolerance.
Immediate
The identified body is remeshed before performing any analysis. This control can also be used to change element types
from quadrilateral elements to triangular elements or from hexahedral elements to tetrahedral elements. For example, it is
possible to use this control to change an initially defined mesh using element type 7 (an 8 node-hexahedral element) to one
using element type 157, 241 or 247 (5 node-tetrahedral element) in order to use the tetrahedral remeshing capability.
Strain Change
Equivalent strain measures element deformation. This criterion keeps a record of the strain change after remeshing for each
element. When any element of the body has a strain change greater than the control limit, the remeshing will start.
Meshing Techniques
2-D Solid Meshers
2-D meshing can be done for both triangles and quadrilaterals. The main mesh generator is the MoM (Mesh on Mesh)
mesher, which supports both triangles and quadrilaterals. For quadrilateral meshes it may create some triangles, and they
will be treated as collapsed quadrilaterals. This mesher has full support for the general mesh density control described below,
as well as the geometry preservation including hard and soft entities. It also has full support of loads and boundary
conditions, which can also be applied to the interior of the mesh. Current limitations are that it does not support trimming
and crack propagation.
The Advancing Front and Delaunay meshers support loads and boundary conditions on the outline, but they have
limitations in the geometry preservation and they do not support the general mesh density control. There are some limited
capabilities for mesh density control, with curvature control and refinement boxes. They are currently the only 2-D meshers
that support trimming and crack propagation.
Marc has a so-called overlay mesher, but it only has limited capabilities. The mesh quality is often poor, and it does not
support holes in the body or boundary conditions.
There are a number of mesh generators primarily targeted for the analysis of recession in a thermal analysis of pyrolysis and
similar phenomena. These are as follows:
 Relax mesher: The mesh is improved by shifting nodes. Also available for 3-D solids. A user subroutine
UREMESH_RELAX is available for shifting the nodes. Marc Volume D: User Subroutines and Special Routines gives details
about the usage of this routine.
 Shave mesher: Elements are deleted by shaving off layer by layer of a region where recession takes place. Also
available for 3-D solids.
 Stretch mesher: Also available for 3-D solids.
 Multi-zone mesher: Special mesher for streamline regions.
Shell Meshers
Shell meshing is done using the Patran MOM (Mesh on Mesh) surface mesher. This mesher supports triangles and
quadrilaterals and has full support for mesh density control.
3-D Solid Meshers

There is an overlay hexahedral mesher available in Marc but it has limited capabilities. Boundary conditions are not
supported, the mesh quality is not always good and it has limited mesh density control.
For tetrahedral meshing, Marc uses the adaptive Hybrid mesher. It has full support for boundary conditions and mesh
density control. This is usually the mesher of choice for 3-D solid remeshing.
Mesh Density Control

The user must define how fine of a mesh to create during meshing. Mesh density control refers to the variation of element
sizes in the new mesh.
2-D Meshing
The MoM mesher in 2-D has full support of the general mesh density control as described below.
The advancing front and Delauney meshers have limited possibilities to control the mesh size variations. The user specifies
either a target element edge length, a target number of elements or specifies that the existing number of elements should
be preserved. When a target element size is given, the user can also give a minimum element size. The default is 1/3 of the
target element size. The maximum element size is 3 times the target. The meshers cannot exactly satisfy the user
suggestions. A higher mesh density can be obtained in areas where the curvature of the outline is high. Finally, it is possible
to define refinement boxes, where the mesh is refined a user specified number of levels.
Shell Meshing
Shell meshing using the so-called Simplified option has the same type of controls as for advancing front and Delaunay as
described above. The more general mesh density control is described below.
3-D Solid Meshing

Similar to shell meshing, the Simplified option offers simple ways for controlling the mesh density. In addition to the
options available for 2-D it is also possible to specify a coarsening of the mesh towards the interior of the volume.
The hexahedral mesher is based upon an initial regular mesh which is reshaped in order to fill the body with elements. The
user specifies the element size of this initial mesh and optionally refinement levels inside boxes.
CHAPTER 7 635
Mesh Coarsening
To reduce the computational costs, but ensure a fine mesh on the exteriors, one can request that the mesh is coarsened on
the interior region.
The more general mesh density control is described below.
General Mesh Density Control

The more general mesh density control is available for 2-D MoM, shell and tetrahedral meshing, and offers general and
flexible ways to control how the mesh density varies in the remeshed body.
The procedure is based upon target mesh density defined in the current mesh, which acts as a background mesh in the
mehing process. Every element in the current mesh is given a target edge length, and this data is passed on to the mesh
generator. It tries to honor the specified mesh density variation while making sure the mesh has smooth transitions between
regions of different mesh density.
The user defines one or more so-called mesh density controls for the background mesh. A density control is defined for
some or all of the elements in the background mesh (or none of the elements, in which case Marc will give a warning but
no error). Multiple overlapping controls can be given, and Marc will use the values that specify the smallest target edge
length. Exceptions are region control and user subroutine control which may override other definitions. If a target length
is not defined for a certain element of the background mesh, a value will be calculated based upon the current size of the
element.
Global Density Controls
These mesh density controls are specified for the whole background mesh. A constant target edge length can be given for
all elements, or a length can be calculated based upon the current mesh. The latter case does not guarantee that the same
number of elements as before will be obtained, only that the target edge length is based upon the edge lengths in the current
mesh. A scale factor for the edge lengths is available for this option.
Curvature
Here, the curvature of the body controls the target edge length. It comes in two flavors that can be defined separately. For
shells and tetrahedrals, it is defined as follows:
 Surface curvature is defined on the surface of the body.
 Edge curvature is defined on the boundary of a shell and on sharp edges of a solid.
For example, a flat shell with a hole will have an infinite surface curvature, and the edge curvature is defined by the radius
of the hole.
For 2-D there is no surface curvature. The edge curvature is given separately for convex and concave parts of the boundary.
Region
This defines a refinement (or coarsening) region. If any element node is inside the refinement region, the element will get
the value defined for the current region. If an element is in multiple regions, it will get the smallest edge length defined
except if a region uses coarsening. Then, the value is determined by the last defined region with coarsening.
CHAPTER 7 637
Three types of regions can be defined: box, cylinder (3-D and shells only), and sphere (for 2-D a circle). They are visualized
during preprocessing. The regions can move during the analysis, following the motion of a given node or a rigid contact
body or by defining a velocity. If the node is specified and this node is part of a remeshing body, it will be preserved when
the body is remeshed.
Distance
The target edge length is specified as the distance to an entity. A radius of influence is defined, and if an element node is
within this distance from the entity, it will get the target edge length defined. The edge length is specified at the entity and
at the radius of influence, which allows a linear variation of the edge length within the radius of influence.
The supported entity types are: node, point, position, curve, and cracks. The node is a node anywhere in the mesh. If this
node is part of a body that is being remeshed, then the node will be preserved during the remeshing. The point and curve
options are defined on geometric entities during pre-processing. These entities do not move during the analysis. The
distance from crack allows a refined mesh around a crack tip (shells) or crack front (tetrahedrals). This includes cracks that
are stationary, growing and initiated. Currently, cracks are only supported for shells and tetrahedrals.
Table
This option allows a variation of the target edge length with a table, where the table typically is a function of the spatial
coordinates. It is also possible to get a edge length variation along an arbitrary direction by using a curve.
Element or Nodal Quantity

This option uses results quantities for determining the edge length for all elements. An element or nodal quantity is
selected, and element edge lengths will be assigned depending on the results value for each element.
CHAPTER 7 639
User Subroutine
With this option, the mesh density is given by UMESHDENS user subroutine for shells and tetrahedrals, and UREMESH_TARGET
user subroutine for 2-D. Marc Volume D: User Subroutines and Special Routines gives details about the usage of this routine.
Geometry Preservation
For 2-D using MoM, shell and tetrahedral meshing, there are special procedures and options for preserving the geometry
of the remeshed body.
A representation of a geometry with lower-order finite elements gives a piecewise linear (faceted) description of a curved
geometry. A simple example with a shell structure is given in the left-hand side of Figure 7-33. If we would use the current
mesh directly as a base for the new mesh, we would obtain new nodes on the old mesh as opposed to on the geometry the
mesh is representing. Repeated remeshing would lead to a shrinking geometry.
Hence, the current mesh can not directly be used as is as a base for the remeshing. The geometry that is used in the meshing
is obtained from the current mesh by smoothing the surface. Cubic splines are generated along the boundary of 2-D and
shells and along sharp edges. Coons surfaces are defined on curved surfaces. This gives a better preservation of the actual
geometry, as seen in the right-hand side of Figure 7-33.
Figure 7-33 Preservation of Geometry in a Shell Case
During remeshing there are model features that need to be preserved. Corners and sharp edges need to remain so after
remeshing. There are two general controls for corners and edges. The Feature Vertex Angle defines what should be
considered a corner. The default value is 120 degrees. If the angle between edges meeting at a point is larger than that, then
the point is not considered a corner and can be smeared out during remeshing. For edges, there is a Feature Edge Angle
available with default 30 degrees. If the angle between normals of faces defining an edge is smaller than this value, then the
edge is not considered sharp and may be smeared out during remeshing.
There are cases where it is not possible to set these global angles so that all features are preserved as expected. In Figure 7-33
for example, we may want to preserve the sharp corner along the bottom. However, that vertex angle is smaller than the
vertex angles on curved part. If the feature vertex angle is increased, then we would preserve the corners along the curved
part which may not be what we want. Another example is in Figure 7-34 where we have a cylindrical region with a cut. The
initial mesh is rather irregular but it adequately describes the geometry. It is not possible to define feature edge and vertex
angles so that the geometry is properly preserved.
Figure 7-34 Example of a Cylinder with a Cut
Marc has options to allow the user to select portions of the mesh to be preserved. Special entities can be defined on the
initial mesh. The supported entities are as follows:
 Hard Nodes: Nodes defined as hard nodes are treated as corners.
 Hard Edges: These edges are preserved as is during remeshing.
CHAPTER 7 641
 Hard Faces: Similar to hard edges, these faces are preserved as is during remeshing.
 Soft Edges: These edges are not smeared out during remeshing. The mesh density along a set of soft edges is
allowed to change.
 Soft Faces: A set of soft faces defines a region for which the boundary is preserved and the mesh inside the region
and along the region boundary is allowed to change. The edges that define the boundary of the soft face region are
treated as soft edges.
In the cylinder example, we define the edges along the cut as soft edges indicated in blue in Figure 7-34. If necessary, the
corners where the cut meets the end of the cylinder could be marked up as hard nodes to make sure they are preserved. The
resulting mesh is shown in Figure 7-35. The geometry is quite well preserved, including the cut. In the example, curvature
control was used for the mesh density. Towards the end of the cut there is a higher edge curvature along the soft edges, and
the curvature control picks this up and gives a finer mesh in this region.
Figure 7-35 New Mesh for the Cylinder with a Cut
Remeshing with Self Contact

In 3D solid meshing it is not possible to mesh a body which has coincident or overlapping surface segments. Marc will
detect this situation, and nodes will be shifted to create an appropriately large gap before performing the meshing. This gap
is supposed to be large enough to avoid mesh overlaps during the meshing operation. In special situations this gap may not
be large enough, so there is an option in Marc to scale this gap, with the so-called self contact shift factor. After meshing
the gap is taken out so the final geometry should be the same as before meshing.
The procedure to check for coincident or overlapping mesh does not depend on self contact being active for the remeshed
contact body, but in general it is recommended to use this in order to minimize the mesh overlap. The segment-to-segment
contact procedure is typically better than the node-to-segment procedure in avoiding self contact mesh overlaps.
Remeshing with Boundary Conditions

There are two types of boundary conditions that global remeshing needs to transfer from old mesh to the new one:
 Contact boundary conditions:
This type of boundary conditions includes friction, heat convection, radiation, symmetry or cyclic symmetry
conditions. After a new mesh is created, these contact boundary conditions are automatically updated based on
the new contact detection using the new mesh.
 User defined boundary conditions:
If the remeshing body has user defined boundary conditions, such as, point loads, distributed loads or fixed nodal
displacements, the boundary conditions are transferred to the new mesh with two different approaches as follows:
a. Defined Set Approach:
This approach allows users to directly apply boundary conditions to the mesh entities, such as, element faces,
edges, or nodes. Each boundary condition is arranged in a defined set with a set name and a set type. Currently,
the set type can be either a node or an element edge or face set. The element edge sets in 2-D and face sets in
3-D are typically those of distributed boundary conditions. Edges or faces in the set can be continuous or
discontinuous. During the remeshing stage, the boundaries of these edges or faces are marked and preserved.
This allows the new edges or faces to be created coincident with the old edges or faces in the set. After
remeshing, the element edges or faces in the set are replaced with the new element edges or faces (see Figure 7-36).
The node ids in nodal sets are automatically replaced with the new node ids (see Figure 7-37).
Figure 7-36 Distributed Loads Transfer
Figure 7-37 Point Loads Transfer

CHAPTER 7 643
The nodal displacement or temperature conditions applying to all the nodes of an element edge or face are
treated as an element edge or face boundary condition. This means the new nodes created on the same element
edge or face will have the same displacement or temperature boundary conditions. In 2-D, an element edge or
face can take up to four boundary condition sets for the remeshing. In 3-D, the number of sets on the same
element face is two. There is no limitation on the node set.
b. Geometry Attachment Approach
Boundary conditions can be assigned to geometry entities, such as points, curves, surfaces, CAD body vertices,
CAD body edges, CAD body faces and CAD bodies. In this approach, these geometry entities are attached to
the mesh entities of a remeshing body, such as, a point attached to a node and a curve attached to a set of
element edges. During remeshing, mesh entities associated with these geometry entities are marked and
preserved. After remeshing, the geometry entities are re-attached to the new mesh entities. By doing so,
boundary conditions assigned to these geometry entities are automatically applied to the new mesh (see
Figure 7-38). In 3-D, the attached surface ID number cannot exceed 99 and only one surface can be attached to
an element face. In 3-D, curves can also be attached to the element edges but only nodal displacement
conditions are supported (see Figure 7-39). Geometry attachments are shown.
Figure 7-38 Distributed Load with Curve Attachment
Figure 7-39 Geometry Attachment Boundary Conditions in 3-D

History Data Mapping Technique

History data such as stress, strain and temperature need to be transferred from the old mesh (solid black) to the new mesh
(dashed red), see Figure 7-40. The data to be transferred consists of both nodal data and element data stored at integration
points. The mapping of these quantities is done separately.
The data mapping only involves the part of the model which is being remeshed, and the rest of the model is untouched by
the data mapping.
For the mapping of element data, all integration points of the old mesh are collected in a point cloud. For each integration
point of the new mesh, the closest points in the point cloud are found, and the new values are interpolated from these
points. If an integration point in the new mesh is at exactly the same position as a point in the old mesh, then it will obtain
exactly the same value. Hence, in the special case that the new mesh is identical to the old mesh – the results will not change.
Mapping of nodal data is performed in the same way. A point cloud is collected from the nodes of the old mesh, and the
closest points for each new node are found. Nodes on the boundary use a special mapping. (For shells, the nodes on free
edges). They only take values from the boundary nodes of the old mesh.
After the data mapping, the model is in general not in equilibrium. This state is the starting point for the following load
increment, and the mapped state is not available on the post file.
Figure 7-40 Mapping from Old Mesh (Black) to New Mesh (Red)
Trimming
Trimming is an option to allow the model to be changed during remeshing. It is done using a special loadcase called TRIM.
In such a trimming loadcase there is no analysis performed. It automatically triggers a remeshing, and during this remeshing
operation the remeshed body is modified. Just as with any remeshing in Marc, the results data are mapped from the old
mesh to the new. These results are shown on the results file. It should be noted that these results do not represent the effect
of the model changes, since no analysis was performed. In order to see the effect of the model changes it is necessary to use
a regular loadcase following the trimming loadcase.
The type of model modification that can be done during trimming is that the remeshed body is intersected by a so-called
trimmer. The trimmer is a surface (curve in 2-D) which can be open or closed. The trimmer makes a cut in the mesh of
the remeshed body, and the material inside the trimmer can be deleted – modelling a simplified trimming operation. For
a definition of what is inside or outside the trimmer, see below. It is also possible to keep both parts, using the option to
separate a model into multiple parts.
CHAPTER 7 645
There are currently less options in 2-D than in 3-D for defining the trimmer. It can only be a geometric (rigid) contact
body using a piecewise linear geometry description (called discrete in Mentat). The part which is deleted is determined by
the orientation of the contact body, and it is not possible to keep both parts and only make a cut. The use in 2-D is
illustrated by the following example.
Trimming option: punch a hole
Trimming option: retain a circular disk
3-D hexahedral meshing does not support trimming.

3-D trimming using shells or tetrahedrons is more general than 2-D. The trimmer can be a general faceted surface, defined
from a set of elements, or a rigid or deformable contact body.
A faceted surface in Mentat is a special surface type, defining the surface with a mesh of triangles. A faceted surface can be
imported in the STL format and it can be created from a solid or a NURBS surface. When the trimmer is defined as a
faceted surface, it is not part of contact.
With the trimmer defined from a set of elements, a surface of triangles is constructed from the outline surface of the set of
elements. As these elements would typically overlap the elements of the trimmed body, care must be taken to avoid
unwanted contact.
When using a rigid contact body as a trimmer, the contact body must use either a boundary description of type NURBS
with the discrete option or the Facets type of boundary description. The analytical NURBS description is not supported.
For a trimmer defined as a deformable contact body, the supported element types in the deformable body are all higher and
lower order solid and shell elements supported by contact. Beam elements cannot be used as a trimmer. The trimmer
surface is extracted from the surface boundary of the deformable contact body. Shells are special here. If the shell thickness
is taken into account in contact, then the trimming surface includes the thickness. If the inside part of the trimmer is
removed, then it will create a cavity in which the shell contact body will fit. If the shell thickness is not taken into account,
then the trimmer will only make a cut with the mid-surface.
When using a contact body as a trimmer it is important to remember that the body is a regular contact body in all loadcases
except in the trimming loadcase where it is used as a trimmer. As in any contact analysis it is important to use proper contact
tables. During the trimming loadcase, all contact bodies except the one used as trimmer will be active according to the
currently active contact table. Hence it may be necessary to use a contact table for the initial contact if the trimming
loadcase is the first loadcase.
Using a contact body as a trimmer has the advantage that the contact body can be embedded into the remeshed body during
the analysis. This makes it possible to trim out a cavity in the deformed structure and for instance glue the contact body
to the remeshed body. For this case it may be advantageous to use the imprint option. This option takes the triangular mesh
of the trimmer and makes an imprint of it onto the surface of the remeshed body. This way the new mesh after trimming
will match with the mesh of the trimmer. Subsequent contact analysis will give a good fit without spurious stresses due to
mesh mismatch. This option should be used with some caution. Suppose it is used with a faceted surface, and the surface
has very small facets. Then the new mesh will have the same size at the surface. The imprint of the mesh is exact for triangles
that are inside the remeshed body. If the trimmer intersects the surface of the remeshed body, then unless the meshes match
up there will be a slight mismatch. The new mesh in this region is adjusted to avoid small or distorted elements.
An example of an open trimmer is shown below. A bar in axial tension has a trimmer at the center, shown in red below.
When the option to not remove material is used, the bar is split in two parts as shown below.
Chapter 8: Boundary Conditions
8 Boundary Conditions
 Loading 649

Kinematic Constraints 679
 Mesh Independent Connection Methods 717
Marc is based on the stiffness method and deals primarily with force-displacement relations. In a linear elastic system, force
and displacement are related through the constant stiffness of the system; the governing equation of such a system can be
expressed as
Ku = F (8-1)
where K is the stiffness matrix and u and F are nodal displacement and nodal force vectors, respectively. Equation (8-1)
can be solved either for unknown displacements subjected to prescribed forces or for unknown forces (reactions) subjected
to prescribed displacements. In general, the system is subjected to mixed (prescribed displacement and force) boundary
conditions, and Marc computes both the unknown displacements and reactions. Obviously, at any nodal point, the nodal
forces and nodal displacements cannot be simultaneously prescribed as boundary conditions for the same degree of
freedom.
Note: You must prescribe at least a minimum number of boundary conditions to insure that rigid body motion does
not occur.
The prescribed force boundary conditions are often referred to as loads and the prescribed displacement boundary
conditions as boundary conditions.
Note: Boundary conditions can be prescribed in either the global or a local coordinate system. A nodal transformation
between the global and the local coordinate systems must be carried out if the boundary condition is prescribed
in a local system.
In a nonlinear stress analysis problem, Marc carries out the analysis incrementally and expresses the governing equation in
terms of the incremental displacement vector u and the incremental force vector F .
Ku = f (8-2)
Consequently, you must also define both the loads and the prescribed nodal displacements incrementally.
In addition to the prescribed displacement boundary conditions, constraint relations can exist among the nodal
displacements. For example, the first degree of freedom of node i is equal to that of node j at all times. The expression of
this constraint relation is
ui = uj (8-3)
Generally, a homogenous linear constraint equation can be expressed as

ut = a1 u1 + a2 u2 +  + an un (8-4)
where u represents the degrees of freedom to be constrained, u 1  u n are other retained degrees of freedom in the
structure, and a 1  a n are constants.
You can enter constraints through either the TYING, SERVO LINK, RBE2, or RBE3 options.
CHAPTER 8 649
Boundary Conditions
You can use linear/nonlinear springs and foundations to provide special support to the structure and the gap and friction
element (or CONTACT option) to simulate the contact problem.
Loading
Different types of analyses require different kinds of loading. For example, the loads in stress analysis are forces; those in
heat transfer analysis are heat fluxes. Force is a vector quantity defined by magnitude and direction; heat flux is a scalar
quantity defined by magnitude only. Loading can be time invariant (constant value) or time dependent.
Loading Types
You can categorize a particular type of load as either a point (concentrated) load or surface/volumetric (distributed) load,
depending on application conditions. The spatial distribution of the load can be uniform or nonuniform. Special loading
types also exist in various analyses. For example, centrifugal loading exists in stress analysis, and convection and radiation
exist in heat transfer analysis. You can add point loads directly to the nodal force vector, but equivalent nodal forces first
must be calculated by Marc from distributed loads and then added to the nodal force vector. These distinguishing features
are described below.
A point (or nodal) load of either a vector (force, moment) or a scalar (heat flux) quantity is a concentrated load that is
applied directly to a nodal point (see Figure 8-1). Mechanical point loads can be defined as fixed direction forces or as
follower forces. In a global or local coordinate system, a force vector must be defined in terms of its vector components (see
Figure 8-2). If the force vector is defined in a local coordinate system, then a global-to-local coordinate transformation
matrix must be defined for the nodal point (see Figure 8-3 and Figure 8-4). For axisymmetric elements, the magnitude of the
point load must correspond to the ring load integrated around the circumferences.
Q F
y y
x x
Point (a) Point (b)
Heat Flux Q (Scalar) Force F (Vector)
Figure 8-1 Schematic of a Point Load
y’
Fy F F
Fy’
Fx Fx’
y y x’
x x
Figure 8-2 Force Components
y” x”
Fy” = 0
y F = Fx”
x
Figure 8-3 Special Selection of Local (x”, y”) Coordinate System Force Components: F y'' = 0
Surface/volumetric loads, such as pressure, distributed heat flux, and body force, are distributed loads that are applied to the
surface (volume) of various elements (see Figure 8-4). A surface/volumetric load is characterized by the distribution
(uniform/nonuniform) and the magnitude of the load, as well as the surface to which the load is applied (surface/volume
identification). The total load applied to the surface (volume) is, therefore, dependent on the area (interior) of the
surface (volume).
Equivalent nodal forces first must be calculated from surface/volumetric loads and then added to the nodal force vector.
Marc carries out this computation through numerical integration. (See Marc Volume B: Element Library for the numbers and
locations of these integration points for different elements). The calculated equivalent nodal forces for lower-order
elements are the same as those obtained by equally lumping the uniformly distributed loads onto the nodes. However, for
high-order elements, the lumping is no longer simple (see Figure 8-5). As a result, the surface/volumetric loads should not
be lumped arbitrarily.
P1 P2 P1 P2
4 3 4 3
Px
Q
Py
1 2 1 2
y y
P q
x x
(a) Distributed Mechanical Load (b) Distributed Heat Flux
Surface 2-3:
Uniform Normal Pressure p Uniform Heat flux q
Surface 3-4:
Nonuniform Normal Pressure p1 - p2 Nonuniform Heat flux q1 = q2
Whole Volume:
Volumetric Loads Px’Py Volumetric Heat Flux Q
Figure 8-4 Schematic of Surface/Volumetric Load

CHAPTER 8 651
Boundary Conditions
1/2 1/2 1/4

1/4 1/4 1/4
1/4
1/4
1/4 1/4 1/6 2/3 1/6 -1/12

-1/12 1/3 -1/12 1/3
-1/12 1/3
-1/12
1/3 1/3
1/3
1/3
-1/12
-1/12 1/3 -1/12
Figure 8-5 Allocation of a Uniform Body Force to Nodes for a Rectangular Element Family
Using the table driven input format, boundary conditions may be applied to finite element entities, nodes, elements, edges
or faces, or geometric entities such as points, curves, surfaces, CAD body vertices, CAD body edges, CAD body faces and
CAD bodies. The latter is advantageous because it improves compatibility with CAD and because it facilitates adaptive
meshing. The application of boundary conditions on geometric entities is only effective if finite elements have been
associated with these geometric entities using the ATTACH NODE, ATTACH EDGE, and ATTACH FACE model definition options. As
shown in Figure 8-6, Mentat shows a 2-D model consisting of four curves around the perimeter and a curve where the circle
is. The mesh was created using the automatic mesh generator which automatically attaches the edges to the curves. Those
nodes which are attached to points are shown using the circle symbol as opposed to the square symbol. Now, if a boundary
condition, such as an internal pressure, is applied on a curve during the analysis, it will automatically be applied to those
edges attached to the curve as shown in Figures 8-8 and 8-9. In a similar manner, if a displacement is applied to a point,
then it will be applied to the node. For 3-D, the analogous situation exists; namely, a distributed load applied to a surface
will be applied to those elements attached to this face as shown in Figure 8-10. And nodal boundary conditions will be
attached to nodes attached to face.
Figure 8-6 Geometric Representation with Five Curves
Figure 8-7 Original Mesh, Red Edges Indicate that they are Attached to Curve
CHAPTER 8 653
Boundary Conditions
Figure 8-8 Boundary Conditions Applied to Curves
Figure 8-9 Boundary Conditions Applied to Attached Element Edges

Figure 8-10 Distributed Load Applied to Surfaces. In Mentat, Faces Attached to Surfaces are Indicated by a Darker Color
(usually Blue and Midnight Blue)
Now, if global or local adaptive meshing is used, then, as shown in Figures 8-11 and 8-12, the load automatically is applied
to the new element edges which are attached to the curve. If a point load of prescribed displacement is applied to a
geometric point, the mesh generator will make sure that a point/node is in this exact location after remeshing occurs. This
allows all boundary conditions to be used in conjunction with global or local adaptive meshing in 2-D.
For the case of shells, this is complicated because, for certain types of load, there is a difference between application to the
top and bottom surfaces. In mechanical analysis, a positive pressure on the top surface is equivalent to a negative pressure
applied to the bottom surface. This is because the degrees of freedom of a mechanical shell are considered to be at the
neutral surface. For thermal analysis, distributed fluxes, convective boundary conditions and radiation, there is a difference
between application of thermal boundary conditions on the top or bottom surfaces. This is particularly relevant for
radiation, as the choice of surfaces will strongly influence the viewfactor calculation. Figure 8-13, shows a surface, with the
top (yellow) and bottom (blue) surfaces identified. To select either the top or bottom surface, use the FILTER command.
A boundary condition applied to the “top” and “bottom” surface is shown in Figures 8-14 and 8-15, respectively. If the shell
mesh is automatically generated from the surface, the top face of the element is associated with the top surface of the shell.
It is also possible to display the shell elements to indicate the thickness as shown in Figure 8-16, in which case, one can easily
select the top or bottom faces for which the boundary condition is to be applied.
CHAPTER 8 655
Boundary Conditions
Figure 8-11 Distributed Load on Top Surface after Global Adaptive Remeshing
Figure 8-12 Distributed Load to Top Surface after Local Adaptive Remeshing
Figure 8-13 Identification of Top and Bottom Surface of a Shell

Figure 8-14 Boundary Condition on Top Surface
Figure 8-15 Boundary Condition on Bottom Surface
Figure 8-16 Shell Elements shown in Expanded Mode

Figure 8-17 is an example of what can go wrong if either geometric entities or elements are not consistently defined. Two
curves are used to represent an axisymmetric shell, but as shown by the orientation vector, they are not continuous
in direction.
CHAPTER 8 657
Boundary Conditions
Figure 8-17 Oriented Curves
Figure 8-18 shows the finite element mesh obtained from the convert option, which does not indicate any potential problem.
If one displays the shell in expanded mode and identifies backfaces, then one will see an inconsistency between the top
(blue) and the bottom (brown) surfaces as shown in Figure 8-19.
Figure 8-18 Finite Element Mesh
Figure 8-20 shows the boundary condition applied to the “top” curve, which is not what is desired. To get a consistent
applied boundary condition, one needs to flip both the curves and elements of one of the sections.
Figure 8-19 Shell Elements Expanded showing Top (Blue) and Bottom (Brown) Surfaces
Figure 8-20 Incorrect Boundary Condition on Top Surface
Face ID for Distributed Loads, Fluxes, Charge, Current, Source, Films, and Foundations
When using the table driven input procedure, there are three ways to identify the type of load and which edge or face it is
applied to.
In the first method. an IBODY, as defined for each element type defined in Marc Volume B: Element Library , is associated with
each distributed load list. This IBODY effectively defines the type (normal pressure, shear, gravity, pressure, centrifugal,
Coriolis) and the edge or face to which the load is applied.
For the second method, one needs to enter the type of load and the location separately, but the location of the load is
consistent for all element types of the same class. This is called the face ID and is shown in the following figures. It should
be noted that when input files are created using Mentat, the convention is to use the same ID as used in the Marc
convention minus 1.
1-D 2-Node Elements

y FACE ID NODES
2
1 1–2
1
1-D 3-Node Elements

3 FACE ID NODES
1 1–2–3
2
1
CHAPTER 8 659
Boundary Conditions
2-D 4-Node Quadrilateral Elements

4 3 FACE ID NODES
1 1–2
2 2–3
3 3–4
4 4–1
1 2
Load shown on FACE ID 1
2-D 8-Node Quadrilateral Elements

4 7 3 FACE ID NODES
1 1–5–2
2 2–6–3
8 6
3 3–7–4
4 4–8–1
1 5 2
2-D 3-Node Triangle

3 FACE ID NODES
1 1–2
2 2–3
3 3–1
1 2
2-D 6-Node Triangle

3 FACE ID NODES
1 1–4–2
6 5 2 2–5–3
3 3–6–1
1 4 2
3-D 3-Node Shell

z 3 FACE ID NODES
1 1–2–3 (top)
1 2 1–3–2 (bottom)
y
x 2
3-D 4-Node Shell/Membrane

4 P FACE ID NODES
1 1–2–3–4 (top)
3 2 1–4–3–2 (bottom)
1
3-D 6-Node Shell

P 3 FACE ID NODES
1 1–2–3–4–5–6 (top)
6 5 2 1–3–2–6–5–4 (bottom)
1 4 2
3-D 8-Node Shell

4 7 3 FACE ID NODES
1 1–2–3–4–5–6–7–8 (top)
2 1–4–3–2–8–7–6–5 (bottom)
8 6
1 5 2
CHAPTER 8 661
Boundary Conditions
3-D 4-Node Tetrahedral

4 FACE ID NODES
3 1 1–2–4
2 2–3–4
3 3–1–4
1
4 1–2–3
2
3-D 6-Node Pentahedral

6 FACE ID NODES
4 1 1–2–5–4
2 2–3–6–5
5 3 3–1–4–6
4 1–2–3
3
5 4 – 6– 5
1
3-D 15-Node Pentahedral

3 FACE ID NODES
15
1 1 – 2 – 5 – 4 – 7 – 14 – 10 – 13
6
2 2 – 3 – 6 – 5 – 8 – 15 – 11 – 14
3 3 – 1 – 4 – 6 – 9 – 13 – 12 – 15
9 8
4 3–2–1–8–7–9
5 4 – 5– 6 – 10 – 11 – 12
12 11
1 7 2
13 14
4 10 5
3-D 8-Node Brick

8 FACE ID NODES
7 1 1–2–6–5
5
2 2–3–7–6
6
3 3–4–8–7
4
4 4–1–5–8
5 1–2–3–4
1 3
6 6–5–8–7
3-D 10-Node Tetrahedral

4 FACE ID NODES
10
1 1–2–4–5– 9– 8
3
8 9 2 2 – 3 – 4 – 6 – 10 – 9
7
3 3 – 1 – 4 – 7 – 8 – 10
6
1 4 1–2–3–5– 6– 7
5
2
3-D 20-Node Brick

8 FACE ID NODES
16 15
1 1 – 2 – 6 – 5 – 9 – 18 – 13 – 17
5 7
13 20 2 2 – 3 – 7 – 6 – 10 – 19 – 14 – 18
6 14
3 3 – 4 – 8 – 7 – 11 – 20 – 15 – 19
17 19
4 4 4 – 1 – 5 – 8 – 12 – 17 – 16 – 20
12 11
18 5 1 – 2 – 3 – 4 – 9 – 10 – 11 – 12
1 3
6 6 – 5 – 8 – 7 – 13 – 16 – 15 – 14
9 10
2
CHAPTER 8 663
Boundary Conditions
3-D 5-Node Pyramid

FACE ID NODES
1 1–2–5
2 2–3–5
3 3–4–5
4 4–1–5
5 1–2–3–4
3-D 13-Node Pyramid

FACE ID NODES
1 1 – 2 – 5 – 6 – 11 – 10
2 2 – 3 – 5 – 7 – 12 – 11
3 3 – 4 – 5 – 8 – 13 – 12
4 4 – 1 – 5 – 9 – 10 – 13
5 1–2–3–4–6–7–8–9
For the third method, the element edges or faces are attached to a curve or surface, respectively. In this case, one only needs
to give the type of load as the location is inferred from the ATTACH EDGE and ATTACH FACE data.
Mechanical Loads
Marc allows you to enter mechanical loads in various forms for stress analysis. These loads can be concentrated forces and
moments, uniformly and nonuniformly distributed pressures, body forces, gravity or centrifugal loads. Table 8-1 lists input
options for mechanical loads.
Table 8-1 Input Options for Mechanical Loads

Input Options
Load Description Model Definition History Definition User Subroutine
Point Load: Concentrated Force/Moment POINT LOAD POINT LOAD FORCDT
Surface Load Pressure Shearing Forces, and DIST LOADS DIST LOADS FORCEM
Distributed Moment
(Uniform/Nonuniform Distribution)
Volumetric Load Body Forces and DIST LOADS DIST LOADS FORCEM
Acceleration Forces
Centrifugal Loading DIST LOADS DIST LOADS
ROTATION A
Coriolis Loading DIST LOADS DIST LOADS
ROTATION A
Fluid Loading DIST LOADS
FLUID DRAG
Application of centrifugal and Coriolis loadings require the ROTATION A model definition option, which defines the data
corresponding to the axis of rotation. If multiple parts are rotating, then different rotation axes may be defined. The actual
load can be invoked by specifying an IBODY load types 100, 103,104, or 105 for centrifugal and Coriolis loadings,
2
respectively. For load types 100 or 103, enter the values  and d  dtime for the magnitude of the distributed load
2
where  is the angular velocity in radians/time. For load types 104 or 105, enter the values  and d  dtime for the
magnitude of the distributed load where  is the angular velocity in cycles/time. The mass density must also be defined
in the ISOTROPIC, ORTHOTROPIC, or other material options.
Application of gravity load (load per unit mass) is achieved by using IBODY load type 102. The mass density must be
defined in the ISOTROPIC, ORTHOTROPIC, or other material options. The acceleration can be given independently in the x, y,
and z direction through the DIST LOADS option.
Volumetric loads (load per unit volume) may be entered by using IBODY type 106 or 107.
The FOLLOW FOR parameter is used when the direction of Distributed and/or Point Loads are not fixed and need to follow
the geometry. FOLLOW FOR automatically invokes the LOAD COR parameter, so stresses should be stored at all integration
points. Separate flags under this parameter are used to control follower forces for distributed loads and point loads
respectively. When the FOLLOW FOR parameter is flagged for distributed loads, equivalent nodal loads due to pressures are
calculated based on the current geometry. Any change of surface area or orientation results in a change of load. You can
also specify that the follower force stiffness matrix due to distributed loads can be included. The FOLLOW FOR parameter for
distributed loads is typically used when a shell structure, which can undergo large deformation and rotations, is subjected
to a pressure load.
A separate flag on the FOLLOW FOR parameter allows follower forces to be specified for point loads. When this parameter
option is flagged, the user still has the choice of specifying individual point loads as fixed direction load vectors or as
CHAPTER 8 665
Boundary Conditions
follower forces. Fixed direction load vectors are specified by the usual vector components. When defined as a follower force,
there are two techniques available for specifying the load:
 The first technique is similar to the option available in MSC Nastran. The scalar magnitudes of the load and/or
moment are specified instead of the vector components. The direction of the force is then specified through 2 or 4
independent nodes in a manner similar to the FORCE1/MOMENT1 or FORCE2/MOMENT2 options in MSC Nastran. In
the former, two nodes A and B are to be specified by the user and the direction of the load follows the vector
directed from node A to node B, as shown in Figure 8-21(a). In the latter, four nodes A, B and C, D are specified by
the user. The direction of the load follows the cross product of the vector directed from node A to node B and the
vector directed from node C to node D, as shown in Figure 8-21(b).
P (Load Magnitude)
P (Load
A Magnitude)
Load
Direction Load D
Direction
B
A,C
B
(a) FORCE1 Style (b) FORCE2 Style
Figure 8-21 MSC Nastran Style Follower Force Definition for Point Load
 The second technique offers an automated way of defining the follower force. The load/moment is specified
through vector components as usual. The initial orientation of the load with respect to an optimal mesh location
is noted. Then, as the structure deforms, the load direction is updated such that the relative orientation of the load
with respect to the mesh is maintained, as shown in Figure 8-22. Optimal nodes in the mesh are automatically
chosen for calculating and maintaining the relative orientation. One or two optimal nodes can be used. The
optimal nodes are chosen based on the following criteria: The nodes should belong to the same element as the
loaded node. If the initial vector from the loaded node to any node matches the initial load direction, then that
node is chosen as the optimal node. Else, a node whose vector from the loaded node has a component in the load
direction is chosen. If this also does not provide a suitable node, the closest node to the loaded node is chosen. For
3-D shells, membranes, and 3-D solids, two optimal nodes are chosen if the cross product of the vectors of the
loaded node to these two nodes matches the load direction. Else, two nodes whose vectors from the loaded node
have a cross product with a component in the load direction are chosen. If none of the above rules provide suitable
nodes, the two closest nodes to the loaded node are chosen.
Initial Load
Direction

 Updated Load
Optimal Direction
Node
Figure 8-22 Geometry Based Automated Follower Force Definition for Point Load
The second technique is quite convenient and can be generally applied to 2-D and 3-D beams, shells and solids. In
situations where the loaded node does not belong to an element (for example, load controlled rigid dies, etc.), the first
technique should be used. It should be noted that a follower force stiffness matrix is not currently available for the point
loads. Also, the follower force capability is not supported for a point load specified through the FORCDT user subroutine.
A number of history definition options are available for input of multiple load increments. For example, the AUTO LOAD
option generates a specified number of increments, all having the same load increment, and is useful for nonlinear analysis
with proportional loads; the PROPORTIONAL INCREMENT option allows the previous load increment to be scaled up or down
for use in the current load increment. The AUTO INCREMENT option allows automatic load stepping in a quasi-static analysis
and is useful for both geometrically and materially nonlinear problems.
Fluid Drag and Wave Loads

Marc provides a fluid drag and wave load capabilities that can be applied on beam type structures that are partially or fully
submerged in fluid (see Figure 8-23). Morison’s equation is used to evaluate the fluid drag loads that are associated with
steady currents. Only distributed drag and buoyancy effects are considered. Marc employs Airy wave theory to evaluate
wave velocities that can be invoked for dynamic analysis option.
Fluid drag and wave loads are invoked using the DIST LOADS model definition with an IBODY load type of 11. These loads
also require the FLUID DRAG model definition to input the relevant information regarding the fluid elevation and its flow.
Table 8-2 lists input options for fluid drag and wave loads.
Table 8-2 Input Options for Fluid Drag and Wave Loads
Input Options
Fluid Drag Load DIST LOADS DIST LOADS
FLUID DRAG
Wave Load DIST LOADS DIST LOADS
FLUID DRAG
CHAPTER 8 667
Boundary Conditions
Steady Velocity
Wave Load
Flow Velocity Pipe

Distribution
Buoyancy
Force
Velocity Outside
Gradient Fluid
Fluid Drag
Force
Inside
Fluid
Sea Bed
Figure 8-23 Fluid Drag and Wave Loads
Cavity Pressure Loading

Marc allows the modeling of structures enclosing cavities by updating the cavity internal pressure as the cavity volume
change. For ideal gas-filled cavities, the equation of state relating the cavity pressure and volume can be written as:
pV = nR o T (8-5)
where p is the cavity total pressure, V is the cavity volume, n is the number of molecules of the gas, R o is the Universal
Gas Constant ( R o = 8.31447 J/(mol°K) = 1545.35 ft.lbf/(mol°R) ), and T is the absolute gas temperature. The gas
mass, M , is related to the number of molecules of the gas by:
 = Mn (8-6)
where  is the molar mass of the gas. Substituting by n from Equation (8-6) into Equation (8-5) gives:
pV = MRT (8-7)
where R = R 0   is the Specific Gas Constant. With the gas density,  , defined as:
 = MV (8-8)
the equation of state of an ideal gas can be finally written as:

p = RT (8-9)
The Specific Gas Constant is calculated from:
R = pr  r Tr (8-10)
where p r is the gas reference pressure, T r is the gas reference temperature, and  r is the gas reference density. The user
must ensure that the values entered for the gas reference pressure, temperature and density are consistent. The cavity total
pressure is given by:
p = pa + pg (8-11)
where p a is the ambient pressure and p g is the cavity gage pressure. Only the cavity gage pressure is applied to the structure
forming the cavity.
The following loading scenarios are available for cavities:
Closed Cavity
If the cavity is closed, the mass of the gas is preserved
M = Mo (8-12)
where M o is the cavity mass from the previous increment. The gas is assumed to undergo a general polytropic process
represented by:

pV = constant (8-13)
where  is the polytropic exponent. The gas pressure can thus be updated using:
p Vo 
----- =  ------- (8-14)
po  V
where p o and V o are the cavity pressure and volume from the previous increment, respectively. Using Equation (8-7) and
Equation (8-14), the gas temperature can be updated using
T Vo –1
------ =  ------- (8-15)
To  V
where T o is the cavity temperature from the previous increment. The gas density is calculated from:
 = MV (8-16)
CHAPTER 8 669
Boundary Conditions
Closed cavity processes can be set to occur:

 At constant pressure, isobaric, using  = 0 .
 At constant temperature, isothermal, using  = 1 .
 With no heat transfer to the surroundings, adiabatic, using  = k , where k is the adiabatic exponent. For ideal
gases, k is a constant that depends only on the number of atoms in the gas molecule (monoatomic gases:
k = 1.67 , diatomic gases: k = 1.4 , triatomic gases: k = 1.33 ).
Applied Pressure
In this case the new cavity pressure is updated using:
p = p o + p (8-17)
The gas temperature is assumed to be constant

T = To (8-18)
The gas density is updated using:

 = p  RT (8-19)
and the gas mass is recalculated as:
M = V (8-20)
Applied Mass
In this case, the new cavity mass is updated using:
M = M o + M (8-21)
If M  0 , gas is pumped in or out of the cavity. The gas density is first updated as:
 = MV (8-22)
the gas temperature is assumed to be constant
T = To (8-23)
and the cavity pressure is calculated as:

p = RT (8-24)
If M = 0 , the cavity is assumed to be closed and Equations 8-12 to 8-16 apply with  = 1 .
User Defined Loading

The UCAV user routine allows the user to define and control the cavity pressure for loading scenarios other than the ones
specified above.
Gas Cavity Stabilization

Marc automatically turns on stabilization for gas filled cavities if it finds that the relative change in volume of the cavity is
large. Stabilization limits the iterative cavity volume change thus aiding the overall convergence of the problem.
Liquid Filled Cavity

Marc allows the analysis of cavities filled with liquid. The analysis is based on the assumption that such a cavity is
completely filled with liquid, and, hence, effects like splashing inside the cavity cannot be modelled. Both pressure and
mass load can be applied to the liquid filled cavity.
The user needs to specify the bulk modulus and density of the enclosed liquid, along with a stabilization factor (which has
a default value of 0.001) can either be manually specified or is automatically determined. The automatic option (default)
is the recommended approach, and in this case, Marc internally controls the value of the stabilization factor throughout
the analysis. This stabilization factor is used to improve the convergence behavior in case of large values of the bulk
modulus. The user may also specify the bulk modulus to be a function of temperature by using a table.
The pressure p inside a closed cavity is given by:
Vo
p =   T In  -------
 V
where   T  is the bulk modulus which can be a function of temperature, V o is the original volume and V is the current
volume of the cavity.
The cavity may be subject to certain boundary conditions. The cavity may be subject to either a cavity pressure load or a
cavity mass load, both of which can be specified using tables. Further, the cavity can also be modeled as a closed cavity in
which case the mass inside the cavity is preserved.
Incremental vs Iterative Approach

Marc allows for both an incremental as well as an iterative approach for cavity calculations. This is applicable to both gas
as well as fluid filled cavities. The default is an incremental approach and this can be changed to an iterative approach in
the Cavity Parameters section under the Jobs menu in Mentat. In an incremental approach, the volumes used for cavity
calculations are from the start of the increment whereas for an iterative approach they are updated at every iteration. Hence,
results for the same problem may slightly vary depending upon which approach is chosen.
Cavity Modeling
The CAVITY parameter is used to enter the number of cavities in the model (maximum 1000), an upper bound to the number
of segments in each cavity and an upper bound to the number of nodes per segment of cavity boundary.
The CAVITY model definition option is used to enter the ambient pressure, the polytropic process exponent, and the
reference pressure, temperature and density for each cavity.
The applied pressure or mass is entered through the DIST LOADS model definition and history definition options. The distributed
load type, entered on the DIST LOADS option of the elements forming the cavity, is modified for cavity loading according to
the following relation
ibody_cavity = icavity * 10000 + icavity_type * 1000 + ibody (8-25)
CHAPTER 8 671
Boundary Conditions
where
ibody_cavity is the cavity-modified value for the distributed load type.

icavity is the cavity ID.
icavity_type is the cavity load type:
0: cavity is closed.
1: cavity is loaded with an applied pressure.
2: cavity is loaded with an applied mass.
9: cavity load is defined by the UCAV user subroutine
ibody is the original value for the distributed load type (see library element description in Marc Volume B: Element
Library .)
In the first increment of the analysis, increment zero, the cavity temperature is calculated by averaging the temperatures of
the elements forming the cavity. If there are no nodal temperatures defined, the cavity temperature is taken equal to the
cavity reference temperature. For AXITO3D (or PRE STATE) analysis, the cavity pressure, mass, and temperature for increment
zero are read from the post file of the corresponding axisymmetric problem. If the cavity is closed in increment zero, the
cavity is assumed to be initially at ambient conditions. In this case, the initial gage pressure is equal to zero and the initial
mass is based on the ambient pressure and initial volume. Moreover, if the ambient pressure is equal to zero, the cavity is
assumed to be initially unloaded. Tables 8-3 and 8-4 summarize the cavity functioning for the different cavity load types
and incremental load values during the initial and subsequent load increments, respectively.
Table 8-3 Cavity Functioning for Increment Zero

Cavity Load Type Incremental Load Value Function
0 Ignored Cavity is initially at ambient conditions
1 p = 0.0 Cavity is initially empty
1 p  0.0 Cavity is initially at the applied pressure
2 M = 0.0 Cavity is initially empty
2 M  0.0 Cavity initially contains the applied mass
9 Passed to UCAV Call UCAV
Table 8-4 Cavity Functioning for Subsequent Increments

0 Ignored Cavity is closed, constant mass, polytropic process
1 p = 0.0 Cavity is open, constant pressure and temperature
Table 8-4 Cavity Functioning for Subsequent Increments

1 p  0.0 Cavity is open, pressure is applied, constant temperature
2 M = 0.0 Cavity is closed, constant mass, isothermal process
2 M  0.0 Cavity is open, mass is applied, constant temperature
9 Passed to UCAV Call UCAV
In general, standard structural elements are used to define the boundaries of cavities and no extra elements are required.
However, to model the boundaries of cavities in regions where standard finite elements are not present (for example, along
rigid boundaries) cavity surface elements (elements 171-174) can be used. These elements can also be glued to moving rigid
surfaces. They are for volume calculation purposes only and do not contribute to the stiffness equations of the model.
For 2-D problems, the volume of the cavity is calculated as the sum of the areas of all the triangles formed by the cavity
segments as bases and the coordinate system origin as the apex multiplied by the cavity thickness. Elements forming the
cavity are assumed to be of equal thickness. For 3-D problems, the volume of the cavity is calculated as the sum of the
volumes of all the tetrahedrons formed by the cavity patches as bases and the coordinate system origin as the apex. For
axisymmetric problems, the volume of the cavity is calculated as the sum of the volumes of all the cone frustums formed
by the cavity segment as the cone slant height and two parallel base circles and with an axis along the axis of symmetry of
the problem. In the axisymmetric case, it is not necessary to use cavity surface elements along lines that are perpendicular
to the axis of symmetry to close cavities.
In coupled thermo-mechanical analysis, heat transfer between the gas inside the cavity and the surrounding structure is not
supported.
Cavities are not allowed to split or to join due to deformation or self-contact. If self-contact occurs within a cavity, the
resulting cavities are still treated as a single cavity.
If the cavity is formed out of membrane and/or shell elements, the user must ensure that all elements defining the cavity
are aligned; that is, the cavity surface is defined by either all top or all bottom element faces.
Cyclic Loading
Application of cyclic loading is important when investigating fatigue issues in a structure. There are two ways of easily doing
this in Marc. In the first method, one associates a table that is time dependent with each boundary condition. This table
can be either a sawtooth function or a sinusoidal function. The boundary condition will be repeated for as long as one has
a time period assigned. note that the AUTO STEP option can either exactly reach the peaks or not be constrained to reach the
peaks.
The alternative is to use the BEGIN SEQUENCE and END SEQUENCE options. All history definition options placed between these
two options will be repeated as often as requested. One should make sure that the value of the boundary conditions at the
beginning of the sequence is the same as the value at the end of the sequence. Multiple loadcases may be repeated.
CHAPTER 8 673
Boundary Conditions
Thermal Loads
Element integration point temperatures are used in a thermal stress analysis to generate thermal load. The AUTO THERM
history definition option allows automatic application of temperature increments based on a set of temperatures defined
throughout the mesh as a function of time. The CHANGE STATE option presents the temperatures to Marc which then creates
its own set of temperature steps based on a temperature change tolerance provided through this option. Table 8-5 lists input
options for thermal loads. You can input either the incremental temperature or the total temperatures as thermal loads.
Table 8-5 Input Options for Thermal Loads

Input Options
Load Description Parameter Model Definition History Definition User Subroutine
Incremental Temperatures* THERMAL THERMAL LOADS** THERMAL LOADS** CREDE
Total Temperatures* THERMAL CHANGE STATE CHANGE STATE NEWSV
Initial Temperature* INITIAL STATE INITSV
Total Nodal Temperatures POINT TEMP POINT TEMP
Initial Nodal Temperature INITIAL TEMP USINC
Total Nodal Temperature MAP TEMP
Alternative Mesh
* Temperatures must be specified at each integration point (or at the centroid if the CENTROID parameter is used) of
each element in the mesh.
** The THERMAL LOADS option and CREDE user subroutine should not be used with the table driven input procedure.
The thermal strain increment is defined as

th
 =  T (8-26)
th
where  is thermal strain increment,  is the coefficient of thermal expansion, and T is the temperature increment.
Equivalent nodal forces are calculated from the thermal strain increment and then added to the nodal force vector for the
solution of the problem.
You can input the coefficient of thermal expansion through the ISOTROPIC, ORTHOTROPIC, or other material options and the
temperature increment through various options.
Initial Stress and Initial Plastic Strain

Marc allows you to enter a set of initial stresses that simulate the stress state in the structure at the beginning of an analysis.
A typical example is prestress in a tensioned fabric roof. The set of initial stresses must be self-equilibrating and should not
exceed the yield stress of the material. Table 8-6 shows the input options for initial stress.
Table 8-6 Input Options for Initial Stress and Initial Plastic Strain
Input Options
Load Description Parameter Model Definition User Subroutine
Initial Stress Prestress ISTRESS INIT STRESS UINSTR
Initial Plastic Strain INITIAL PLASTIC STRAIN INITPL
Marc also provides various ways of initializing the equivalent plastic strain throughout the model. This is useful in metal
forming analysis in which the previous amount of equivalent plastic strain is often required. This history dependent
variable represents the amount of plastic deformation that the model was subjected to, and is used in the work (strain)
hardening model. The input option is shown in Table 8-5.
Inherent strains
Predicting residual stresses and distortion in welding and additive manufacturing (AM) is important to ensure quality of
the fabricated component. But it becomes very time consuming to simulate large weld assemblies or multi layered AM parts
using coupled thermo-mechanical analysis. Inherent strains method is an approximate alternative for the simulation of
welding and additive manufacturing simulations. It is suitable to capture the quick structural response with some
permissible error. In this method, the combined effect of plastic strains, thermal strains, creep strains and phase
transformation strains is simulated using inherent strains as a structural boundary condition. The portion of the assembly
which is subjected to high temperature during welding process is known as the Heat Affected Zone (HAZ). Based on the
application needs, there are two ways in which the HAZ can be defined:
1. User defined
2. Weld kinematics
User defined: This option is typically used to capture the structural response at the end of the manufacturing phase. In this
option, the user precisely selects the list of entities for the given set of inherent strain values. The structure is usually treated
as linear elastic with the inherent strains applied either over a number of increments or as a one-increment total value. The
inherent strain values can be varied using tables. Table independent variables supported are: time, normalized time,
increment number, normalized increment, X, Y, Z, temperature, arc length, X0, Y0, Z0. The inherent strain values can be
defined in the element system or the material coordinate system. Note that for continuum elements, the element system is
the global coordinate system. Additionally, user subroutine UINHSTRN can be used to redefine / modify incremental
inherent strain values.
Weld kinematics: As the name suggests, this option for heat affected zone entity selection is based on the weld torch
trajectory. This option is typically used to capture incremental structural response during the welding phase. It uses the
existing Marc welding setup for a precise and incremental selection of elements for inherent strain assignment. For the weld
kinematics option, three additional sets of data are required.
1. Weld pool dimensions- Depending on the application, the user can choose between double ellipsoid, cylindrical or
conical shape for the weld pool and define appropriate dimensions. The weld pool dimensions can be optionally
varied with the help of tables. Shape and size of the pool is only used for entity selection purpose. It does not change
the input inherent strain values.
Note that these dimensions are also used for the “Node in Weld Pool” adaptive meshing criterion.
CHAPTER 8 675
Boundary Conditions
2. Weld pool motion - To define the pool motion, the user needs to define the weld path and the velocity of the torch.
Weld velocity can be varied using table data. The actual position of a weld heat source can be offset by a user
specified amount from the associated weld path. The offsets are specified in the local x, y directions. Refer to Figure
6-28 for a definition of the local x, y and z directions. These offsets allow the same weld path to be used for multiple
weld pools and is particularly useful for multilayered additive manufacturing where the weld pool repeatedly
moves in a similar pattern, but offset from the original path by a small amount. User can also define the z offset. In
this case, the weld pool will move offset z distance along the weld path before selecting entities. User can also define
the initial position of the weld pool by specifying X, Y and Z coordinates. The weld motion along the path is
governed by the velocity of the torch and the loadcase time. In addition, the user can define the distance travelled
by the weld pool. After this distance has been traversed, no more entities will be selected for inherent strain
assignment.
3. Weld filler definition – To model metal deposition, the user can refer to WELD FILL data. , Existing weld filler
options for “Deactivated” and “Quiet” are available for the inherent strain boundary condition. Similar to welding
process, filler elements falling in the activation box are activated. The filler activation box can be symmetric or
asymmetric – more details are provided in the Weld Filler Section of Volume A. Weld filler features like “Melt
Point Temperature” and “Temperature activation time” are not supported with inherent strains boundary
condition.
Refer to Marc Volume A Theory and User Information, Chapter 10: Nonstructural Procedure Library, WELD FLUX,
WELD FILLER and WELD PATH for more details.
The inherent strain boundary condition for weld kinematics follows a similar logic as the one for welding. The steps are as
follows:
a. A list of possible base material and filler material elements are provided by the user. Inherent strain values and
associated tables are separately specified for base and filler material. For each material, the user can choose to
define the inherent strains in one of three systems: global coordinate system OR material coordinate system OR
weld local coordinate system for inherent strain values. User subroutine UINHSTRN can be used to redefine
/ modify incremental inherent strain values.
At present, analysis with inherent strains boundary condition is only supported with additive plasticity method. If it is
necessary to use multiplicative decomposition for some element groups in the model then feature 20601 can be used in
conjunction with large strain Automatic option and suitable large strain preferences.
Heat Fluxes
In a heat transfer analysis, you can enter heat fluxes in various forms. Heat transfer analysis requires entering the total values
of flux. Table 8-7 lists the input options for heat fluxes.
Table 8-7 Input Options for Heat Fluxes

Input Options
Point Heat Flux (Sink or Source) POINT FLUX POINT FLUX FORCDT
Surface Heat Flux, Convection, Radiation DIST FLUXES DIST FLUXES FLUX*
QVECT QVECT UQVECT
FILMS FILMS FILM*
SURFACE ENERGY
Volumetric Flux Load Body Flux DIST FLUXES DIST FLUXES FLUX*
FILMS
* Can be used for complicated flux loadings, convection, and radiation, allowing the input of nonuniform temperature-
and time-dependent boundary conditions.
There are three special heat flux conditions representing insulation, convection, and radiation as discussed in Chapter 6:
Nonstructural and Coupled Procedure Library in the Boundary Conditions section.
1. Insulation
q = 0 (8-27)
No input is required for the insulated case.
2. Convection
Nondirected:
q = H  Ts – T  (8-28)
You must enter the film coefficient H and ambient temperature T through the FILMS model definition option or
the FILM user subroutine. You can also directly input the heat flux q using the FLUX user subroutine.
Directed
q = –q0    n  ns (8-29)
˜
where n is the orientation defined in the QVECT option and n s is the outward unit normal.
˜
3. Radiation
4 4
q =     Ts – T  (8-30)
CHAPTER 8 677
Boundary Conditions
You must enter either the heat flux q using the FLUX user subroutine or the temperature dependent film coefficient
H  T s T    and ambient temperature T using the FILM user subroutine. These relationships are shown in
Equation (8-31). Using the table driven input format, you can directly specify radiation to the environment through
the FILMS model definition option. The use of the FILM user subroutine is recommended to ensure a stable solution.
4 4
q =   T s – T  
3 2 2 3 (8-31)
=     Ts + Ts T + Ts T + T   Ts – T 
= H  T s T    T s – T  
where  is the Stefan-Boltzmann constant,  is emissivity, T s and T  are unknown surface and ambient
temperatures, respectively.
As an alternative for radiation between surfaces, Marc will calculate the viewfactors using the CAVITY DEFINITION and
RAD-CAVITY model definition options or the VIEW FACTOR model definition option can be used to read in viewfactors
calculated by Mentat. The RADIATION parameter is also required.
Mass Fluxes and Restrictors

In a hydrodynamic bearing analysis, you can enter mass fluxes, restrictors, and pump pressures as loads. Table 8-8 lists input
options for these quantities.
Table 8-8 Input Options for Mass Fluxes and Restrictors

Input Options
Load Description Parameter Model Definition History Definition User Subroutine
Nodal Mass Fluxes BEARING POINT MASS POINT MASS FORCDT
Distributed Mass Fluxes BEARING DIST MASS DIST MASS FLUX*
Restrictors BEARING RESTRICTOR URESTR**
RESTRICTOR
* Can be used for nonuniform mass fluxes
** Can be used for nonuniform restrictions or pump pressures.
Electrical Currents
In coupled thermo-electrical (Joule heating) analysis and coupled electrical-thermal-mechanical analysis (Joule-
mechanical), you can prescribe electrical currents as loads for the calculation of unknown nodal voltages. In magnetostatic
analysis, you can also define the current. In such cases, as a steady state analysis is performed, there is no time variation of
the currents. Table 8-9 lists input options for electrical currents.
Table 8-9 Input Options for Electrical Currents

Input Options
Nodal Current POINT CURRENT POINT CURRENT FORCDT
Surface and Body Currents DIST CURRENT DIST CURRENT FLUX
Electrostatic Charges
In an electrostatic analysis, the charge can be entered, noting that a steady state analysis is performed so there is no time
variation of charge. In a coupled electrostatic structural analysis there can be a variation of charge in time, but this variation
is considered to be quasi-static. Table 8-10 summarized the input options.
Table 8-10 Input Options for Electrostatic Charge

Input Options
Nodal Charge POINT CHARGE POINT CHARGE FORCDT
Distributed and Body Charges DIST CHARGES DIST CHARGE FLUX
Acoustic Sources
In acoustic analysis, you can enter a source pressure if a transient analysis by modal superposition is being performed.
Table 8-11 summarized the input options.
Table 8-11 Input Options for Acoustic Sources

Input Options
Nodal Source POINT SOURCE POINT SOURCE FORCDT
Distributed Source DIST SOURCES DIST SOURCES FLUX
Nodal Source FIXED PRESSURE PRESS CHANGE FORCDT
Piezoelectric Loads
In a piezoelectric analysis, both mechanical loads and electrostatic charges can be entered. These values can have time
variation if a transient analysis is performed or a harmonic excitation can be applied. Table 8-1 gives a summery of the
mechanical input options, and Table 8-10 gives a summary of the electrostatic input options.
CHAPTER 8 679
Boundary Conditions
Electrostatic-Structural Loads
In a coupled electrostatic-structural analysis, both mechanical loads and electrostatic charges can be entered. Table 8-1 gives
a summery of the mechanical input options, and Table 8-10 gives a summary of the electrostatic input options.
Magnetostatic Currents
In a magnetostatic analysis, the current can be entered, noting that a steady state analysis is performed so there is no time
variation of current. Table 8-12 summarizes the input options.
Table 8-12 Input Options for Magnetostatic Current

Input Options
Distributed and Body Current 2-D DIST CURRENT DIST CURRENT FLUX
Distributed and Body Current 3-D DIST CURRENT DIST CURRENT FORCEM
Magnetodynamic Currents and Charges

In an magnetodynamic analysis, the current can be entered. These values can have time variation if a transient analysis is
performed or a harmonic excitation can be applied. Table 8-13 summarizes the input options.
Table 8-13 Input Options for Currents and Charges

Input Options
Distributed and Body Current DIST CURRENT DIST CURRENT FLUX
Nodal Charge POINT CURRENT POINT CURRENT FORCDT
POINT CHARGE
Distributed and Body Charge DIST CHARGE DIST CHARGE
Kinematic Constraints
Marc allows you to input kinematic constraints through various options that include
 Boundary Conditions (prescribed nodal values)
 Transformation of Degree of Freedom
 Shell Transformation
 Tying Constraint
 Rigid Link Constraint
 Shell-to-Solid Tying
 Insert
 AUTOMSET
 Support Conditions
 Bushings
 Cyclic Symmetry
 MSC Nastran RBE2 and RBE3
 Beam - Shell Offsets
 Pin Code for Beam Elements
Boundary Conditions
Marc allows you to specify the nodal value for a particular degree of freedom. If you do not give a nodal value when you
specify the boundary condition, Marc sets the fixed nodal value to zero. An option allows boundary conditions to be
specified at the time of two-dimensional mesh generation with MESH2D. You can apply a different set of boundary
conditions for each load increment.
Table 8-14 gives the nodal values for the various analyses, and Table 8-15 lists the input options for boundary conditions in
different analyses.
A special boundary condition type is used in a co-simulation analysis, combining the multi-body dynamics program Adams
and the finite element program Marc. In such a co-simulation analysis, the displacements and rotations of the interface
nodes (connecting the finite element and multi-body dynamics models) are prescribed by Adams. See the FIXED DISP option
(with table input) in the Marc Volume C: Program Input manual for more details.
Table 8-14 Analyses with Corresponding Nodal Values

Analysis Nodal Values
Acoustics Pressure
Coupled Fluid Thermal Velocity, Pressure, and Temperature
Coupled Thermo-electrical Voltage and Temperature
Coupled Thermo-mechanical Displacement and Temperature
Coupled Electrical-thermo-mechanical Voltage, Temperature, and Displacement
Electrostatic Potential
Piezoelectric Displacement and Potential
Fluid Velocity and Pressure
Heat transfer Temperature
Hydrodynamic Bearing Pressure
CHAPTER 8 681
Boundary Conditions
Table 8-14 Analyses with Corresponding Nodal Values (continued)

Analysis Nodal Values
Magnetostatic Potential
Magnetodynamic Magnetic Vector Potential and Electric Scalar Potential
Rigid Plastic Flow Velocity
Stress Displacements
Table 8-15 Input Options for Boundary Conditions

Input Options
Displacement FIXED DISP DISP CHANGE FORCDT
Temperature FIXED TEMPERATURE TEMP CHANGE FORCDT
Voltage FIXED VOLTAGE VOLTAGE CHANGE FORCDT
Pressure FIXED PRESSURE PRESS CHANGE FORCDT
Potential (Electrostatic) FIXED EL-POT FORCDT
FIXED POTENTIAL
Potential (Magnetostatics) FIXED MG-POT
FIXED POTENTIAL
Potential (Magnetodynamics) FIXED EL-POT
FIXED POTENTIAL
Velocity FIXED VELOCITY VELOCITY CHANGE FORCDT
Transformation of Degree of Freedom

Marc allows transformation of individual nodal degrees of freedom from the global direction to a local direction through
an orthogonal transformation that facilitates the application of boundary conditions and the tying together of shell and
solid elements.
Transformations are assumed to be orthogonal. Once you invoke a transformation on a node, you must input all loads and
kinematic conditions for the node in the transformed system.Nodal output is in the transformed system. This option is
invoked using the TRANSFORMATION option.
The UTRANFORM option allows transformations to be entered via the UTRANS user subroutine. This allows you to transform
the degrees of freedom at an individual node from global directions to a local direction through an orthogonal
transformation. UTRANFORM allows you to change the transformation with each increment. When you invoke this option,
the nodal output is in both the local and the global system.
Shell Transformation
The SHELL TRANSFORMATION option allows you to transform the global degree of freedom of doubly curved shells to shell
degrees of freedom in order to facilitate application of forces in the shell directions, edge moments, and clamped or simply
supported boundary conditions. There are four types of shell transformations.
The SHELL TRANSFORMATION model definition option specifies information on the shell transformation. For Types 1 and 3,
only the node number has to be specified. For Types 2 and 4, a boundary direction (the direction cosine of t as shown in
Figure 8-24) also has to be specified in the  1  2 surface.
2
1
Figure 8-24 Boundary Directions in Shell Transformation
After transformation, the following definitions apply.

Type 1: Transformation for two-dimensional beams and shell nodes (Element Types 15, 16, and 17). The
transformation defines the degrees of freedom with respect to a local coordinate system (s, n) (see
Figure 8-32). The four degrees of freedom after transformation are:
1 = us tangential displacement
2 = u n normal displacement
3=  rotation
4=  Meridional stretch
x2
n s
Positive
Direction
x1
Figure 8-25 Type 1: Shell Transformation
CHAPTER 8 683
Boundary Conditions
Type 2: Transformations for doubly curved shell nodes with nine degrees of freedom (Element Type 4 corner
nodes of Element Type 24). The transformation defines a local coordinate system ( t s n ). (See
Figure 8-26).
The nine degrees of freedom after transformation are:
1 = ut displacement in specified (boundary) direction
2 = us displacement normal to (boundary) direction but tangential to shell surface
3 = un displacement normal to shell surface
4 = t rotation of shell around boundary
5 = s rotation of shell around normal to boundary tangential to the shell surface
6 = n rotation of boundary around normal to shell surface
7 = t stretch tangential to specified (boundary) direction
8 = s stretch normal to specified (boundary) direction
9 = t s shear stretch in t-s direction
n 2
t
s
1
x3
x2
x1
Figure 8-26 Types 2 and 4: Shell Transformations
Type 3: Transformations at midside nodes for doubly curved shell nodes with three degrees of freedom (Element
Type 24). The transformation defines a local coordinate system ( t s n ), (see Figure 8-27). The degrees of
freedom after transformation are:
1 = u t  s rotation of normal to boundary around normal to shell

2 = u s  s stretch normal to boundary
3 = u n  s rotation around boundary
n t
x3 s
x2
x1
Figure 8-27 Type 3: Shell Transformation
Type 4: Transformation for doubly curved nodes with 12 degrees of freedom (Element Type 4). The
transformation defines a local coordinate system ( t s n ). The first 9 degrees of freedom after
transformation are the same as 1 through 9 in Type 2 (Figure 8-26), and the remaining three are:
2
10 =  u t   1  2 variation of g along the boundary (8-32)
2
11 =  u s   1  2 variation of e along the boundary (8-33)
2
12 =  u n   1  2 variation of f along the boundary (8-34)
When using the SHELL TRANSFORMATION option, the displacement increments and reaction forces are output in the local
directions immediately after solution of the equations. At the end of the increment, you can print out the global
displacement increments and total displacements in the global coordinate directions.
If you invoke the FOLLOW FOR parameter, the SHELL TRANSFORMATION option defines a local coordinate system in the current
(updated) geometry of the structure. This additional feature is especially useful with the LARGE STRAIN parameter, because
you can then specify edge moments and/or large edge rotations of shells and beams.
Caution: If you apply a shell transformation to a node, do not apply a standard transformation or shell tying type to that
node.
Tying Constraint
Marc contains a generalized tying (constraint) condition option. Any constraint involving linear dependence of nodal
degrees of freedom can be included in the stiffness equations.
A tying constraint involves one tied node and one or more retained nodes, and a tying (constraint) condition between the
tied and retained nodes. The degrees of freedom (for example, displacements, temperatures) of the tied node are dependent
on the degrees of freedom of the retained nodes through the tying condition. In some special tying conditions, the tied
node can also be a retained node. The tying condition can be represented by a tying (constraint) matrix. Note that if the
tying constraint involves only one retained node, the choice of which node is to be tied to retained is arbitrary.
As a simple example, impose the constraint that the first degree of freedom of node I be equal to that of node J at all times
(see Figure 8-28).
CHAPTER 8 685
Boundary Conditions
As a second example, the simulation of a sliding boundary condition requires the input of both the boundary conditions
and the tying constraints (see Figure 8-29).
I UI
Constraint Equation:
UI = UJ
2,V
J UJ
Tied node I, retained node J, or
1,U
Tied node J, retained node I
Figure 8-28 Simple Tying Constraint
Local axes
X
Y (v) i Y
j
k
l
X (u)
Figure 8-29 Tying Constraint Illustration (Sliding Boundary Conditions)
The example illustrated in Figure 8-29 enforces rigid sliding on the boundary in the local coordinates defined above.
vi = vj = vk = v1 = 0 (8-35)
ui = uj = uk = u1 (8-36)
The first equation is a set of fixed boundary conditions. The second equation is a constraint equation and can be rewritten
as three constraint equations:
ui = uj (8-37)
uk = uj (8-38)
u1 = uj (8-39)
These equations express all the u displacements in terms of u j . In this example, node j is chosen to be the retained node;
nodes i, k, and l are tied nodes. You can use the TYING option to enter this information.
Marc has a number of standard tying constraints that can be used for mesh refinement, shell-to-shell, shell-to-beam, beam-
to-beam, and shell-to-solid intersections. Tables 8-16 through 8-19 describe these options. Tables Table 8-21 through
Table 8-23 show the tying constraints for pipe elements, shell stiffeners, and nodal degrees of freedom. Table 8-24 summarizes
the rigid link constraint.
The SERVO LINK option uses the homogeneous linear constraint capability (tying) to input simple constraints of the form
u t = a 1 u r1 + a 2 u r 2 +  (8-40)
where u t is a degree of freedom to be constrained; u r 1 u r 2  are the other retained degrees of freedom in this structure;
and a 1 a 2  are constants provided in this option.
You can use the TYING or SERVO LINK model definition option to enter standard tying constraint information, and the TYING
CHANGE option to change tying constraints during load incrementation.
The UFORMSN user subroutine is a powerful method to specify a user-defined constraint equation. This constraint can be
nonlinear; for example, it can be dependent on time or previous deformation.
Table 8-16 Tying Constraints for Mesh Refinement

Number of
Tying Code Retained Nodes Purpose Remarks
31 2 Refine mesh of first order Tie interior nodes on refined side to corner
(linear displacement) elements in two nodes in coarse side (see Figure 8-30)
dimensions
32 3 Refine mesh of second order (quadratic Tie interior nodes on refined side to edge
displacement in two dimensions of element of coarse side (see Figure 8-31)
33 4 Refine mesh of 8-node bricks Tie interior node on refined side to four
corner nodes of an element face on coarse
side (see Figure 8-32)
34 8 Refine mesh of 20-node bricks Tie interior nodes on refined side to eight
(four corner, four midside) nodes of
element face on coarse side (see Figure 8-33)
T – Tied Node
T R – Retained Node
R
Figure 8-30 Mesh Refinement for 4-Node Quad
CHAPTER 8 687
Boundary Conditions
T
R T – Tied Node
R – Retained Node
T
R
Figure 8-31 Mesh Refinement for 8-Node Quad
T
R
R
Figure 8-32 Mesh Refinement for 8-Node Brick
R
T
R
T
R
T
R
R
Figure 8-33 Mesh Refinement for 20-Node Brick
Table 8-17 Tying Constraints for Shell-to-Shell Intersection

Tying Number of Retained
Code Nodes Purpose Remarks
22 2 Join intersecting shells, Element Type 4, 8, Tied node is also the second retained
or 24; fully moment-carrying joint node*
18 2 Join intersecting shells, Element Type 4, 8, Tied node is also the second retained node
or 24; fully moment-carrying joint
28 2 Join intersecting shells, Element Type 4, 8, Tied node is also the second retained node
or 24; pinned point
24** 2 Join intersecting shells or beams, Element Tied node is also the second retained node
Types 15-17
* Thickness vector must be specified at tied nodes and retained nodes.
** See Figure 8-34.
A
B
S
S
1
Figure 8-34 Standard Tying Type 24, Tie Shell-to-Shell or Beam-to-Beam; Moment-Carrying
Table 8-18 Tying Constraints for Beam-to-Beam Intersection

Number of
13 2 Join Two Elements Type 13 under an Tied node also the second retained node
arbitrary angle; full moment-carrying
joint
52 1 Pin joint for Beam Type 14, 25, 52, 76-79,
98
53 1 Full moment-carrying joint for Beam Tie interior node on refined side to four
Types 14, 25, 52, 76-79, 98 corner nodes of an element face on coarse
side (see Figure 8-32)
CHAPTER 8 689
Boundary Conditions
Table 8-19 Tying Constraints for Shell-to-Solid Intersections

Number of
23 1 Tie axisymmetric solid node to Tied and retained nodes must be
axisymmetric shell transformed to local system and
(Element Type 1 or 89) node TRANSFORMATION option invoked
(see Figure 8-35).
25 2 Join solid mesh to beam (Type 16) Tied node also second retained node. (see
Join solid mesh to axisymmetric shell Figure 8-36)
(Type 15)
26 2 Tie axisymmetric solid node to Similar to tying type 23 but no
axisymmetric shell transformation is required. Tied node also
(Element Type 1 or 89) node second retained node. (see Figure 8-36)
T1
y’
R
T2 x’
Figure 8-35 Standard Tying Type 23, Tie Solid-to-Shell (Element Type 1 or 89)
T1
T2
R
 T3
B T4
A 
Figure 8-36 Standard Tying Type 25, Tie Solid-to-Beam (Element Type 16)
Standard Tying Type 25, Tie Solid-to-Axisymmetric Shell (Element Type 15)
Standard Tying Type 26, Tie Solid-to-Axisymmetric Shell (Element Type 1 or 89)
Table 8-20 Tying Constraints for Beams to Element Edge or Face or Axisymmetric Shell to Element Edge
Tying Code Number of Retained Nodes Purpose Remarks
44 2 is 2-D or axisymmetric Rigidly tie a node with The number of retained nodes
lower-order element edge displacements and rotations to a is required. This tying type fully
surface patch. This is internally supports large
3 is 2-D or axisymmetric
used for CWELD and CFAST option. deformation/rotations. No
higher-order element edge
transformations are required. See
3 if 3-D lower-order triangular Figure 8-37.
face
4 if 3-D lower-order
quadrilateral face
6 if 3-D higher-order
triangular face
8 if 3-D higher-order
quadrilateral face
CHAPTER 8 691
Boundary Conditions
R
T T
R
R
2-D Application Lower-order 2-D Application Higher-order
R R
R R R R
T
T R
R
R
3-D Application Lower-order Quadrilateral Face 3-D Application Higher-order Triangular Face
Figure 8-37 Beam to Edge Moment Carrying Constraints

Table 8-21 Tying Constraints for Pipe Bend Element (Elements 14 and 17)
Tying Code Number of Retained Nodes Purpose
15 One less than the number of shell nodes Special tying types for pipe bend Element 17 to remove rigid
in the z-r plane of the section body modes
16 Number of shell nodes in the z-r plane of Special tying types for pipe bend Element 17 to remove rigid
the section body modes
17 2 Special tying types for pipe bend Element 17 to couple bend
section into pipe bend
Table 8-22 Tying Constraints for Shell Stiffener (Element 13 as a Stiffener on Shell Elements 4 or 8)
Number of
19 2 Use beam Element 13 as a stiffener on shell Tied node also second
Elements 4 or 8. Tied node is beam node; first retained node
retained node is shell node, second is beam node
again. Beam node should be on or close to the
normal to the shell at the shell node
20 3 Create an extra node in a shell Type 8 Element Always used after tying Type 21
tied to the interpolated shell displacements with
tying Type 21 to tie a beam Element 13 or a
stiffener across a shell element
21 2 Same as Type 19, but tying beam to an Must be followed by tying Type 20
interpolated shell node not a vertex of an
element (Element Type 8 only) must be
followed by Type 20 to tie interpolated shell
node into shell mesh
CHAPTER 8 693
Boundary Conditions
Table 8-23 Tying Constraints for Nodal Degrees of Freedom

Number of
INDEG 1 Tie the Ith degree of freedom at the tied NDEG = number of degrees of freedom
node to the Ith degree of freedom at the per node
retained node
100 1 Tie all degrees of freedom at the tied node
to the corresponding degrees of freedom
at the retained node
>100 1 Generate several tyings of type NDEG Tying code is first degree of freedom
multiplied by 100 added to last degree of
freedom (209 means tie second through
ninth degree of freedom at tied node to
respective second through ninth degree of
freedom at retained node)
<0 User-defined User-generated tying type through the
(negative UFORMSN user subroutine
integer)
Rigid Link Constraint

Tying type 80 can be used to define a rigid link between nodes. This capability can be used for both small deformation and
large deformation, large rotation problems. In small deformations, a linear constraint equation is used. In addition to the
end points of the link, a second retained node must be given. This node is used to calculate and store the rigid body
rotations. For two-dimensional problems, this represents a rotation about the Z-axis and this is stored as the first degree of
freedom for the node. In 3-D, it should be noted that the rotations/moments about the x, y, and z axes for the second
retained node are stored and output as its first, second, and third degrees of freedom, respectively.
A complete rigid region can be modeled by using multiple tying type 80’s. In this case, the same two nodes are used for the
retained nodes in all of the constraints. Transformations can be applied to the retained nodes so that any kinematic or
force boundary conditions can be defined in a local coordinate system. It is not necessary that the same transformations
should be applied to both retained nodes. It is also optional to transform the tied nodes to the same or a different local
coordinate system.
The rigid links can be used with all elements except types 4, 8, 24, 15, 16, and 17. In addition, it should not be used in
rigid plastic analysis.
To reduce linearization errors in large deformation, large rotation problems using tying type 80, an optional recycling
feature is offered through the CONTROL model and history definition options. Recycling occurs if the maximum change over
the retained nodes of rigid link 80 exceed a rotation tolerance (default is 0.001 radians). This check can be circumvented
by setting the rigid link rotation tolerance to zero (in conjunction with the FEATURE,5701 parameter) or a negative number.
Table 8-24 Rigid Link Constraint

Number of
80 2 Define a rigid link between nodes The second retained node is an unattached node
which contains the rotation
Shell-to-Solid Tying
In many problems, a region exists that is modeled with both brick elements and shell elements. A particular case of this is
shown is Figure 8-38 and Figure 8-39.
Figure 8-38 4-Node Shell-to-Solid Automatic Constraint
Figure 8-39 8-Node Shell-to-Solid Automatic Constraint
In the first case, an 8-node brick which has been reduced to a triangular prism is connected to a 4-node shell. In the second
case, a 20-node brick is connected to an 8-node shell. An automatic constraint equation is developed between the elements.
Note that the thickness of the shell must be entered as the brick thickness.
Rigid Tying to a Surface Patch

Tying type 44 rigidly ties a node with its translation and rotation degrees of freedom to a surface patch. There are four
different tying types for 3-D analyses tying the displacements and rotations of a node (assuming it has 6 degrees of freedom)
to the motion of a 3- or 6-node triangular patch or a 4- or 8-node quadrilateral patch. There are two different tying types
for 2-D analyses, tying the displacements and rotation of a node (assuming it has three degrees of freedom) to the motion
of a 2- or 3-node line patch. The tied node can be a node of a beam or a shell element and will generally have translation
and rotation degrees of freedom. The retained nodes are the nodes of the patch and the only retained degrees of freedom
involved in the tying are the patch node translations. The exact tying type is determined from the tying type number (44)
CHAPTER 8 695
Boundary Conditions
and the number of retained nodes and, therefore, the latter is required in the input. The location where the tied node pierces
the patch is computed in terms of the local parametric coordinates that define locations inside a patch using the usual shape
functions for such a patch. The translations of this piercing point are computed from the nodal displacements of the patch,
again using the usual shape functions of the patch and these define the translations of the tied node. The rotations can be
computed by constructing a local triad to the patch at the piercing point. The translations of the patch nodes define the
change in orientation of the triad. The orientation change of the triad defines the rotations of the tied node. The tied node
must have a proper projection on the patch inside its boundary. When it does not lie exactly on the patch, the distance will
be treated as an offset. The patch does not necessarily have to coincide with a shell or brick element face in a 3-D model
or a beam or quad element edge in a 2-D model, but the nodes constituting a patch must adhere to the conventions that
are generally required for the particular patch. The patch also does not have to be a flat surface in 3-D or a straight line in
2-D, but it is recommended to avoid overly distorted patches. A patch is considered overly distorted when the patch
normals at two of its nodes make an angle larger than 10 degrees.
In case of a heat transfer or a thermo-mechanically coupled analysis the temperatures may be tied as well. The temperature
constraints are similar to the mechanical displacement constraints, but only one degree of freedom is involved. When the
retained nodes are shell nodes, the temperature is always tied to the first degree of freedom of the shell nodes (i.e., the top
surface of the shell), regardless of the orientation in space. When the tied node is a shell node, its first degree of freedom is
tied. In other analysis procedures, tying type 44 is not supported.
When no geometric nonlinearities are flagged in the model (i.e., neither the LARGE DISP nor the LARGE STRAIN parameter is
activated), Marc uses the small displacement method to apply the constraints. In all other cases, the large displacement
method is used fully accounting for finite rotations. The small displacement method linearizes the constraint equations
w.r.t. the reference configuration. The large displacement method takes account of all geometric nonlinearities caused by
finite displacements and rotations.
e3 e2
 e1

GA
Patch at end
of increment
Patch at start
of increment
Figure 8-40 Motion of a Patch From Start to End of an Increment
The large displacement method (Figure 8-40) can be summarized as follows. The point GA is the node of, for example, a
beam and it pierces the patch at some point for which the parametric coordinates  G A ,  G A have been computed. A local
triad is fixed to the beam cross section and another triad to the piercing point on the patch. For simplicity, it is assumed
that both triads have the same orientation in the initial configuration. The tying constraints are that the displacements of
the beam node and the displacements of the piercing point on the patch are equal and that the rotations of the triad
connected to the beam cross-section and the triad connected to the patch are equal. The rotations of the triad fixed to the
beam cross-section are expressed in terms of the rotational degrees of freedom of the beam node. The rotations of the triad
fixed to the patch are expressed in terms of the displacement degrees of freedom of the patch nodes. In this way, the
nonlinear relations are established between the beam rotations and the patch displacements. The relations between the
displacements of the beam node and the displacements of the patch nodes are linear when the beam node lies exactly on
the patch. In case of an offset, nonlinear contributions due to the rotations are added to the constraints.
The tied node generally has translation and rotation degrees of freedom, but this is not a requirement. In case it has only
translation and rotation degrees of freedom, the displacements are tied to the patch, possibly accounting for an offset. No
constraints are applied for rotations; hence, adequate constraints must be applied to prevent a singular system.
The connectivity conventions for the patches are shown in Figure 8-41 for 2-D and 3-D models:
Line2
1 2
2-D Model
Line3
1 2 3
(a) Line Patch in 2-D Analysis
4 3
Quad4
1 2
3-D Model
4 7 3
Quad8 8 6
1 5 2
(b) Quadrilateral Patch in 3-D Analysis
3
Tria3
1 2
3-D Model
3
Tria6 6 5
1 4 2
(c) Triangular Patch in 3-D Analysis
Figure 8-41 Patch Connectivity Convention
Overclosure Tying
In various engineering applications, it is necessary to define a pre-stress in, for example, bolts or rivets before applying any
other structural loading. Although such a pre-stressed state is often simulated using a temperature loading, it is rather
CHAPTER 8 697
Boundary Conditions
difficult to arrive at a desired net force in the bolt or rivet. An easier way is to use overclosure tyings (type 69). These tyings
create overlaps or gaps between two parts of a model. If the motion of these parts is somehow constrained in the direction
in which the gap or overlap is being created, then an overlap will introduce an tensile (pre-) stress in each of the parts and
a gap will result in a compressive stress.
Overclosure tyings have one tied node and two retained nodes. The tied node and the first retained node are usually nodes
on the boundaries of the respective parts (see Figure 8-42). The second retained node is very often a free node and is usually
shared by all overclosure tyings which connect the parts. This node is also called the control node of the tying, since it can
be used to apply load or control the size of the gap or overlap between the parts.
top part top part Fcontrol

top nodes mesh split F1,bot F2,bot
(first retained) overclosure tyings
u1,bot u2,bot ucontrol
(overlap) ucontrol
control node
(second retained) u1,top u2,top
bottom nodes F1,top F2,top
(tied)
bottom part bottom part
undeformed deformed
Figure 8-42 Pre-stressing a Structure by Creating an Overlap Between the top and the Bottom Part Using Overclosure
Tyings
An overclosure tying imposes the following constraint on the model:
u tied = u retained + u control , (8-41)
in which u tied , u retained and u control are the displacement and rotation degrees of freedom (if any) of, respectively, the
tied node, the first retained node, and the control node of the tying. It immediately follows from this equation that
u control is the displacement difference of the tied and the retained node of the tying and is equal to the size of the overlap
or gap between the parts. Hence, by prescribing this displacement using the FIXED DISP and DISP CHANGE options, gaps or
overlaps of a particular size can be created.
Instead of prescribing the size, gaps or overlaps can also be created by prescribing the total force in the model. This follows
immediately from the fact that the work done by a constraint equation is zero. To demonstrate this, consider the model
displayed in Figure 8-42. The model is split into two disjoint parts: the top part and the bottom part. At the split,
corresponding nodes of the respective parts are connected by overclosure tyings, in which the nodes of the bottom part are
the tied nodes and the nodes of the top part are the first retained nodes of the tyings. Both tyings share a common (free)
control node. If u 1 bot and u 2 bot are the displacements of the bottom nodes and u 1 top and u 2 top are the
displacements of the top nodes, then the two overclosure tyings impose the following two constraints on this model:
u 1 bot = u 1 top + u control , (8-42)
u 2 bot = u 2 top + u control . (8-43)

Introducing the vectors,

T T
u = u 1 bot u 2 bot u 1 top u 2 top and û = u 1 top u 2 top u control ,
and the constraint matrix

T
I 0 I 0
T = 0 I 0 I ,
I I 0 0
in which I and 0 are the unit and null matrices, Equations (8-42) and (8-43) can be summarized into the matrix equation:
u = Tû . (8-44)
Let
T
F = F 1 bot F 2 bot F 1 top F 2 top and T
F̂ = F̂ 1 top F̂ 2 top F̂ control
be the force vectors which are work conjugate to the displacements u and û , respectively. Then the zero-work principle
states that
T T
F u = F̂ û . (8-45)
Substitution of Equation (8-44) into Equation (8-45) and requiring that the result is valid for all û , it follows that
T
T F = F̂ , or
F 1 bot + F 1 top = F̂ 1 top , (8-46)
F 2 bot + F 2 top = F̂ 2 top , (8-47)
F 1 bot + F 2 bot = F̂ control . (8-48)
Equations (8-46) and (8-47) express the equilibrium of forces across the split, in which F̂ 1 top and F̂ 2 top can be viewed
as externally applied forces. While Equation (8-48) enables overclosure tyings to pre-stress a model with a certain total force.
It states that if the constraints described by Equations (8-42) and (8-43) are applied, the force on the control node is the sum
of the forces on the tied nodes of all overclosure constraints which share that control node. Hence, the total force on the
bottom part of the model is prescribed by applying a force to the control node using the POINT LOAD option. Obviously (by
CHAPTER 8 699
Boundary Conditions
the equilibrium Equations (8-46) and (8-47)) if no external force is applied to the nodes on the split, the total force on the
top part is equal but opposite is sign to the force on the control node.
Notes:  Both methods to control the size of overlaps or gaps can be combined in one analysis. For example, in
the first part of an analysis, a pre-stress can be defined by prescribing the net force in the structure
using the POINT LOAD option on the control node of the overclosure tyings. Then in the second part,
the size of the gap or the amount of overlap which is the result of this load, can be fixed by suppressing
the displacement changes of the control node using the DISP CHANGE option.
 If similar arguments are applied to tying 100 (which yields Equation (8-41) without the dependence on
the control node), then only the equilibrium Equations (8-46) and (8-47) are obtained. Equation (8-48)
is solely due to the dependence of Equation (8-41) on the control node.
The control node of an overclosure tying has the same displacement and rotation degrees of freedom as the tied and the
first retained node as given by Equation (8-41). This allows creation of overlaps or gaps in any particular direction or
combination of directions. Local coordinate systems can be defined at any of the nodes of an overclosure tying using the
TRANSFORMATION option. Since the constraint Equation (8-42) is always applied in the global coordinate system, a local
coordinate system at the control node can be used to pre-stress the model in directions other than the global coordinate
directions.
Sufficient boundary conditions must be applied on the control node to suppress any rigid body modes, if the two parts of
the structure are not constrained otherwise. Total forces on the split are available for postprocessing as reaction forces on
the control node for all suppressed or prescribed displacements and rotations.
Overclosure tyings can be used in combination with the CONTACT option, that is, the tied and the first retained node may
be nodes on the boundary of a contact body.
In non-mechanical passes of a coupled analysis, the overclosure tying reduces to tying type 100 between the tied node and
the first retained node. There is no dependence on the control node in that case. This guarantees continuity of the primary
field variable (for example, temperature) across the split.
Table 8-25 Overclosure Tying

Number of
Tying Retained
Code Nodes Purpose Remarks
69 2 Create gaps or overlaps between two parts of a The second retained node is the control node
model either by prescribing the total force on of the tying. The force on this node is equal to
the nodes on either side of the gap or overlap the total force on the tied nodes of all tyings
or by prescribing the size of the gap or overlap. that share this control node. The displacement
of the node is equal to size of the gap or
overlap between the parts. In non-mechanical
passes, the tying reduces to tying type 100
between the tied and the first retained node.
Cross Section
In various engineering applications, it is necessary to define a pre-stress in for example bolts or rivets before applying any
other structural loading. Although such a pre-stressed state is often simulated using a temperature loading, it is rather
difficult to arrive at a desired net force in the bolt or rivet. An easier way is to use the option CROSS-SECTION.
A cross section in a model is defined by (see Figure 8-43):
1. All the nodes in the cross section;
2. The normal vector to the cross section;
3. All the elements sharing nodes in the cross section and lying on the side of the cross section corresponding to the
opposite normal vector.
Figure 8-43 Cross section examples: regular mesh (left) and irregular mesh (right)
Notice that this option can be used for continuum as well as for beam and shell stress elements.
An additional node, called nc, has to be assigned to every cross section defined in a model. Such a node has only one degree
of freedom, namely the displacement in the direction of the cross section normal vector, which is indicated as unc. Elements
corresponding to the cross section definition now get a modified displacement field, depending on which nodes of the
elements are part of the cross section. If the third and fourth node of a 4-node element are part of the cross section, the
element displacement vector is changed according to (see Figure 8-44):
T T T T T T T
U e = [u 1 , u 2  u 3 + u n c n  u 4 + u n c n ] (8-49)
Similarly, if only the fourth node is part of the cross section, the element displacement vector is given by:
T T T T T T
U e = [u 1 , u 2  u 3  u 4 + u n c n ] (8-50)
CHAPTER 8 701
Boundary Conditions
Figure 8-44 Cross section with 4-node elements: two nodes in cross section (left), one node in cross section (right)
As a result, the following is possible:

1. A force in the cross section can be defined by specifying a POINT LOAD on the first degree of freedom of node nc.
Notice that this force works in the direction of the cross section normal vector. The resulting displacement of node
nc can be seen as the shortening of the cross section elements.
2. A constant overclosure in the cross section can be obtained by specifying the first degree of freedom of node nc via
the FIXED DISP or DISP CHANGE option. Then, the reaction force in node nc is the section force in the direction of the
cross section normal vector.
During an analysis, you can switch from one of the above options to the other. So, to simulate the behavior of a pre-stressed
bolt under additional loading conditions, you should first define the pre-tension force via the POINT LOAD option and
subsequently use the DISP CHANGE option to “lock” the bolt by setting the incremental displacement u n c to zero.
If you are only interested in the total force in the cross section, the displacement u n c should be set to zero. Then, the total
force in the cross section is given by the reaction force in node nc.
Insert
Marc provides an INSERT model definition option which allows the definition of host bodies and lists of elements or nodes
to be inserted in the host bodies. The degrees of freedom of the nodes in the inserted node list or element list are
automatically tied using the corresponding degrees of freedom of the nodes in host body elements based on their
isoparametric location in the elements.
The INSERT model definition option can be used to place reinforcing cords or rods, such as 2-D rebar membrane elements,
into solid elements.
The INSERT model definition option can be used to apply point loads in some specific locations other than element nodes.
It also can be used to link two different meshes.
If a node to be inserted is also a node of a host body element, no tying is applied to the node.
Transformation must not be used at any nodes of host body elements and at inserted nodes, unless the same set of local
coordinate system is used for all nodes involved.
AUTOMSET
This parameter automatically re-writes the constraint equations such that a tied degree of freedom is not used as a retained
degree of freedom in another constraint equation. It also modifies constraint equations where a constrained degree of
freedom is used in a prescribed boundary condition. Because of chaining of constraint equations, this is not always possible,
in which case the analysis will terminate. The constraint equations that are considered are those obtained from user-
specified options:
TYING
SERVO LINK
RBE2
RBE3
RROD
CONRAD GAP
Constraint equations internally generated are not considered from the following options:
CONTACT
INERTIA RELIEF
INSERT
If both the MPC-CHECK and AUTOMSET parameters are included in the model, the MPC-CHECK parameter is ignored.
For models with a large number of RBE2 links associated with a single retained node or with RBE3 associated with a large
number of reference nodes, AUTOMSET may estimate a very large memory requirement for the stiffness matrix since it
provides for a possible bandwidth increase associated with shuffling the equations. In order to avoid this, FEATURE,20701
can be used in conjunction with the AUTOMSET parameter. With FEATURE,20701, the program makes the initial
assumption that AUTOMSET will not be active for RBE2 and RBE3 links. An automatic check is then made on the tied
and retained nodes of the RBE2 and RBE3 to see if AUTOMSET really needs to be active for RBE2 / RBE3 as well. Only
if the check indicates that AUTOMSET needs to be active for RBE2 / RBE3 will it be turned on for these rigid links. Refer
to Volume C (FEATURE) for more details.
CHAPTER 8 703
Boundary Conditions
Support Conditions
Marc provides linear and nonlinear springs and foundations for the modeling of support conditions. For dynamic analysis,
a dashpot can also be included. Table 8-26 lists input options for linear and nonlinear springs and elastic foundations.
Table 8-26 Input Options for Springs and Elastic Foundations

Input Options
Springs SPRINGS USPRNG
Elastic Foundation FOUNDATION FOUNDATION USPRNG
The force in the spring is:

F = K  u2 – u1  + D  v2 – v1  (8-51)
where K is the spring stiffness, u 2 is the displacement of the degree of freedom at the second end of the spring, and u 1
is the displacement of the degree of freedom at the first end of the spring. In a dynamic analysis, D is the damping factor
and v 2 and v 1 are the velocities of the nodes.
You can specify the elements in Marc to be supported on a frictionless, elastic foundation. The foundation supports the
structure with a force per unit area (force per unit length, for beams) given by:
p n = Ku n (8-52)
where K is the equivalent spring stiffness of the foundation, and u n is the normal displacement of the surface at a point
in the same direction as p n .
The same conventions apply to the elastic foundation specification as for pressure specification, in terms of the face of the
element that is used. The force is applied whether the displacement is tensile or compressive.
Nonlinear spring stiffness for mechanical, thermal, and electrical analyses can be specified through tables using the
parameter and model definition option. For more details, refer to Chapters 2 and 3 in Marc Volume C: Program Input.
The USPRNG user subroutine allows you to supply or further modify a nonlinear spring stiffness or specify a nonlinear
foundation stiffness as a function of prior displacement and force.
Bushings
A generalized spring-damper capability is provided in Marc to model bushings between various components. They are
modeled using element types 194 and 195 for 2-D and 3-D applications, respectively. The connectivity of these elements is
generally comprised of two nodes which are specified through the CONNECTIVITY model definition option. Grounded
bushing elements can be defined by providing only the first end-node for connectivity. The properties of these bushing
elements are specified through the PBUSH or PFAST model definition option. These elements may be automatically generated
using the CFAST option.
Bushing Coordinate System

A unique coordinate system is used to specify the bushing properties. A number of options are available to define this
coordinate system:
 Local element coordinate system along the bushing element
 Global cartesian coordinate system, or
 A user-specified coordinate system specified through the COORD SYSTEM option.
The bushing coordinate system defined through global cartesian coordinates or through the user-specified COORD SYSTEM
option are maintained constant and are not updated for large displacement analysis.
The local element coordinate system can only be used for 2-node bushing elements. In this case, the local X axis is along
the element length and forms a perpendicular triad with the local Y and Z axes. In 2-D, the local Z axis is always taken as
(0,0,1) and the local Y axis is then perpendicular to the local X axis in the global x-y plane. In 3-D, the local Y-Z axes are
located by using an extra node or an orientation vector. If both are specified, the extra node takes precedence. The local Y
axis is first estimated by using the orientation vector or using the vector from the first bushing node to the extra node. The
local Z axis is then formed by taking the cross product of the local X and estimated local Y axis. The final local Y axis is
obtained by taking the cross product of the local Z and local X axes. For large displacement analysis, the local element
coordinate system is updated continuously. Mid-increment direction cosines are used to calculate the strains while end-
increment direction cosines are used to calculate the stiffnesses and internal bushing forces.
Nodal Offsets for Bushing Elements

The exact position of the bushing element can be modified by specifying nodal offsets. This offset point is generally defined
from the first end-node of the bushing element. It serves two purposes: It allows convenient modeling of the bushing at a
location away from the actual physical location. Also, it allows coupling between translational and rotational stiffnesses.
The offset point can be located along the bushing element axis, in the global cartesian coordinate system, or in a user-
defined coordinate system. In the first case, the offset point is located along the line joining the two end nodes of the
bushing element. The distance of the offset point from the first node S is specified by the user. In the second case, the offset
vector to the bushing location from the first end-node is specified.
The treatment of offsets for bushing elements closely follows the treatment of offsets for beam elements (described later in
Beam - Shell Offsets). Once the offset point location is specified, the offset vector from each node is determined. The
equations governing the treatment and update of the offsets are given by Equations (8-64) to (8-67). Bushing properties are
used to define the stiffness/damping/mass matrix at the offset position and are then transformed to the geometrical
position.
CHAPTER 8 705
Boundary Conditions
zelem
yelem
zelem Bush
Location CID
xelem
(S1,S2,S3) OCID
1 v
S.l yelem
2
(1-S) . l
Bush Location
1
2
xelem
Figure 8-45 Offset location Specification for Bushing Element
Bushing Properties
Five sets of properties can be provided for bushing elements depending on the analysis type. All properties are provided
through the PBUSH or PFAST model definition option:
 Mechanical stiffness - for static/dynamic/harmonic analysis
 Mechanical dashpot - for dynamic/harmonic analysis
 Mass - for dynamic/harmonic analysis
 Thermal conduction coefficient - for heat transfer/coupled analysis involving a thermal pass
 Electrical resistance coefficient - for electrical/coupled analysis involving a joule pass
Mechanical stiffness for bushing elements can be specified either directly in coefficient form or through force-displacement
curves. Note that 2-D bushing elements have two translational and one rotational degree of freedom and separate stiffnesses
can be specified for the X, Y and RZ degrees of freedom. Three-dimensional bushing elements have three translational and
three rotational degrees of freedom and separate stiffnesses can be specified for X, Y, Z and RX, RY, RZ degrees of freedom.
For nonlinear bushing elements, these stiffnesses can be varied through tables. These tables can be a function of time,
increment number, position coordinates, frequency (only for harmonic analysis), temperature, displacement (this variable
is mandatory when stiffness is specified via forces). Up to four independent variables can be used in the tables. Based on
the current stiffness value and the change in displacement between the end-nodes, the bushing forces are updated in the
specified bushing coordinate system.
Two forms of damping can be specified for the bushing elements: Nominal damping B i for the ith degree of freedom of
the bushing elements where the damping is specified either directly in coefficient form or through force-velocity curves.
For nonlinear dampers, the nominal damping can be varied through tables as a function of time, increment number,
position coordinates, frequency (only for harmonic analysis), temperature or velocity (this variable is mandatory when
nominal damping is specified via forces). The second form of damping is structural damping Si for the ith degree of
freedom of the bushing elements. The damping is specified in coefficient form only in this case and can be varied through
tables with similar independent variables as the nominal damping case. A composite damping coefficient is formed for the
bushing element as follows:
G 1
C i = B i + -------- K i + -------- S i K i
W3 W4
where C i is the total damping coefficient; G, W3, and W4 are structural coefficients that are specified through the
COEFFICIENT model definition option; B i is the nominal damping coefficient; S i is the structural damping coefficient; and
K i is the current stiffness coefficient.
For a linear bushing element, the total force is given by Equation (8-51). Note that for bushing elements defined in the local
element coordinate system, the force is positive if the bushing element is in tension and negative if it is in compression. For
bushing elements defined via the global coordinate or user-coordinate systems, the force along a particular degree of
freedom is positive if the displacements of end-node 2 along the specified degree of freedom is greater than the displacement
of end-node 1 along the specified degree of freedom. Note that if tables are used to define the bushing properties,
care should be taken to include both positive and negative values of independent variables like displacements/velocities
since these would influence the calculated stiffnesses.
For dynamic and harmonic analyses, lumped translational and rotational masses can be provided at the end-nodes of the
cbush elements. Uniform masses apply to translational and rotational degrees of freedom at the end-nodes and are
maintained constant through the analysis.
Stresses in the bushing elements can be related to the associated forces through recovery coefficients. Similarly, strains in
the bushing elements can be related to the associated deformations through recovery coefficients. Separate coefficients are
allowed for the translational components and the rotational components through the PBUSH option. If set to 1.0, the stresses
are equal to the forces and the strains are equal to the difference in displacements between the end-nodes.
The bushing elements can also be used for thermal or electrical field analyses. The coefficients of thermal conduction and
electrical resistance are specified via the PBUSH option. The bushing element behaves like a general link in this case and
dashpot properties are not considered. Tables can be used to vary the properties as a function of temperature (for thermal
conduction), voltage (for electrical resistance), time, increment number, coordinates.
Cyclic Symmetry
A special set of tying constraints for continuum elements can be automatically generated by the Marc program to effectively
analyze structures with a geometry and a loading varying periodically about a symmetry axis. Figure 8-46 shows an example
where, on the left-hand side, the complete structure is given and, on the right-hand side, a sector to be modeled.
CHAPTER 8 707
Boundary Conditions
Y Y
y’ x’
B

X X
A
Figure 8-46 Cyclic Symmetric Structure: Complete Model (Left) and Modeled Sector (Right)
Looking at points A and B on this segment, the displacement vectors should fulfill:
u' B = u A (8-53)
which can also be written as:
u B = Ru A (8-54)
where the transformation matrix R depends on the symmetry axis (which, in the example above, coincides with the global
Z-axis) and the sector angle  (see Figure 8-46). In Marc, the input for the CYCLIC SYMMETRY option consists of the direction
vector of the symmetry axis, a point on the symmetry axis and the sector angle  . The following items should be noted:
1. The meshes do not need to line up on both sides of a sector (for example, see Figure 8-47).
Figure 8-47 Finite Element Mesh for Cyclic Symmetric Structure with Different Mesh Densities on the Sector Sides
2. Any shape of the sector sides is allowed provided that upon rotating the sector 360   times about the symmetry
axis over the sector angle  will result in the complete model.
3. The CYCLIC SYMMETRY option can be combined with the CONTACT option. In this case, both sides of the cyclic
symmetry sector must belong to a contact body and corresponding parts of the sector sides must belong to the same
body (see Figure 8-48). Note that the contact bodies do not need to span between the sector sides.
Figure 8-48 Cyclic Symmetry and Contact Bodies; upper left: one body (correct); upper right: two bodies (correct); lower
left: two bodies (right); lower right: two bodies (wrong; corresponding parts of the sector sides are not in the same body).
4. The CYCLIC SYMMETRY option can be combined with global remeshing.

5. For Poisson type physics passes, the degrees of freedom are forced to be cyclic symmetric, i.e. (see
Figure 8-46) T A = T B for heat transfer,  A =  B for electrostatics, and A A = A B for 2-D magnetostatics.
6. For structural, magnetostatic and magnetodynamic passes, the nodal points on the cyclic symmetry axis are
automatically constrained in the plane perpendicular to this axis.
7. For structural passes, a possible rigid body mode about the symmetry axis can be automatically suppressed.
8. Cyclic Symmetry is valid for:
a. Only the continuum elements. However, the presence of beams and shells is allowed, but there is no connection
of shells to shells, so the shell part can, for example, be a turbine blade and the volume part is a turbine rotor.
The blade is connected to the rotor and if there are 20 blades, 1/20 of the rotor is modeled and one complete
blade.
b. Nonlinear static analysis including remeshing.
c. Analyses which consist of one or a combination of the following physics passes: Structural, Thermal,
Electrostatic, Magnetostatic, Magnetodynamic, and Electrical (from joule heating).
d. Eigenvalue analyses such as buckling or modal analyses, harmonic analyses, and transient dynamic analyses.
However, there are restrictions in the case of modal analyses which are described in more detail in the following
paragraphs.
9. In order to check which nodes get tying constraints due to cyclic symmetry, nodal post code 38 (contact status) can
be used; its value will be 2.
To prevent confusion, it must be emphasized that the cyclic symmetry feature described above is different than linear cyclic
symmetry commonly used in modal analysis where physical quantities such as, x n , displacements, forces, stresses and
CHAPTER 8 709
Boundary Conditions
k
temperature in the n-th segment are expanded in a Fourier series with terms of the cyclic components, u , in the
fundamental region, like:
K N
n – 1 ----
1 0 2 k c k s  –1  2
xn = ---- u + ----  u cos  n – 1 k + u sin  n – 1  k  + ---------------------- u (8-55)
N N N
k = 1
where k is the harmonic order; N is the total number of sectors;  is the fundamental inter-sector phase shift defined as
2  N ; and K is defined as:
N –1
-------------- if N is odd
 2
K =  (8-56)
 N –2
-------------- if N is even
 2
There are considerable savings in both computing time and data storage associated with the use of the linear cyclic
symmetry concept. Assuming a finite element model with a sector size of m degrees of freedom, a real-valued cyclic
symmetry approach leads, in the worst case, to one eigenvalue problem of size m and  N – 1   2 eigenvalue problems of
size 2m ; a complex approach leads to N eigenvalue problems of size m; while the full analysis leads to a single, but very
costly, eigenvalue problem of size Nm .
Although linear cyclic symmetry can reduce the problem size greatly, it is restricted to linear analysis, and the sector must
have its surface mesh on the symmetry planes to be identical on each side of the sector. The nonlinear cyclic symmetry
implemented in Marc can be used for nonlinear problems, such as contact, and the nodes do not need to line up on both
symmetry planes of the sector (for example, see Figure 8-47).
MSC Nastran RBE2 and RBE3

The basic formulation of the MSC Nastran rigid body elements RBE2 and RBE3 is similar to tying type 80, which is also
known as a rigid link. However, they offer additional flexibility because the degrees of freedom may be partly unconstrained
and a list of tied/retained nodes can be entered with RBE2 and RBE3.
RBE2
The nonlinear relation of an RBE2 can be expressed as follows:
x t = R  r   xt – xr  (8-57)
where x t is the coordinate of the tied node, x r is the coordinate of the reference node, and R   r  is the rotation matrix
of the retained node. The linearized relation between the incremental displacement of a tied node to a retained node and
its correction term is expressed as follows:
i i
u t = Su r + C non (8-58)
where the tying matrix S is derived from the linearized rigid body relation. For linear analysis, C n o n is zero. For large
i
displacement analysis, Su r may not match the nonlinear constraint exactly. Therefore, an error vector, C n o n , is
required to meet the constraint. It is defined as the difference between the expected coordinates ( x t e ), using the rigid body
kinematics, and the current coordinates ( x t ) of the tied node.
n+i n n+i 0 0
xt e = xr + R  r   xt – xr 
(8-59)
n+i n+i
C non = x te – xt
where n is increment number, i is iteration number, and zero is the original value. In order to reduce linearization errors,
an optional recycling feature is offered through the CONTROL model and history definition options. Recycling occurs if the
maximum iterative rotation change over the retained nodes of the RBE2 exceeds a rotation tolerance (default is 0.001
radians). This check can be bypassed by setting the rigid link rotation to zero (in conjunction with the FEATURE,5701
parameter) or a negative number.
When the rotation is finite, then the coordinate system attached to the tied node will be co-rotated according to the
rotation of the retained node. The degrees of freedom of the tied node will be assigned with this co-rotated system. As an
illustration, consider a slider link shown in Figure 8-49. In this case, a local transformation must be defined for the tied node
in which the x-direction is coincident with the line connecting the tied and the retained node. Then, an RBE2 is defined
where y-translational degree of freedom is tied but not the x-translational and any of the rotation degrees of freedom.
.
y x
Y E2 x
RB
RBE2
X (a) Increment i (b) Increment i + 1

Figure 8-49 Co-rotated Coordinate System at the Tied Node
As another example of the use of RBE2 options, consider Figure 8-50 where a cylindrical tube is shown. The tube is clamped
at one end and loaded by a torque at the other end.
If it is assumed that the loaded end remains circular with a constant radius, then a single RBE2 option can be easily used
instead of a number of tyings of type 80. With the RBE2 option, node 1 is defined as the retained or reference node and
nodes 2 to 13 are defined as the tied nodes, with all the degrees of freedom constrained.
However, if the tube is allowed to expand in radial direction, then tying type 80 cannot be used. However, RBE2 option can
still be used, but now the radial displacement is not included in the list of degrees of freedom to be constrained.
CHAPTER 8 711
Boundary Conditions
2
13
3
12
11 
4
Mz Z 1
10 5
R
Clamped end
6
9
7
8
Figure 8-50 RBE2 Example: Cylindrical Tube loaded in Torsion
RBE3
The RBE3 option is a powerful tool to distribute applied loads and massed in a finite element model. Forces and moments
applied to reference or tied node are distributed to a set of independent degrees of freedom based on the RBE3 geometry
and local weight factors.
The way in which the forces are distributed is analogous to the classical bolt pattern analysis. Consider the bolt pattern
shown in Figure 8-51 with a force F A and moment M A acting at reference point A. The force and moment can be
transferred directly to the weighted center of gravity along with the moment produced by the force offset e .
FA FA
Reference point A
CG MA
MCG = MA + FAe
e
Figure 8-51 Transfer of Force and Moment on a Reference Point to the Weighted Center of Gravity (CG)
The force is distributed on the bolts proportional to their weighting factors. The moment is distributed as forces
proportional to the product of their distance from the center of gravity with corresponding weighting factors, as shown in
Figure 8-52. The total force acting on the bolts is equal to the sum of the two forces. These results apply to both in-plane
and out-of-plane loadings.
Fi
Force Distribution:
 i 
F i = F A  -------------
   k
Fi Moment Distribution:
 M C G  i r i
F i =  -----------------------
ri    k r 2k 
where F i = force at DOF i

 i = weighting factor for DOF i
r i = radius from the weighted
center of gravity to point i
Figure 8-52 RBE3 Force and Moment Distribution
As an example, consider the cantilever plate model with a single quadrilateral shell element shown in Figure 8-53. The plate
is subjected to nonuniform pressure represented by a resultant force acting at a distance of 10 mm from the center of gravity.
The easiest way to apply the pressure is to use an RBE3 option, with a reference node 5, to distribute the resultant load to
each of the four corner points. Setting the weighting factors to be equal for all nodes, results in a force distribution based
exclusively on the spatial location of the connected nodes. In this way, nodes 3 and 4 will get 3N each and nodes 1 and 2
will get 2N each.
10N
2
3
5
m
100 m
1
50 mm
4
40 mm
100 mm
Figure 8-53 Using the RBE3 Option to Represent a Nonuniform Pressure Load
CHAPTER 8 713
Boundary Conditions
The formulation of the RBE3 is based on the idea that the motion at a reference node is the weighted average of the so-called
error terms of the generalized displacements at a set of connected nodes. The error terms x el , and  el are defined as
follows:
xl = R  t   xl – xt  + xe l
(8-60)
 l =  t +  el
where x l and  l are the coordinates and rotations of the l-th retained node, respectively; x t and  t are the coordinates
and rotations of the reference node, respectively; R is a rotation matrix of the reference node. The error measure for
translational and rotation degrees of freedom are weighted with weighting factors, W l .
N
T
e tra n =  x el W l x el
l = 1
(8-61)
N
T
e ro t =   el W l  el
l = 1
The final nonlinear constraint relation is derived by minimizing the error terms with regard to the reference node
coordinates and rotations using least square fit.
The linearized relation between the reference node and the connected nodes and its correction term can be expressed as:
i i
u t = Su r + C non (8-62)
with C non being zero for linear analyses. For large rotation analysis, the error vector, C n o n , is required to meet the
r ot
constraint. The rotational error, C n on , is assumed to be zero. In other words, the expected incremental rotation of the
reference node is the same with the current value.
tran
The translational error, C n on , is derived by evaluating the difference between the expected coordinate ( x t e ) of the
reference node with the current one ( x t ). The expected coordinates are calculated by introducing the coordinates of the
connected nodes at the previous increment and the incremental rotation of the reference node into the nonlinear constraint
equations.
–1 N
 N 
x te =   W l  W l  x l + R l   t  
 
l = 1  l = 1
(8-63)
tran
C no n = x t e – x t
Beam - Shell Offsets

In many problems, it is convenient and sometimes necessary to model beam and shell elements at a geometric location that
does not match the actual physical location. Such cases are common when flanges or surfaces for shell and beam elements
with varying cross-section properties have to be aligned. Another common instance is when beams are used as stiffeners for
shells. It is most convenient to model the beam elements by sharing the shell nodes at the midsurface of the shell, as shown
in part (a) of Figure 8-54. The fact that the beam is actually offset by half the plate thickness and half the height of the beam
section will have to be achieved by providing a suitable beam offset as shown in part (b) of Figure 8-54.
(a) Geometric model with the (b) Physical model with the beam
beam element sharing the offset by half-shell thickness
nodes of the shell elements. + half-beam height.
Figure 8-54 Beam-stiffened Shell Structure
There are two methods by which beam/shell offsets can be modeled:

The first method is to place the beams and shells at the actual offset position and then tie the nodes of these
elements back to the original position through manually defined RBE2 links. While this method is accurate, it is
difficult to used for large models. Furthermore, it is not possible for the offset elements to contact other bodies
since the nodes of the offset elements are already tied.
The second method involves the beam/shell offset capability in Marc. The elements are specified at the original
position in the model and the offset vectors at the nodes of the element are specified in a variety of ways through
the GEOMETRY option. The current nodal offset is applied to each node of the offset element to calculate the offset
element position.
n+i 0 n+i
x offs et = x + voff (8-64)
n+i
where n is the increment number, i is the iteration number, x o f f se t is the current coordinate vector of the node
0
at the offset position, x is the original user-specified coordinate vector of the node at the original position, and
n+i
voff is the current offset vector. For small displacement analysis, the initial user-specified offset vector,
0
voff , is maintained through the analysis. For large displacement analysis, the offset vector is rotated based
upon the rotations at the node.
n+i n+i 0
voff = R voff (8-65)
n+i n+i
where R    is the current rotation matrix at the node and  is the current rotation vector at the node.
CHAPTER 8 715
Boundary Conditions
The tying relationship between the offset nodal position and the original nodal position is then established. The
basic formulation for this tying relationship is similar to the RBE2 link described in the previous section.The
relation between the incremental displacement of the node in the offset position to the corresponding
displacement at the original position is expressed as follows:
i i
u offs et = Su o r i gi n a l + C n on (8-66)
where the tying matrix S is based on a linearized rigid body relation and its terms are derived from offset vector
i
components. For linear analysis, C n on is zero. For large displacement analysis, Su r may not match the
nonlinear constraint exactly. Therefore, an error vector, C n o n , is required to meet the constraint. It is defined in a
manner similar to Equation (8-59).
In order to reduce linearization errors, an optional recycling feature is offered through the CONTROL model and
history definition options. Recycling occurs if the maximum iterative rotation change over the offsets nodes exceeds
a rotation tolerance (default is 0.0001 radians). This check can be by passed by setting the rigid link rotation
tolerance to zero (in conjunction with the FEATURE,5701 parameter) or a negative number.
The element stiffness matrix is formed on the offset element geometry. It is then converted to the original element
geometry through a suitable tying relationship.
T
K origin al = S K o f f s et S (8-67)
All distributed loads on the offset element are also treated in the same way. Note that point load vectors are not
offset and are always calculated at the original nodal positions. The nodal displacements at the original positions
are calculated during the solution phase. Finally, during the recovery phase, the displacements at the offset position
are calculated from Equation (8-66) and element strains, stresses are calculated on the offset element geometry.
Additional details for beam/shell offsets in Marc are provided below:
Offset Vector Specification

For beams, offset nodal vectors can be specified in four different ways:
1. Offset nodal vector specified in the global coordinate system.
2. Offset nodal vector specified in the local beam coordinate system:
For 2-D beams (element types 5, 45), the x-axis is along the beam axis, the z-axis is (0,0,1) and the y axis is
perpendicular to both.
For 3-D beams, the z-axis is along the beam, the x-axis is user-specified and the y-axis is perpendicular to both. This
option is not available for heat transfer links.
3. Offset nodal vector along the associated shell normal:
This option is particularly useful when beams are used as stiffeners for shells. In the case, only the offset magnitude
is provided by the user. The associated shell normal vector at the node is automatically calculated and is used for
defining the offset vector. This option is not available for 2-D beams.
4. Offset nodal vector specified in the local coordinate system at the node:
The local coordinates at the node can be specified using the TRANSFORMATION, CYLINDRICAL, or similar options.
For shells, the offset magnitude is provided by the user and the element normal calculated at the centroid of the shell
element is used to define the nodal offset vector.
Supported Elements
Beam/Shell offset support is provided for all standard beam and shell elements with global translational and rotational
degrees of freedom at the corner nodes.
Supported mechanical beam elements include element types 5, 14, 25, 45, 52, 76, 77, 78, 79, 98. Heat transfer links, element
type 36, 65, 99, and 100 are also supported (see section on Field Analysis below). Non-standard beam elements that are not
supported include element types 13, 16, 51. Supported mechanical shell elements include element types 1, 22, 75, 89, 138,
139, 140. Heat transfer shell elements, element types 50, 85, 86, 87, 88 are also supported (see section on Field Analysis below).
Non-standard shell elements that are not supported include elements types 4, 8, 15, 24, 49, 72.
For higher order beam/shell elements (element types 45, 22, 89), there is a choice to use interpolated values of the corner
node offset vectors at the midside nodes. If this is not chosen, the offset is taken as 0 at the mid-side nodes. For beam
elements 76, 77, where the midside node degree of freedom is not compatible with the corner node degree of freedom, only
the nodal position is updated for the offset and no other tying is undertaken at the midside node.
Post File
All nodal vectors on the post file are calculated and depicted at the original nodal positions. This allows visual compatibility
when elements with offsets and elements without offsets are in the model. Element quantities like strains and stresses on
the post file are calculated based on the offset element geometries.
Adaptive Meshing
There is no support for global remeshing with beam/shell offsets. Local adaptive meshing is supported. Nodal offsets are
automatically interpolated and transferred over to the new nodes created during the adaptive meshing process.
Field Analysis
Beam/shell offsets are predominantly supported for mechanical analysis. Since offset elements can also be used in thermo-
mechanical analysis and other field analyses, offset support is extended for field analyses other than purely mechanical
analysis. In this case, only the coordinates of the element are adjusted for the offsets, as per Equation (8-64). No other tying
of the field variables or element matrices is undertaken.
Contact
The average nodal offset vector is calculated at every node of all beam/shell contact bodies. All touching nodes and touched
patches are then updated using the average nodal offset vectors.
Dynamics
The stiffness matrix, mass matrix and damping matrix for the offset element are all calculated in a manner similar to
Equation (8-67). Implicit in this calculation is the assumption that the displacements as well as velocities and accelerations
between the original and offset positions are tied in a manner similar to Equation (8-66).
CHAPTER 8 717
Boundary Conditions
Pin Code for Beam Elements

Pin code is used to remove connections between the node and selected degrees of freedom of the beam. It can only be
applied on the nodes at the ends of the beam. It is not allowed to use it for the middle node of 3-noded beams. The degrees
of freedom are defined in the element’s coordinate system (see Cross-section Orientation in Marc Volume B: Element Library for
more detail description on how the local coordinate system of beam element is set up), and the pin code is applied at the
offset ends of the beam.
Marc manipulates the pin code in two different ways. The first one, by default, is done by internally generating a new node
which is tied to the flagged node. The tied degrees of freedom are the negation of the pinned ones. The tying process is
carried out at element level. The element stiffness matrix and internal forces of the element are calculated using a standard
procedure. During the assembly process they are “expanded” to include the degrees of freedom of the new internally
generated node. This procedure works fine both for quasi static and dynamics analysis.
The second one, using FEATURE, 6901, is done by using the so-called static condensation technique. In this way, no internal
node is generated. The pinned degrees of freedoms are condensed out from the element stiffness and internal forces before
they are assembled into the global system. This feature is added to allow Marc pin code fully compatible with MSC Nastran
implementation. For quasi-static analysis, both ways give exactly the same behavior. For dynamics analysis, the coefficients
used to condensed out the element stiffness matrix is reused for mass matrix. This is an approximation to the correct
procedure where the static condensation has to be done on the element operator matrix before being assembled into global
system.
The pin code is activated using PIN CODE or the GEOMETRY option. For postprocessing purposes, the element connectivity is
always presented with the original nodes. For example, in a stress-free large displacement analysis for a beam element with
pin code that simulate slider, the beam length will become shorter or longer although the element is stress free (see
Figure 8-55; node 4 is generated internally. If FEATURE, 6901 is activated, node 4 is not generated). The pin code is defined
on element 10 at node 2. The axial-displacement and out-of-plane rotation degrees of freedom are flagged.
2
10 20 4
10
20
1 1
3 3
Figure 8-55 Slider Connection using Pin Code
Mesh Independent Connection Methods

In Marc, there are various ways to model connections between structural components in a finite element model. These
connections may represent spot welds, seam welds, bolts, screws, rivets, and so on. Depending on the modeling goals, they
can be modeled using line elements (like trusses and beams), linear or nonlinear springs, rigid body connections (RBE2,
RBE3), and tyings or servo links (MPC constraints). In this way, general connections can be made between parts, but they
often require a lot of modeling effort in defining the connections or in assuring that meshes of the parts involved are
congruent so that they match in those points where connections are desired. Data and input preparation for these
connections can be a formidable task in real-world applications. Often, it is desired to model certain stiffness characteristics
in the connections. An easy way to establish this is by making point-to-point connections between the parts and using beam
elements or springs as the connector elements. Apart from the high meshing effort, this modeling technique may also
produce poor results. Increasing mesh refinement can introduce stiffness errors; point-to-point connections in which the
effective cross-sectional areas are larger than 20% of the characteristic element face areas can often lead to significant
underestimation of connector stiffness. Point-to-point connections can also cause numerical problems in finite element
shell models. The stiffness of the rotations normal to the shell plane may not exist or may be represented by a penalty
stiffness to suppress the associated singularity, but in the point-to-point connections, such rotations may be excited. If the
surface of one of the parts is defined by the faces of continuum elements, there are no rotational degrees of freedom for the
nodes of these elements and point-to-point connections may introduce undesired singularities.
The CWELD and CFAST connection methods that are available in MSC Nastran have also been implemented in Marc to
address and solve the aforementioned modeling issues. Connections can now be established with ease between points,
elements, patches, or any of their combinations and load transfer between the parts can be established over multiple nodes
on each side of a connection providing a more accurate stiffness representation of the connection. The following
connections between two parts can be made: patch-to-patch, point-to-patch and point-to-point as illustrated in Figure 8-56.
Patch-to-patch Connection Point-to-patch Connection Point-to-point Connection

For noncongruent meshes, For noncongruent meshes, For nearly congruent meshes,
area wise connection point to area wise connection point wise connection
Figure 8-56 Supported Connections Between Different Parts
Each connection involves the two end nodes of a connector element and two surfaces that are to be connected at a certain
location. The connector element can be any Marc beam element, a truss element or a cbush element. The surfaces can be
defined by shell elements, the faces of continuum elements or by nodal points. These surface element faces are referred to
as patches, can be triangular or quadrilateral, and may be linear or quadratic. On each side of the connection, either a patch
connection or a point connection is made thus giving the three connections shown in Figure 8-56. A patch-to-patch
connection can be shell-to-shell, shell-to-solid, or solid-to-solid. A point-to-patch connection can be point-to-shell or
point-to-solid. A point connection is the simplest connection since the end node of the connector element is directly
connected to a node of the surface; the connector element and the surface share a common node and no additional
constraints are required to establish the connection. A patch connection connects the end node of a connector element to
one or more element faces and is more general as it eliminates the need of congruent meshes. The location where the end
node pierces the surface is automatically determined. The element face being pierced is called the master patch. Internal
constraints are generated to assure a rigid (i.e., a force and moment carrying) connection of the connector end node to the
surface. Two types of patch connections can be made: Direct Patch Connection which connects the connector end node
directly to the nodes of the master patch with tying type 44; the constraints involved in this tying are also known as
Kirchhoff constraints, (see Rigid Tying to a Surface Patch in this chapter) or the Indirect Patch Connection which connects the
connector end node indirectly to the nodes of the surface by constructing a four-node auxiliary patch at the end node
normal to the connector. The size of this auxiliary patch is determined from the diameter or area of the connector. The end
CHAPTER 8 719
Boundary Conditions
node of the connector makes a rigid (tying type 44) connection to this auxiliary patch. The four nodes of this auxiliary
patch (the auxiliary nodes) are connected to the surface by additional constraints. The location where each auxiliary node
pierces the surface is automatically determined. The patch being pierced is called the secondary patch. By default, RBE3
constraints are generated between each auxiliary node and the nodes of the secondary patch it pierces, but other constraints
may be chosen as well. Note that, depending on the mesh density, two or more auxiliary nodes may connect to the same
secondary patch. All piercing points are determined by normal projections to the surface to which the connection is made.
The two patch connection types are illustrated in Figure 8-57 which displays a top view of one side of the connection
involving regular four-node quadrilateral element faces. Other element face types and element faces with collapsed edges
may be present as well. All constraints are applied internally as tying constraints and they are automatically generated when
processing the input.
master patch
auxiliary patch
master patch
secondary patch
Direct Patch Connection Indirect Patch Connection

Figure 8-57 Patch Connection Types
The direct patch connection is the simpler type, but may lead to asymmetrical load transfer between the surfaces when the
projections are near the element face boundaries and the element faces are not nicely facing each other. It may also lead to
a very localized load transfer in case of fine meshes, resulting again in a poor stiffness representation of the connection. The
indirect patch connection improves the load transfer by incorporating more element faces, and thus more nodes, in the
connection, but this will also lead to more constraint equations.
In general, a connection between two surfaces can be defined by entering a point that identifies the approximate location
of the connection and the two surfaces that are to be connected. This point can be entered by referencing a node or by
entering its coordinates when defining the connection and it is referred to as the GS-node or GS-point. These GS reference
points determine the spacing of the connections and they can be defined beforehand, independently of the surface meshes.
The surfaces can be identified by entering the element faces to which the connection will be made or by entering element
or face sets from which the element faces to connect to will be sought. For indirect patch connections, only the master
patches need to be entered. The secondary patches are automatically found in the vicinity of the master patches. If the
surfaces are entered as sets, the master patches and the secondary patches are searched from these sets. The connection
between the two surfaces is established as follows. The GS-node is projected to the first (side A) and second (side B) surface
by a normal projection. These two projections determine the end nodes of the connector and therefore its final position
and are referred to as the GA-node for side A and the GB-node for side B. The nodes on side A and B of the master patches
are referred to as GAi and GBi respectively, where i ranges from one to the number of nodes in the patch. The auxiliary
nodes are referred to as GAHi and GBHi respectively, where i always ranges from one to four. The nodes of a secondary
patch are referred to as GAij and GBij respectively, where j ranges from one to the number of nodes in the patch and i refers
to the ith auxiliary node. The definitions of the connection are made on the CWELD or CFAST model definition options. The
properties of the connector elements are defined on the PWELD or PFAST model definition options. These properties are
material and geometric properties in case of CWELD connections and stiffness properties in case of CFAST connections. A
number of control parameters can be entered on the SWLDPRM option to enhance the search and projection process and to
control its output.
CWELD patch-to-patch connections

There are several methods to define CWELD patch-to-patch connections. The methods available are ELEMID, GRIDID,
ELPAT, and PARTPAT. In each of these methods, the surfaces on both sides have to be defined together with the GS
reference point of the connection. The ELEMID and GRIDID methods make the simpler direct patch connections and the
ELPAT and PARTPAT methods make the more advanced indirect patch connections. All methods can be used to connect
noncongruent meshes of any type and they are recommended when the cross-sectional area of the connector is larger than
20% of the characteristic element face area. The connector element can be any Marc beam element. The default type used
is 98 which is a straight beam in space that includes transverse shear effects. By default, it assumes elastic behavior and a
solid circular cross-section. It is always required to include the ELEMENTS parameter for each element type used as a
connector, even when it does not appear in the CONNECTIVITY input. The data for each connection are defined on the CWELD
model definition input. Some data inputs are always required, because no defaults exist for them. Other data inputs are
optional and if not defined, Marc assumes appropriate default values for them. These optional data inputs concern the
connector element type, the way the projections are carried out, the type of constraints and the orientations of specific
CWELD connections.
Direct Patch-to-patch Connections

The ELEMID and GRIDID methods define the connection directly between two master patches. For ELEMID the patches
are entered as shell elements or the faces of continuum elements (SHIDA and SHIDB). For GRIDID, the patches are entered
as nodes (GAi and GBi) following the same connectivity conventions as for corresponding shell patches. The two methods
are illustrated in Figure 8-58. The GS-node is projected to both patches and, in case it is already positioned between the two
patches, these projections determine the connector end node locations. But it is not a requirement to position the GS-node
between the two patches, and if it is not, the projection is done in two steps as illustrated in Figure 8-59. First, a normal
projection to patch A and patch B is made resulting in the points GA’ and GB’. Then the midpoint GC’ between these
two points is projected to both patches and this determines the connector end point locations GA and GB. The nodal
coordinates of the connector element are updated after this projection. The GS-node must have a proper projection to the
patch within its boundaries. The PROJTOL parameter can be defined to accept projections outside but near patch
boundaries.
CHAPTER 8 721
Boundary Conditions
GS GS
SHIDA GA4 SHIDA

GB3
GA3
GA1 GA2
GB2
SHIDB
GB1
ELEMID GRIDID
Figure 8-58 ELEMID and GRIDID Connection Methods
GS GS
PIDB Shell
B
GB GB
GB
SHIDB
L PIDA
GA GC GC
SHIDA GA
D A GA
Shell
Figure 8-59 Determination of the Connector End Node Locations GA and GB
The data needed to define a connection with ELEMID are:

1. The reference node GS or its coordinates (XS,YS,ZS).
2. The patches SHIDA and SHIDB on both sides, where for shell elements these are the element numbers and for
continuum elements these are element and face number pairs.
3. A PWID, which is the identification number of a corresponding PWELD property entry.
The data needed to define a connection with GRIDID are:
2. The patch nodes GAi and GBi on both sides. The number of nodes determines the patch type, which can be quad4,
quad8, tria3 or tria6. No mixed patches with partly linear and partly quadratic edges are supported.
Indirect Patch-to-patch Connections

The ELPAT and PARTPAT methods make the connection of the connector end nodes to the surfaces involved in two phases,
as illustrated in Figure 8-60 and Figure 8-61. First the reference GS-node is projected to a master patch. For the ELPAT
method, the master patches are specified in the input. For the PARTPAT method, the master patches are searched from sets
that have been specified in the input. Around each projection point, a square four-node auxiliary patch is constructed. This
auxiliary patch is normal to the line connecting the two projection points. The dimensions of this auxiliary patch follow
from the diameter or the area of the connector as displayed in the right part of Figure 8-60. For the four auxiliary nodes, new
projections are computed on the master patches or on patches in their vicinity. The patches to which these new projections
are found are called secondary patches and each auxiliary node is connected by default with RBE3 constraints to the nodes
of its respective secondary patch.
GBH4 GBi
GBH3
GB D/2
Shell B D/2
GAH4 GAH3
GBH1
GBH2
a
GA D
a =  ----
GAH4 4
a
GAH1 GAi
GAH3
GA GAH1 GAH2
a a
GAH2
Shell A
Figure 8-60 ELPAT or PARTPAT Connections
The master patches and the secondary patches do not have to be distinct as can be seen from Figure 8-60 and Figure 8-61. For
a coarse mesh, they may even be one and the same patch but for fine meshes they may all be distinct. The requirement is
that the secondary patches do not span a larger area of the mesh than a 3 x 3 area as shown in the right part of Figure 8-61
otherwise a number of nodes not involved in the constraints defining the connection is present under the area covered by
the auxiliary patch. If the NREDIA parameter is set and the secondary patches violate the 3 x 3 requirement, the auxiliary
patch size is reduced a number of times in an attempt to fulfill this requirement. The MSC Nastran 2016 version has
removed this constraint on a 3 x 3 area; this has not been implemented in the Marc product.
CHAPTER 8 723
Boundary Conditions
GAij GAij
GAH4 GAH3
GAH4 GAH3
GA
GAH1 GAH2
GAH1 GAH2
Coarse Mesh Fine Mesh

Figure 8-61 Connections in Coarse and Fine Meshes for the ELPAT and PARTPAT Methods
The main purpose of the ELPAT and PARTPAT methods is to improve the load transfer between the two surfaces. These
methods allow for a more symmetrical load transfer than the simpler ELEMID or GRIDID methods as illustrated in
Figure 8-62. On side A in the left part of the figure, the shaded patch is the master patch and the constraints are only applied
between the connector node and the nodes of this patch. On side A in the right part of the figure the neighboring patches
also become involved in the connection and the load at the connector node is transferred through the four auxiliary nodes
to all the nodes of the secondary patches in the surface, leading to a better load transfer. It also avoids a too localized load
transfer as may be clear from the right part in Figure 8-61. If this connection were of the ELEMID type, the only patch
involved in the load transfer would be the one entirely covered by the shaded area leading to a much more localized load
transfer.
The PARTPAT method defines a connection between two sets. In case the part is meshed with shell elements, the set can
be an element set identifying the elements that make up the surface. In case the part is meshed with continuum elements,
the set must be a face set identifying the faces that make up the surface. A shell surface can also be identified by a face set,
but the face information is not essential and therefore element sets can be used as well.
The ELPAT method is similar to the PARTPAT method. The only difference is that the ELPAT method defines the master
patches, whereas the PARTPAT method searches them from a set of elements. A further difference is that for shell meshes
the ELPAT method searches its secondary patches from the whole mesh, whereas the PARTPAT method searches them from
the same sets from which the master patches were found, but sets may be used to limit the search regions. Sets are only
required for the ELPAT method if the surfaces contain faces of continuum elements.
SHIDB SHIDB
SHIDA
SHIDA
Connected Elements for ELEMID Format Connected Elements for ELPAT Format
Figure 8-62 Comparison of Load Transfer in the ELEMID and the ELPAT Methods
The data needed to define a connection with PARTPAT are:

2. The sets SetA and SetB for both sides, where, for shell elements, these sets may be element sets, and for continuum
elements, they must be face sets containing element and face number pairs. The two sets SetA and SetB do not have
to be disjoint, but the search procedure is facilitated if they are.
The data needed to define a connection with ELPAT are:
2. The patches SHIDA and SHIDB on both sides, where for shell elements these are the element numbers and for
continuum elements these are element and face number pairs.
3. The sets SetA and SetB for both sides, where, for shell elements, these sets may be element sets and for continuum
elements they must be face sets containing element and face number pairs. The two sets SetA and SetB do not have
to be disjoint. For surfaces defined by shell elements, these sets are optional, but for surfaces defined by faces of
continuum elements, they must be specified.
Parameters for the Projection Process

When searching master or secondary patches, it may happen that no projection is found at all to any of the candidates
because the surface is not flat and the GS-node or an auxiliary node only has a projection outside two neighboring patches
as illustrated in the left part of Figure 8-63 or because one or more of the auxiliary nodes lie outside a free edge as illustrated
in the right part of Figure 8-63.
CHAPTER 8 725
Boundary Conditions
GS Free Edge
Figure 8-63 Exceptional Cases needing Special Consideration
In the first case, a projection is accepted to the common edge of the two patches and the nearest patch is selected as the
master or secondary patch. In the second case, two parameters may be used to improve the situation. The GSMOVE
parameter allows the GS-node to be moved in the direction of the auxiliary nodes that have found successful projections.
The NREDIA parameter allows the area of the auxiliary patch to be reduced without moving the GS-node. Both parameters
may be used simultaneously. In that case, it is first tried to move the GS-node a number of times and if that is not successful
the area is reduced a number of times. If that still fails, it is once more tried to move the GS-node a number of times, now
using the reduced area for the auxiliary patch. Successful projections of the auxiliary nodes are counted in pairs. A successful
pair means that the two corresponding auxiliary nodes on both sides of the connection have found a successful projection.
So in the situation in the right part of Figure 8-63, two pairs of auxiliary nodes are successful and two pairs have failed,
because their projections fall over a free edge of one or both of the surfaces. In addition, the PROJTOL parameter may be
used to accept projections outside but near patch boundaries. If GSMOVE is nonzero, the new GS location is determined
from the successfully projected GAHi and GBHi pairs as follows:
 If one pair of GAHi and GBHi finds a successful projection, GS is moved to the middle position between that pair
and the original GS.
 If two pairs of GAHi and GBHi find a successful projection, GS is moved to the middle position between those
two pairs and the original GS.
 If three pairs of GAHi and GBHi find a successful projection, GS is moved closer to the second pair found. The
move is one quarter of the distance from GS to the second pair.
 If no pairs are found at all, moving GS is abandoned.
If the number of moves has been exhausted and NREDIA is nonzero, the characteristic diameter of the CWELD connector is
halved a number of times reducing the auxiliary patch size, thus bringing its nodes closer to each other. If one or more pairs
still fail after this diameter reduction, a last attempt is made to move the GS node while maintaining the reduced sized of
the auxiliary patch. The size of this movement is twice that of the movement before the size reduction which, in general,
is still a small movement, because of the reduced auxiliary patch size. This second attempt to move the GS-node is only
carried out when NREDIA is nonzero.
In some occasions, it may happen that a proper projection is found far away from the GS point (e.g., when the surface is
curved and the situation displayed in the left part of Figure 8-63 occurs for patches near to the GS point. In that case, the
GSTOL parameter may be used to reject projections that are further away than specified by the parameter, so projections
will only be accepted that are relatively near to the GS point.
It is assumed that the two surfaces that are being connected are reasonably parallel. The GSPROJ parameter may be used
to check this explicitly as illustrated in Figure 8-64. If the master patch normal vectors on both sides make an angle larger
than specified by the parameter, the projection is rejected.
It is also assumed that master patches and associated secondary patches are reasonably parallel. It is not desirable that
elements almost normal to the master patch are selected as secondary patches. The GSCURV parameter may be used to reject
secondary patches that make a large angle with the master patches as illustrated in Figure 8-64. This may occur in sharp
corners of the surface which in practice may represent edges of the structure.
The parameters discussed in this section are defined on the SWLDPRM option.
GSCURV
GSPROJ
nB
GSCURV
nA n A H1
nA
GAH1 GA GAH2
Figure 8-64 Angle Parameter GSPROJ and GSCURV to Control the Patch Selection Process
Internal Constraints to Connect Two Surfaces

Each connector end node, that is not making a point connection, makes a rigid connection to its master or auxiliary patch
by means of a Kirchhoff constraint (tying type 44). This constraint rigidly constrains the displacements and rotations of
the end node to the motion of the patch it connects to. The possible patches to which such a connection can be made are
linear and quadratic quadrilateral and linear and quadratic triangular patches. With the ELPAT and PARTPAT methods,
the auxiliary patch nodes are connected to their partnering secondary patches by additional constraints. By default, these
will be RBE3 constraints, but alternatively Kirchhoff constraints may be used or a combination of RBE2 and RBE3 constraints.
The RBE2/RBE3 combination makes a RBE2 constraint between the auxiliary patch node and its projection point on the
secondary patch and RBE3 constraints between this projection point and the nodes of this patch.
When a surface is curved, the auxiliary patch nodes will in general not lie on the surface. A special parameter may be set in
the CWELD input to force a repositioning of these nodes onto the surface. This slightly alters the orientation of the auxiliary
patch and in some cases this might allow the connector to follow the curvature of the surface somewhat better. By default,
this repositioning is not done.
For the RBE3 constraints involved in the CWELD connection, a weighting other than uniform weighting can be used by
specifying a distance weighting exponent. The weight factor for each retained node in the RBE3 is:
1
f i = -----n-
di
where f i is the weighting factor for retained node i, d i is the distance from the tied node to retained node i and n is the
weighting exponent. Negative values for n are not recommended, since they will result in heavier weighting for nodes
further away. The default results in uniform weighting  f i = 1  .
CHAPTER 8 727
Boundary Conditions
The RBE2/RBE3 combination is only useful when the auxiliary nodes are not lying on the secondary patches and when a
nonuniform distance weighting is used. It also comes with a cost penalty because it involves more constraint equations in
each connection.
If the CONNECTIVITY input contains no elements with nodes having displacements and rotations as degrees of freedom, it is
necessary to set the RBE parameter.
Multiple Surface Connections

The patch-to-patch connection can be used to connect more than two layers of shell elements by using the same GS-node
in multiple CWELD definitions. For example, if three layers need to be connected, a second CWELD definition is made, that
refers to the same GS-node as the first CWELD definition. Surface B on the first CWELD definition is repeated as surface A in
the second CWELD definition. If the surfaces are not curved, the locations of the connector end nodes connecting from the
top and bottom side to the surface in the middle automatically coincide, but for curved surfaces, this is not guaranteed.
For all connection types except PARTPAT, there is no real difference between repeating the GS-node or defining it a
multiple number of times.
When the PARTPAT method is used and a set pair is repeated in a number of connections that use the same GS-node (e.g.,
the same set is used for side A and side B for each of these connections), this repeated usage of the GS-node results in
searching through a stack of patches that can be found from the two sets, in order to connect multiple plate surfaces.
However, when the set pairs are different, i.e., at least one side in a new connection refers to a set with a name different
from the name referred to by this side in a previous connection using the same GS-node, the effect is as in all other
connection types.
The definition for a search through a stack of surfaces can only be made by repeated usage of a GS-node (i.e., a repetition
of its node ID, not just a repetition of its coordinates XS, YS, and ZS) in combination with the same sets for side A and
side B for all the connections that must be present in the stack. It only applies to the PARTPAT connection type and in
that case the sets necessarily have an overlap, i.e., they are not disjoint. The stack is limited to 10 connections, i.e., 11 plate
surfaces.
Coupled Thermo-mechanical Analysis

Cweld connections are be used in a coupled thermo-mechanical analysis. In the heat transfer pass of an increment, the
Kirchhoff constraints will tie the temperature of the tied node to the temperature of the projection point on its
corresponding patch. For RBE3 constraints, the temperature of the tied node will be the least squared weighted average of
the temperatures of its retained nodes. The same weighting factors apply as in the mechanical pass. If the retained nodes
are the nodes of a shell element, all temperature degrees of freedom of the shell node are used in the weighting process. For
RBE2 constraints, the temperature of the tied node is set equal to the temperature of its corresponding retained node, so
the constraint in the heat pass is identical to typing type 1.
Rules for usage of the GS-node or the GA-, GB-node Pair in the Projection Process
In most cases, the GS-node will be used to find the projections determining the end nodes of the connector element. The
GS-node and its XS,YS,ZS-coordinates are ignored on a CWELD definition if both the GA and the GB end nodes of the
connector are specified. In that case, GA and GB should, at least, be approximately in the right locations and GA is
projected to side A and GB is projected to side B of the connection. If GS and GA are not specified, but GB is, then GB
is used as GS and if GS and GB are not specified, but GA is, then GA is used as GS. If GS, GA, and GB are not specified,
then the XS,YS,ZS coordinates must be specified. If GS is specified and maximally one of the GA or GB, then GS is used
for the projections on both sides and the GA or GB is ignored for this purpose. If one side of the connector makes a point
connection and the connector node for that side has been left blank and the GS-node has been specified, the point
connection will be made to the GS-node and the GS-node will be projected onto the opposite side to determine the
location of the other connector node. If the opposite side also makes a point connection, the node for that side must have
been specified.
CWELD Point-to-patch Connections

It is possible to connect one end of the connector element to the surface by a point connection as is shown in Figure 8-65.
This may even be a free node to which certain boundary conditions may be applied. In such cases the surface information
that identifies the master patch for the particular side is omitted. This means that for the ELEMID and ELPAT method, the
input for SHIDA or SHIDB is omitted, for the GRIDID method, the nodes GAi or GBi are omitted, and for the PARTPAT
method, the set SetA or SetB is omitted. The GS-node is projected to the surface on the other side according to the
specifications made in the input.
GS
GA4 GA3
GA
GA1
GA2
Figure 8-65 Point-to-Patch Connection on Side A using GRIDID
CWELD Point-to-point Connections

All methods may be used to define point-to-point connections by omitting the surface information that identifies the
master patches and by specifying the connector end nodes on the CWELD input. These must be existing nodes in the
structure, because they can no longer be generated internally. Using any of the patch connection methods would lead to
cumbersome definitions of such point-to-point connections and the ALIGN method was designed to simplify their
definition. It is shown in Figure 8-66. The GS-node is no longer required, because the connector end nodes in general will
be nodes belonging to the two surfaces. Point-to-point connections are only recommended for nearly congruent mesh, so
CHAPTER 8 729
Boundary Conditions
the connector element will be approximately normal to the surfaces. However, they are not recommended when the
connector cross sectional area exceeds 20% of the characteristic element face area.
GA Upper Shell Midsurface
GB
Lower Shell Midsurface
Figure 8-66 Point-to-point Connection using ALIGN
CWELD Properties
PWELD Property Definition
The properties for the connector element can be entered on the PWELD model definition option. The material properties
are defined by entering the material identification number of a valid material for the element. The default cross section is
a solid circular section for which the diameter can be entered on the PWELD input. If other geometric properties than those
of a circular section are required, a complete set of geometric input data can be entered here as well. In that case, the
diameter is only used to determine the size of the auxiliary patches. If no diameter is entered but a set of geometric data is,
an attempt is made to estimate the auxiliary patch size from the geometric data. The geometric data input also specifies the
local directions of the cross section. If their corresponding data values are all zero, the local directions are determined by
the procedure outlined in the Connector Orientation section in this chapter, otherwise the local directions are determined from
the geometric data input according to the usual Marc convention. The default length of the connector is given by the
distance between the two end points GA and GB. Two control values on this length can be entered to prevent overly stiff
or overly flexible behavior of the connector element. Additionally, a spot weld connection type can be selected as an
alternative to the default general weld connection type. In both cases, the effective stiffness of the element is changed by
making certain length modifications. Their details are discussed later in the Connector Length Modifications and Spot Connections
section.
Connector Orientation
The connector element is positioned between the two projections GA and GB and these points determine the end nodes
of the element and thus its direction. The section forces for the default beam type are displayed in Figure 8-67 and the local
section directions are determined as follows. The element z-axis points from GA to GB.
The first procedure outlined in this section is used when the 4th, 5th, and 6th fields of the appropriate data blocks in the
PWELD or GEOMETRY options are all zero and no coordinate system (MCID) is specified in the CWELD option. This procedure
is as follows:
xB – xA
e 3 = ------------------------- is the local z-axis
xB – xA
In case of zero length, the average normal of the two surfaces is taken. Find the index j of smallest component of e 3 .
j i
e 3 = min  e 3  i = 1 2 3
In case of two equal components, the smaller index i is taken. Define the corresponding base vector as follows:
 1j 0
bj =  2j , where  kj = 1 when k = j and 0 otherwise, e.g. for j = 3 , b 3 = 0
 3j 1
mz
fz
ze my B
fy
GB fz axial force
fx shear force plane 1
fx fy shear force plane 2
mx B mz torque
m xA
plane 2 mx A bending moment end A, plane 2
fx mx B bending moment end B, plane 2
GA ye my B bending moment end A, plane 1
plane 1 my A bending moment end B, plane 1
fy
xe
m yA
fz
mz
Figure 8-67 Connector Orientation, Local Cross-section Directions and Forces
It provides a good directional choice for e 1 . In addition, the vector e 1 must be orthogonal to e 3 .
ẽ 1
ẽ 1 = b j –  e 3  b j e 3 and ẽ 1 = ----------
- is the local x-axis
ẽ 1
The cross product of local z and local x provides the local y-axis
CHAPTER 8 731
Boundary Conditions
e2 = e3  e1
The second procedure outlined in this section is used when the 4th, 5th, and 6th fields of the appropriate data blocks in
the PWELD or GEOMETRY options are all zero, but a coordinate system (MCID) is specified in the CWELD option. In that case
the local directions are determined by a procedure that uses a coordinate system defined in the COORD SYSTEM option. This
procedure is as follows:
xB – xA
e 3 = ------------------------- is again the local z-axis.
xB – xA
In case of zero length, the average normal of the two surfaces is taken. The 2nd direction (T2) defined by MCID is used to
define the orientation vector v of the connector element.
The local y-axis is then defined as:
e3  v
e 2 = --------------------
-
e3  v
The cross product of local y and local z provides the local x-axis:
e1 = e2  e3
If any of the 4th, 5th, or 6th fields of the appropriate data blocks in the PWELD or GEOMETRY options is nonzero, the
conventional Marc procedure is used to determine the local element directions, irrespective of the presence of a MCID input
in the CWELD option. A coordinate system may be associated with the first beam node (GA), but offsets are not possible and
are ignored.
An angle  may be specified for each connector element, which rotates the local axes about the line connecting the two
end points. This rotation takes effect after the orientation has been found by either one of the procedures outlined above
or by the geometric specifications made in the PWELD or GEOMETRY options.
Connector Length Modifications and Spot Connections

In the default general connection, the distance between the points GA and GB determines the length of the connector
element. To prevent the connector element from overly stiff or overly flexible behavior two tolerance values, LDMIN and
LDMAX can be entered on the PWELD input. LDMIN and LDMAX are the minimum and maximum length to diameter ratio
of the connection. If the length to diameter ratio is less than LDMIN, the behavior of the connection may become overly
stiff and if it is more than LDMAX, it may become overly flexible. In either case, a length modification of the CWELD is made
such that its behavior remains within the two limits. There are three methods available to account for this
length modification.
The first method is stiffness scaling. For linear elastic material behavior, the length modification could be accounted for by
scaling the Young’s modulus such that the axial, shear, and torsional stiffness of the beam with its actual length (i.e., the
length determined by the distance between its end nodes) is identical to the axial, shear, and torsional stiffness of a beam
with the modified length using the original Young’s modulus. This scaling can only be done when the beam material
behavior is linear elastic and the cross-section properties are not entered through the BEAM SECT parameter. Internally, the
cross-section geometric properties of the beam are scaled and not the Young’s modulus.
The second method is repositioning of the end nodes of the beam. Each beam node is moved in a direction normal to its
patch such that the beam obtains the desired length. The distance to the patch is treated as an offset in the Kirchhoff
constraints. If there is a point connection on one side, then that node is not repositioned.
The third method is repositioning of the end nodes and the auxiliary nodes. For methods ELPAT and PARTPAT, the
auxiliary nodes can be repositioned together with the beam nodes such that the beam obtains the desired length. This
requires updating the projections of the auxiliary nodes which may fail if these nodes become located far away from their
respective surfaces and these surfaces are curved. In that case, they may no longer have a projection inside any of the element
face boundaries.
If the element and its material behavior allow it, the default method is the first method, stiffness scaling, otherwise, no
length modification is made. The length modification method can be activated by defining the CWSPOT parameter on the
SWLDPRM option. If a CWELD has a point connection on both sides (like in ALIGN) or the distance between the two end nodes
of the beam is zero, no length modification will be made.
Alternatively, a spot weld connection can be defined. In this case, the length of the connecting beam is modified to be
0.5  t A + t B  where t A is the patch shell thickness on side A and t B is the patch shell thickness on side B. If the surface
on one side is defined by the faces of a solid element, its thickness contribution is zero. If both sides are faces of solid
elements, the connection type is automatically reset to a general connection.
The type of connection, general connection or spot connection, can be selected on the PWELD input. Spot connections are
only possible when one of the methods ELEMID, ELPAT, or PARTPAT is used.
If the distance between the connector end nodes GA and GB is zero, the two nodes will be connected rigidly using tying
type 100 and no beam element is activate in the connection.
If stiffness scaling is requested as the length modification method and the material is nonlinear or the cross section
properties are entered through the BEAM SECT parameter, the scaling is not done and a warning message is written to the
output.
CWELD Input Styles

For CWELD connections, two different input styles can be used to define the connections and these styles may be mixed in
one model.
The first style is the most natural style, where only the GS-node and the two surfaces are defined in the CWELD input. This
style was used in the foregoing discussion. The beam connector elements and its nodes are generated internally and are not
present in the CONNECTIVITY and COORDINATES inputs. All necessary geometric properties of the connection are defined in
the PWELD input and its material properties are defined by entering a valid material identification number in the PWELD
input.
With the second style, the connector elements are entered explicitly in the CONNECTIVITY option and their nodes in the
COORDINATES option. The connector nodes do not have to be in the right location as long as the GS-node is in the
approximate location. The material properties of the connectors may then be entered through the usual material options
and their cross-section properties through the GEOMETRY option. All these properties are assigned to the elements directly.
The CWELD option still defines the connection, but only refers to the connector elements and a PWELD input is generally no
longer required. The second style requires that the CWELD input is made before the CONNECTIVITY input of the element.
CHAPTER 8 733
Boundary Conditions
However, it is not required to rigidly adhere to the input styles mentioned above. A connector element with its ID, type,
and nodes may be defined entirely on the CWELD option. If no element ID is specified, Marc generates it internally. If no
node IDs are specified, Marc generates them internally. It no type is specified, type 98 is chosen for a 3-D analysis and type
5 for a 2-D analysis. If the CONNECTIVITY for an element is entered after its CWELD input, the CONNECTIVITY input determines
the element type and the nodes of the element. If the CWELD input for an element is entered after its CONNECTIVITY input,
the CWELD input determines the element type and nodes of the element. If the CWELD input specifies the element ID, a
CONNECTIVITY input for the element is not required. It is optional and can be used to redefine the nodes if these were left
undefined in the CWELD input.
If the connector element is only defined through the CWELD input, it is necessary to enter an ELEMENTS parameter for the
type of the connector element as it is necessary to enter an ELEMENTS parameter for each element type that is used in the
CONNECTIVITY input.
If a side of a connection makes a point connection, the node to which this connection is made cannot be an automatically
generated node. It must be a node defined by the user in the input. The user has no control over automatically generated
nodes; i.e., how their ID’s are assigned to them. Therefore, it is not possible to load or constrain such nodes or to connect
them to other parts of the model.
CWELD Error and Warning Messages

A number of error and warning messages can be written to the output file for any of the connector end nodes or the
auxiliary nodes. Error messages obtain a positive code and terminate the job with exit number 13. Warning messages obtain
a negative code and allow the job to continue. The possible messages are listed in Table 8-27.
Table 8-27 Codes for CWELD Error and Warning Messages

Value Description
1 Projection algorithm did not converge within desired tolerance.
2 Projection converged, but not inside the patch boundary.
-2 Projection converged, but near the patch boundary (inside: 0.01, outside: PROJTOL).
3 The patch is severely distorted or inside out (normal vectors in different locations make larger than 90°
angles).
-3 The patch is OK, but still has much distortion (normals in different locations make larger than 10°
angles).
4 The angle of an element/face normal to its auxiliary or master patch normal is too large; the patches are
almost normal to each other.
-4 The angle of an element/face normal to its auxiliary or master patch normal is rather large.
5 GS projects OK, but is located too far away from the projection.
-5 GS or an auxiliary node projects OK, but is located far away from the projection.
For GS nodes we use GSTOL as the tolerance.
For auxiliary nodes, we use XDIM (characteristic cweld dimension) as the tolerance.
XDIM = 1  4D , where D is characteristic cweld diameter.
Table 8-27 Codes for CWELD Error and Warning Messages (continued)
Value Description
6 The element we want to project to is unknown (i.e., useless, because element type = 0).
-6 A projecting is being made to a deactivated element, but accepted anyway.
7 The beam is almost parallel to the patch. (its angle with the patch is less than GSPROJ)
-7 The beam is not very normal to the patch (it makes an angle larger than 2*GSPROJ to the patch normal).
8 The patches to which the auxiliary nodes tie are more than 3 x 3.
9 The patch nodes have invalid degrees of freedom.
10 The patch has an invalid number of retained nodes.
11 A search over a group of patches resulted in no success at all; no projection has been even attempted to
any of the patches.
12 A search over a group of patches resulted in no success at all; although projections were tried, they were
all unsuccessful.
13 The projection is outside the patch boundary and there are no neighbors; so, there is no positioning to a
common edge or a common corner.
-14/14 Warning/Error: Position the cweld node to corner 1 of patch.
-18/18 Warning/Error: Position the cweld node to edge 1-2 of patch.
-20/20 Warning/Error: Position the cweld node to edge 3-4 (quad) or 3-1 (tria) of patch.
22 No face id was specified for the continuum element.
24 No viable patches in the set (PARTPAT, ELPAT).
26 No viable elements in the whole model (PARTPAT, ELPAT).
-27 To account for a length modification, the beam node on this side of the patch has been offset with regard
to the patch it is connecting with.
28 The beam node is shared with other elements, while not making a point connection.
Or the GS-node is shared with elements on the side that makes a point connection.
-28 The GS-node is shared with other elements.
29 The face ID identifying a patch is missing for a continuum element.
30 The connector node cannot connect to this element because it is not a valid connection type.
CHAPTER 8 735
Boundary Conditions
Table 8-27 Codes for CWELD Error and Warning Messages (continued)
Value Description
32 In 3-D, a connection to a line type of element was attempted; this is not possible.
-33/33 The connector node made a point connection but is not connected to any other element than the
connector element. If this node is an automatically generated node, the result is an error; otherwise, a
warning.
34 The projection is to an element/face sharing the point connection node.
35 An element/face of another connection lies between the current element/face pair.
CWELD Output
A CWELD connection defines a beam element as the connector element and generates internal tyings to define the connector
constraints. Output to the .t16 or .t19 post file can be requested for the beam elements by using the standard element post
codes available for them. Tying forces can be requested by using the standard nodal post codes available for them. The
displacements of all nodes involved in the connection will by default be on the post file.
Four sets are automatically generated and available on the post file when the CWSETS parameter on the SWLDPRM option is
set to one. The first set contains all the beam elements used in the connections. The second and third sets contain all patches
defining side A and side B of all connections, respectively. The fourth set contains all elements that have caused warnings.
An error set is not available because, upon encountering an error, no attempt is made to complete a connection. This means
that, in case of an error, the information at best is incomplete, but because of its intermediate character it may often be
misleading and often it is not yet available at all.
The sets are called:
 Set 1: fastener_all_beams_inc0000
 Set 2: fastener_all_faces_sidea_inc0000
 Set 3: fastener_all_faces_sideb_inc0000
 Set 4: fastener_all_warnings_inc0000
Set definitions defining sets with any of these names should not be made in the input.
The PRTSW parameter in the SWLDPRM input can be used to control the amount of output of each CWELD connection. If
PRTSW = 0 or is left undefined, no diagnostic messages are printed to the output file. If PRTSW = 1, only errors are
printed. If PRTSW = 2, errors and warnings are printed. if PRTSW = 3, all projection diagnostics are printed for each
CWELD connection. If PRTSW = 4, all projection diagnostics and all specifics of the tying constraints involved in each
connection are printed.
Two Dimensional Models

The previous sections described the CWELD connections for three dimensional models. However, CWELD connections can
also be made in two dimensional models. In that case, the patches consist of beams, trusses, or edges of continuum
elements. The same connection methods that are available for three dimensional models can be used in two dimensional
models. The connector element must be a two dimensional beam or an axisymmetric shell depending on the type of two
dimensional analysis. The number of auxiliary nodes is two instead of four; one to the left and one to right of the projection
point on the master patch. Two patch types are available: linear (with two nodes) and quadratic (with three nodes) patches.
The 3 x 3 requirement only applies in one direction now, so a connection on one side can not span more than three patches
of the surface.
CFAST Connections
CFAST connections can only be defined using indirect patch-to-patch connections. Two methods, ELEM or PROP, are
available to create them. They are equivalent with the CWELD ELPAT and PARTPAT methods, respectively (see Figures 8-59
and 8-60). In each of these methods, the surfaces on both sides have to be defined together with the GS reference point of
the connection. Both methods can be used to connect noncongruent meshes of any type. The constraints between the
connector nodes and the auxiliary patches are applied through Kirchhoff constraints (typing type 44) and the constraints
between the auxiliary nodes and the secondary patches are applied through RBE3 constraints.
The connector element is a cbush element. In a two dimensional model, it is element type 194 and in a three dimensional
model, it is element type 195. For more details about the cbush elements, refer to Marc Volume B: Element Library . It is always
required to include the ELEMENTS parameter for each element type used as a connector even when it does not appear in the
CONNECTIVITY input. The data for each connection is defined on the CFAST model definition input. Some data inputs are
always required because no defaults exist for them. Other data inputs are optional and if not defined, Marc assumes
appropriate defaults values for them. These optional data inputs concern the connector element type, the way the
projections are carried out, the type of constraints, and the orientations of the specific CFAST connections. For a more
detailed description of the CFAST option, refer to Marc Volume C: Program Input.
CFAST Properties
PFAST Property Definition
The properties of the CFAST connection are defined using the PFAST model definition option. It contains the diameter of
the fastener, the stiffness matrix, damping coefficient, and mass. The diameter data is only used to compute the locations
of the auxiliary nodes involved in the CFAST connection. It will not affect the element stiffness as the latter has to be
explicitly entered. For a more detailed description of the PFAST option, refer to Marc Volume C: Program Input.
CFAST Element Coordinate System

The stiffness and damping values are given is a specified coordinate system  e 1 ,e 2 ,e 3  . This coordinate system is used as
the element coordinate system. The output of CFAST is given in this system. There are two options to define it:
1. Using the connector element orientation. The x-axis points from GA to GB.
xB – xA
e 1 = -------------------------
xB – xA
In case of zero length, the normal of the first patch is taken. Find the index of j of smallest component of e 1
j i
e 1 = min  e 1 
i = 1 ,2 ,3
CHAPTER 8 737
Boundary Conditions
In case of two equal components, the smaller index i is taken. Define an auxiliary vector:
 1j 0
b =  2j , where  k j = 1 when k = j and 0 otherwise (for example, for j = 3 , b = 0 ) e 3 is created
 3j 1
by the normalized cross product of e 1 and b . And, finally, e 2 is the normalized cross product of e 3 and e 1 . For
geometrically nonlinear analysis, this coordinate system co-rotates with the element.
u u u
2. Using a global or user-defined coordinate system  e 1 ,e 2 ,e 3  . For this option, you may choose between a relative
and an absolute orientation.
a. the absolute orientation:
u u u
 e 1 ,e 2 ,e 3  =  e 1 ,e 2 ,e 3 
For a geometrically non-linear analysis, this coordinate system is fixed in time.

b. The relative orientation: the x-axis points from GA to GB.
xB – xA
e 1 = -------------------------
xB – xA
In case of zero length, the normal of the first patch is taken. e 3 is defined by the normalized cross product of
u
e 1 and e 2 . e 2 is defined by the normalized cross product of e 3 and e 1 . For a geometrically nonlinear analysis,
this coordinate system co-rotates with the element.
SWLDPRM, Special CWELD/CFAST Parameters

A number of global parameters that control the behavior of CWELD and CFAST connections and their output to the jobid.out
file can be entered through the SWLDPRM model definition option. These parameters and their descriptions are summarized
in Table 8-28.
Table 8-28 SWLDPRM Parameter Names and Descriptions
Name Type Default Description

CHKRUN Integer > 0 0 This parameter is available in MSC Nastran but has no meaning in Marc
(0 or 1) and is ignored.
GSMOVE Integer > 0 0 Maximum number of times GS is moved in case a complete projection
of all points has not been found.
NREDIA 0 < Integer < 4 0 Maximum number of times the characteristic diameter D is reduced in
half in case a complete projection of all points has not been found.
PRTSW 0 < Integer < 4 0 Parameter to control the CWELD/CFAST diagnostic output to the Marc
output file (jobid.out).
0= no diagnostic output
1= print errors only
2= print errors and warnings only
3= print projection diagnostics with no tying details
4= print all diagnostics
GSPROJ -90 < Real < 90 20.0 Maximum angle allowed between the normal vectors of master patch A
and master patch B. The connection will not be generated if the angle
between these two normal vectors is greater than the value of GSPROJ.
If GSPROJ is negative, the program will always accept the connection
and will only issue a warning if the angle is larger than |GSPROJ| (see
Figure 8-64).
CHAPTER 8 739
Boundary Conditions
Table 8-28 SWLDPRM Parameter Names and Descriptions (continued)

GSCURV -90 < Real < 90 20.0 Maximum angle allowed between the normal vectors of a patch to which
an auxiliary node projects and its corresponding auxiliary and master
patches. It provides a measure to monitor the curvature of a surface and
to recognize patches that belong to, for example, stiffeners. A connection
is not generated if the angle between the normal vectors is greater than
90-GSCURV meaning that the patches are almost normal to each other.
In that case, the patch is rejected and the search proceeds to the next
patch in the list. If the angle is between zero and GSCURV, no message is
displayed. If the angle is between GSCURV and 90-GSCURV, a large angle
warning is displayed. The following three tests are performed in the
order given below when GSCURV is positive:
If 0 < angle < GSCURV => OK
If GSCURVE < angle < 90-GSCURV => trigger a warning.
If angle > 90-GSCURV => reject.
Note that the warning condition is never triggered when GSCURV > 45
as it is overruled by the reject condition.
If GSCURV is negative, the projection is always accepted and a warning
is issued when the angle is larger than |GSCURV| (see Figure 8-64)
GSTOL Real 0.0 Maximum allowable distance of the node GS to its projection on a
patch. IF GSTOL is positive, the distance is relative to the characteristic
CWELD/CFAST diameter D, (the tolerance is GSTOL*D). If GSTOL is
negative, the distance is absolute (i.e., the tolerance is -GSTOL). If GS is
used for the projection together with one of the methods
PARTPAT/PROP or ELPAT/ELEM, an error is issued if the distance is too
large. If GA and GB are used for the projection or if one of the ELEMID
or GRIDID methods is used, the test only issues a warning if the distance
is too large. If GSTOL is zero, any distance is accepted.
PROJTOL 0.0 < Real < 0.2 0.0 Tolerance to accept the projected point GA or GB if the computed
coordinates of the projection point lie outside the patch boundary, but
are located within PROJTOL*(dimension of the patch).

ACTVTOL Integer > 0 1111 Parameter controlling the behavior of PROJTOL for the different
Integer < 2211 CWELD/CFAST connection methods. This parameter is entered as an
integer and is converted to a four-character string. If its value is less than
1000, the string is prepended with zeros. The first character (from the
left) controls the behavior when the PARTPAT/PROP method is used.
The second controls the behavior when the ELPAT/ELEM method is
used. The third controls the behavior when the ELEMID method is used
and the fourth controls the behavior when the GRIDID method is used.
For ALIGN, the PROJTOL tolerance has no significance. Each digit  d i 
in the string can have the value 0 or 1 or 2, where the value 2 only has
significance for the ELPAT/ELEM or PARTPAT/PROP methods. The
values have the following meaning:
0= PROJTOL is completely deactivated
1= PROJTOL is activated for ELEMID and GRIDID, PROJTOL is
activated in initial projections for ELPAT/ELEM, PROJTOL is only
activated over free edges of the patch in auxiliary projections for
ELPAT/ELEM, and in initial and auxiliary projections for
PARTPAT/PROP. Free edges have no neighbors within the set that
defines the complete surface.
2= PROJTOL is always activated
CWSETS Integer > 0 0 Parameter to control the automatic creation of four element sets with the
(0 or 1) elements involved in the CWELD/CFAST connections.
0= the sets are not created
1= four sets are created automatically:
fastener_all_beams_inc000”, the set containing all
connector beam elements.
fastener_all_faces_sidea_inc0000, the set
containing all elements with patches on side A of the
connection.
fastener_all_faces_sideb_inc0000, the set
containing all elements with patches on side B of the
connection.
fastener_all_warnings_inc0000, the set containing all
elements involved in CWELD/CFAST warning messages.
Defining sets with any of these names must be avoided and are
considered an error.
CHAPTER 8 741
Boundary Conditions

MAXEXP Integer > 0 2 Parameter to control the maximum number of expansions in the search
for projections of the auxiliary nodes. First, the master patch is tried. If
no projection is found on the master patch, a first expansion is made
including all neighboring patches of the master patch. If no projection
is found on any of the new patches, a second expansion is made
including all neighbors of the patches tried so far. This process continues
until the number of expansions exceeds MAXEXP. Two patches are
neighbors if they share at least one node in their connectivities.
DLDMIN Real > 0.0 0.2 Default value for LDMIN; the smallest ratio of length to characteristic
diameter.
DLDMAX Real > 0.0 5.0 Default value for LDMAX; the largest ratio of length to characteristic
diameter.
MAXITR Integer > 0 25 The maximum number of iterations allowed in the iteration process for
finding the projection on a patch.
EPSITR Real > 0.0 1.0E-5 Tolerance to terminate the iteration process for finding the projection on
a patch. If the parametric coordinate change in an iteration is less than
EPSITR, the projection is accepted as converged.
DELMAX Real > 0.0 0.1 Maximum allowable parametric coordinate change during the iteration
process for finding the projection on a patch. At first DELMAX is not
activated (i.e., the parametric coordinate change is not limited during
the iteration process). The parameter is only activated when the full
Newton Raphson iteration process for a projection did not converge. In
that case, the iteration process is restarted with DELMAX activated.
CWSPOT 0 < Integer < 3 1 Parameter to choose the method for modifying the beam length.
1= scale the stiffness of the beam
2= reposition the end nodes of the beam
3= reposition the auxiliary cpath nodes and the end nodes of the
beam.

RBE3WT Real 0.0 Default RBE3 distance weighting exponent.
The weight factor for each retained node in a RBE3 involved in a
1
CWELD/CFAST connection is: f i = -----
n
where di
fi is the weighting factor for retained node i.

di is the distance from the tied node to retained node i
n is the weighting exponent RBE3WT.
Negative values for RBE3WT are not recommended, since they result in
heavier weighting for nodes further away. The default results in uniform
weighting  f i  = 1 .
BOXING -1 < Integer < 1 0 Parameter to control the boxing algorithm used to speed up the search
for master patches when connection method PARTPAT/PROP is used.
-1 = The boxing algorithm is always deactivated
0 = The boxing algorithm may or may not be activated depending on
the number of elements in the sets.
1 = The boxing algorithm is always activated
Chapter 9: Element Library
9 Element Library
 Truss Elements 739


Membrane Elements 740
 Continuum Elements 740
 Beam Elements 741
 Plate Elements 742
 Shell Elements 742
 Heat Transfer Elements 742

Magnetodynamic Analysis 744
 Soil Analysis 744
 Fluid Analysis 744

Piezoelectric Analysis 744
 Special Elements 744
 Incompressible Elements 746
Marc includes an extensive element library. The element library allows you to model various types of one-, two-, and three-
dimensional structures, such as plane stress and plane strain structures, axisymmetric structures, full three-dimensional
solid structures, and shell-type structures. (See Marc Volume B: Element Library for a detailed description of each element, a
reference to the use of each element, and recommendations concerning the selection of element types for analysis).
After you select an element type or a combination of several element types for your analysis with the ELEMENTS or SIZING
parameter, you must prepare the necessary input data for the element(s). In general, the data consist of element
connectivity, thickness for two-dimensional beam, plate and shell elements, cross section for three-dimensional beam
elements, coordinates of nodal points, and face identifications for distributed loadings.
You can choose different element types to represent various parts of the structure in an analysis. If there is an incompatibility
between the nodal degrees of freedom of the elements, you have to provide appropriate tying constraints to ensure the
compatibility of the displacement field in the structure. Marc assists you by providing many standard tying constraint
options, but you are responsible for the consistency of your analysis.
You can use almost all of the Marc elements for both linear and nonlinear analyses, with the following exceptions.
 No plasticity or cracking is allowed in the elastic beams, the shear-panel, and the Fourier elements.
 The updated Lagrange and finite strain plasticity features are not available for all elements.
 Only heat links and two-dimensional heat transfer elements can be used for hydrodynamic bearing analysis.
 Fourier analysis can be carried out only for a limited number of axisymmetric elements.
Marc defines all continuum elements in the global coordinate system. Truss, beam, plate, and shell elements are defined in
a local coordinate system and the resulting output must be interpreted accordingly. You should give special attention to the
use of these elements if the material properties have preferred orientations. The ORIENTATION option is available to define
the preferred directions.
All Marc elements are numerically integrated. Element quantities, such as stresses, strains, and temperatures, are calculated
at each integration point of the element if you use the ALL POINTS parameter. This is the default in Marc. These quantities
are computed only at the centroid of the element, if the CENTROID parameter is used.
Distributed loads can be applied along element edges, over element surfaces, or in the volume of the element. Marc
automatically evaluates consistent nodal forces through numerical integration. (See Marc Volume B: Element Library for details
on this process). Concentrated forces must be applied at the nodal points.
All plate and shell elements can be used in a composite analysis. You can have a variable thickness shell, and control the
thickness and material property and orientation for each layer. For the thick shell elements (types 22, 45, 75, or 140), the
interlaminar shear can also be calculated.
Five concepts differentiate the various elements types. These concepts are listed invalid for pure heat transfer. below.
1. The type of geometric domain that the element is modeling. These geometric domains are:
 Truss
 Membrane
 Beam
 Plate
 Shell
 Plane stress
CHAPTER 9 739
Element Library
 Plane strain
 Generalized plane strain
 Axisymmetric
 Three-dimensional solid
 Special
2. The type of interpolation (shape) functions used in the element. These functions are:
 Linear
 Quadratic
 Cubic
 Hermitian
 Special
The interpolation function is used to describe the displacement at an arbitrary point in the body.
u i  x  = N i  x u i (9-1)
where u  x  is the displacement at x, N are the interpolation (shape) functions, and u are the generalized nodal
displacements.
Engineering strain is:
u i  x  u j  x 
 ij  x  = 1  2  ---------------
- + ---------------- (9-2)
 x x i 
j
Therefore, the computational evaluation is:
N i u i N j u j
 ij  x  = 1  2  -------------
- + -------------- =  i j u i (9-3)
 x x i 
j
Hence,
N
 ij = ---------i (9-4)
x j
3. The number of nodes in a particular element.

4. The number of degrees of freedom associated with each node, and the type of degrees of freedom.
5. The integration method used to evaluate the stiffness matrix. Marc contains elements which use full integration and
reduced integration.
Truss Elements
Marc contains 2- and 3-node isoparametric truss elements that can be used in three dimensions. These elements have only
displacement degrees of freedom.
Since truss elements have no shear resistance, you must ensure that there are no rigid body modes in the system.
Membrane Elements
Marc contains 3-, 4-, 6-, and 8-node isoparametric membrane elements that can be used in three dimensions. These
elements have only displacement degrees of freedom.
Since membrane elements have no bending resistance, you must ensure that there are no rigid body modes in the system.
Membrane elements are often used in conjunction with beam or truss elements.
Continuum Elements
Marc contains continuum elements that can be used to model plane stress, plane strain, generalized plane strain,
axisymmetric and three-dimensional solids. These elements have only displacement degrees of freedom. As a result, solid
elements are not efficient for modeling thin structures dominated by bending. Either beam or shell elements should be used
in these cases.
The solid elements that are available in Marc have either linear or quadratic interpolation functions.
They include:
 3-, 4-, 6-, and 8-node plane stress elements
 3-, 4-, 6-, and 8-node plane strain elements
 6 (4 plus 2)- and 10 (8 plus 2)-node generalized plane strain elements
 3-, 4-, 6-, and 8-node axisymmetric ring elements
 8-, 10-, and 20-node brick elements
 4- and 10-node tetrahedron
 6-node and 15-node pentahedral
 5-node and 13-node pyramids
 12- and 27-node semi-infinite brick elements
In general, the elements in Marc use a full-integration procedure. Some elements use reduced integration. The lower-order
reduced integration elements include an hourglass stabilization procedure to eliminate the singular modes.
Continuum elements are widely used for thermal stress analysis. For each of these elements, there is a corresponding
element available for heat transfer analysis in Marc. As a result, you can use the same mesh for the heat transfer and thermal
stress analyses.
Marc has no singular element for fracture mechanics analysis. However, the simulation of stress singularities can be
accomplished by moving the midside nodes of 8-node quadrilateral and 20-node brick elements to quarter-point locations
near the crack tip. Many fracture mechanics analyses have used this quarter-point technique successfully.
The 4- and 8-node quadrilateral elements can be degenerated into triangles, and the 8-and 20-node solid brick elements
can be degenerated into wedges and tetrahedra by collapsing the appropriate corner and midside nodes. The number of
nodes per element is not reduced for degenerated elements. The same node number is used repeatedly for collapsed sides
or faces. When degenerating incompressible elements, exercise caution to ensure that a proper number of Lagrange
CHAPTER 9 741
Element Library
multipliers remain. You are advised to use the higher-order triangular or tetrahedron elements wherever possible, as
opposed to using collapsed quadrilaterals and hexahedra.
Beam Elements
Marc’s beam elements are 2- and 3-node, two- and three-dimensional elements that can model straight and curved-beam
structures and framed structures and can serve as stiffeners in plate and shell structures. A straight beam of a circular cross
section can be used for modeling the straight portion of piping systems. Translational and rotational degrees of freedom
are included in beam elements. The cross section of the beam can be solid or thin-walled. The solid sections can be standard
elliptical, rectangular, trapezoidal, or hexagonal sections or arbitrary solid sections. The thin-walled sections can be standard
closed circular sections or arbitrary closed or open sections. The BEAM SECT parameter is used to define all solid sections and
to define arbitrary thin-walled open or closed sections.
The beam elements are numerically integrated along their axial direction using Gaussian integration. The stress strain law
is integrated through thin-walled sections using a Simpson rule and through solid sections using a Simpson, Newton-Cotes,
or Gauss rule depending on the input specifications. Stresses and strains are evaluated at each integration point through
the thickness. This allows an accurate calculation if nonlinear material behavior is present. In elastic beam elements, only
the total axial force and moments are computed at the integration points.
If element 52 or 98 is used with numerical section integration, six stiffness factors may be used to modify its stiffness
behavior in axial tension, bending, torsion and shear. The modified internal virtual work of beam element type 98 when
using any of the stiffness factors is given as:
L
W int =   f 1 N z  z + f 2 M z  z + f 3 M y  y + f 4 D x  x + f 5 D y  y + f 6 T z  z  dz
0
For element type 52 the transverse shear terms are absent. In this expression, f 1 is the normal stiffness factor, f 2 and f 3
are the bending stiffness factors, f 4 and f 5 are the transverse shear stiffness factors and f 6 is the torsional stiffness factor.
All stiffness factors have a default value of 1 and must be positive. If stiffness factors with other values than 1 are used, the
stress output must be interpreted as follows:
 Generalized stresses will be scaled with their respective stiffness factors (i.e., f 1 N z , f 2 M x , etc. will be written to
the output or post file)
 Layer stresses are not scaled with their respective stiffness factors, meaning that their section integration will result
in the unscaled generalized stresses (i.e., N z , M x , etc.)
If the section is used as a pre-integrated section all stiffness factors can have arbitrary positive values, but if the section is
used with numerical integration throughout the analysis the stiffness contributions resulting from the direct stresses must
use identical positive stiffness factors (i.e., f 1 = f 2 = f 3 ) and the stiffness contributions resulting from the shear stresses
must use identical positive stiffness factors (i.e. f 4 = f 5 = f 6 ) in order to guarantee symmetry of the stiffness matrix. If
these conditions are not met, f 2 and f 3 will silently be set to f 1 and f 4 and f 5 will silently be set to f 6 .
Plate Elements
The linear shell elements (Types 49, 72, 75, 138, 139, or 140) or the quadratic element (Type 22) can be used effectively to
model plates and have the advantage that tying is unnecessary. For element type 49, you can indicate on the GEOMETRY
option that the flat plate formulation is to be used. This reduces computational costs. To further reduce computational
costs for linear elastic plate analysis, the number of points through the thickness can be reduced to one by use of the
SHELL SECT parameter (please refer to the description of this option in Marc Volume C: Program Input for more information
about the optimal choice of the number of points through the thickness).
Shell Elements
Marc contains three isoparametric, doubly curved, thin shell elements: 3-, 4-, and 8-node elements (Types 4, 8, and 24,
respectively) based on Koiter-Sanders shell theory. These elements are C1 continuous and exactly represent rigid-body
modes. The program defines a mesh of these elements with respect to a surface curvilinear coordinate system. You can use
the FXORD model definition option to generate the nodal coordinates. Tying constraints must be used at shell intersections.
Thin shell analysis can be performed using either the 3-node discrete Kirchhoff theory (DKT) based element (Type 138),
the 4-node bilinear DKT element (Type 139), the 6-node bilinear semi-loof element (Type 49), or the 8-node bilinear semi-
loof element (Type 72). Thick shell analysis can be performed using the 4-node bilinear Mindlin element (Type 75), the 4-
node reduced integration with hourglass Mindlin element (Type 140), or the 8-node quadratic Mindlin element (Type 22).
The thick shell elements have been developed so that there is no locking when used for thin shell applications.
The global coordinate system defines the nodal degrees of freedom of these elements. These elements are convenient for
modeling intersecting shell structures since tying constraints at the shell intersections are not needed.
Marc contains three axisymmetric shell elements: 2-node straight, 2-node curved, and 3-node curved. You can use these
elements to model axisymmetric shells; combined with axisymmetric ring elements, they can be used to simulate the thin
and thick portions of the structure. The program provides standard tying constraints for the transition between shell and
axisymmetric ring elements.
Heat Transfer Elements

The heat transfer elements in Marc consist of the following:
 2- and 3-node three-dimensional links
 3-, 4-, 6-, and 8-node planar and axisymmetric elements
 6- and 9-node planar and axisymmetric semi-infinite elements
 8-, 10-, and 20-node solid elements
 4- and 10-node tetrahedral
 6-, 15-node pentahedral
 5-node and 13-node pyramids
 12- and 27-node semi-infinite brick elements
 2-, 3-, 4-, 6-, and 8-node shell elements.
 3-, 4-, 5-, and 8-node 3-D membrane elements
 4-, 8-node planar interface elements
CHAPTER 9 743
Element Library
 4-, 8-node axisymmetric interface elements

 6-, 8-, 15-, 20-node solid interface elements.
For each heat transfer element, there is at least one corresponding stress element, enabling you to use the same mesh for
both the heat transfer and thermal stress analyses. Heat transfer elements are also employed to analyze coupled
thermo-electrical (Joule heating) problems.
In heat transfer continuum elements or link elements, temperature is the only nodal degree of freedom.
The heat transfer shell elements can be used in two modes. In the first mode, the number of degrees of freedom is two or
three for linear or quadratic temperature distribution across the complete thickness. In the second mode, the number of
degrees of freedom are M + 1 or 2 * M + 1 for linear or quadratic temperature distribution for each M layer. This is
defined through the HEAT parameter.
The heat transfer membrane elements have only one degree of freedom per node, so there is no thermal gradient through
the thickness.
In Joule heating, the voltage and temperature are the nodal degrees of freedom. Shell elements are not available for Joule
heating.
Acoustic Analysis
Heat transfer elements are also used to model the compressible media in acoustic analysis. In this case, the pressure is the
single nodal degree of freedom.
Heat transfer elements are also used for electrostatic analysis. The scalar potential is the degree of freedom.
Coupled Electrostatic-Structural
Stress elements are used for a coupled electrostatic-structural analysis. For each of these elements, there is a corresponding
heat transfer element available for the electrostatic pass. When a medium has no stiffness, like air, heat transfer elements
can be chosen for this region. Then this region will be inactive in the stress pass.
Fluid/Solid Interaction
Heat transfer elements are used to model the inviscid, incompressible fluid/solid interaction problems. The hydrostatic
pressure is the degree of freedom.

The three-dimensional heat transfer links and planar elements are used to model the lubricant film. As no variation occurs
through the thickness of the film, two-dimensional problems are reduced to one-dimensional, and three-dimensional
problems are reduced to two-dimensional. The pressure is the degree of freedom.
For two-dimensional problems, a scalar potential can be used; hence, the heat transfer planar and axisymmetric elements
are employed. The single degree of freedom is the potential. For three-dimensional analyses, magnetostatic elements are
available. In such cases, there are three degrees of freedom to represent the vector potential.
In magnetodynamic problems, a vector potential, augmented with a scalar potential, is used. There is a full family of lower-
and higher-order, planar, axisymmetric, and 3-D elements available for these analyses.
Soil Analysis
There are three types of soil/pore pressure analysis. If a pore pressure analysis only is performed, “heat transfer” elements
(41, 42, or 44) are used. If an uncoupled soil analysis is performed, the standard elements (21, 27, or 28) are used. If a coupled
analysis is performed, the Herrmann elements (32, 33, or 35) are used. In this case, the last degree of freedom is the pore
pressure.
Fluid Analysis
When performing fluid analysis, any planar, axisymmetric or solid continuum element can be used. These elements either
used a mixed formulation or a penalty formulation based upon the value of the FLUID parameter. For the mixed formulation,
each node of lower- or higher-order continuum element has velocities as well as pressure degrees of freedom. The penalty
method yields elements with only the velocities as the nodal variables.
In a piezoelectric analysis, a strong coupling exist between stress and electric field. There are lower-order elements available
for this analysis. The first two or three degrees of freedom (in resp. 2-D or 3-D) are available for displacement components,
and the last (3rd or 4th) degree of freedom is available for the electric potential.
Special Elements
Marc contains unique features in the program.
Gap-and-Friction Elements
The gap-and-friction elements (12, 97) are based on the imposition of gap closure constraint and frictional-stick or
frictional-slip through Lagrange multipliers. These elements provide frictional and gapping connection between any two
nodes of a structure; they can be used in several variations, depending on the application. In the default formulation, the
elements simulate a gap in a fixed direction, such that a body does not penetrate a given flat surface. Using the optional
formulation, you can constrain the true distance between two end-points of the gap to be greater than some arbitrary
distance. This ability is useful for an analysis in which a body does not penetrate a given two-dimensional circular or three-
CHAPTER 9 745
Element Library
dimensional spherical surface. Finally, you can update the gap direction and closure distance during analysis for the
modeling of sliding along a curved surface.
Pipe-bend Element
The pipe-bend (3-node elbow) element 17 is designed for linear and nonlinear analysis of piping systems. It is a modified
axisymmetric shell element for modeling the bends in a piping system. The element has a beam mode superposed on the
axisymmetric shell modes so that ovalization of the cross-section is admitted. The twisting of a pipe-bend section is ignored
because the beam has no flexibility in torsion. Built-in tying constraints are used extensively for coupling the pipe-bend
sections to each other and to straight beam elements.
Curved-pipe Element
The curved-pipe (2-node) element 31 is designed for linear analysis only. The stiffness matrix is based upon the analytic
elastic solution of a curved pipe.
Shear Panel Element

The shear panel (4-node) element 68 is an elastic element of arbitrary shape. It is an idealized model of an elastic sheet. This
element only provides shearing resistance. It must be used with beam stiffeners to ensure any normal or bending resistance.
The shear-panel element is restricted to linear material and small displacement analysis.
Cable Element
The cable element (51) is an element which exactly represents the catenary behavior of a cable. It is an elastic element only.
Rebar Elements
The rebar elements (23, 46, 47, 48, 142, 143, 144, 145, 146, 147, 148, 165, 166, 167, 168, 169, and 170) are hollow elements in
which you can place single-direction strain members (reinforcing cords or rods). The rebar elements are used in
conjunction with other solid elements (filler) to represent a reinforced material such as reinforced concrete. The reinforcing
members and the filler are accurately represented by embedding rebar elements into solid elements. The rebar elements can
be used for small as well as large strain behavior of the reinforcing cords. Also, any kind of material behavior can be
simulated by the rebar elements in Marc. Note that for nonlinear applications, the rebar elements are supported only in the
total Lagrange formulation.
Interface Elements
Elements 186 through 193 are special elements which are used to model the delamination in composite structures where
each ply is modeled with solid elements. The material behavior of these elements is defined with the COHESIVE model
definition option.
Elements 220 through 227 are analogous elements for thermal analysis of interfaces. The material of these elements is
defined with the COHESIVE thermal option.
Note that these interface elements do not have a mass matrix or a specific heat matrix and distributed loads, fluxes, or films
are not applied to these elements. One can use either a Guass or Newton-Cotes intergration scheme. The latter may be
advantageous in cases where the interface has significantly different stiffness compared to the surrounding material.
Incompressible Elements
Incompressible and nearly incompressible materials can be modeled by using a special group of elements in the program.
These elements, based on modified Herrmann variational principles, are capable of handling large deformation effects as
well as creep and thermal strains. The incompressibility constraint is imposed by using Lagrange multipliers. Generally, the
low (linear) order elements have a single additional node which contains the Lagrange multiplier, while the high-order
(quadratic) elements have Lagrange multipliers at each corner node.
Elements 155 (plane strain triangle), 156 (axisymmetric triangle), and 157 (3-D tetrahedron) are low-order elements and are
exceptions within the incompressible element group. They have an additional node located at the center of the elements
and have Lagrange multipliers at each corner node. The shape function of the center node is a bubble function. See
Figure 9-1 for an element 155 case. The degrees of freedom of the center node is condensed out on the element level before
the assembly of the global matrix.
3 Nodes with both displacements and pressure as
degrees of freedom
4
x Center node with only displacement degrees of
x freedom. The shape function for this node is a
bubble function.
2
Figure 9-1 Element 155
Large Strain Elasticity

The incompressible elements based on Herrmann formulations can be used for large strain analysis of rubber-like materials,
using either total Lagrangian formulation or updated Lagrangian formulation.
Large Strain Plasticity

The incompressible elements can be used for large strain analysis of elasto-plastic materials using updated Lagrangian
formulation, based on additive or multiplicative decomposition of deformation gradient.
Rigid-Plastic Flow
In rigid-plastic flow analysis, with effectively no elasticity, incompressible elements must be used. For such analyses, you
have a choice of using either the elements with Lagrange multipliers discussed above, or standard displacement continuum
elements. In the latter case, a penalty function is used to satisfy the incompressibility constraint.
CHAPTER 9 747
Element Library
Constant Dilatation Elements

You can choose an integration scheme option, which makes the dilatational strain constant throughout the element. This
can be accomplished by setting the second field of the GEOMETRY option to one. Constant dilatational elements are
recommended for use in approximately incompressible, inelastic analysis, such as large strain plasticity, because
conventional elements can produce volumetric locking due to overconstraints for nearly incompressible behavior. This
option is only available for elements of lower order (Types 7, 10, 11, 19, 20, 136, 149, 151, 152, and 216). For the lower-order
reduced integration elements (114 to 117) with hourglass control, as only one integration point is used, these elements do
not lock, and effectively, a constant dilatation formulation is used. (The constant dilatation approach has also been referred
to mean dilatation and B-bar approach in the literature.)
Elements with Strain Smoothing

The popularity of triangular (2-D) and tetrahedral (3-D) elements is mainly driven by automated mesh generators. A
known drawback of standard linear triangular and tetrahedral elements is the poor performance in cases of nearly
incompressible materials and in bending dominated problems with relatively coarse meshes. Where the Herrmann
triangular and tetrahedral elements 155, 156, and 157 have the advantage that they can handle nearly incompressible
materials, the disadvantage is that they do not support all material models and in bending dominated problems a relatively
fine mesh may be required. To reduce these shortcomings, elements with strain smoothing have been introduced. Elements
239 (plane strain triangle), 240 (axisymmetric triangle), and 241(3-D tetrahedron) are lower order elements and similar to
the element types 155, 156 and 157, an extra node is used. However, unlike element types 155, 156 and 157, no pressure
degrees of freedom are used; all nodes only have displacement degrees.
5
Nodes defining the connectivity of a standard
3 linear triangle
6
Center node; used to define sub-triangles
Domain where the strain is continuous

4
2
1
Figure 9-2 Element 239
The extra nodes of elements 239, 240 and 241, located in the centroid of the elements, are used to construct a domain over
which the strain is continuous (called a strain smoothing procedure; see Figure 9-2). Such a domain spans two sub-elements
(sub-triangles in 2-D and sub-tetrahedrons in 3-D), one in the current and one in the neighboring element. So if an
element has neighboring elements at all of its edges (2-D) or faces (3-D), then the total number of nodes involved in the
final element formulation is 7 (2-D) or 9 (3-D). This increased number of nodes affects the bandwidth of the global
stiffness matrix and makes the elements relatively expensive. It should also be noted that the strain smoothing procedure
only works at shared element edges or faces. Consequently, using symmetry boundary conditions causes the results to be
slightly different compared to using a full finite element mesh.
Reduced Integration Elements

Marc uses a reduced integration scheme to evaluate the stiffness matrix or the thermal conductivity matrix for a number of
isoparametric elements. The mass matrix and the specific heat matrix of the element are always fully integrated.
For lower-order, 4-node quadrilateral elements, the number of numerical integration points is reduced from 4 to 1; in 8-
node solid elements, the number of numerical integration points is reduced from 8 to 1. For 8-node quadrilateral elements,
the number of numerical integration points is reduced from 9 to 4; for 20-node solid elements, the number of numerical
integration points is reduced from 27 to 8.
The energy due to the higher-order deformation mode(s) associated with high-order elements is not included in the
analysis. Reduced integration elements and fully integrated elements can be used together in an analysis. Reduced
integration elements are more economical than fully integrated elements and they can improve analysis accuracy. However,
with near singularities and in regions of high-strain gradients, the use of reduced integration elements can lead to
oscillations in the displacements and produce inaccurate results. Using reduced integration elements results in zero-energy
modes, or breathing nodes. In the lower-order elements, an additional stabilization stiffness is added which eliminates these
so-called “hourglass” modes.
Continuum Composite Elements

There is a group of isoparametric elements in Marc which can be used to model composite materials. Different material
properties can be used for different layers within these elements. The number of continuum layers within an element, the
thickness of each layer, and the material identification number for each layer are input via the COMPOSITE option. A
maximum number of 1020 layers can be used in one 2-D continuum composite element. A 3-D continuum composite
element only allows a maximum of 510 layers within the element.
These structural elements are available as both lower- and higher-order and can be used for plane strain (element types 151
and 153), axisymmetric (element types 152 and 154) or 3-D solid analysis (element types 149 and 150). Corresponding heat
transfer elements are available for lower- and higher-order and can be used for planar (element types 177 and 179),
axisymmetric (element types 178 and 180) or 3-D solid analysis (element types 175 and 176).
Fourier Elements
A special class of elements exists which allows the analysis of axisymmetric structures with nonaxisymmetric loads. The
geometry and material properties of these structures do not change in the circumferential direction, and the displacement
can be represented by a Fourier series. This representation allows a three-dimensional problem to be decoupled into a series
of two-dimensional problems. Both solid and shell Fourier elements exist in Marc. These elements can only be used for
linear elastic analyses.
Semi-infinite Elements
This group of elements can be used to model unbounded domains. In the semi-infinite direction, the interpolation
functions are exponential, such that the function (displacement) is zero at the far domain. The rate of decay of the function
is dependent upon the location of the midside nodes. The interpolation function in the non semi-infinite directions are
either linear or quadratic.
CHAPTER 9 749
Element Library
These elements can be used for static plane strain, axisymmetric, or 3-D solid analysis. They can also be used for heat
transfer, electrostatic, and magnetostatic analyses.
Cavity Surface Elements

This group of elements can be used to define the boundaries of cavities where standard finite elements are not used; for
example, along rigid boundaries. These elements are for volume calculation purposes only and do not contribute to the
stiffness equations of the model. They can be used for plane strain, plane stress, axisymmetric, and 3-D analyses. No
material properties are needed for these elements. They do not undergo any deformation except if they are attached or glued
to other elements or surfaces.
Assumed Strain Formulation

Conventional isoparametric four-node plane stress and plane strain, and eight-node brick elements behave poorly in
bending. The reason is that these elements do not capture a linear variation in shear strain which is present in bending when
a single element is used in the bending direction. For the six elements (3, 7, 11, 160, 161, and 163), the interpolation
functions have been modified such that shear strain variation can be better represented. For the lower-order reduced
integration elements (114 to 123), an assumed strain formulation written with respect to the natural coordinates is used.
For elastic isotropic bending problems, this allows the exact displacements to be obtained with only a single element
through the thickness. This is invoked by using the ASSUMED STRAIN parameter or by setting the third field of the GEOMETRY
option to one.
Follow Force Stiffness Contribution

When activating the FOLLOW FOR parameter, the distributed loads are calculated based upon the current deformed
configuration. It is possible to activate an additional contribution which goes into the stiffness matrix. This improves the
convergence. This capability is available for element types 3, 7, 10, 11, 18, 72, 75, 80, 81, 82, 83, 84, 114, 115, 116, 117, 118,
119, 120, 139, 140, 149, 151, 152, 160, 161, 162 and 163. Inclusion of the follower force stiffness can result in nonsymmetric
stiffness for nonenclosed volumes, thereby, resulting in increased computational times. You can flag the nonsymmetric
solver in the SOLVER option.
Explicit Dynamics
The explicit dynamics formulation IDYN=5 model is restricted to the following elements:
1, 2, 3, 5, 6, 7, 9, 10, 11, 18, 19, 20, 52, 64, 75, 89, 98, 114, 115, 116, 117, 118, 119, 120, 130, 139, and 140.
When using this formulation, the mass matrix is defined semi-analytically; for example, no numerical integration is
performed. In addition, a quick method is used to calculate the Courant stability limit associated with each element. For
these reasons, this capability is limited to the elements mentioned above.
Adaptive Mesh Refinement

Marc has a capability to perform local adaptive mesh refinement to improve the accuracy of the solution. This capability
is invoked by using the ADAPTIVE parameter and model definition option. The adaptive meshing is available for the following
2-D and 3-D elements:
2, 3, 6, 7, 10, 11, 18, 19, 20, 37, 38, 39, 40, 43, 75, 80, 81, 82, 83, 84, 111, 112, 113, 114, 115, 116, 117, 118, 119, 120,
121, 122, 123, 138, 139, 140, 155, 156, 157, 160, 161, 162, 163, 164, 196, 198, 201 and 247.
Chapter 10: Nonstructural Procedure Library
10 Nonstructural Procedure
Library
 Overview 758

Heat Transfer 758

Diffusion 831
 Hydrodynamic Bearing 835
 Electrostatic Analysis 839

Magnetostatic Analysis 844

Magnetodynamic Analysis 856
 Piezoelectric Analysis 873

Acoustic Analysis 877
 Fluid Mechanics 879
 Nonlocal Analyses 887

Coupled Analyses 888

References 925
Overview
This chapter describes the nonstructural analysis procedures available in Marc. These are comprised of several areas
including heat transfer, hydrodynamic bearing, electrostatic, magnetostatic, electromagnetic, piezoelectric, fluid
mechanics, and coupled analysis. This chapter provides the technical background information as well as usage information
about these capabilities.
Heat Transfer
Marc contains a solid body heat transfer capability for one-, two-, and three-dimensional, steady-state and transient
analyses. This capability allows you to obtain temperature distributions in a structure for linear and nonlinear heat transfer
problems. The nonlinearities in the problem may include temperature-dependent properties, latent heat (phase change)
effect, heat convection in the flow direction, and nonlinear boundary conditions (convection and radiation). The
temperature distributions can, in turn, be used to generate thermal loads in a stress analysis.
Marc can be applied to solve the full range of two- and three-dimensional transient and steady-state heat conduction and
heat convection problems. Marc provides heat transfer elements that are compatible with stress elements. Consequently,
the same mesh can be used for both the heat transfer and stress analyses. Transient heat transfer is an initial- boundary value
problem, so proper initial and boundary conditions must be prescribed to the problem in order to obtain a realistic
solution. Marc accepts nonuniform nodal temperature distribution as the initial condition, and can handle
temperature/time-dependent boundary conditions. The thermal conductivity can be isotropic, orthotropic, or anisotropic.
Both the thermal conductivity and the specific heat in the problem can be dependent on temperature; however, for
conventional heat transfer, the mass density remains constant at all times. Latent heat effects (solid-to-solid, solid-to-liquid
phase changes) can be included in the analysis. A time-stepping procedure is available for transient heat transfer analysis.
Temperature histories can be stored on a post file and used directly as thermal loads in subsequent stress analysis. User
subroutines are available for complex boundary conditions such as nonlinear heat flux, directional heat flux, convection,
and radiation.
A summary of Marc capabilities for transient and steady-state analysis is given below.
 Selection of the following elements that are compatible with stress analysis:
• 1-D: three-dimensional link (2-node, 3-node)
• 2-D: planar and axisymmetric element (3-, 4-, and 8-node)
• 2-D: axisymmetric shells
• 3-D: solid elements (4-, 5-, 6-, 8-, 10-, 13-, 15-, and 20-node)
• 3-D: membrane elements (3-,4-, 6-, and 8-node)
• 3-D: shell elements (4- and 8-node)
 Specification of temperature-dependent materials (including latent heat effects) is performed with the ISOTROPIC,
ORTHOTROPIC, ANISOTROPIC, TEMPERATURE EFFECTS, ORTHO TEMP, TABLE, and LATENT HEAT model definition options.
 Selection of initial conditions is done using the INITIAL TEMP option.
CHAPTER 10 759
Nonstructural Procedure Library
 Selection of the temperature and time-dependent boundary conditions (prescribed temperature history,
volumetric flux, surface flux, film coefficients, radiation, change of prescribed temperature boundary conditions
during analyses) is done using the FIXED TEMPERATURE, TEMP CHANGE, DIST FLUXES, POINT FLUX, QVECT, and FILMS
options. Moving heat sources due to welding can be specified using the WELD FLUX, WELD PATH, and
WELD FILL options.
 Import of viewfactors calculated by Mentat for radiation analyses.
 Selection of time steps using the TRANSIENT or AUTO STEP history definition option.
 Application of a tying constraints on nodal temperatures using the TYING model definition option.
 Generation of a thermal load (temperature) file using the POST option, which can be directly interfaced with the
stress analysis using the CHANGE STATE model definition option.
 Use of the ANKOND user subroutines for anisotropic thermal conductivity, FILM or UFILM for convective and
radiative boundary conditions or FLUX or QVECT for heat flux boundary conditions.
 Selection of nodal velocity vectors for heat convection is done using VELOCITY and VELOCITY CHANGE options.
 Use of the UVELOC user subroutines for heat convection.
In addition, a number of thermal contact gap and fluid channel elements are available in Marc. These elements can be used
for heat transfer problems involving thermal contact gap and fluid channel conditions.
The CONRAD GAP option, used in conjunction with 4- and 8-node 2-D continuum elements or 8- and 20-node 3-D
continuum elements, provides a mechanism for perfect conduction or radiation/convection between surfaces, depending
on the surface temperatures. The perfect conduction capability in the elements allows for the enforcement of equal
temperatures at nodal pairs and the radiation/convection capability allows for nonlinear heat conduction between surfaces,
depending on film coefficient and emissivity. The perfect conduction is simulated by applying a tying constraint on
temperatures of the corresponding nodal points. An automatic tying procedure has been developed for such elements. The
radiation/convection capabilities in the elements are modeled by one-dimensional heat transfer in the thickness direction
of the elements with variable thermal conductivity.
For the purpose of cooling, the CHANNEL option allows coolant to flow through passage ways that often appear in the solid.
The fluid channel elements are designed for the simulation of one-dimensional fluid/solid convection conditions based on
the following assumptions:
1. Heat conduction in the flow direction can be neglected compared to heat convection;
2. The heat flux associated with transient effects in the fluid (changes in fluid temperature at a fixed point in space)
can be neglected.
For high velocity air flow, these assumptions are reasonable and the following approach is used:
 Each cooling channel is modeled using channel elements.
 On the sides of the channel elements, convection is applied automatically.
 The film coefficient is equal to the film coefficient between fluid and solid, whereas the sink temperature
represents the temperature of the fluid.
 A two-step staggered solution procedure is used to solve for the weakly coupled fluid and solid temperatures.
Thermal Contact
The CONTACT and THERMAL CONTACT options may also be used to define the fluxes entering the surfaces. For more
information, see Chapter 6: Contact in this volume.
Use the CONTROL model definition option to input convergence controls for heat transfer analysis. These options are also
used for all coupled thermal systems. These options are:
1. Maximum allowable nodal temperature change. This is used only for transient heat transfer analysis in conjunction
with an automatic time stepping scheme. It controls the time step size.
2. Maximum allowable nodal temperature change before properties are re-evaluated and matrices reassembled. This
is used only for transient heat transfer analysis in conjunction with an automatic time stepping scheme. For mildly
nonlinear problems, if the time step remains constant, the operator matrix is not reassembled until this value is
reached.
3. Maximum error in temperature estimate used for property evaluation. This control provides a recycling capability
to improve accuracy in highly nonlinear heat transfer problems (for instance, for latent heat or radiation boundary
conditions). When nonzero, this control is used for both steady state and transient heat transfer analysis when either
the fixed or adaptive stepping procedures are used. For the first iteration, the temperature estimate is based upon
the temperature at the beginning of the increment, the rate of change of temperature T, and the increment in time.
For subsequent iterations, the last calculated temperature is used for the estimate. If the time step is reduced, then
a fraction of the delta temperature will be used in determining the estimate for the next iteration.

For steady state problems, use the STEADY STATE history definition option. If the problem is nonlinear, use the tolerance for
temperature estimate error on the CONTROL model definition set to obtain an accurate solution.
You must distinguish between two cases of nonlinearity in steady-state solutions: mild nonlinearities and severe
nonlinearities. In the case of mild nonlinearities, variations are small in properties, film coefficients, etc., with respect to
temperatures. The steady-state solution can be obtained by iteration. After a small number of iterations, the solution should
converge.
The technique described above is not suitable for severe nonlinearities. Examples of severe nonlinearities are radiation
boundary conditions and internal phase-change boundaries. In these cases, you must track the transient with a sufficiently
small time step ( t ) to retain stability until the steady-state solution is reached. Clearly, the choice of t is dependent on
the severity of the nonlinearity. The number of steps necessary to obtain the steady-state solution can often be reduced by
judicious choice of initial conditions. The closer the initial temperatures are to the steady-state, the fewer the number of
increments necessary to reach steady-state.
Transient Analysis
Both fixed stepping and adaptive stepping schemes are available for transient heat transfer analysis. The fixed stepping
scheme is TRANSIENT NON AUTO. The adaptive stepping schemes are TRANSIENT and AUTO STEP.
CHAPTER 10 761
In the fixed stepping scheme, the program is forced to step through the transient with a fixed time step that is user specified.
Only the error in temperature estimate (convergence control 3) is used with the fixed stepping scheme and the first two
convergence controls are not checked.
The scheme for the TRANSIENT option is as follows:
1. After the program obtains a solution for a step, it calculates the maximum temperature change in the step and
checks this value against the allowable temperature change per step given on the CONTROL option (convergence
control 1).
2. If the actual maximum change exceeds the specified value, the program repeats the step with a smaller time step
and continues repeating this step until the maximum temperature change is smaller than the specified value or until
the maximum number of recycles given on the control option is reached (in which case the program stops).
3. If the actual maximum temperature change is:
a. between 80 percent and 100 percent of the specified value, the program goes on to the next step, using the same
time step.
b. between 65 percent and 80 percent of the specified value, the program tries the next step with a time step of
1.25 times the current step.If the actual maximum is below 65 percent of the specified value, the program tries
the next step with a time step that is 1.5 times the current step.
The objective of the scheme is to increase the time step as the analysis proceeds.
The TRANSIENT option can be used for both heat transfer as well as thermo-mechanical coupled analysis. The time step for
coupled analyses is only based on the convergence characteristics of the heat transfer pass.
The AUTO STEP option is a unified time-stepping scheme that is available for thermal, mechanical and thermo-mechanically
coupled analysis. The time step for coupled analyses is based on the convergence characteristics of each of the heat transfer
and structural passes. Details of the AUTO STEP scheme are provided in Chapter 11.
Let the temperature T  x  within an element be interpolated from the nodal values T of the element through the
interpolation functions N  x  ,
T  x  = N  x T (10-1)
The governing equation of the heat transfer problem is
dT
C  T  ------- + K  T T = Q (10-2)
dt
In Equation (10-2), C  T  and K  T  are the temperature-dependent heat capacity and thermal conductivity matrices,
respectively, T is the nodal temperature vector, T· is the time derivative of the temperature vector, and Q is the heat flux
vector.
The selection of the backward difference scheme for the discretization of the time variable in Equation (10-2) yields the
following expression:
1 1
----- C  T  + K  T  T n = Q n + ----- C  T T n – 1 (10-3)
t t
Equation (10-3) computes nodal temperatures for each time increment   t  .

For the evaluation of temperature-dependent matrices, the temperatures at two previous steps provide a linear
(extrapolated) temperature description over the desired interval:

T    = T  t – t  + -----  T  t – t  – T  t – 2t   (10-4)
t
This temperature is then used to obtain an average property of the material f over the interval to be used in Equation (10-3),
such that:
t
1
f = -----  f  T    d (10-5)
t
t – t
During iteration, the average property is obtained based on the results of the previous iteration:

T    = T  t – t  + -----  T*  t  – T  t – t   (10-6)
t
T*  t  are the results of the previous iteration.
Temperature Effects
The thermal conductivity, specific heat, and emissivity in a heat transfer analysis can depend on temperatures; however, the
mass density remains constant. Specify reference temperature values of thermal conductivity, specific heat, mass density,
and emissivity with the ISOTROPIC option. Enter temperature variations of both the thermal conductivity and specific heat
using the TEMPERATURE EFFECTS or the TABLE option. The temperature-dependent data can be entered using either the slope-
break point representation or the property versus temperature representation. During the simulation, the temperature
dependent properties are either at the estimated temperature if a steady-state simulation or based upon the mid-increment
value of the temperature for conductivity, specific heat, and convective (film) coefficients.
Latent heat can be induced because of a phase change that can be characterized as solid-to-solid, solid-to-fluid, fluid-to-
solid, or a combination of the above, depending on the nature of the process.
Phase change is a complex material behavior. Thus, a detailed modeling of this change of material characteristic is generally
very difficult. The use of numerical models to simulate these important phenomena is possible; several major factors
associated with phase change of certain materials have been studied numerically.
The basic assumption of the latent heat option in Marc is that the latent heat is uniformly released in a temperature range
between solidus and liquidus temperatures of the materials (see Figure 10-1). Marc uses a modified specific heat to model
the latent heat effect. If the experimental data is sufficient and available, a direct input of the temperature-dependent
CHAPTER 10 763
specific heat data (see Figure 10-2) can be used. Results of both approaches are comparable if the temperature increments
are relatively small.
Latent Heat Uniformly Release
Between ST and LT
Specific Heat ST – Solidus Temperature

LT – Liquidus Temperature
ST LT
Temperature
Figure 10-1 Modeling Phase Changes with the Latent Heat Option
Specific Heat
Temperature
Figure 10-2 Modeling Phase Changes with the Specific Heat Option
Initial Conditions
In a transient heat transfer analysis, Marc accepts nonuniform nodal temperature distribution as the initial condition. Enter
the initial condition through the INITIAL TEMP model definition option or through the USINC user subroutine. Initial
conditions are not required in steady-state heat transfer analysis, even though they can improve convergence when
temperature-dependent properties are included.
Boundary Conditions
There are two types of boundary conditions in transient/steady-state heat transfer analysis: prescribed nodal temperatures
and nodal/element heat fluxes. These boundary conditions are entered directly through input or through user subroutines.
Prescribe nodal temperatures using the FIXED TEMPERATURE model definition option. These prescribed temperatures are
constant with time unless a time dependent table is referenced, the FORCDT user subroutine is invoked, or a TEMP CHANGE
history definition option is encountered.
The POINT FLUX model definition option allows you to enter concentrated (nodal) heat fluxes such as heat source and heat
sink. These applied fluxes are constant with time unless a time dependent table is referenced, the FORCDT user subroutine
or a POINT FLUX history definition option is encountered.
The DIST FLUXES or QVECT model definition option allow the definition of distributed (surface or volumetric) heat flux.
These applied fluxes may be spatially varying by an invoked spatial table or by the use of the FLUX or UQVECT user subroutine.
These applied fluxes are constant with time unless a time dependent table is referenced, the FLUX or UQVECT user subroutine
is invoked, or a DIST FLUXES history definition option is encountered. A special case is the volumetric heat generated by
inelastic loads. This will be calculated by the program when the distributed flux type is 101. It may also be necessary to
include the CONVERT option. For spatially varying fluxes associated with a welding heat source, it is more convenient to use
the WELD FLUX model definition option. Specialized welding related options are discussed in the next section.
Notes: 1. In heat transfer analysis, you must always specify total values; for example, total temperature boundary
conditions or total fluxes. This specification is to be used consistently for the heat transfer portion of
analysis on coupled thermal-mechanical (thermal-solid), fluid-thermal, and fluid-thermal-solid.
2. The time variation used for thermal boundary conditions is a function of the stepping procedure that is
used. If adaptive stepping (either TRANSIENT or AUTO STEP) is used, thermal boundary conditions are
applied instantaneously. If fixed stepping (TRANSIENT NON AUTO) is used, the magnitudes of the
thermal boundary conditions are applied in accordance with the tables used to define them.
Use the FILMS model definition option to input the constant film coefficient and the ambient temperature associated with
the convective boundary conditions. Use the FILM user subroutine for time/temperature-dependent convective boundary
conditions. When not using the table input option, the expression of the convective boundary condition is
q = H  Ts – T  (10-7)
where q , H , T s , and T  are heat flux, film coefficient, unknown surface temperature, and ambient temperature,
respectively.
The radiative boundary condition can be expressed as:
q =   T s4a – T 
4 
a (10-8)
where q is the heat flux,  is the Stefan-Boltzmann coefficient,  is emissivity, and T sa and T  a are unknown surface
and ambient temperatures, respectively. The radiative boundary condition can be rewritten as:
q =   T s3a + T s2a T a + T s a T a
2 + T 3 T – T 
a s 
(10-9)
= H      T s T    T s – T  
This shows that the radiative boundary condition is equivalent to a nonlinear convective boundary condition, in which the
equivalent film coefficient H    T s T   depends on the unknown surface temperature T s . When using the nontable
driven input, this case requires the FILM user subroutine.
CHAPTER 10 765
When using the table driven input, the distributed thermal flux into an element may be defined as:
 4 4
q = H  T s – T   + H n v  T s – T   + e v f    T s a – T  a  + f (10-10)
where:
H is the conventional film coefficient

Ts is the unknown surface temperature
T is the sink temperature
T sa is the unknown surface temperature in absolute units
T a is the sink temperature in absolute units
H n v is the natural convection coefficient
 is the natural convection exponent
e vf in the effective view factor to the environment; usually equal to 1.0
 is the emissivity
 is the Stefan Boltzman constant
f is the distributed flux from other sources
The absolute temperature is the temperature in user units plus the offset, which is defined in the PARAMETERS model
definition option.
A surface flux may be entered on the FILMS model definition option using the table driven input or the DIST FLUXES model
definition option. If both convective behavior and other terms contributing to the flux are present it is computationally
efficient to do it here.
When using table driven input, all coefficients given above may be temperature dependent by referencing a table. If more
complex physical boundary conditions exist, then the UFLIM user subroutine should be used.
When using the table driven input, an alternative method is available for defining the sink temperature through the use of
the SINK POINTS model definition option. This allows the user to interface to exterior fluid/thermal programs to input
spatially varying environment temperatures. In Figure 10-3, temperatures may be known at the points marked with an X.
Using the SINK POINTS model definition option when evaluating the flux into an element, the program will first determine
the closest sink point to the surface integration point, and then use this sink point temperature in the calculation.
x
x x x x
x x x
Figure 10-3 Temperature Points

If the heat flux defined in the FLUX or FORCDT user subroutine is temperature dependent, then the convergence behavior of
the solution process of the set of nonlinear equations can be improved by defining not only the current flux value, but also
the derivative of the flux with respect to temperature. Based on this derivative, Marc adapts the conductivity matrix K  T 
as well as the heat flux vector Q . This is done by rewriting the heat flux using a Taylor series expansion. For a temperature
dependent point flux on node i , the flux during iteration n is thus given by:
i
i i dq i i
q n = q n – 1 +  -------  T – Tn – 1  (10-11)
 dT  n – 1 n
i
Since all the terms in Equation (10-11) except for T n are known from iteration n – 1 , the corresponding entries in the
conductivity matrix and heat vector are updated as:
i
ii ii dq
K n  K n –  ------- (10-12)
 dT  n – 1
i
i i dq i
Q n  Q n –  ------- T (10-13)
 dT  n – 1 n – 1
The temperature dependent distributed fluxes are treated in a similar way and require the FILM user subroutine.
If a film coefficient is temperature dependent, then, similar to temperature dependent fluxes, defining the derivative of the
film coefficient with respect to temperature may speed up the convergence of the iterative process. This derivative should
be defined in the FILM user subroutine.
The QVECT option may be used to define a directed flux into the surface. The user specifies a magnitude q 0 and an
orientation n . The magnitude of the flux applied is then Q = – q 0  A  n  n s where A is the area of the surface, 
˜ ˜
is the face absorptivity, and n s in the outward unit normal into the surface. The absorptivity is defined through the
EMISSIVITY option. Note that when using heat transfer membranes or shell element, the normal is based upon the right-hand
rule.
For both the FILMS and QVECT option using table driven input, the coefficients used in the expressions may be temperature
dependent or dependent upon other quantities.
For FILMS, this lead to:
q = H  T eval    T s – T  
and for QVECT, this leads to:

q = q 0  T eval     T e v al   A  n  n s 
˜
The user can specify that the evaluation temperature be either at the surface (TEMTYP=0), the average of the surface and
environment temperature (TEMPTYP=10), or the environment temperature (TEMPTYP=20).
Furthermore, for the FILMS option, one can specify that the environment temperature at a control node; in which case,
q = H  T eval    T s – T   T c nt r ln d  
CHAPTER 10 767
Control Systems
Often in heat transfer, one desires that a boundary condition is dependent upon the temperature at another point. In
HVAC systems, this is driven by the temperature at the thermostat (multiple thermostats). If multiple thermostats are
involved, an effective temperature should be obtained by using a Servo Link.
If a control node is specified, then the following boundary conditions are applied:
Point Flux
q = Q  T cntrlnd  FLUXTYPE=6
q = Q  T cntrlnd FLUXTYPE=4
Distributed Flux
q = q a  T cntrlnd   A FLUXTYPE=6
q = q a  T cntrlnd  A FLUXTYPE=4
Film
q = H  T eval    T s – T   T c n t r l nd   FLUXTYPE=6
q = H   T eval   T cn t r l n d   T s – T   FLUXTYPE=4
QVECT
q = q o  T cntrlnd     T e v al   A   n – n s  FLUXTYPE=6
˜ ˜
q = q o  T cntrlnd    T e v al   A   n – n s  FLUXTYPE=4
˜ ˜
The FLUXTYPE is specified in the respective options. FLUXTYPE=4 is compatible with MSC Nastran definition of control
node.
Velocity and Pressure Dependent Convection

In many applications, the convective coefficients are dependent upon the external fluid boundary conditions. In such cases,
one can express:
q = H  T eval ,V ,P    T s – T  
where, P and V are the pressure and the velocity of the flow field. This simulation can be done by using the VELOCITY
option to define the velocity and the FIXED PRESSURE option to define the pressure at the nodal points on the boundary. This
data is then interpolated to the integration points where it is used to evaluate the function/table describing H. Furthermore,
in a transient simulation where the velocity and pressure are a function of time, the VELOCITY and FIXED PRESSURE option
may be used to define the temporal variation. In the most general case, one then has:
q = H  T eval ,V  t  ,P  t  ,t    T s – T   t  
If spatial variation is also required, then the UFILM user subroutine is required.
Surface Energy
The SURFACE ENERGY option calculates contributions to the thermal boundary conditions, including the recession due to
thermochemical ablation at the surface of a material which is subjected to very high thermal fluxes. This includes the effects
of convective heat flux with a possible blowing effect by mass injection, an enthalpy flux due to molecular diffusion, a
thermochemical ablation due to heterogeneous chemical reactions with gases, a possible thermal internal decomposition,
a thermochemical ablation and a mechanical erosion by liquid or solid particles impacts. Surface energy allows one to
obtain the recession due to thermochemical ablation by gases or particles. The mechanical erosion must be calculated by
use of a UFLUXMEC user subroutine, and added to the thermochemical ablation in order to obtain the total surface recession.
Surface energy is coupled with the heat transfer into the material by the conductive heat flux, and possibly with the mass
flow rate towards the surface due to water evaporation and/or thermal decomposition gases generated inside the material.
The SURFACE ENERGY option may be used in any heat transfer analysis, but depending on the model, some terms could be
null. When used in conjunction with the THERMO-PORE option for materials undergoing pyrolysis, most of the terms
are nonzero.
The PRINT,15 parameter allows the printing of each contribution to the surface energy.
Physical Presentation
particles
radiative blowing impact
convection diffusion
balance
flow
surface
wall
conduction decomposition ablation by ablation by

gases particles
Figure 10-4 Schema of the Surface Energy Balance
The surface energy balance is summarized by this equation:

Convection + Enthalpy flux due to external chemical species diffusion effects + Enthalpy flux due to
thermochemical ablation by gases + Enthalpy flux due to internal thermal decomposition products flow – Enthalpy
flux due to blowing of gaseous products + Enthalpy flux due to thermochemical ablation by impacting particles +
Thermal effects of mechanical erosion by impacting particles – Enthalpy flux due to mechanical removal of liquid
phases at the surface + Radiative heat transfer balance – Thermal conduction = 0
CHAPTER 10 769
Mathematical Presentation
The equation below is noted
 c onv  x ,t ,T s  +  diff  x ,t ,T s  +  s ,t h ,g  x ,t ,T s ,m· · ·
s ,t h  +  g  x ,t ,T s ,m g ,p ,m g ,w 
· · · · ·
–  blo w  x ,t ,T s ,m g ,p ,m g ,w ,m s ,t h  –  l i q  x ,t ,T s ,m l ,j  +  s ,t h ,p  x ,t ,T s ,m p ,j , 
+  s.m.p  x ,t ,T s ,m· ,  +   x ,t ,T  – 
p ,j rad s c on d  x ,t ,T s  = 0
It is homogeneous with a power per surface unit  W.m – 2  .

Details on Terms:
Convection: enthalpy flux at the surface due to processes of heat conduction in the
boundary layer.
Diffusion: enthalpy flux at the surface due to processes of chemical species diffusion in
the boundary layer.
Thermochemical ablation by gases: enthalpy flow associated to mass flow of thermochemically ablated material
by gases.
Decomposition products flow: enthalpy flow from inside of the material and ejected to the boundary layer
without change of phase (generally pyrolysis gases and water).
Blowing: enthalpy flux due to evacuation of gases freed both by thermochemical
ablation (by both gaseous products and chemical effects of particle impacts)
at the surface, and internal thermal decomposition (including possibly
water evaporation).
Liquid phases removal: enthalpy flow associated with the blowing of liquid phases possibly
appearing at the surface, by external actions (flow, …)
Thermochemical ablation by external enthalpic flow associated with mass flow of thermochemically ablated
particles: material by surface impacting particles.
Thermal effects of mechanical erosion by enthalpy flow associated with energy given to the surface by impacting
impacts of particles: particles (fraction of conversed kinetic energy).
Radiative balance: balance of radiative heat flux, resulting from absorption and emission.
Conduction: apparent conductive flow within the material.
Each time that a upper-case H is used for an enthalpy, it means that is the total enthalpy, including both the formation
enthalpy and the so-called sensible enthalpy. If we refer only to the sensible part, a lower-case h is used. In all cases, we
refer to specific enthalpies J.kg – 1 .
Equation of Each Term of the Energy Balance:
T
s
Convection:  H  H rec – H e 
 
with:
 H =  H  x ,t ,T s  heat transfer coefficient (homogeneous with surface mass flow rate

 kg.m – 2 .s – 1  /specific enthalpy  J.kg – 1  ). In some papers,  H is sometimes written
as  e u e C H , with C H being the Stanton number for heat transfer, and  e u e the mass
flow rate of the external flow parallel to the surface at the edge e of the boundary layer.
H re c = H rec  x ,t  specific recovery enthalpy of the external flow   J.kg – 1   . This quantity depends also
implicitly on the external pressure. But this dependence is hidden behind the time
dependence (because the pressure is itself a function of the time and the space variable).
Ts Ts
He = H e  x ,t ,T s  specific enthalpy of the external flow, calculated for the frozen chemical composition
existing at the edge e of the boundary layer, but evaluated at the surface temperature T s .
This term is calculated by neglecting the variation of the chemical composition of the
external flow between the edge of the boundary layer and the wall, due to both chemical
reactions in the gas phase, and molecular diffusion of chemical species in the boundary
layer with possibly unequal mass diffusion coefficients for the species. Other diffusion
effects of the second-order are neglected (Soret effect, Dufour effect). The non-neglected
diffusion effects are taken into account in the diffusion term. This quantity depends also
implicitly on the external pressure. But this dependence is hidden behind the time
dependence (because the pressure is itself a function of the time and the space variable).
Often, a 'blowing correction' modifies the heat transfer coefficient. In fact, there are two possibilities:
1. the CFD software allows one to take into account the blowing in some manner, for instance by using a simplified
subroutine representing the material behavior as a boundary condition. In that case, the coefficient  H  x ,t ,T s  is
obtained directly.
2. the CFD software allows one to calculate only a heat transfer coefficient  H 0 for 'inert' wall (no blowing). So, this
coefficient must be corrected in the thermal software. The usual correction is:
2B' ; Ln  1 + 2B' ; 
CH H v0  v
---------- = ---------- = --------------------------------------------- = -----------------------------------------
CH H
0 0 exp  2B' ; – 1 2B' ;
 v0  v
with:
 v  w  v  w
B' ;  --------------- B' ;  ---------------
v0 H v H
0
 is the ‘transpiration factor’. Usual values are  = 0.5 for laminar flow and  = 0.4 for turbulent flow.
Marc has two options to enter the heat transfer coefficient: either directly  H or  H with the above correction. For the
0
latter, there is both the standard form above with  as an entry in the data deck, and a UFAH user subroutine allowing one
to enter another kind of correction.
CHAPTER 10 771
Ts
Diffusion:  M   Z i ;* – Z i ;* H
 e s i
i
with:
 M =  M  x ,t ,T s  mass transfer coefficient (homogeneous with surface mass flow rate  kg.m – 2 .s – 1  /specific
enthalpy  J.kg – 1  ). In some papers,  M is sometimes written as  e u e C M ; C M being the
Stanton number for mass transfer and  e u e the mass flow rate of the external flow parallel to
the surface at the edge e of the boundary layer. Very often, C M (or  M ) is supposed to be
correlated to C H (or  H ) by C M  C H  Le  n , Le being the Lewis number for the frozen
external flow. This Lewis number is often supposed to be equal to 1, so that C M  C H and n
is meaningless Otherwise, n depends on the configuration of the external flow, a value
n = 2  3 is generally accepted. For the sake of generality, one can enter separate entry tables
for  H and  M . If the heat transfer coefficient  H is corrected for blowing then the same
blowing correction is applied to the mass transfer coefficient  M . Therefore,
Ln  1 + 2kB' v 
 M =  M 0 -------------------------------------
2kB' v
where  M 0 is the mass transfer coefficient entered in the input file and B' v as defined in the
covection term.
T T specific enthalpy of the chemical component i in the external flow, evaluated at the surface
H i s = H i s  x ,t ,T s  temperature T s   J.kg – 1   .
special fraction characterizing the component i in the chemical composition of the external
Z i ;*e = Z i ;*e  x ,t 
flow at the edge e of the boundary layer. Z i ;*e has no dimension (no unit). It is a rather
complicated algebraic combination of the mole fractions and the mass fractions of all the
chemical species existing in the flow. These quantities depend also implicitly on the external
pressure and on the temperature. But these dependencies are hidden behind the time
dependence (because the pressure and the temperature of the flow are themselves functions of
the time and the space variable).
special fraction characterizing the component i in the chemical composition of the external
Z i ;*s = Z i ;*s  x ,t 
flow near the surface s of the material. The difference between Z i ;*e and Z i ;*s characterizes
the diffusion effects across the boundary layer. These quantities depend also implicitly on the
external pressure and on the temperature. But these dependencies are hidden behind the time
dependence (because the pressure and the temperature of the flow are themselves functions of
the time and the space variable).
Ts Ts
The quantities  Z i ;*e H i  x ,t ,T s  and  Z i ;*s H i  x ,t ,T s  are tabulated as a whole.
i i
Thermochemical Ablation by External Gases: B' C  M H S
B' C normalized surface mass flow rate of thermochemically ablated surface material by the
action of external gases (dimensionless).
M mass transfer coefficient ([kg.m-2.s-1]). If the heat transfer coefficient  H is corrected
for blowing then the same correction is applied to the mass transfer coefficient  M .
Therefore,
Ln  1 + 2kB' v 
 M =  M 0 -------------------------------------
2kB' v
where  M 0 is the mass transfer coefficient entered in the input file and B' v as defined
in the covection term.
HS total enthalpy of the solid material evaluated at the surface temperature. If the
material undergoes pyrolysis the enthalpy of the solid material is either the total
enthalpy of the charred material or the total enthalpy of the material in course of
pyrolysis (weighted average of total enthalpies of the virgin and charred material),
based on the user choice. The total enthalpies are calculated from the specific heats
and enthalpies of formation are entered in the material properties.
·
Decomposition Products Flow: m ·
g p H g p + m g w H g w
·
m surface mass flow rate of gaseous products produced internally by pyrolysis,  kg.m – 2 .s – 1  ).
g p
Note that if there is no pyrolysis in the model defined by the THERMO-PORE option, this term
is zero.
H g p total enthalpy of the pyrolysis gas calculated from the specific heat and the enthalpy of formation
of the pyrolysis gas entered in the material properties and evaluated at the surface temperature
T s   J.kg – 1   .
·
m surface mass flow rate of gaseous products produced internally by water drying  kg.m – 2 .s – 1  ).
g W
H g W total enthalpy of the gas issued from water drying, calculated from the specific heat and the
enthalpy of formation of vapor entered in the material properties and evaluated at the surface
temperature  J.kg – 1  .
Blowing:  v  w H w
CHAPTER 10 773
with:
·
 v  w = m · ·
g p + B' C  M + m s t h p – m l surface mass flow rate of all gaseous products leaving the surface of
the material (including gas coming from internal thermal
decomposition [pyrolysis, water, …], thermochemical ablation by
external gases, thermochemical ablation by impacting particles)
 kg.m – 2 .s – 1  .  M is the mass transfer coefficient
 kg.m – 2 .s – 1  . If the heat transfer coefficient  H is corrected for
blowing then the same correction is applied to the mass transfer
coefficient M .
Ln  1 + 2kB' 
v
--------------------------------------
Therefore,  = 
M M0 2kB'
v
where B' v depends on different mass flow rates of ablated or eroded

solid material as defined in the covection contribution.
· m
H w = H w x ,t , T s m · ·
g s ,t h ,g m s ,t h ,p  specific enthalpy of the mix of all gaseous products existing at the
surface, including blowed gases and chemical species from the
external flow, evaluated at the surface temperature T s   J.kg – 1   .
For each point x and time t (discrete values), a table gives H w as a function of the mass flow rate of gaseous products
coming from the inside (pyrolysis, water evaporation), the total mass flow rate coming from thermochemical ablation, and
the surface temperature T s .
Impacting Particles:
Particles impacting the surface can also contribute to the heat flux. The user can define this contribution globally for all
particles families or based on each particles family.
If the impacting particles contribution is defined globally for all particles families, then the user enters directly a heat flux.
This flux can represent the heat due to the thermochemical ablation triggered by the impacting particles  s ,t h ,p or the
heat due to the mechanical effect of the impacting particles  s ,m ,p . Therefore,  s ,t h ,p is the enthalpy flow associated
with mass flow of thermochemically ablated material by surface impacting particles.  s ,m ,p is the enthalpy flow associated
with mass flow of material eroded by surface impacting particles.
The heat flux due to impacting particles is:  s ,t h ,p +  s ,m ,p . These two fluxes can be entered in the input file or via the
UTIMP user subroutine.
·
The mass flow rate of material ablated thermochemically by the impacting particles m s ,t h ,p also contributes to the heat
·
flux though the blowing contribution:  v  w = m · ·
g p + B' C  M + m s t h p – m l .
If the impacting particles contribution is based on particles families, then the heat flux is either calculated by the G-law
described below or via the UTIMP user subroutine.
· H
Input based on families of particles: f t h ,p  T s   G t h ,p ,j m p ,j r ,p ,j
with:
· m
m · surface mass flow rate of particles for the j family   kg.m – 2 .s – 1   .
p ,j p ,g  x ,t 
G th ,p ,j = G th ,p ,j  V p ,j  x ,t  ,D p ,j , p ,j  x ,t  ,  empiric law for thermochemical ablation by impacting particles
(without unit)
H r ,p ,j = H r ,p ,j  T s ,J  specific enthalpy of reaction for the interaction between the surface
material and the j family of particles   J.kg – 1   .
f th ,p = f th ,p  T s  empiric correction for the effect of surface temperature (without
unit)
The summation is based on families of particles. Each family j is characterized by an impacting mass flow rate, a velocity
of impact V p ,j , an angle of impact  p ,j , and the mean diameter of the family D p ,j . The three first quantities are generally
dependent on time, and on the location on the surface. The mass flow rate of the surface material ablated thermochemically
by the family j is generally given by a correlation G t h ,p ,j  V p ,j ,D p ,j , p ,j  of these parameters (often called G-Law):
·
m ·
s ,th ,p ,j  x ,t  = G th ,p ,j  V p ,j  x ,t  ,D p ,j , p ,j  x ,t  m p ,j  x ,t 
The correlation is established by experiments. The G-Law can be entered by a UGLAW user subroutine or a table.The total
ablated mass flow rate m·
s ,t h ,p is obtained by summing the above formula for the j families.For the sake of generality, all
the families are not obliged to be composed by particles of the same chemical species. So it is more general to consider an
enthalpy of reaction different for each family j : H r ,p ,j .
The enthalpy of reaction depends on the temperature of the surface T s This enthalpy of reaction is a property of both the
surface material (which can be different depending on the location on the part to be calculated), and the family of particles
p under consideration. It is attached to the material, and the family of particle j is considered as a parameter.
This model does not take into account the effect of the temperature of the surface, as far as the ablated mass flow rate is
concerned. An empiric correction has been proposed by several authors. The value given above (after summation on j )
must be corrected by multiplying it by a function of the surface temperature. Several functions have been proposed:
 Ts – T1 
f 1  T s  =  0 if T s  T 1 ; ------------------- if T 1  T s  T 2 ; 1 if T 2  T s 
 T 2 – T 1 
1 Ts – T0
f 2  T s  = --- tan h  ------------------- + 1
2  T 
CHAPTER 10 775
1
with T 0 = ---  T 1 + T 2  and T = n  T 2 – T 1  .
2
These two corrections are implemented in standard input, and an alternate UFTHP user subroutine allows one to enter other
kinds of corrections.
·  x ,t  generally comes from an aerothermal analysis (performed with a CFD software) at discrete
A last point is that m p ,j
times and are calculated on a mesh which does not match with the thermal mesh. The analysis provides as well other
parameters such as V p ,j  x ,t  , D p ,j ,  p ,j  x ,t  . For 3-D analysis, the proper angle  p ,j  x ,t  will be calculated within the
CFD software (angle between the normal direction at the surface and the velocity vector, in the common plan defined by
these two vectors).
Liquid Phases Removal:
The removal of the liquid phases can also contribute to the heat flux. The user can define this contribution globally for all
phases or based on each phase.
If the liquid phase removal contribution is defined globally then the user enters a heat flux  l i q and a mass flow rate.
l i q enthalpy flow associated with the blowing of liquid phases possibly appearing at the surface, by
external actions (flow, …)
·
m1 surface mass flow rate of liquid products formed at the surface of the material, and blowed out by
the action of the external flow or other mechanisms.
·
The mass flow rate m 1 contributes to the heat flux though the blowing contribution:
·
 v  w = m · ·
g p + B' C  M + m s t h p – m l .
If the liquid phase removal contribution is based on different liquid phases:  m· 1 H 1

· ·  x ,t 
m1 = m 1 surface mass flow rate of liquid products formed at the surface of the material, and blowed out by
the action of the external flow or other mechanisms   kg.m – 2 .s – 1   .
H1 = H1  Ts  specific enthalpy of the liquid phase 'l' taken into account in, evaluated at the surface temperature
T s   J.kg – 1   .
This term takes into account possible liquid phases appearing at the surface of the ablating material, by several mechanisms
(phase change of a component of the material [silica, …], oxidation at the surface of gaseous products coming from the
inside forming liquid phase  SiO g  SiO 2  , for instance), and the blowing of these phases (gravity, shear stress, …).
Moreover, this kind of phenomena would require a more complex modeling, for example with the explicit modeling of the
liquid film running at the surface. Finally, there could exist a liquid film at the surface produced by other mechanisms such
as impacts of many liquid particles.
Radiative Balance:  r – r
ab em
This is defined through the FILMS or RAD-CAVITY option.
Mass Conservation Equation

Mass flow blown (injected in the outflow)  mass flow of removed liquid phases = mass flow of internally produced gaseous
products + mass flow of the gaseous products of thermochemical ablation reactions with external gases + mass flow of
gaseous products of thermochemical ablation reactions with impacting particles.
 v  w +  m· 1 · +m
= B' C  M + m g
· · ·
s ,t h ,p = B' C  M + m g + f t h ,g  G t h ,p ,j m p ,j
1
Surface Recession Rate

When the material undergoes surface recession a recession rate must be provided. There are 3 ways to do it.
The user can:
 Directly enter a recession rate (simple method) or
 Retrieve the recession rate from Surface Energy Balance data (advanced method) or
 Enter a recession rate via the UABLATE user subroutine (user sub. uablate method)
In all cases a recession rate S· is calculated.

The recession rate can be evaluated either at the nodes or at the surface integration points (SFIP). When the evaluation of
the recession rate is done at the SFIP, the code calculates a recession rate at the nodes by a weighted average of the SFIP
values of the elements adjacent to the node.
Wherever the calculation of the recession rate takes place, a recession rate at the nodes is calculated and a displacement
normal to the surface is provided.
This normal displacement can then be:
 Kept normal to the surface: the normal considered to move a node is the average of edge/surface normal of all
elements this node belongs to. The surface node is moved along this average normal with a distance equal to
S·  t , with time step t .
 Projected to the streamlines: the surface node is moved along the streamline with a magnitude equal to the
projection of the normal displacement to the streamline.
 Projected to the side edges of the surface element: the surface nodes are moved in the direction of the side edges
with a magnitude equal to the projection of the normal displacement to the side edge.
CHAPTER 10 777
The method to move the surface nodes as well as the location of the calculation of the surface recession are done in the Jobs
-> Analysis Option menu via the 'Normal' and 'Recession Calculation' buttons respectively. Note that the recession rate
calculation at nodes is recommended and prevents from interpolation/extrapolation errors.
Simple Model
The recession rate S· is entered. It is unit consistent with [ms-1]
Advanced model: data from Surface Energy Balance

In the advanced model the recession is represented by:
 A contribution representing the recession due to thermochemical effects called ablation S· t h , and
 A contribution representing the recession due to mechanical effects called erosion S· m ec .

· · ·
• The total recession rate is the sum of the ablation and erosion: S = S t h + S m ec [ms-1].
• The ablation can be due to external gases and due to impacting particles.
• The erosion can be due to impacting particles or can be described by what is called in Marc as the simple
erosion.
Surface Recession Due to Thermochemical Ablation
The recession rate due to ablation S· t h can come from mass flow rates of solid material caused by external gases or
impacting particles.
Ablation due to external gases
The mass flow rate of material thermochemically ablated by the action of impacting particles is equal to
m·
s ,th ,g = B' c  M ; where B' c is the normalized mass flow rate of solid material thermochemically ablated by external
gases and  M is the diffusion coefficient.
If the blowing correction is not active then where  M =  M0 where  M0 is the input for the diffusion coefficient.
Ln  1 + 2KB' V 
If the blowing correction is active then  M =  M 0 ---------------------------------------- where B' V depends on different mass flow rates of
2KB' V
ablated or eroded solid material as defined in the 'convection' contribution of the Surface Energy Balance.
Ablation due to impacting particles
·
The mass flow rate of material thermochemically ablated by the action of impacting particles is equal to m s ,t h ,p .
·
If this contribution depends on the family of particles, the mass flow rate m s ,t h ,p can either be entered as:
· ·
 m s ,th ,p = f th p  G t h p j m p j
 or be entered via the UTIMP user subroutine.

·
If this contribution doesn't depend on the family of particles, the mass flow rate m s ,t h ,p can either be entered directly as:
 An input
 Or be entered via the UTIMP user subroutine
All the terms of this contribution are explained in details in the Surface Energy.
CHAPTER 10 779
The recession rate due to ablation is obtained by dividing the mass flow rates by the mass density of the solid at the surface
̂ S . If the material undergoes pyrolysis (Streamline Pyrolysis Model or Darcy Pyrolysis Model), the density is the density
of the material in course of pyrolysis. The recession due to ablation is
S· th =  B' C  M + m
·
s ,th ,p   ̂ s
Surface Recession due to mechanical erosion

Erosion due to impacting particles
The recession rate due to erosion S· m e c can come from mass flow rates of solid material eroded by impacting particles.
If this contribution depends on families of impacting particles and if the flag triggering the use of the UTIMP user
subroutine is not active, then
· ·
m s ,me c ,p = f mec p  G m e c p j m p j ; where,
 The empirical correction f m e c p is entered via the UFMEC user subroutine.
 The flux correlation G m e c p j is entered via the UGMEC user subroutine

·
The mass flow rate of particles for the j family m
 p j is entered in the Surface Energy menu.
If this contribution doesn't depend on families of impacting particles or if the flag triggering the use of the UTIMP user
·
subroutine is active then m s ,m e c ,p is entered via the UFLUXMEC user subroutine.
Simple Erosion
The simple erosion is defined by a recession rate and a threshold defining a critical mass density of the solid material allowed
to be eroded. The idea is to control the erosion by the mass density inside the material (inside the surface element) and
only allow (enough) pyrolyzed material to erode. Therefore, the recession calculated from the input recession rate cannot
always be fulfilled, as explained below. Though, if the material is pyrolyzed enough then the recession is evaluated from the
recession rate input.
The code evaluates:
 The recession obtained from the recession rate S· intense erosion by the formula:
x rec rate = S·  t with time step t .
intense erosion
 The mass density of the solid ̂ r e c rate inside the material at the potentially eroded distance x r ec rate by linear
interpolation between the density at the surface node and the density at the node at the back of the surface
element. In case of quadratic elements, the linear interpolation is done either between the surface and the mid-
edge nodes or between the mid-edge and the back edge nodes.
 The critical mass density is given by the formula: ̂ cr i t =  c +  1.0 – pyr_threshold     v –  c 
 The distance x c r i t , which is the distance between the surface node and the location where the mass density of
the solid is equal to ̂ c r i t .

Then the recession is:
 Given by the recession rate , S· intense erosion if the potential recession x r ec rate leads to erode material with a mass
density ̂ rec rate lower than the critical mass density ̂ cr i t .

Otherwise, the recession is
 Calculated by Msc.Marc so that the recession reaches the material when the mass density is exactly equal to the
critical mass density ̂ c r i t .
In equations, it becomes:
if ̂ r e c rate < ̂ cr i t
S· mec = S· intense erosion
if ̂ r e c rate > ̂ cr i t
x c r i t
S· mec = --------------
t
The two situations are illustrated in the below diagrams where a surface element is represented.
In the figure below, the recession is calculated from the simple erosion recession rate that fulfills the mass density criteria.
In this case, the recession is given by the recession rate.
CHAPTER 10 781
The recession calculated from the simple erosion recession rate doesn't fulfill the mass density criteria is shown in the
below figure. In this case, the recession is driven by the mass density inside the element.
The recession rate due to erosion is: S· m e c = m

·
s ,m e c ,p  ̂ s or defined by simple erosion.
User Subroutine UABLATE

The user can define a recession rate S· . It must be unit consistent with [ms-1].
Surface Recession and Remeshing

In many mechanical and thermal processes, there is the loss or removal of surface material by an erosive process such as
melting or vaporization. Mechanical processes, such as particle impact or friction, may be involved as well. This process is
particularly significant in high temperature applications, such as spacecraft reentry, where the intentional removal of
material contributes to the thermodynamic cooling. For high temperature applications, special ablative materials have been
created which may be defined using the THERMO-PORE option. In the current release, ablation has been designed to be used
in thermal analyses only. The capability exists for 2-D, axisymmetric and 3-D solid elements.
There are seven aspects that control the receding surface calculation in Marc.
1. Specifying which surfaces are to undergo recession.
The RECEDING SURFACE option is used to specify which surface is to be subjected to recession. For nodes that reside
on elements edges (2-D) of faces (3-D) specified by the RECEDING SURFACE, the incremental displacement will be
calculated during the recovery phase.
2. Specifying the magnitude of the recession rate.
The recession rate is based upon either:
a. Simple model, that is either a constant, or evaluated by a table.
b. Based upon the data given in the SURFACE ENERGY model definition block.
c. Specified in UABLATE user subroutine.
3. Specifying the direction of the movement of the nodes on the surface.
The direction of the recession is based upon the normal evaluated at the node. If the ABLATION,1 parameter is used,
then the normal is based upon applying a unit distributed load to all external surfaces. The magnitude of the
equivalent force is the effective area, and the direction is the normal. This procedure is based upon lower order
elements. The result is that at corners of the mesh the calculated normal is not perpendicular to either surface, but
a weighted average (based upon the area of the edges).
This procedure may lead to more physically reasonable results, as corners tend to be rounded off, but leads to more
meshing problems.
If the ABLATION,2 parameter is used, then the normal is based upon applying a unit distributed load to those element
edges/faces specified by the ABLATION option. This results in the movement of the node lying at the corner to be
perpendicular to the surface.
If the ABLATION,3 parameter is used, then the direction of recession is based upon the direction of the streamline,
assuming that the streamline method is specified on the PYROLYSIS parameter.
If the ABLATION,4 parameter is used, then the direction of recession is based upon the direction of the streamline,
assuming that the streamline method is specified on the PYROLYSIS parameter. The magnitude of the recession is
equal to the projection of the velocity to the streamline.
If the ABLATION,5 parameter is used, then the direction of recession is based upon the direction of the side edge. The
magnitude of the recession is equal to the projection of the velocity to the side edge. With this parameter the
recession behaves as the parameter ablation, 4 but the streamlines here are not required.
4. Specifying when remeshing is to occur.
Remeshing is based upon three criteria specified via the ADAPT GLOBAL option. The tolerance values are relevant, but
not critical unless the recession rate is very high, the time step is large, or the elements are small. The two key issues
are:
a. The element should not be so distorted at the beginning of the increment that the recession during the
increment would lead to an inside-out element. This may lead to an Exit 1005.
b. The ablation criteria should only find at most one element along each streamline that should be shaved off. This
may lead to an Exit 1111.
With this in mind, the following remeshing criteria have been developed:
a. Element size reduction – two ratios are calculated. The first is the ratio of the current element edge length
divided by the original element edge length. The second ratio is the ratio of the projected current element edge
length with the surface normal divided by the projected original element edge length with the surface normal.
For a 2-D quadrilateral mesh for each element on the receding surface, two edges are checked; while for 3-D
hexahedral mesh, four edges are checked. If any edge satisfies the criteria, remeshing occurs. The recommended
value of the tolerance is 0.4.
b. Motion of a node. The magnitude of the displacement since the last remeshing is compared with the user-
specified tolerance. This method is not recommended.
c. The program also uses the previous displacements to predict if in the next increment the element will go inside
out. As long as the next increment is similar to the previous increment this is a reasonable procedure. If may fail
if the either the recession rate suddenly increases, or if the time step is increased. The user does not need to
provide any tolerance value. While in theory, this method could be used exclusively, without the ones
CHAPTER 10 783
mentioned above, this is not recommended, because it leads to very thin elements. Two procedures are used to
determine if an element has gone inside out. The first is based upon the Jacobian of the isoparametric shape
functions. This method was determined not to be conservative. Hence additionally the area of four triangles (2-
D) or volume of eight tetrahedrals is used.
5. Specifying the mesher used to create a new mesh
Currently the user needs to specify whether the shaver mesher, relax mesher, stretch mesher, or multi-zone mesher
is to be used. The shaver, relax, and stretch meshers may be used with quadrilateral and hexahedral elements. The
multi-zone mesher can only be used with quadrilateral elements.
a. The shaver (type 9) is advantageous because rezoning is required in only a smaller region of the mesh. Elements
are only modified if they have failed any of the criteria specified above, or they are adjacent to a modified
element. The shaved mesher is also advantageous because the original biased mesh is retained during the
process. The shaver mesher results in degenerated quadrilateral elements (triangles), or degenerated hexahedral
elements (wedges), which may not be esthetically pleasing.
b. The relax mesher (type 7) modifies every element in a body as soon as any one of them fails the mesh criteria
specified above. The relax mesher itself is simpler than the shaver mesher. The relax mesher also requires the
use of the SPLINE option to identify edges that are to be fixed.
c. The stretch mesher (type 8) modifies every element in a body as soon as any one of them fails the mesh criteria
specified above. The stretch mesher is similar to the relax mesher, but it only works in one direction, and it
reasonably assures that all elements have a uniform thickness in this direction.
d. The multi-zone mesher (type 20) checks along the node columns of the streamlines (see Figure 10-23) for the
threshold values defined as mesh criteria. In the initial mesh, the elements define a single region called D 3 . If
the D 2 – D 3 threshold value is reached, the mesh along the streamline is split into two zones: D 2 and D 3 . If
later on during the analysis, the D 2 – D 1 threshold is reached, the mesh along the streamline is split into three
zones: D 1 , D 2 , and D 3 . As soon as a threshold value is reached, the nodes are relocated using a geometric
series, such that in the region of a large solution gradient, a relatively fine mesh can be obtained. The number
of nodes in the model does not change.
Figure 10-5 Multi-zone Meshing
6. Which element type is being used and control of midside nodes

Either mesher may be used with either lower order or higher order elements for both 2-D and 3-D problems. When
using higher order elements, two considerations are made.
a. During the recession itself the midside node on the surface is positioned such that the surface geometry is linear
within and element, bi-linear for 3-D geometries. While this is not strictly consistent with the idea of higher
order elements, it resulted in less distortion problems. Temperature fields are still quadratic. The midside nodes
on the edges perpendicular to the surface that are in the interior of the element are adjusted such that they are
mid-edge. This was done to make sure the quadratic shape functions did not break down.
b. The remeshing process itself is based upon using lower order elements, and the midside nodes are then
repositioned to the midpoints.
7. Region geometry.
For the streamline model the element edges normal to the surface in theory should follow the streamline behavior;
i.e., be normal to the surface. This is also advantageous from a surface recession perspective. When the recession,
which by default is normal to the surface, is not aligned with the element edges, addition mesh distortion occurs.
To overcome this problem a projection procedure can be requested. This is only available in conjunction with the
shaver mesher, and is still not as good as creating a good mesh to begin with. It is highly advantageous to have
initially a good mesh.
Examples
1. Axisymmetric analysis with edges not perpendicular to surface). This is an example of a nonoptimal meshing
technique because the element edges are not perpendicular to the surface.
CHAPTER 10 785
Figure 10-6 Original Mesh – Note vertical edges are not normal to the surface
Figure 10-7 Time=44.0 seconds
Figure 10-8 Close Up Showing Effect of Edges Not Aligned Perpendicular to Surface
2. Axisymmetric analysis of a nozzle using four node element and shaver mesher. Here the mesh is well aligned with
the geometry, which leads to the ability to ablate further, and improves the remeshing.
Figure 10-9 Initial Mesh
Figure 10-10 Time = 220. seconds
Figure 10-11 Close Up at Leading Edge

CHAPTER 10 787
Figure 10-12 Final Temperatures
3. Example of nozzle using 2-D quadratic elements tolerance = 0.2
Figure 10-13 Initial Mesh, Showing Nodes
Figure 10-14 Time = 220 Seconds
Figure 10-18 is a close-up showing nodes, including those shaved off. Note that outer edges remain straight, and that
the midside nodes for the elements on the receding surface also remain at the midside of the edges. The element
distortion is due to the fact that the element edges are not aligned perpendicular to the surface.
Figure 10-15 Time = 220 Seconds
4. Example of 3-D nozzle.
Figure 10-16 Initial Mesh has 6480 Elements, 7657 8-node Bricks
CHAPTER 10 789
Figure 10-17 Time = 44.0 seconds
Figure 10-18 Close Up of Leading Edge

5. Example using linear elements and relax mesher.
Figure 10-19 Final Configuration
Note that the number of elements using the relaxed mesher does not change so elements become thinner during
the ablation process.
6. Example – 3-D – Linear hexahedral elements – Relax mesher – extruded nozzle.
Figure 10-20 Final Configuration - Linear 3-D Elements

CHAPTER 10 791
7. Example – 3-D – Quadratic hexahedral elements – Relax mesher – extruded nozzle.
Figure 10-21 Time = 44.0 Seconds
Note that in Figure 10-25, the outer edges remain straight and that the midside nodes for the elements on the
receding surface also remain at the midside of the edges.
Figure 10-22 Close Up of Leading Edge – Showing nodes

Pyrolysis
Pyrolysis involves the decomposition of materials due to thermal processes. There are two aspects in simulations of these
processes: the thermochemical decomposition of the material creating gaseous products and the transport of these gaseous
products. The creation of the gas is based on kinematic laws modeling the reaction kinetics of the decomposition and is
the same for the two pyrolysis model present in Marc. The transport of gas differentiates the two models called in Marc the
streamline model and the Darcy flow model. The choice of these two models is made on the PYROLYSIS parameter.
We focus here on the Streamline model. The Darcy flow model is explained in the Coupled Diffusion/Thermal analysis
chapter.
In the streamline model, the gas moves along streamlines that are aligned with the mesh, as shown in Figure 10-23. This
procedure is available for planar, axisymmetric and solid elements using either linear or quadratic interpolation functions.
Because the need for regular meshes, triangular and tetrahedral elements are not available. The user must identify which
element contain streamlines.
The streamline method introduces two new terms: streamlines and streamline integration points (SIP). For lower order
elements, the SIP are at the intersection of the streamlines and the element edges. For higher order elements, an additional
SIP exists midway between the two. The pyrolysis calculation occurs at the SIP. An extrapolation and interpolation
procedure is used to move data from the SIP to conventional integration points (CIP). The pyrolysis gas m · moves from
g
SIPi to SIPi+1. If the temperature becomes hotter in the interior, such that flow of gases would reverse, an error would occur.
When using the Darcy law model, any solid heat transfer element may be used. When this method is used, the pyrolysis
calculation occurs at the CIP. The pyrolysis gas may move in any direction.
SFIP
node
SIP Streamline Integration Point

CIP Conventional Integration Point
SIPi+1 SFIP Surface Integration Point
SIPi
CIP
Regular Mesh
Figure 10-23 Streamlines

CHAPTER 10 793
Presentation of the Mass Equation

The conservation of mass equation for the streamline model yields:
̂ s ,p* ̂ l ̂ s ,c ,c*

· = – ---------------
.m g – -------- – -------------------
t t t
where the first term is the conventional generation (source) term due to a change in density of the solid, the second term
represent the change in density of initial water vapor, and the final term is due to change in density due to coking.
·
m mass flow rate of the gases of decomposition = ̂ g  g .
g
̂ g mass density of the pyrolysis gas.
̂ s ,p * mass density of the solid undergoing pyrolysis.
̂ l mass density of the liquid vapor.
̂ s ,c ,c* mass density of the coked solid.
g velocity of the pyrolysis gas.
̂ s ,p*
– --------------- source term of decomposition.
t
̂ s ,c ,c*
– ------------------
- source term of carbon deposit.
t
̂
– --------l source term of water drying.
t
The drying of the initial water vapor occurs before the pyrolysis occurs so:
· ̂ l
.m g ,w = – -------
t
-
The global mass equation is defined as the sum of these. The global mass flow rate m ·
· is equal to m ·
g g ,p + m g ,w .
Principle of Pyrolysis
When pyrolysis occurs, the solid material is degraded due to the effect of the temperature. This degradation is done in steps
(Figure 10-24), according to the temperature. We represent these steps on the drawing below. With each step a loss of density
corresponds: ̂ s ,p ,j .
̂ s
̂ s ,p ,v
̂ s ,p ,1
̂ s ,p ,3
̂ s ,p ,c
Temperature
Figure 10-24 Degradation Steps
The density of material ̂ s thus varies from ̂ s ,p ,v , density of virgin material to ̂ s ,p ,c , the density of charred (entirely
pyrolyzed material). If coking occurs the density increases because carbon is redeposited in the solid
We define a dimensionless variable  j which varies from 1 to 0 during pyrolysis (it represents the advancement of
pyrolysis).
It is calculated by the law of Arrhenius:
 – T a ,j 
--------j = – B j exp  ------------  j j .
t  Ts 
The coefficients B j , T a ,j , and  j are determined by thermogravimetry (TGA). The material properties associated with
each phase (virgin, charred, coked, liquid, gas) are entered through the ISOTROPIC or ORTHOTROPIC options. The THERMO-
PORE option is used to reference these phases and provide additional material data including the terms of the Arrhenius
series.
̂ s ,p *
This variable  j facilitates the calculation of the variation in solid density due to pyrolysis, --------------- , which is the sum of
t
the preceding  j law of Arrhenius:
Nd Nd
̂ s ,p *  j – T a ,j j
---------------
t
=   ̂  s ,p ,j -------- = –
t   ̂  s ,p ,j B j exp  -----------
Ts  j
- 
j = 1 j–1
or
Nd Nd
̂ s ,p *  j – E a ,j 
--------------- =   ̂  s ,p ,j -------- = –   ̂  s ,p ,j B j exp  ------------  j j
t t  R * T
j = 1 j = 1
CHAPTER 10 795
where
Nd is the number of terms in the Arrhenius series.

T a ,j is the temperature of activation of the reactions of decomposition
j is the order of the reaction
j is the dimensionless variable of decomposition varying from 1 to 0 during pyrolysis
Bj is a factor to the exponential
E a ,j is the energy of activation of the reaction j of decomposition,
 ̂  s ,p ,j is the representation of the reduction of density (counted in positive quantity)
Ts is the local temperature of the solid phase.
R is the perfect gas constant.
It is assumed that at a point thermal equilibrium occurs and the temperature of the solid and the gas are the same.
Alternative Formulation
We see in the literature an alternative writing of the Arrhenius law know as the rho law:
Nd j
̂ s ,p T a ,j  ̂ s ,p ,j – ̂ s c ,j
------------ =  –  j B' j exp  – --------
- ̂ s ,v ,j  --------------------------------
t  T   ̂ 
s ,v ,j
j = 1
where:
̂ s ,v ,j is the virgin j material density.

̂ s ,c ,j is the charred j material density.
̂ s ,p ,j is the current j material density.
The user can enter the input for Arrhenius law for the both laws. If he chooses the rho law, then he must give:
Nd  j B' j E a ,j  j ̂ s ,v ,j ̂ s ,c ,j . If he chooses the  j law, then he must give: Nd ̂ s ,p ,j B j E a ,j  j .
One can define an alternate model for pyrolysis using the UPYROLSL user subroutine.
Coking
The processes considered when the material is heated are very complex. For instance, we can consider the primary pyrolysis
chemical reactions producing gases, and the subsequent secondary chemical reactions, between gases, between gases and
solid, and between constituents in the solid phase. We focus here on one class of secondary chemical reactions, namely the
heterogeneous reactions between gases and porous solid leading to a carbon deposit, known as coking phenomenon.
The processes mentioned above are governed by complex sets of kinetically controlled chemical reactions, which are neither
well known, nor the constants associated with them.
One Term Arrhenius Model

In this model, we consider the total mass fraction of atomic element carbon K̃ cg in the gaseous phase, independently of
the molecular configuration (chemical specie) in which it is present. The initial value K̃ cg 0 , corresponding to the gases
generated by the initial primary pyrolysis chemical reactions, is data that can be determined experimentally by calculating
the difference in carbon content between the virgin material and the one of the char before coking. We should note that
· is the mass flow rate of atomic element carbon independently of the molecular configuration.
K̃ c g m g
This model is inspired by a paper written by R.A. Rindal [Ref. 20].

During the subsequent process of secondary gas/gas chemical reactions, this quantity does not change, because only the
molecular configurations are changing in the gaseous phase, but not the content in any atomic element: K̃ cg = K̃ cg 0 .
It is known that the gaseous phase at lower temperature contains more carbon than it would contain in the case of chemical
equilibrium. When the gases flow towards the heated surface, the temperature is rising; the rate of reactions is rising too,
and the composition of gases approaches the one corresponding to chemical equilibrium. It results in a deposit of solid
carbon (coking), and then the quantity K̃ c g is changing in the zone where coking occurs.
From the considerations above, it seems reasonable to assume that the potential of coking is driven by the difference
between the local value K̃ c g and the value of K̃ c g E at chemical equilibrium, and that the rate with which chemical
equilibrium is approached is of Arrhenius kind. Furthermore, the homogenous chemical reactions in the gaseous phase are
generally not equimolar, so that a pressure dependence and an order of reaction different from the unity are assumed. So,
the following kinetic equation for the evolution of the quantity K̃ cg is proposed:
˜ –Ea c n
K c g n
- = – k c exp  ------------  P  c  K̃ c g – K̃ c gE  c
------------
t  RT 
with:
k c E ac n c : kinetic coefficients for coking reactions.

K̃ c g : total mass fraction of the atomic element carbon in the pyrolysis gas.
K̃ c gE : total mass fraction of the atomic element carbon in the pyrolysis gas when chemical equilibrium is
achieved.
The quantity K̃ cgE can be calculated using standard thermochemical calculation software.
The streamline model does not give access to the value of the local pressure P . There are two assumptions that can be made:
 Either neglect the pressure dependence, by setting P equals to 1

 Assume that P is equal to the external pressure.
In the current implementation the pressure  P  = 1.0

CHAPTER 10 797
When using the Darcy law model, the pressure to be considered will be the calculated pore pressure.
Remarks
The quantity K̃ cg will always be between K̃ c g 0 and K̃ c gE , and its variation rate will approach zero when K̃ cg becomes
close to K̃ cgE .
Based upon the assumption made in this analysis that the coking only begins when the material is completely pyrolysed,
we control the coking area by the advancement of the pyrolysis. Coking occurs if the advancement variable of pyrolysis at
this location is bigger than a critical value, named “pyrmax”. It is usually just smaller than 1., which means that pyrolysis
is over. This may be defined through the THERMO-PORE option; the default is 0.96.
Beside this Arrhenius model to calculate the mass fraction of carbon in the pyrolysis gas, a linear model is proposed.
Linear Model
This model determines the mass fraction of carbon in the pyrolysis gas as a function of the temperature. The evolution is
linear from a low temperature where the pyrolysis gas has an excess of carbon to a higher temperature where it is in
equilibrium:
1. For temperatures below the low temperature T l , there is no coking: K̃ cg = K̃ cg  T l  .
2. For temperatures between the low temperature and the equilibrium one:
• If the temperature is increasing, K̃ c g is a linear function of temperature. ((2))
• If the temperature is decreasing, the coking is stopped and K̃ cg retains its last value. This forbids the inverse
coking reaction that would lead to an increase of carbon in the pyrolysis gas.
For temperatures larger than the equilibrium temperature, there is no more chemical reaction: K̃ cg = K̃ cg E .
Figure 10-25 illustrates this:
The user can also specify a coking model through the UCOKSL user subroutine.
Kcg
~
Kcg(Tl)= K cg 0
KcgE
Tlow Thigh=TE Temperature

Figure 10-25 Calculation of Mass Fraction of Carbon in Pyrolysis Gas by Linear Function of Temperature
Water Drying Model

The water drying model chosen for the streamline model is based upon the same principle than the pyrolysis one, namely
an Arrhenius law.
Written in function of the liquid mass density variable ̂ l , the Arrhenius law for water drying is:
w
̂ Ew   ̂ l 
--------l = – B w * exp  – ----------
- *̂ ---------
t  R*T l ,0  ̂ 
l ,0
̂ l
If we set  w = 1 – --------
- , we have:
̂ l ,0
  1 – w  Ew  
̂ l ,0 ------------------------- = – B* exp  – ----------
- *̂ l ,0  1 –  w  w ,
t  R*T 
So the temporal gradient of  w is:
 w Ew  
---------- = B w * exp  – ----------
- *  1 – w  w
t  R*T 
CHAPTER 10 799
where:
̂ 1
w is defined by  w = 1 – ---------- , and represents the drying state. It varies from 0 to 1 during evaporation,
̂ 1 ,0
Bw is a pre-exponential factor,
Ew is the energy of activation of the reaction of water drying,
w is the order of the reaction,
T is the local temperature of the solid phase.
R is the perfect gas constant.
Moreover, we have:
̂  w
--------l = – ̂ l ,0 * ----------
t t
where ̂ l ,0 is the mass density of liquid at the beginning of the analysis.
̂ l
The user can define its own law to calculate the term -------- via a UWATERSL user subroutine.
t
Monitoring Thermal Degradation

A special parameter can be evaluated and made available on the POST file providing a measure for the total thermal
degradation of a material. This parameter is evaluated during the heat transfer pass of a coupled diffusion-thermal analysis,
that uses the Darcy fluid pyrolysis model, but it has no effect on the heat transfer process. It merely provides a measure for
the thermal degradation of the material, which may affect its mechanical properties, so it is primarily an output quantity,
that gets written to the post file when post code 260 is used. This parameter can be evaluated by either an Arrhenius law
for which the material constants are entered on the THERMO-PORE model definition option or by the UTHERDEG user
subroutine. The Arrhenius degradation law evaluates the degradation parameter  m by means of the following rate
equation
d m
---------- = 0 if T  T thre s
dt
Ta
d m n m  – ------
---------- = – B m e  T  if T  T t h r e s
dt
The value range of  m is 0   m  1
m = 1 : no degradation, i.e. fully intact material

m = 0 : full degradation
The degradation only gets evaluated when the temperature rises above a certain threshold temperature T t h r es ; below this
temperature, it remains at its current value. The constant B , the exponent n m , and the activation temperature T a are the
material constants in the Arrhenius law. The default initial value for  m is  m = 1 , but other values can be entered on
the INITIAL PYROLYSIS model definition option. This degradation parameter is only available in the Darcy fluid pyrolysis
model, but not in the streamline pyrolysis model.
The integration of the rate law for the thermal degradation is done as follows. The rate of degradation for temperatures
above the threshold is
Ta
d m n  – ------
- = – B mm e  T 
---------
dt
The incremental value  m is determined in the thermal pass of the coupled diffusion-thermal analysis by a mid-
increment integration.
 Ta 
n  – ------------------
-
d m  mid  m  T m i d 

---------- = – B m e t
dt
where
 start   e nd 
mid m + m
m = ----------------------------------------
2
and
T  s tart  + T  e nd 
T  m i d  = -----------------------------------------
2
The superscripts “start” and “end” refer to the start and end of the current increment, respectively, and t is the time step
of the increment. The degradation parameter is now updated as:
 en d   s tar t 
m = m +  m
CHAPTER 10 801
Presentation of the Energy Equation

To take into account pyrolysis and coking and water evaporation in the calculation of the temperature, a convective and
three enthalpic terms are added to the standard equation. The phenomena is included in:
 · . The latter decreases during coking because of the mass loss of
The convective term by the mass flow rate m g
gaseous carbon to solid carbon.
̂ s ,p*
 An enthalpic term ---------------  H g ,p – H s ,p ,v c  that represents the energy consumed by the pyrolysis reaction.
t
̂ s ,c ,c*
 An enthalpic term – -------------------  H c * – H g ,p  that represents the energy consumed by the coking reaction.
t
̂ l
 An enthalpic term --------  H v – H l  that represents the energy consumed by the evaporation of the water vapor.
t
The equation becomes:
T · .T =
 ̂ s ,i* c pi + ̂ s ,p* c s ,p* + ̂ s ,c ,c* c c *  ------- + c p g ,p m g
t
̂ s ,p* ̂ s ,c ,c * ̂ l
.  *T  + ---------------  H g ,p – H s ,p ,v c  – -------------------  H c * – H g ,p  + --------  H v – H l 
t t t
where,
 ̂c p  eff is effective heat capacity of the material during the analysis.
c p ,g is the effective specific heat of the gas (pyrolysis and vapor).
* is the effective conductivity of the material during the analysis.
H g ,p is the enthalpy of the gas, including only gas of decomposition.
H s ,p ,vc is the enthalpy of the solid in course of pyrolysis (no liquid influence).
Hv is the enthalpy of the water vapor.
Hl is the enthalpy of the liquid water.
The effective specific heat c p ,g is based upon a mixture law between heat capacity of pyrolysis gas  ̂ g ,p c p ,g ,p  and the
one of vapor ̂ g ,w c p ,g ,w :
We have ̂ g c p ,g = ̂ g ,p c p ,g ,p + ̂ g ,w c p ,g ,w
m · c ·
g ,p p ,g ,p + m g ,w c p ,g ,w
And so c p ,g = -----------------------------------------------------------
· -
m g
To take into account the effect of water drying in the material properties, we use the drying state  w defined above by
̂ l
 w = 1 – --------
-.
̂ l ,0
The effective heat capacity is defined by:
 ̂c p  eff =  1 –  p   ̂c p  v +  p  1 –  c   ̂c p  c +  p  c  ̂c p  cd +  1 –  w ̂ l ,0 c p ,l .
The effective conductivity * =  1 –  p  v +  p  1 –  c  c +  p  c  cd
The heat capacity of the virgin material, charred material, deposit by coking effect and the liquid part
 ̂c p  v  ̂c p  c  ̂c p  c d  ̂ c 
l ,0 p ,l
The conductivity of the virgin material, charred material, deposit by coking effect  v  c  cd
We can note that:

 The vapor is not taking into account, which is consistent with the fact that we neglect also the effect of the
pyrolysis gas,
 The conductivity does not depend on the liquid water. Indeed, we can consider that the liquid belongs to small
porosity, and that water evaporates before the beginning of the pyrolysis.
We have, respectively:
 The rate of pyrolysis, coking  p  c
 The drying state  w
 1 ̂ s ,c ,c*
--------c = ----------------------------------- -------------------
t ̂ c ,d* – ̂ s ,p ,c * t
The enthalpy of material in the course of pyrolysis function of the densities and enthalpies of virgin and charred material:
̂ s ,p ,v* H s ,p ,v * – ̂ s ,p ,c * H s ,p ,c *
H s ,p ,vc = --------------------------------------------------------------------------
-.
̂ s ,p ,v* – ̂ s ,p ,c *
The enthalpies, written in upper case, are the absolute enthalpies. They are equal to:
T
H = H f0 + h = H f0 + T c p dT .
0
H f0 is the enthalpy of formation, is an input of the program.

T0 is the temperature of reference for enthalpy of formation, is an input of the program.
h is the sensible enthalpy.
CHAPTER 10 803
Welding
Welding is a thermal process with specialized boundary conditions. These boundary conditions are specified through the
WELD FLUX model definition option in conjunction with the WELD PATH and WELD FILL model definition options. Using these
options, the welding heat input can be specified through two different techniques:
 Modeling the heat input as a spatially varying distributed flux through the WELD FLUX model definition option
applied to the base metal and filler elements. No temperatures are specified on the WELD FILL model definition
option in this case.
 Modeling the heat input as a spatially varying temperature boundary condition through the WELD FILL model
definition option applied to the nodes of filler elements. No flux value (zero power) is specified on the WELD FLUX
model definition option in this case.
It is also possible to combine a nonzero flux applied to base metal elements through the WELD FLUX model definition option
with a nonzero temperature applied to filler element nodes through the WELD FILL model definition option. In all these cases,
the heat input path for the welding heat source or for the filler element deposition is defined by the WELD PATH model
definition option.
Weld Flux
A disc shaped surface/edge heating source, a double ellipsoidal shaped volume heating source, or a cylindrical shaped
volumetric heating source can be specified using the WELD FLUX model definition option. For more complex shapes, the
heating source can be specified through the UWELDFLUX user subroutine. The associated weld path (defined through the
WELD PATH model definition option) and the associated filler element set (defined through the WELD FILL model definition
option) are also referred to through the WELD FLUX model definition option.
Pavelic’s disc shaped weld heat source is expressed as
3Q – 3x 2  – 3z 2 
q  x y z  = -------- exp  ------------ exp ----------------- [Ref. 16] (10-14)
r 2  r2  r2
where q is the heat flux rate per unit area; Q = VI is the applied power with  = efficiency, V = Applied Voltage and
I = Applied Current; r is the radius of the disc; z is the local coordinate along the weld path and x is the local coordinate
along the tangent to the weld path. The disc shaped model can be used for specifying edge fluxes in 2-D and surface fluxes
in 3-D. It is particularly useful for situations where the depth of weld penetration is not significant.
Goldak’s double ellipsoidal shaped weld heat source can be used to specify volume fluxes in 2-D and 3-D [Ref. 16]:
6 3f f Q – 3x 2 – 3y 2 – 3z 2
q f  x y z  = ----------------------- exp  ------------ exp  ------------ exp  ------------
abc f    a2   b2   c2 
6 3f r Q (10-15)
– 3x 2 – 3y 2 – 3z 2
q r  x y z  = ----------------------- exp  ------------ exp  ------------ exp  ------------
abc r    a2   b2   c2 
where q f and q r are the weld flux rates per unit volume in the front and rear weld pools respectively; Q = VI is the
applied power; a is the weld width along the tangent direction x; b is the weld penetration depth along the arc direction
y; c f and c r are the forward and rear weld pool lengths in the weld path direction z; f f and f r are dimensionless factors
given by
2
f f = -----------------------------
 1 + cr  cf 
(10-16)
2
f r = -----------------------------
 1 + cf  cr 
The double ellipsoidal source is also shown below in Figure 10-26.
cr cf z
a
x
b
y
Figure 10-26 Double Ellipsoidal Weld Flux showing the Local Coordinate Dimensions
The weld pool dimensions in Equations (10-14) and (10-15) are to be estimated and provided by the user. The width, depth,
forward length and rear length are used for the volume weld flux. The surface radius is used for the edge or face weld flux.
The cylindral shaped weld surface is used for laser heating applications as the heat source is more focused and penetrates
deeper into the material. The shape is described by Figure 10-27. The dimensions are mandatory when the standard disc or
ellipsoidal models are used and are optional when the UWELDFLUX user subroutine is used. The weld pool dimensions are
used in three different ways by the program:
 For defining the weld flux rate,
 For defining a bounding box for filler element detection, and
 For defining box dimensions when a box adaptive meshing criterion is used.
The weld pool dimensions can be optionally varied with the help of tables. The tables can be functions of time or arc length.
The arc length is measured along the associated weld path, from the first point of the path to the current position of the
weld flux.
The weld flux rate in Equations (10-14) and (10-15) can be further scaled by a factor. This scale factor can be automatically
calculated by the program or optionally set by the user. For 3-D problems, the automated scale factor is set to 1. For 2-D
problems (both planar and axisymmetric), the automated scale factor s is calculated by equating the integral of the flux
rate over the entire x-y plane and the out-of plane thickness t to the applied power Q .
s   q  x y 0 tdxdy = Q
CHAPTER 10 805
Heat Source in total:

Cylindrical part with surface part
Surface part with 1-Cylinder Surface Depth only for

heat fraction) x Qtotal as search area for faces
Gaussian-Surface distribution with heat input
in [W/m2]
Cylinder part with

cylinder heat fraction x Qtotal
as constant heat input q
in [W/m2]
Surface radius cut-off at 2 x ,

above q = 0

Surface Depth
Point Position , 
(Origin of local
CO-System)
Cylinder Depth
Cylinder Radius
Figure 10-27 Cylindral Shaped Weld Surface

For Equations (10-14) and (10-15), the scale factor can be shown to be
 r
s = --- -
3 t
(10-17)
  cr + cf 
s = --- ---------------------
3 2t
It should be noted that the automated scale factor only ensures that the applied fluxes in the 2-D case matches the 3-D
case. In general, the temperature solution comparisons between 2-D and 3-D also depend on the velocity of the weld
source, the thermal properties, other thermal and mechanical boundary conditions, and heating effects before and after the
heat source has crossed the 2-D plane.
The total heat input applied through the weld flux should, theoretically, be equal to the power Q . However, the integration
of the weld flux over the applied finite element domain will typically not match Q due to two reasons. First, there is an
error due to the numerical integration over the element integration points. Second, there is an error due to the integration
over the finite dimensions specified for the weld flux (note that only integrating the weld flux over an infinite domain would
yield Q ). Even for a very fine mesh for which the numerical integration error is negligible, due to the second reason, the
applied weld flux will be in the range of 0.85 to 0.9 of Q .
For a 3-D analysis, it is possible to optionally adjust the applied weld flux such that the total weld flux matches Q . This is
activated through the WELD FLUX model definition option for all the welds in the model. When activated, an auto scaling
factor is computed and applied such that the total heat input on the finite element domain equals Q .
In a restart run, existing WELD FLUX boundary conditions from the original run can be used without any modifications. One
cannot modify and reuse an existing WELD FLUX boundary condition. A new WELD FLUX boundary condition can, however,
be defined and activated for a new loadcase. Note, however, that only direct flux parameters (e.g. power of the weld flux,
dimensions of the weld flux, position and velocity of the weld flux, etc.) can be changed; references can only be made to
WELD PATH and Weld Filler options from the original run.
Weld Path
The path followed by the welding heat source and the orientation of the weld arc are specified by the WELD PATH model
definition option. This information is used to determine a local coordinate system for the moving source, as shown in
Figure 10-28. The weld path vector is taken as the local z axis. The arc orientation vector is taken as the local y axis. The
tangent vector that is perpendicular to both y and z axes is taken as the local x axis.
It is possible to specify the weld path and orientation information through the Marc Input, Text Input or the UWELDPATH
user subroutine.
CHAPTER 10 807
Y (weld path) z
X
x (weld width)
Z
y (weld orientation)
Figure 10-28 Local Coordinate System for Moving Source obtained by Translation and Rotation of Global Coordinate System
Marc Input
Two options are available for specifying the weld path – Ordered List of Nodes or Point Coordinates of Polyline Curves.
The weld path vector at point N is defined as the vector from point N to point N+1. For a closed weld path, the first and
last node should be identical, or, the first and last point should have the same coordinates. An update path option is
provided for the nodes option for a coupled analysis which allows the path to be updated based on the current
displacements of the nodes along the path. This option is used in conjunction with the FOLLOWER FOR parameter. When
the latter flag is used, all distributed fluxes are evaluated on the updated geometry and the path followed by the weld source
is also updated in this case.
A variety of options are available for specifying the arc orientation: Ordered List of Nodes, Point coordinates of Polyline
Curves, Vector Components, or Euler Angles. When nodes or points are used, the arc vector at point N is defined as the
vector from the weld path point at N to the arc orientation point at N. It is further possible to rotate the arc vector through
an angle specified in degrees about the weld path axis. When vector components are used, they can be specified at distinct
points and can be optionally interpolated as a function of arc length (length of the path from the first point) or of position
coordinates along the path. When Euler angles are used, a unit vector in the global X direction is rotated through the global
X-axis, Y-axis, and Z-axis by the specified angles to yield the arc orientation vector. The rotation angle, arc orientation
vectors or Euler angles can be further varied along the weld path using tables as a function of arc length or point coordinates.
For a point that is in between two given path points, Marc can automatically interpolate to find out the current
arc orientation.
Text Input
In many industrial applications, weld paths for robots are specified via text files. A generic text file capability is provided in
Marc in order to allow the specification of the weld path and the weld orientation. The weld path is specified through point
coordinates in the columns 1-3 of the text file. The weld path vector at point N is defined as the vector from point N to
point N+1. The arc orientation, defined in columns 4 - 6 of the text file for each point N, can be defined via point
coordinates, vector components or Euler angles. The entry in each column is a real number of width 10. The columns can
be in free or fixed format, with commas being used to separate the columns in the free format mode.
User Routine Input

The weld path can also be defined using the UWELDPATH user subroutine. The final position of the associated weld flux, the
weld vector at that point, and the orientation vector at that point have to be defined in the subroutine. Refer to Marc Volume
D: User Subroutines and Special Routines for more details.
In some cases, it is possible that the given/computed local y axis is not perpendicular to the local z axis. It is however
important for purpose of weld flux computation to have a set of mutually perpendicular local axes. Since weld flux
dimensions are typically provided in directions along and perpendicular to the weld path, the weld path vector (z axis) is
taken as fixed and an equivalent set of perpendicular x and y axes are computed from the given information.
The actual position of a weld heat source can be offset by a user specified amount from the associated weld path. The offsets
are specified in the local x and y directions. This offset allows the same weld path to be used for multiple weld heat sources
and is particularly useful for multi-pass welding where the weld heat source repeatedly moves in a similar pattern, but offset
from the original path by a small amount.
Weld Filler
The dynamic creation of filler elements and the generation of associated boundary conditions is defined by the WELD FILL
model definition option. Two methods are supported for the creation of filler elements:
 The quiet element method and
 The deactivated element method.
In the quiet element method, filler elements are initially used with scaled down material properties. The scale factor is 10-
5
by default. The coefficient of thermal expansion is set to zero and all other material properties (except yield stress, specific
heat, and thermal mass density) are scaled down. The scaling is applied to reference temperature material properties and
temperature dependencies are ignored. When temperature dependent material properties have been entered through tables,
the reference temperature properties are evaluated at the lowest temperature (i.e, the first table value). Weld fluxes and other
point/distributed thermal loads on the quiet elements are ignored. When the filler elements are physically created by the
moving heat source, the thermal properties are fully restored while the filler elements are still quiet on the mechanical side.
If the thermal activation time is zero, the mechanical properties are also fully restored in the next increment. If the thermal
activation time is nonzero, the filler elements remain mechanically quiet till the thermal activation time duration is passed.
During this period, all stresses and strains in the elements are reset to zero. This disjoint thermal and mechanical treatment
of the weld filler prevents the build up of mechanical strains and stresses during the temperature ramp up period. The quiet
element method allows the elements to move with the model. This is particularly useful for accommodating large
deformations during the welding process. However, the quiet element method is prone to ill-conditioning due to the large
discrepancy in material stiffnesses.
In the deactivated element method, filler elements are initially deactivated in the analysis and are not shown on the post
file. When the elements are physically created by the moving heat source, they are activated in the model and appear on
the post file. The filler elements are only thermally activated initially and remain inactive on the mechanical side. If the
thermal activation time is zero, the filler elements are activated on the mechanical side in the next increment. If the thermal
activation time is nonzero, the filler elements remain mechanically inactive until the thermal activation time is passed. This
disjoint thermal and mechanical activation of the weld filler prevents the build up of mechanical strains and stresses during
the temperature ramp up period. The deactivated element method does not suffer from any ill-conditioning problems, but
the interior nodes of the deactivated elements do not move with the rest of the model. This can cause distorted filler
elements when used in a large-deformation welding analysis.
CHAPTER 10 809
The initial temperatures for the deposited filler elements can be directly specified on the WELD FILL model definition option.
Alternately, heat input from the weld torch can be modeled as distributed fluxes acting on for the base metal and deposited
weld filler through the WELD FLUX model definition option. When the temperature option is used, special thermal boundary
conditions are dynamically specified on the filler elements as they are created. These melting point temperatures are directly
specified on the nodes of the created filler elements as long as the latter remain in the weld pool. Once the heat source
moves on, the nodal temperature boundary conditions are removed and the filler elements are allowed to cool. By default,
the melting point temperatures are applied instantaneously. This sudden boundary condition can cause convergence
difficulties, especially with the AUTO STEP time stepping algorithm which bases the time step on the allowable temperature
increase. A small ramp time (thermal activation time) can be optionally provided by the user to alleviate this problem. The
temperature of the filler elements is increased from the current value to the melting point value over the ramp time specified
by the user. If the ramp time is left at 0, the melting point temperature is applied instantaneously.
User can define filler bounding box in two ways:
1. Symmetric:
In this option, The identification of the filler elements to be created in any particular increment is given by the
following algorithm: Let X g i , Y g i , Z g i be the global position of the weld heat source origin at the beginning of
the increment. Let X g f , Y g f , Z g f be the global position of the weld heat source origin at the end of the increment.
The global coordinates of the final position and of the filler element nodal points are transformed to local
coordinates using the local welding coordinate system based on the weld path and arc orientation. Let X l f , Y l f ,
Z lf be the local coordinates of the final source position and X l n , Y l n , Z l n be the local coordinates of the nth
node of a filler element. A bounding box in the local coordinate system is then used to determine if the nodal point
fallsZwithin
+ Z the<weld
0 pool. Let
Z lnthe
– Zdimensions
lf – Z bf < 0
of the bounding box be defined as X b , Y b , Z b f and Z b r , where
ln br
X b is the dimension in the weld width direction, Y b is the dimension in the weld depth direction, Z b f is the
dimension in the forward weld path direction and Z b r is the dimension in the rear weld path direction. If not
specified by the user, X b is taken as 1.5 times the weld pool width, Y b is taken as 2 times the weld pool width,
Z bf is taken as the forward weld pool length and Z b r is taken as the rear weld pool length.
Nodal point n is said to be inside the weld pool if the following relations are satisfied:
X ln – X b  0 Y ln – Y b  0
Z ln + Z br  0 Z ln – Z lf – Z bf  0
2. Non-symmetric:
In this option user defines 4 points (X1b, Y1b, X2b, Y2b, X3b, Y3b, X4b, Y4b) in the weld local XY plane. Refer to
Figure 6-28 for a definition of the weld local XY plane. These points should be in a clockwise or anticlockwise
sequence. User also defines dimension in the forward weld path direction (Zbf) and the rear weld path direction
(Zbr). In this case following tests are performed to identify the filler elements to be created in any particular
increment. The global coordinates of the final position of the filler element nodes are transformed to local
coordinates using the local welding coordinate system based on the weld path and arc orientation. Let Xln, Yln and
Zln be the local coordinates of the nth node of a filler element.
Nodal point n is said to be inside the weld pool if the following relations are satisfied:
a. Point position Xln, Yln is inside box defined by four points X1b, Y1b, X2b, Y2b, X3b, Y3b, X4b, Y4b.
b.
If all nodes of a particular element are inside the weld pool, then the element is activated. It should be noted that in case
the weld heat source moves across multiple segments of the weld path in a given increment, a number of sub-steps are used
within the increment and the above algorithm is used for a linear segment in each substep.
Filler elements can be set up as a homogeneous mesh with the base structure or as separate contact bodies. In the former
case, thermal boundary conditions are automatically created at the filler-base interface. In the latter case, an appropriate
CHAPTER 10 811
contact heat transfer coefficient needs to be specified to ensure that the heat from the filler is transferred to the base
structure.
Additional items to facilitate welding analysis in Marc are:
Time-Stepping and Convergence Checking

Fixed stepping or adaptive stepping procedures can be used for welding problems. The fixed stepping scheme that should
be used is TRANSIENT NON AUTO. The adaptive stepping scheme that should be used is AUTO STEP.
For fixed stepping, the only temperature control that can be used is the error in temperature estimate. It is very important
that this is defined with welding problems because of the temperature dependent material properties and phase changes
that occur. For adaptive stepping, the temperature controls that can be used include the temperature error in estimate and
the allowable temperature change per increment. The default temperature change per increment is 20. Time-step cutback
is used in the AUTO STEP algorithm if the temperature controls are not satisfied. The TRANSIENT time stepping scheme is
obsolete and is not recommended for welding problems since time step cutback is not available with that scheme when the
time step is reduced. When the time step is reduced, time step cutback is necessary to ensure that the weld motion, filler
element activation, etc. work properly.
Adaptive Meshing
Two of the most useful adaptive meshing criteria for welding are:
 Temperature Gradient Based criterion and
 The Node in Box criterion.
The general Box criterion requires the UADAPBOX user subroutine. However, for welding, the box definition has been
automated in the program. The current location of the weld source and the weld pool dimensions are used to define the
box. Nodes that fall in the box and the associated elements are adapted. An unrefine capability is also provided so that when
the box moves away, the subdivided element mesh become coarse. It is important that the user provides weld pool
dimensions through the WELD FLUX model definition option for the box criterion to work properly. It is also important to
note that if the weld filler elements are identified as candidates for adaptive meshing, a suitable upper bound for the number
of weld filler elements should be provided on the WELDING parameter.
Material Properties
Nonlinear temperature dependence for mechanical and thermal material properties is automatically taken into account by
the program. Latent heat of solidification can be modeled by specifying the solidus/liquidus temperatures or by varying the
specific heat as a highly nonlinear function of temperature. An appropriate temperature error in estimate control should be
used to ensure that the material properties at the current temperatures are used. Phase transformations are supported
through the MICROSTRUCTURE model definition option.
Parallel Computation
The welding capability is fully supported in DDM. Each data file should contain the complete weld path. When nodes are
used to define the weld path/weld orientation, this may require nodes belonging to other domains to be written out to the
domain file. Also, if a weld flux belonging to domain A refers to a weld filler F, the elements of which are in domain B, then
weld filler F with an empty set of elements needs to be available in domain A.
Radiation
There are six approaches to solve radiation problems in Marc with different levels of sophistication. They include:
1. Viewfactor calculation by direct adaptive integral method.
2. View factor calculation by Monte Carlo method.
3. Viewfactor calculation by Pixel Based Modified Hemi-cube method.
4. Radiation to Space using the FILMS model definition option.
5. Radiation to Space using any of the CONTACT or THERMAL CONTACT options.
6. Radiation into the body using the QVECT option.
There are several aspects in performing a radiation viewfactor calculation including:
1. Defining the edges or faces involved in the viewfactor calculation and determining if the region (cavity) is open or
closed.
2. Calculation of the viewfactors.
3. If the region is open defining the environment temperature. This temperature may be constant or varying with
time.
4. Definition of the surface emissivity and absorptivity, including temperature dependence and frequency dependence
(spectral behavior). By default, the absorptivity is equal to the emissivity.
5. Redefinition of the viewfactors due to large deformation or other phenomena.
6. Redefinition of the viewfactors due to either local or global adaptive meshing.
7. Radiation between surfaces results in increasing the size of the operator (stiffness) matrix, which results in greater
memory requirements and increased computational times.
All of these aspects are discussed in the following paragraphs.
In the radiation calculations in Marc, there are several assumptions made:
 Each surface is a diffuse emitter and reflector; i.e., the thermal behavior is independent of the orientation of the
radiation.
 Each surface is black; i.e., is a perfect absorber for all incident radiation.
 The surfaces are isothermal.
The third assumption requires either that an “adaptive” procedure is used to insure accuracy or that the finite elements are
sufficiently small for each surface to be assumed to be isothermal.
Using modern mesh generation techniques, there is a tendency to over-refine the finite element mesh, so the need for these
adaptive techniques may be less significant.
In theory, the viewfactors would form a symmetric matrix, the size of which is dependent upon the number of radiating
surfaces. If one has a closed cavity, in theory, the summation of all viewfactors emitting from a surface should equal one. If
desired, the numerically evaluated viewfactors can be scaled such that the sum is one.
Computational Approaches
There are several methods available for the viewfactor calculation in Marc, namely Direct Adaptive Integration, Monte
Carlo, and Pixel Based Modified Hemi-Cube Method. The choice is made on the RADIATION parameter. While this is the
CHAPTER 10 813
historical order that they have been implemented, the Hemi-Cube method runs fastest, gives the most accurate viewfactors,
and is the most flexible, and hence, is discussed first.
1. Defining the Radiating Surfaces
Depending upon which technique is used for calculating the viewfactors, this can be summarized as follows:
a. Pixel Based Modified Hemi-Cube
The CAVITY DEFINITION option is used to define the edges (2-D) or faces (3-D). Using Mentat, this is in the
MODELING TOOLS CAVITIES menu. Shells may be included in the model, in which case, one needs to
indicate if it is the top or bottom or both surfaces. Symmetry surfaces may be defined. If physically there are
multiple cavities in the problem, they should be placed in separate cavities to reduce the computational time.
b. Monte Carlo
The definition of the edges (2-D) or faces (3-D) is done via Mentat or Patran; see BOUNDARY
CONDITIONS>THERMAL>EDGE RADIATION OR FACE RADIATION menu. The calculation of the
viewfactors is also performed in Mentat or Patran via the BOUNDARY
CONDITIONS>THERMAL>COMPUTE RADIATION menu.
c. Direct Integration
In Marc, this is only available for axisymmetric solid elements. The geometry is defined using the RADIATING
CAVITY option by specifying the node numbers around the cavity. It is not supported by Mentat.
2. Radiation Viewfactors
In many analyses, the radiative transfer of heat between surfaces plays a significant role. To properly model this
effect, it is necessary to compute the proportion of one surface which is visible from a second surface known as the
formfactor or viewfactor. The viewfactor, defined by a fourth order integral, presents many difficulties in its
computation. Primary among these is the large amount of computing power needed, especially when shadowing
effects are included.
The radiative flow of hear from surface 1 to surface 2 is given by:
4 4
q 12 = F 12  T 1 – T 2  (10-18)
in which, F 12 is the viewfactor and is calculated as:
1 cos  1 cos  2
F 12 = -------   -----------------------------
- dA 2 dA 1
A1 r
2
A1 A2
Pixel Based Modified Hemi-Cube Method

The foundation of radiation is based upon radiation being received from multiple directions and/or being emitted in
multiple directions. This leads to the concept of a hemisphere, where one considers an element being projected onto the
hemisphere as shown in Figure 10-29.
This geometric projection can be “easily” calculated for 2-D or axisymmetric, but is problematic in 3-D. To overcome this
problem of exact projections, one can consider dividing the hemisphere into pixels as shown in Figure 10-30, and the simple
“count” the pixels. This procedure is easier to implement in 3-D, though it is still expensive. As an alternative one can
approximate the hemisphere with a Hemi-Cube as shown in Figure 10-31.
Figure 10-29 Element Projected onto Hemisphere
Figure 10-30 Hemisphere Divided into Pixels
Figure 10-31 Hemisphere with Hemi-Cube
Using the Hemi-Cube method is less expensive, but is still difficult around the edges, and especially the corners in 3-D.
One way to reduce this problem is to change the dimensions of the box, such that there is a smaller possibility that leakage
CHAPTER 10 815
out of the sides is important and then simply neglect this. This is shown in Figure 10-32, where a Hemi-Box now replaces
the Hemi-Cube. To retain accuracy, one must select a good dimension of the box, and as the space is now biased, an equally
spaced pixel grid will lead to inaccuracies. To overcome this problem, a Hemi-Plane will be used with non-equally spaced
pixels, such that the weight of each pixel is the same. This is shown in Figure 10-33
Figure 10-32 Hemisphere with Hemi-Box
Figure 10-33 Hemisphere with Hemi-Plane
The basic algorithm for the viewfactor program is as follows:

a. Read in file created by the analysis program containing the geometry.
b. For axisymmetric geometries, radiating faces are swept over a user controlled angle.
c. In 3-D, subdivide quadrilateral faces into triangles.
d. Create a map of nonuniform pixels and associate a weight with them.
e. Convert coordinates to local system.

f. Loop over receiving surface, lines in 2-D, triangles in 3-D.
g. In 2-D, project line unto line; in 3-D, project triangle onto plane.
h. Identify pixels (number and location).
i. Sum the pixel weights.
j. Output the view factor for this surface.
When symmetry planes are present, the receiving faces are doubled for each symmetry plane. For cyclic symmetry, the
receiving faces are doubled n times.
This method does not guarantee a symmetric Radiation Exchange Matrix  A j  F i j  because all F i j are calculated
separately and independently based upon the above algorithm. There exists a possibility to make the radiation Exchange
Matrix symmetric. For large models, this may be time consuming. The method is based upon an iterative averaging and
column normalization, and for large problems, requires a large amount of memory or disk space.
Monte Carlo Method

In this method, the idea is to randomly emit rays from the surface in question. The percentage of these rays which hit
another surface is the formfactor between the surfaces. The Monte Carlo method computes N formfactors at one time,
providing linear scaling. In fact, the larger the number of surfaces, the faster the formfactors are computed compared to the
Direct Adaptive Integration and Adaptive Contour Integration. Hence, this method is adopted for the viewfactor
calculations.
Once the viewfactors are calculated by Mentat; the resulting output file containing the viewfactors can be read in with the
inclusion of the RADIATION parameter and the VIEW FACTOR model definition option.
Some of the features of the viewfactor calculation in Marc are:
a. You are not required to specify blocking elements. This is embedded into the algorithm completely and, hence,
done automatically. This is specially useful in three-dimensional analysis since, for complex geometries, it is
impractical to predict what surfaces are blocking other surfaces.
b. The cost of calculation is nearly linearly proportional to the number of elements which means that, for big
problems, the cost does not increase significantly.
c. The methodology in Marc guarantees that the sum of viewfactors is always 1.0. This is in contrast to the direct
integration approach which has the requirement of normalization to avoid artificial heat gain and loss, because
the sum of the viewfactors for any one surface is unlikely to add up to 1.0 due to errors in the computation of
individual viewfactors.
d. Shadowing effects (due to two surfaces being hidden from one another by other surfaces) can be modeled.
For a surface to participate in the computations, it must participate in the following operations:
a. Ray Emission: The surface should be able to randomly emit a ray from its surface. The origin of the ray should
be randomly distributed over the area of the surface (see Figure 10-34). The direction of the ray should be
distributed according to the cosine of the angle between the ray and the normal to the surface at the origin of
the ray, thereby emitting more rays normal to the surface than tangent to it.
CHAPTER 10 817
b. Ray Intersection: Given an origin, direction, and length of a ray, the surface should be able to determine whether
it is hit by that ray, and if hit, the length of the ray at the point of intersection.
Determination of viewfactors involves consideration of several desired properties like speed, accuracy, shadowing,
translucence, absorption, nonuniform emission, reciprocity, and efficiency. In light of the properties listed above, the
Monte Carlo approach is adapted in conjunction with the ray tracing and boxing algorithms.
Figure 10-35 depicts the formation of shadows which essentially involves the computation of incident light by tracing rays
from the light source to the point of incidence of the eye rays. These rays do not reflect or refract. Shadows are formed
when light source rays are obstructed, either partially or fully, from reaching an object. Refraction and reflection of light
from light sources is not computed.
Figure 10-34 Random Ray Emission in Monte Carlo Method
Point Light
Eye Ray
Light Source Rays
Shadow
Figure 10-35 Shadowing Effects in Ray Tracing

Finally, to compute the viewfactors, it is not necessary to check the intersection of incident rays with all objects under
consideration. Such a method would be prohibitively computationally expensive and preclude a large scale three
dimensional analysis. An effective technique for fast calculation of intersections is employed.
The method relies on sorting the objects before any intersections are computed. The information computed is used to
eliminate most of the intersection computations. The technique requires that each object to be sorted have a bounding box
which entirely encloses it.
The sorting relies on creation of a binary tree of bounding boxes. Thus, a bounding box is computed for all objects. The
objects are then sorted by the coordinate of the longest dimension of the bounding box. This list of objects is then divided
into two sets, each having an equal number of objects. This process is repeated recursively until each set contains no more
than a given maximum number of objects. This recursive sorting process is depicted in Figure 10-36.
Ray intersections are performed by searching the binary bounding box tree which involves determining whether the ray
hits the bounding box of all the objects. If the intersection with one bounding box is not found, then another bounding
box is considered. In the event an intersection does exist, each object in the node is intersected with the ray if the
intersection is a bottom node of the tree. Otherwise, the bounding boxes of both subtrees are intersected with the search
in the closest subtree conducted first. The process continues until all rays are exhausted.
CHAPTER 10 819
1 2
3 4 5
6 7 8
9
Figure 10-36 Fast Intersection Technique for Sorting Objects
Direct Adaptive Integration

This approach computes viewfactors by directly computing a fourth order integral between every pair of surfaces. This
means that there are N squared integrations to be performed, a quadratic scaling.
Besides being quadratic, to achieve reasonable accuracy, this approach requires a huge number computations for surfaces
which are close together; a situation which frequently occurs.
Using the direct integration approach, Marc calculates the viewfactors automatically for you in each cavity of an
axisymmetric body involving radiative heat transfer. This capability is only available for axisymmetric bodies.
You must subdivide the radiative boundary in this heat transfer problem into one or more unconnected cavities. For each
cavity, you define the outline of the cavity in terms of an ordered sequence of nodes. Usually, the nodes coincide with the
nodes of the finite element mesh. You can add extra nodes, provided you also give the appropriate boundary conditions.
The nodes must be given in counterclockwise order with respect to an axis orthogonal to the plane of the figure and
pointing to the viewer (see Figure 10-37). If the cavity is not closed, the program adds the last side by connecting the last
node with the first one. This side is treated as a black body as far as radiation is concerned; its temperature is taken as the
average between the temperatures of the adjacent nodes.
7
8
6
9
5
10 4
R
Visibility
11 V 3
Z 1 2
Figure 10-37 Radiating Cavity (Approach II: Valid for Axisymmetric Capability Only)
Marc internally computes the viewfactor between every side of the cavity and all other sides. The matrix with the
viewfactors can be stored into a file, and read in again during a subsequent analysis, thus avoiding a new computation.
During a transient heat transfer analysis, for every time step, the program estimates the temperature reached at the end of
the step. From the estimated temperature, the emissivity (temperature dependent) is computed. In addition, the radiating
heat fluxes are computed. The temperatures at the end of the step are computed by solving the finite element equations.
At every node, the difference between estimated and computed temperature is obtained. If the tolerance allowed by the
CONTROL model definition option is exceeded, iterations within the time step take place. Otherwise, the computation of the
step is concluded.
The cavity is defined by its boundary defined by a list of nodes ordered counterclockwise.
Insulated boundary condition (for example, symmetry boundaries) requires that the sum of the heat fluxes at a node be
zero. This requirement is satisfied automatically. Therefore, no input is required for this type of boundary condition.
As previously mentioned, in a heat transfer analysis of axisymmetric body involving radiative boundary conditions, Marc
automatically calculates viewfactors for radiation. A description of the viewfactor calculation follows.
CHAPTER 10 821
The amount of radiation exchanged between two surfaces will depend upon what fraction of the radiation from each
surface impinges the other surfaces. Referring to Figure 10-37, the radiation propagating from surface i to surface j will be:
cos  i cos  j A j
q ij = J i A i  ---------------------------------- = J i A i F i j (10-19)
 r2 
It is noted that the viewfactor, F i j , is solely geometrical in nature. From the definition of F i j , we see that:
A j F j i = A i F ij (10-20)
We are now ready to derive the heat transfer radiation equation. Steady-state (equilibrium) energy conservation requires:
qi = Ai  Ji – Gi  (10-21)
Two independent expressions for G i can be formed:
a. The incoming radiation on a surface must equal the radiation emitted by all other surfaces which strike this
surface,
N N
Ai Gi =  Jj Aj Fj i =  Jj Ai Fi j
j = 1 j = 1
or (10-22)
N
Gi =  Jj Fi j
j = 1
where N is the number of surfaces involved in the computations.

b. The other expression for G i is:
Ji – i En Ji – i En
G i = -----------------------i = -----------------------i (10-23)
i 1 – i
where,
En is the emissive power
 is the emissivity
 is the reflectance.
Substituting Equation (10-23) into Equation (10-21) and rearranging it gives:

1 – i
J i = E n – -------------- q i (10-24)
i Ai i
This expression for J i can be inserted into Equation (10-22). After regrouping terms, you get the governing
equation for gray body diffuse radiation problems
A i  ij –  1 –  j A i F i j  
----------------------------------------------------  q j  =  A i  i j – A i F i j   E n  (10-25)
Aj j  i
For the problem of black bodies, that is,  = 1, we have,

 
 qj  =  Ai i j – Ai Fi j   En  (10-26)
 i
This equation states the obvious; net radiation heat flow from a black surface is the difference between radiation
given off and received; that is, there is no reflection.
Aj
Ai
r j
i
Figure 10-38 Viewfactor Definition
The heat flux radiating from A i to A j is computed as:

A j cos  j
q ij = J i  A i cos  i   -------------------- (10-27)
 r 2 
where J i is the power radiating from A i , the first term within parentheses is the projection of A i normal to
the connecting line, and the second term is the solid angle under which A j is seen from the center of A i
defining the viewfactor:
cos  i cos  j
F ij = ---------------------------- A j
2 (10-28)
r
q ij = J i A i F i j
3. Environment Temperature
If the sum of the viewfactors from an emitting face is not equal to one, then there is a radiation contribution to the
environment. Hence, If the cavity is closed, there is no radiation to the environment and the environment
temperature is not used.
Including a PRINT-30 parameter, one will obtain the sum of the viewfactors from each emitting surface. For a closed,
cavity, comparing this to one will be a measure of the accuracy of the viewfactor calculation.
CHAPTER 10 823

The environment temperature may either be defined by creating an extra node and associating it with a cavity
on the CAVITY DEFINITION model definition option. The temperature of the node is defined using FIXED
TEMPERATURE model definition option. As an alternative one can specify the environment temperature on the
RAD-CAVITY model definition option.
b. Monte Carlo
A constant environment temperature is defined in Mentat or Patran. This information is passed into Marc in
the vfs file.
c. Direct Integration
The cavity is assumed to be closed; there is no radiation to the environment.
4. Definition of the Surface Emissivity
The emissivity may be entered on an element edge or face. Hence, a particular element such as a brick could
have different emissivities because of surface conditions. The emissivity could be temperature and frequency
dependent. This is defined using the EMISSIVITY model definition option and the TABLE model definition option
if required. Is is also possible to define the emissivity via the ISOTROPIC, ORTHOTROPIC, or other material options.
The emissivity of materials as a function of wavelength is discontinuous in nature due to the discrete nature of
particle physics.
Frequency dependent emissivity (spectral) is defined by three types of bands: zero value, narrow, and wide. The
evaluation is also dependent on whether the band is at low or high frequency as shown in Figure 10-39. The value
of emissivity at the endpoints of each band may be a function of temperature. Based upon the table input
procedure, it is possible to enter the emissivity as a function of temperature and wavelength. The emissivity is
assumed to have a linear variation within each band.
Figure 10-39 Band Frequency

Besides entering the emissivity, if wavelength dependent data is used then one also needs to enter Planck’s
second constant and the speed of light. This is done via the PARAMETERS model definition option. The unit of
wavelength used in the table should be consistent with the unit of length used to define the speed of light.
Wavelength dependent emissivity is based upon:
•Narrow bands - use average reciprocal wavelength
•Wide bands at short wavelength - use exponential expansion
•Wide band at long wavelength - use power series in reciprocal wavelength
b. Monte Carlo or Direct Integration
The emissivity is defined via the ISOTROPIC, ORTHOTROPIC, or other material options. Temperature dependent
emissivity is entered through the TEMPERATURE EFFECTS or TABLE model definition options.
5. Redefinition of Viewfactors due to Large Deformation
If the distance between the updated coordinates and the coordinates at which the last viewfactor calculation was
performed is greater than user-specified value, then the viewfactors will be re-evaluated. Note that the motion of
nodes on the cavity is only an approximate measure of the inaccuracies of the previously generated viewfactors. The
theoretically correct way would be to continuously re-evaluate the viewfactors, and then measure the difference
between the two sets of viewfactors. This, of course, would defeat the objective of minimizing calculations. The
critical distance to re-evaluate is specified on the RAD-CAVITY model definition option. This method is only available
if the LARGE STRAIN parameter is used. It is also possible to specify that the viewfactors should be re-evaluated based
upon increment numbers.
Viewfactors are only recalculated if the Pixel Based Modified Hemi-Cube method is used. In the current release,
this is only available if there is one cavity in the model.
6. Redefinition of Viewfactors due to Adaptive Meshing
If the geometry of the cavity has been defined using geometric entities (curves and surfaces) in the CAVITY DEFINITION
model definition option, then when local adaptive meshing (2-D or 3-D) or global adaptive meshing (2-D) occurs,
the viewfactors will be re-evaluated. This is based upon that after remeshing the new edges or faces are automatically
attached to the curves or surfaces. The program will determine the new list of element edges or faces in the cavity
and recalculate the viewfactors.
This capability is only available with the Pixel Based Modified Hemi-Cube method.
7. Computational Costs
Regardless of the method used to calculate the viewfactors, there are two consequences:
a. The number of viewfactors becomes large, growing quadratically with the number of radiating surfaces.
b. The subsequent calculation in the analysis program is dependent upon the number of viewfactors calculated.
The number of viewfactors is reduced because of shadowing effects, but depending upon the procedure used to
calculate the viewfactors, this may lead to higher computational costs in calculating them. Because self-shadowing
or third surface shadowing is likely, sparse storage techniques are beneficial in storing the view factors. As the data
structures required for the view factor calculations are different than what is required for traditional finite element
or finite difference calculations, the computation is often put into stand-alone programs.
CHAPTER 10 825
Using finite difference, contour integral, Gaussian integration or direct integration, the number of operations
increases quadratically with the number of radiating surfaces. All of these procedures (combined with adaptive
approaches or refined meshes) can produce accurate results, but at a clear computational cost. With the Monte
Carlo method, the computational cost increases linearly with the number of radiating surfaces and increases linearly
with the number of emitting rays. This clearly is beneficial for large models. The problem with the Monte Carlo
method is that a large number of rays may be necessary to achieve accurate results (for example, if thin cavities exist).
The advantage of the Pixel Based Modified Hemi-Cube method is that the computational costs are linearly
dependent on the number of radiating surfaces and linearly dependent upon the number of pixels. The accuracy is
dependent upon the number of pixels chosen, and the size/shape of the cube. A very large number of pixels has the
ramification that the amount of memory required is increased and the amount of calculations is increased.
The number of viewfactors increases the cost of the analysis calculation, because the viewfactor file is larger, and
because the bandwidth of the system increases. There are two ways to control the computational costs:
a. One can set a tolerance and any viewfactor less than this value will be ignored. This results in radiation being
neglected.
b. One can set a second tolerance and viewfactors below this number will be treated explicitly as opposed to
implicitly. This means that the radiation contribution will be on the right-hand side and not in the operator
matrix. This may result in more iterations, but lead to a significant reduction in the size of the operator matrix.
This leads to a smaller bandwidth and a reduction in the computational costs to solve the system of equations.
CHAPTER 10 825
Nonstructural and Coupled Procedure Library
Graphical Visualization
Using the Pixel Based Modified Hemi-Cube method, is possible to visualize the relative magnitude of the viewfactors
associated with an emitting surface. This will create a post file, jid_cxx_vfs.t19, where xx is the cavity ID. Each
increment will represent the viewfactors from one surface. Creating a contour plot with Mentat, one will observe a triangle
(2-D) or pyramid (3-D) indicating which face is emitting and the values of the viewfactors to the other faces.
Data, Data Flow, and Data Constraints

The geometric information is created in Marc and written to a file, jid_cxx_t18, where jid is the job ID
and xx is the cavity number.
The stand-alone program then reads this file and calculates the view factors. This is then written to a file,
jid_cxx.vfs.
The analysis program then reads this file and performs the thermal analysis.
If graphical visualization of the viewfactors is requested, this will be written to file, jid_cxx_vfs.t19.
b. Limitations
A radiating cavity must be fully contained within one domain if DDM is used.
There is currently a limitation of 99 cavities.
Conrad Gap
For the thermal contact gap element, in the gap open condition, two surface temperatures T a and T b at the centroid of
the surfaces of a thermal contact element are obtained by interpolation from the nodal temperatures. These two surface
temperatures are used for the computation of an equivalent conductivity for the radiation/convection link. The expression
of the equivalent thermal conductivity k 1 is:
2 2
k 1 = L  T ka + T kb   T k a + T k b  + H L (10-29)
where  is the emissivity,  is the Stefan-Boltzmann constant, L is the length of the element (distance between a and b),
T k a , T kb are absolute temperatures at a and b converted from T a and T b ; and H is the film coefficient. The film
coefficient can be a function of the temperature by referencing a table.
The equivalent thermal conductivity k 1 for the thermal contact element is assumed to be in the gap direction. The thermal
conductivities in other two local directions are all set to zero. A coordinate transformation from the local to the global
coordinate system allows the generation of the thermal conductivity matrix of the thermal contact element in the global
system for assembly.
Similarly, in the gap closed condition, tying constraints are automatically generated by the program for thermal contact
elements. The constraint equation for each pair of nodes can be expressed as:
TI + TJ if  T gap  T c l o se 
(10-30)
where,
1
T g ap = ---  T I + T J 
2
T I T J = nodal temperatures at nodes I and J
T close = gap closure temperature.
Channel
For the fluid channel element, the one-dimensional, steady-state, convective heat transfer in the fluid channel can be
expressed as:
T f
m· c -------- + h  T f – T s  = 0 (10-31)
s
T f  0  = T inlet
· is mass flow rate, c is specific heat, T is fluid temperature, T is solid temperature, s is streamline coordinate,
where m f s
 is circumference of channel, h is film coefficient, and T inlet is inlet temperature. The film coefficient may be a function
of the temperature, and the inlet mass flow rate and temperature may be a function of time.
Similarly, the conductive heat transfer in the solid region is governed by the following equation:
·
CT s + KT s = Q (10-32)
subjected to given initial condition and fixed temperature and/or flux boundary conditions. At the interface between the
fluid and solid, the heat flux estimated from convective heat transfer is
q = h  Ts – Tf  (10-33)
In Equation (10-32), C is the heat capacity matrix, K is the conductivity matrix and Q is the heat flux vector. Equations
(10-32) and (10-33) are coupled equations. The coupling is due to the unknown solid temperature T s appearing in Equation
(10-31) and unknown fluid temperature T f in Equation (10-33) for the solution of Equation (10-32).
The solutions for Equations (10-31) and (10-32) are obtained from the introduction of a backward difference for the
discretization of time variable in Equation (10-32) and of streamline distance in Equation (10-31). Let
· i i–1
T s =  T s – T s    t  (10-34)
we obtain:
1
 ----- i 1 i–1
C + K T s = Q i + ----- CT s (10-35)
 t  t
CHAPTER 10 827
where t = time-step in transient analysis. Similarly, let
dT j–1
--------f =  T f – T f    s  (10-36)
ds
we obtain:
j j–1
T f =  s + T f    1 + s  (10-37)
where s is the streamline increment,

j–1
 = h   m· c  ; and  = hT s   m· c  (10-38)
Output
Marc prints out both the nodal temperatures and the temperatures at the element centroid when the CENTROID parameter
is used, or at the integration points if the ALL POINTS parameter is invoked. You can also indicate on the HEAT parameter for
the program to print out the temperature gradients and the resulting nodal fluxes.
To create a file of element and nodal point temperatures, use the POST model definition option. This file can be used as
temperature input for performing a thermal stress analysis. This file is processed using the CHANGE STATE option in the
subsequent thermal stress analysis. This post file can also interface with Mentat or Patran plot temperature as a function of
time.
Heat Transfer with Convection

Marc has the capability to perform heat transfer with convection if the velocity field is known. The numerical solutions of
the convection-diffusion equation have been developed in recent years. The streamline-upwind Petrov-Galerkin (SUPG)
method has been implemented into the Marc heat transfer capability.
The elements which are available are described in Table 10-1.
Table 10-1 Heat Transfer Convection Elements

36, 65 2-, 3-node link
37, 39, 41, 69, 131 3-, 4-, 6-, 8-node planar
38, 40, 42, 70, 87, 88, 122, 132 3-, 4-, 8-node axisymmetric
43, 44, 71, 123, 133 8-, 20-node hexahedral
135. 133 4-, 10-node tetrahedral
To activate the convection contribution, use the HEAT parameter and set the fifth field to 2. Due to the nonsymmetric nature
of the convection term, the nonsymmetric solver is used automatically. Specify the nodal velocity vectors using the VELOCITY
option. To change velocity, use VELOCITY CHANGE. If nonuniform velocity vectors are required, the UVELOC user subroutine is
used. This capability can be used in conjunction with the Rigid-Plastic Flow section in Chapter 5 of this volume to perform
a coupled analysis, in which the velocity fields are obtained.
The general convection-diffusion equation is:
T
 * c  ------ + v  T =    T  + Q (10-39)
 t 
The perturbation weighting functions are introduced as:
h h
W = N +   ---------  v  N  +   --------- t  v  N 
2 v  4 v 
N is the standard interpolation function in Equation (10-1). The upwinding parameter,  , is the weighting used to
eliminate artificial diffusion of the solution; while the beta term,  , is to avoid numerical dispersion. v is the magnitude
of local velocity vectors. T is the temperature,  is the diffusion tensor. Q is the source term and t is the time increment.
The optimal choice for  and  are:
 = coth  Peclet  2  –  2  Peclet 
 = C  3 –  2  Peclet  *    C  (10-40)
where Peclet is the local Peclet number in the local element and C is the local Courant number:
Peclet = density  specific heat  characteristic length magnitude of the fluid velocity/conductivity
Peclet = *c*h* v  k
and
C = v *  t   h where  t  is the time increment.
The characteristic length h is defined in [Ref. 11] where C  1 is required for numerical stability. When C  1 , the  is
set to be zero and a large time step is recommended to avoid numerical dispersion.
Note: The interpolation function N is not the time-space functions defined in [Ref. 6], so that most Marc heat transfer
elements can be used. The convection contribution of heat transfer shell elements is limited due to the
definitions of the perturbation weighting function and the interpolation function.
CHAPTER 10 829
Molecular Diffusion
Marc has a capability to perform molecular diffusion analyses. This capability is activated by using the MOLECULAR
parameter. The basis equation governing the behavior is the Fick’s equation, including a convective term.
The unknown of this equation is the molecular concentration.
The capability is available for either steady-state or transient behavior. The capability is available for two- or three
dimensional analysis using the standard heat transfer continuum elements.
Marc computes and prints the following information at the element integration points: molar concentration, gradient of
molar concentration, and the molar flow rate. The nodal point data consists of the molar concentration and the equivalent
nodal molar flux.
The material properties, namely the diffusivity, should be entered on either the ISOTROPIC (with TABLE Input - Molecular
Diffusion), ORTHOTROPIC (with TABLE Input - Molecular Diffusion) model definition options. Dependency on molar concentration,
or the temperature should be defined by referring a table.
The boundary conditions associated with molecular diffusion analysis consist on prescribed molar concentrations defined
by the FIXED CONC (with TABLE Input - Molecular Diffusion) model definition option where there is always one degree of freedom
per node. For transient analyses, initial molar concentration should also be defined using the INIT CONC (with TABLE Input -
Molecular Diffusion) model definition option.
Specify nodal point molar fluxes in the POINT MOL (with TABLE Input - Molecular Diffusion) model definition option. Specify
distributed molar fluxes in the DIST MOL (with TABLE Input - Molecular Diffusion) model definition option. Specify film
coefficients in the MOL FILM (with TABLE Input - Molecular Diffusion) model definition option.
Diffusion analysis can be used in conjunction with the contact capabilities. Contact bodies are defined through the
THERMAL CONTACT with TABLES (2-D) option. Molecule transfer to the environment or molecule transfer from a rigid body can
be defined through the THERMAL CONTACT option. Molecule transfer between flexible bodies can be defined through the CON
INTERA model definition option.
This capability can be coupled with the heat transfer problem by using the HEAT parameter and/or the structural problem
by using the STRUCTURAL parameter.
Technical background
The molecular diffusion capability is based upon an implementation of the Fick’s equation upgraded with a convective
term.
The governing equation of the molecular diffusion problem is:
C
C +   D  C  ------
------ C
+  ------ = Q
t  x  x  x
with
C -is the molar concentration [mol.m-3] if SI units are used.

D(C) - is the diffusivit [m2.s-1] if SI units are used
is the Eulerian fluid velocity [m.s-1] if SI units are used
Let be the molar concentration C(x) within an element be interpolated from the nodal values C of the element through
the interpolation functions N(x),
C(x) = N(x) C
The governing equation of the molecular diffusion problem becomes
C + D   ,C  C = Q
------
t
With D being the diffusivity matrix.
The selection of the backward difference scheme for the discretization of the time variable yields the following expression:
1- + D   ,C  C n = Q n + ----
---- 1- + D   ,C  C n – 1
t t
This equation computes the nodal molar concentration for each time increment ∆t.
Initial Conditions
In a transient molecular diffusion analysis, Marc accepts nonuniform nodal molar concentration distribution as the initial
condition. Enter the initial condition through the INIT CONC model definition option or through the USINC user subroutine.
Initial conditions are not required in steady-state analysis, even though they can improve convergence when concentration-
dependent properties are included.
Boundary Conditions
There are two types of boundary conditions in transient/steady-state molecular diffusion analysis: prescribed nodal molar
concentration and nodal/element molar fluxes. These boundary conditions are entered directly through input or through
user subroutines.
Prescribe nodal molar concentration using the FIXED CONC (with TABLE Input - Molecular Diffusion) model definition option.
The POINT MOL (with TABLE Input - Molecular Diffusion) model definition option allows you to enter nodal molar fluxes. The
molar flux (flow rate) represents the product diffusivity times gradient of molar concentration: D C
The DIST MOL (with TABLE Input - Molecular Diffusion) model definition option allows you to enter distributed (surface or
volumetric) fluxes. They represent the product diffusivity times gradient of molar concentration: C.
The MOL FILM (with TABLE Input - Molecular Diffusion) model definition option allows you to enter molecule transfer coefficient
and the ambient molar concentration associated with the convective boundary conditions.
The VELOCITY (with TABLE Input - Convective Heat Transfer) model definition option allows you to enter the velocity.
CHAPTER 10 831
Diffusion
Marc has a capability to perform diffusion analyses using two approaches. One is to use the capability using the PORE
parameter which was developed for poro-elasticity/soil analysis, and the one described here which is activated using the
DIFFUSION parameter. The basic equations governing the behavior are the mass balance equation and Darcy’s law which are
supplemented by a compressibility model for the fluid/gas. Darcy’s law can be derived from an averaged momentum
balance for viscous flows at low Reynolds numbers in porous media.
The capability is available for either steady-state or transient behavior. The capability is available for two- or three-
dimensional analysis using the standard heat transfer continuum elements.
Marc computes and prints the following information at the element integration points: pressure, gradient of pressure, and
the mass flow rate. The nodal point data consists of the pressure and the equivalent nodal mass flux.
The material permeability should be entered on either the ISOTROPIC, ORTHOTROPIC, or ANISOTROPIC model definition
options. Dependency on the pressure, or the temperature should be defined by referencing a table. The porosity or the void
ratio should be entered using the INITIAL POROSITY or INITIAL VOID RATIO model definition options. In a pure diffusion analysis,
those values do not change with time. The fluid/gas data required is the fluid viscosity, the fluid mass density, and the fluid
bulk modulus. This data is also entered in the ISOTROPIC, ORTHOTROPIC, or ANISOTROPIC model definition options.
It should be noted that for a purely incompressible fluid, the rate term is not significant, and a steady-state solution is
immediately obtained.
The boundary conditions associated with diffusion analysis consist of prescribed pressures defined with the FIXED PRESSURE
model definition option where there is always only one degree of freedom per node. For transient analyses, initial pressure
should also be defined using the INITIAL PRESSURE model definition option.
Specify nodal point mass fluxes in the POINT MASS model definition option. Specify distributed mass fluxes in the DIST MASS
model definition option. Nonuniform or time dependent distributed mass fluxes may be defined by the FLUX user
subroutine or by referencing a table.
Diffusion analysis can be used in conjunction with the contact capabilities. Contact bodies are defined through the
THERMAL CONTACT option.
The diffusion capability is based upon an implementation of the mass balance equation for low Reynolds number flow of
a liquid or gaseous fluid through a rigid porous medium. An averaged momentum balance for such a flow results in Darcy’s
law relating the average flow velocity of the fluid to the pressure gradient. This allows the flow problem to be formulated
in terms of the unknown pressures alone, greatly simplifying the formulation and also greatly reducing the cost of the
solution. Usually, the fluid is assumed to be incompressible, but the formulation here is extended to include compressibility
by allowing the density to be a function of pressure.
The balance relations are derived from considering a fixed control volume of the rigid porous material, where the pores are
filled with the flowing fluid, so mass transport is taking place over the control volume boundaries. The total volume of a
volume element dV is the sum of the volume taken by the pores filled with fluid dV g and the remaining volume of the
solid material dV s :
dV = dV g + dV s
The ratio of the pore volume to the total volume defines the porosity  , which is easily seen to be bounded by 0    1
and is a characteristic parameter of the material:
dV
 = ---------g
dV
The porosity may vary spatially, but it is constant over time since the solid material is assumed to be rigid. A related quantity
that is often used is the void ratio e , which is defined as the ratio of the pore volume to the solid volume and is an
unbounded quantity 0  e   :
dV
e = ---------g
dV s
The two are dimensionless quantities and they are related by:
e
 = ------------
1+e
Since the solid material is assumed to be completely rigid its mass inside the control volume is constant and does not
explicitly enter the mass balance equation. The formulation here is for a single fluid with density  g which is assumed to
flow through the volume with a percolation or filtration velocity û g . This percolation velocity is an averaged flow velocity,
averaged with respect to entire volume, and not the real flow velocity of the fluid, since the fluid only occupies the pores
and not the entire volume. The two velocities are related as:
û g = u g
where, u g denotes the real velocity, which is understood as an averaged velocity in the pores of the volume. This relation
easily follows from comparing the average momentum over an entire volume element dV with the average momentum
over its pore volume dV g :
 g û g dV =  g u g dV g =  g u g dV
The mass balance relation states that the rate of change of the fluid mass inside the control volume equals the rate of fluid
mass flow that takes place over the control volume boundary:
d
-----   g dV = –   g û g  n dA
dt V A
Here n represents the outward unit normal to the control volume boundary. Employing the divergence theorem this leads
to the following partial differential equation:
 g
 --------- + div   g û g  = 0
t
where,
CHAPTER 10 833
div is the divergence operator

The rate term in above differential equation represents the local time derivative of the fluid density.
Darcy’s law relates the percolation velocity to the pressure gradient as:

û g = – --- p

where
 is the absolute or intrinsic permeability [m2]

 is the dynamic viscosity [Ns/m2]
p is the pressure [N/m2]
 is the gradient operator
The units here are mentioned in S.I. units, but, of course, other units can be used as well. Darcy’s law can be viewed as an
averaged momentum balance for low Reynolds number flows through a porous medium. The formulation here is for an
isotropic material, but orthotropic or anisotropic behavior for the permeability is supported as well.
The fluid is assumed to be compressible and its density may be expressed as a function of pressure,  g =  g  p  , from
which the density change rate can be expressed in terms of the pressure change rate. It is customary to express the density
change rate in the following way:
  g p
--------g- = ------------
t Kt
where
K is the bulk modulus of the fluid, which is defined as: d g

K =  g  ---------
dp
Combining these relations results in the following partial differential equation that must be met in the interior of the
volume in terms of the unknown pressure:
 g p 
--------- ------ – div   g --- p = 0
K t   
The boundary conditions are either a prescribed pressure on some parts of the exterior boundary of the volume (denoted
by A p ) or a prescribed mass flow rate on other parts of the exterior boundary of the volume (denoted by A m ):
p = p* on A p : prescribed pressure [N/m2]

 p
 g --- ------ = m* on A m : prescribed mass flow rate [kg/sm2]
 n
Here, the asterisked terms denote the prescribed quantities. The union of the two regions represents the entire exterior
boundary, their intersection is empty. If nothing is specified on some part of the boundary, the natural boundary condition
with m * = 0 automatically applies meaning there is no mass exchange with the environment on this part of the
boundary.
The problem description is completed by specifying the initial pressure distribution:
p  x t = 0  = p 0  x 
˜ ˜
To arrive at the finite element equations of the problem the governing differential equation with the accompanying
boundary conditions is cast in a weak form:
 g p  
V --------- ------ – div   g --- p p dV – A m  m* –  g --- p  n p dA = 0
K t   
Here, the weighting functions are denoted by p . By the Galerkin method, they are interpolated spatially in the same way
as the unknown pressures p and they are referred to as the virtual pressures. The virtual pressure field is zero on those
boundary areas where the pressure is prescribed, so the area integral only extends over the boundary area where the mass
flow rate is prescribed. Using partial integration and employing the divergence theorem this can be brought to its final form:
 g p 
V --------
K t
- ------ p dV +   g --- p  p dV
V 
= A m m*p dA
This represents a nonlinear transient equation for the unknown pressure field. By choosing appropriate spatial
interpolations for this pressure field a nonlinear system of ordinary differential equations is obtained:
M g p· + K g p = m g
where,
p is the vector of unknown nodal pressures

Mg is the diffusion capacity matrix
Kg is the diffusion conductivity matrix
mg is the vector of external nodal mass fluxes
and the dot over the pressure vector denotes time differentiation. By further choosing appropriate temporal interpolations,
this becomes a system of nonlinear algebraic equations. As a time integration method, Marc uses the backward Euler
method which is the same method that is being used in heat transfer analysis. Similar controls for automatic time stepping,
updating of pressure dependent matrices, and for governing the iteration behavior within an increment are used as for heat
transfer analysis.
If the variation of fluid density with pressure is small, the density values in above finite element equations may be held at
their initial values and can be treated as constants. If the other material parameters ( K ,  , and  ) are also constant, the
system of equations becomes linear. If the fluid can be assumed to be incompressible ( K   ), the transient term vanishes
and a steady state solution is immediately obtained.
The fluid density (  g ) and the other fluid material parameters ( K ,  , and  ) can all be functions of pressure making the
system of equations nonlinear and these pressure dependences can be entered through tables. If the fluid density is entered
CHAPTER 10 835
through a table, the bulk modulus is evaluated by the functional relation given earlier through a numerical differentiation.
This is done irrespective of the bulk modulus value given in the input, unless the density variation in the table with pressure
is very small, in which case the bulk modulus from the input also gets evaluated.
Hydrodynamic Bearing
Marc has a hydrodynamic bearing analysis capability, which enables you to solve lubrication problems. This capability
makes it possible to model a broad range of practical bearing geometries and to calculate various bearing characteristics
such as load carrying capability, stiffness, and damping properties. It can also be used to analyze elasto-hydrodynamic
problems.
The lubricant flow in hydrodynamic bearing is governed by the Reynolds equation. The bearing analysis capability has been
implemented into Marc to determine the pressure distribution and mass flow in bearing systems.
Marc is capable of solving steady-state lubrication problems; the incremental procedure analyzes a sequence of different
lubricant film profiles. Marc also can be used to solve coupled elasto-hydrodynamic problems. This analysis requires a step-
by-step solution for both the lubrication and the stress problems using separate runs. Because the finite element meshes for
each problem are different, the program does not contain an automated coupling feature.
Only one-dimensional or two-dimensional lubricant flow needs to be modeled, since no pressure gradient exists across the
film height. This modeling is done with the available heat transfer elements. The library elements listed in Table 10-2 can
be used for this purpose.
Table 10-2 Hydrodynamic Bearing Elements

Element Description
36 2-node, three dimensional link
37 3-node, planar triangle
39 4-node, bilinear quadrilateral
41 8-node, planar biquadratic quadrilateral
65 3-node, three-dimensional link
69 8-node, biquadratic quadrilateral with reduced integration
121 4-node bilinear quadrilateral with reduced integration
131 6-node triangle
Marc computes and prints the following elemental quantities: lubricant thickness, pressure, pressure gradient components,
and mass flux components. Each of these is printed at the element integration point.
The nodal point data consists of pressures, equivalent nodal mass flux at fixed boundary points, or residuals at points where
no boundary conditions are applied. In addition, the program automatically integrates the calculated pressure distribution
over the entire region to obtain consistent equivalent nodal forces. This integration is only performed in regions where the
pressure exceeds the cavitation pressure.
The output includes the load carrying capacity (the total force on the bearing). This capacity is calculated by a vectorial
summation of the nodal reaction forces. In addition, the bearing moment components with respect to the origin of the
finite element mesh can be calculated and printed.
To activate the bearing analysis option, use the BEARING parameter. If the analysis requires modeling of flow restrictors, also
include the RESTRICTOR parameter.
The values of the viscosity, mass density, and cavitation pressure must be defined in the ISOTROPIC option. Specify
temperature-dependent viscosity values via TEMPERATURE EFFECTS or TABLE. If thermal effects are included, the STATE VARS
parameter is also required. In hydrodynamic bearing analyses, temperature is the second state variable. Pressure is the first
state variable.
The fluid thickness field can be strongly position-dependent. A flexible specification of the film profile is allowed by using
either the nodal thickness or elemental thickness option. Define nodal thickness values in the THICKNESS option. You may
also redefine the specified values via the UTHICK user subroutine. Elemental values of lubricant thicknesses can be defined in
the GEOMETRY option.
Marc also enables the treatment of grooves. Constant groove depth magnitudes can be specified in the GEOMETRY option.
If the groove depth is position-dependent, the contribution to the thickness field can be defined in the UGROOV user
subroutine.
The relative velocity of the moving surfaces is defined on a nodal basis in the VELOCITY option. In addition, you can redefine
the specified nodal velocity components in the UVELOC user subroutine.
Specify prescribed nodal pressure values in the FIXED PRESSURE option. Define restrictor type boundary conditions in the
RESTRICTOR option. To specify nonuniform restrictor coefficients, use the URESTR user subroutine.
Input nodal point mass fluxes using the POINT MASS option. Specify distributed mass fluxes in the DIST MASS option. If
nonuniform fluxes are necessary, apply this via the FLUX user subroutine. The TABLE option may also be used to define
spatially varying boundary conditions.
Define variations of the previously specified lubrication film thickness field through the THICKNS CHANGE option. The
program adds this variation to the current thickness values and solves the lubrication problem.
Activate the calculation of bearing characteristics (that is, damping and stiffness properties) through the DAMPING
COMPONENTS or STIFFNS COMPONENTS options. Marc evaluates these properties based on the specified change in film
thickness. This evaluation requires the formation of a new right-hand-side, together with a matrix back substitution. This
is performed within so-called subincrements. The bearing force components calculated within these subincrements
represent the bearing characteristics (that is, the change in load carrying capacity for the specified thickness change or
thickness rate). The previously specified total thickness is not updated within subincrements. The calculated bearing
characteristics are passed through to the UTHICK user subroutine. This allows you to define an incremental thickness change
as a function of the previously calculated damping and/or stiffness properties. This procedure can be applied when
analyzing the dynamic behavior of a bearing structure. Mechanical problems can often be represented by simple mass-
damper-spring systems if the bearing structure is nondeformable. A thickness increment can be derived based on the
current damping and stiffness properties by investigating the mechanical equilibrium at each point in time.
The bearing analysis capability deals with only steady-state solution and does not include the analysis of transient
lubrication phenomena. Note that the incrementation procedure is only meant to analyze a sequence of film profiles.
No nonlinearities are involved; each increment is solved in a single step without iteration.
CHAPTER 10 837
To calculate the reaction forces that act on the bearing structure, Marc requires information about the spatial orientation
of the lubricant. This information is not contained in the finite element model because of the planar representation of the
lubricant. Therefore, it is necessary to define the direction cosines of the unit normal vector that is perpendicular to
the lubricant on a nodal basis in the UBEAR user subroutine. The resulting nodal reaction forces are printed.
Marc requires a step-by-step solution of both the lubrication problem and the stress problem in separate runs. The thickness
changes need to be defined within the lubrication analysis based upon the displacements calculated in the stress analysis.
The stress analysis post file and the UTHICK user subroutine can be used for this purpose. The tractions to be applied in the
stress analysis can be read from the bearing analysis post file in the FORCDT user subroutine.
The flow of a lubricant between two surfaces that move relative to each other is governed by the Reynolds equation
h 3   h  1
   ---------- p – --------------- – ---   hu  + M = 0 (10-41)
 12  t 2
where:
p is lubricant pressure
 is mass density
h is film thickness
 is viscosity
t is time
u is the relative velocity vector between moving surfaces
M is the mass flux per unit area added to the lubricant
The following assumptions are involved in the derivation of this equation:
 The lubricant is a Newtonian fluid; that is, the viscosity is constant.
 There is no pressure gradient across the film height.
 There is laminar flow.
 Inertial effects are negligible.
 The lubricant is incompressible; that is, mass density is constant.
 Thermal effects are absent.
By introducing the film constant
h 3
 = ---------- (10-42)
12
Equation (10-42) can be written as:
   p  + M r = 0 (10-43)
r
where M is the reduced mass flux given by:
  h  1
M r = M – --------------- – ---   hu  (10-44)
t 2
In case of a stationary bearing, the transient term in Equation (10-44) vanishes.
Three different kinds of boundary conditions can be distinguished for the lubrication problem: prescribed pressure on
boundary, prescribed mass flux normal to the boundary, and mass flux proportional to pressure.
Prescribed pressure on boundary is specified as:
p = p (10-45)
where p is the value of the prescribed pressure.

Prescribed mass flux normal to the boundary has the form
p 1
–  ------ = m n – --- hu n = m nr (10-46)
n 2
where m nr is the reduced inward mass flux. Here, n refers to the inward normal on the boundary, and m n and u n are the
inward components of total mass flux and relative velocity, respectively.
If a restrictor is used (as in hydrostatic bearings), the total mass flux is a linear function of the pressure on the boundary.
This condition is specified as
p 1
m n = –  ------ + --- hu n = c  p – p  (10-47)
n 2
or, written in a slightly different form:
p
–  ------ = c  p r – p  (10-48)
n
where, c is the restriction coefficient and
hu
p r = p – ------------n (10-49)
2c
is the reduced pressure.
The differential Equation (10-41), together with the boundary conditions (Equations (10-46), (10-47), and (10-49))
completely describe the lubrication problem. This is analogous to a heat conduction problem as shown in Table 10-3.
Table 10-3 Comparison of Lubrication and Heat Conduction

Lubrication Heat Conduction
Pressure p Temperature T
Film constant  Conductivity k
CHAPTER 10 839
Table 10-3 Comparison of Lubrication and Heat Conduction (continued)

Lubrication Heat Conduction
r
Reduced body mass flux M Body heat flux Q
Reduced boundary mass flux mn Boundary heat flux qn
Restriction coefficient c Film coefficient h
r
Reduced reference pressure p Reference temperature Tr
Marc has the capability to perform electrostatic analysis. This allows the program to evaluate the electric fields in a body
or media, where electrical charges are present. This can be solved for one-, two-, or three-dimensional fields. Semi-infinite
elements can be used to represent an infinite domain. The electrostatic problem is governed by the Poisson equation for a
scalar potential. This analysis is purely linear and has been implemented in Marc analogously to the steady state heat
transfer problem. The available elements are described in Table 10-4 below.
Table 10-4 Element Types for Electrostatic Analysis

37, 39, 131, 41 3-, 4-, 6-, 8-node planar
69 8-node reduced integration planar
101, 103 6-, 9-node semi-infinite planar
38, 40, 132, 42 3-, 4-, 6-, 8-node axisymmetric
70 8-node reduced integration axisymmetric
43, 44 8-, 20-node 3-dimensional brick
71 20-node reduced integration brick
105, 106 12-, 27-node semi-infinite brick
123 8-node reduced integration brick
135, 133 4-, 10-node tetrahedral
50, 85, 86 3-, 4-, 8-node shell
87, 88 2-, 3-node axisymmetric shell
137, 203 6-, 15-node pentrahedral
217, 219 5-, 13-node pyramid
Marc computes and prints the following quantities: electric potential field vector ( E ) and electric displacement vector ( D )
at the element integration points. The nodal point data consists of the potential  and the charge Q .
To activate the electrostatic option, use the ELECTRO parameter. The value of the isotropic permittivity constant is given in
the ISOTROPIC option, orthotropic constants can be specified using the ORTHOTROPIC option. Optionally, the UEPS user
subroutine can be used.
Specify nodal constraints using the FIXED POTENTIAL option. Input nodal charges using the POINT CHARGE option. Specify
distributed charges by using the DIST CHARGES option. If nonuniform charges are required, the FLUX user subroutine can be
used for distributed charges and the FORCDT user subroutine for point charges. The TABLE option may also be used to define
spatially varying boundary conditions.
The electrostatic capability deals with linear, steady-state problems only. The STEADY STATE option is used to begin the
analysis. The resultant quantities can be stored on the post file for processing with Marc Mentat.
The Maxwell equations to govern electrostatics are written in terms of the electric displacement vector D and the electric
field vector E such that:
D =  (10-50)
and
E = 0 (10-51)
where,  is a given volume charge density.

The constitutive law is typically given in a form as:
D = E (10-52)
where  is the dielectric constant.
Introducing the scalar potential  such that:

E = –  (10-53)
which satisfied the constraint Equation (10-51) exactly.
Denoting the virtual scalar potential by  , the variational formulation is
     dV =  dV +      – n dA (10-54)

V V 
The natural boundary condition is applied through the surface integral in terms of the normal electric displacement using:
–   n = D  n = D n (10-55)
CHAPTER 10 841
Consider a material interface  12 , separating two materials 1 and 2, and as  is continuous over the material interface,
the tangential electric field constraint is automatically satisfied.
n   E1 – E2  = 0 on  12 (10-56)
If charges are present on the interface, these are applied as distributed loads as follows:
n   D1 – D2  = s on  12 (10-57)
Using the usual finite element interpolation functions N and their derivatives  , we obtain
 = N  = N
K =  T dV (10-58)

V
F =  N T NdV +  N T ND n dA +  N T N s dA (10-59)
V   12
and finally
K = F (10-60)
Capacitance
A capacitor is a device that can store electrostatic charge and hence electrostatic potential energy. This energy can be
thought as the energy stored in the electric field created by the charge on the capacitor.Capacitance is always associated with
conductors that can store charge. A problem can contain a number of conductors each of which will store charge. This gives
rise to self and cross capacitance due to interaction between the conductors. This is expressed by the capacitance matrix.
All conductors in the problem are assumed to be bodies and defined as a set of elements. This is done using the THERMAL
CONTACT model definition option. No two conductors can touch each other or overlap. It is required to model the inside
of the conductor bodies. A sub-set of the conductor bodies can be considered for capacitance computation. This subset is
specified using the EMCAPAC history definition in the THERMAL CONTACT model definition option. Marc performs capacitance
computation only if the above definitions are specified in the input file.
The electrostatic analysis is used for computation of the capacitance matrix. The computation of the capacitance matrix
results from the usual electrostatic analysis with some constraints on boundary conditions:
 The boundary of the problem must satisfy far-field conditions and the electric potential on this boundary must be
zero.
 At problem boundaries where far field is not satisfied, the homogenous Neumann boundary condition must
apply.
 The electrostatic charge boundary condition must not be used on the problem boundary.
 Any boundary condition applied on an individual conductor body is ignored.
When dissimilar meshes are used to model the problem, then the conductors and dielectrics are distinct contact bodies. In
this case it is best that the conductors are the contact bodies which are being touched and the dielectrics are the touching
contact bodies.
Consider a single conducting body placed in an infinite homogenous dielectric medium. This conductor can be charged
by putting a charge on it or applying an electric potential. The charge redistributes on the conductor such that the electric
field inside the conductor is zero. This indicates that these charges reside on the outside surface of the conductor and there
is no charge inside the conductor. A normal electric field is present on the surface of the conductor. The value of the normal
electric field is equal to:

E n = --- (10-61)

where:
 is the surface charge density

 is the permittivity of the medium surrounding the conductor
The normal electric displacement is

D n = E n =  (10-62)
The total charge on the conductor is:
Q =
  dS (10-63)
S
where S defines the surface of the conductor.

The capacitance C is then defined as:
Q
C = ---- (10-64)
V
where V is the constant potential on the conductor.

CHAPTER 10 843
Capacitance Matrix
For multiple conductors the capacitance computation is expressed as a capacitance matrix. For n multiple conductors, the
capacitance matrix is:
Q1 C 11 C 12 . . C 1 n V1
Q2 C 21 C 22 . . C 2 n V2
. = . . . . . . (10-65)
. · ,
. . . . .
Qn C n1 C n2 . , Cn n Vn
where,
C ij is the self capacitance of conductor i , if i = j

C ij is the mutual capacitance between conductor i and j , if i  j
To find the values of C ij , the analysis has to be repeated n times.
To find C ij for any column j and i = 1 ,2 ,n , the procedure is as follows:
1. Specify V j = 1 volt and V 1 = V 2 =  = V i  j =  = V n = 0 .
2. For this run the values of Q i  i = 1 ,2n  are computed and the capacitance values of column j are calculated.
Marc has the capability to perform magnetostatic analysis. This allows Marc to calculate the magnetic field in a media
subjected to steady electrical currents. This can be solved for two- or three-dimensional fields. Semi-infinite elements can
be used to represent an infinite domain.
The magnetostatic analysis for two-dimensional analysis is solved using a scalar potential, while for three-dimensional
problems, a full vector potential is used. The magnetic permeability can be a function of the magnetic field, hence, creating
a nonlinear problem. Only steady-state analyses are performed. The available elements which are described in Table 10-5.
Table 10-5 Elements Types for Magnetostatic Analysis

37, 39, 131, 41 3-, 4-, 6-, 8-node planar
102, 104 6-, 9-node semi-infinite axisymmetric
109 8-node brick
110 12-node semi-infinite brick
181 4-node tetrahedron
183 2-node 3-D link
204 6-node pentahedral
206 20-node brick
Marc computes and prints magnetic induction ( B ), the magnetic field vector ( H ), and the vector potential at the element
integration points. The nodal point data consist of the potential ( A ) and the current ( J ).
To activate the magnetostatic option, use the MAGNETO parameter. The value of the isotropic permeability () is given on
the ISOTROPIC option; orthotropic constants can be specified using the ORTHOTROPIC option. Optionally, the UEPS user
subroutine can be used. Often, it is easier to specify (1/), which is also available through these options. A nonlinear
permeability can be defined using the B-H relation.
Specify nodal constraints using the FIXED POTENTIAL option. Input nodal currents using the POINT CURRENT option. Specify
distributed currents by using the DIST CURRENT option. Permanent magnets can be introduced by using the PERMANENT
option for isotropic materials, or by entering a remanence vector via the B-H RELATION option for orthotropic materials. In
addition, the FLUX user subroutine can be used for nonuniform distributed current and the FORCDT user subroutine for fixed
CHAPTER 10 845
nodal potentials or point current. For convenience line element 183 can be used to define an external loading. The current
on this element will be converted to point currents on the nodes, pointing in the direction of this line element. The TABLE
option may also be used to define spatially varying boundary conditions.
The magnetostatic capability is linear unless a nonlinear B-H relation is defined. In such problems, convergence is reached
when the residual current satisfies the tolerance defined in the CONTROL option. The STEADY STATE option is used to
begin the analysis. The resultant quantities can be stored on the post file for processing with Mentat.
The Maxwell equations for magnetostatics are written in terms of the magnetic flux density vector B such that:
H = J (10-66)
and
B = 0 (10-67)
where J is the current density vectors.
For magnetic materials, the following relation between B , H , and M , the magnetization vector, holds:
B = 0  H + M  (10-68)
with  0 being the permeability of vacuum. Denoting the magnetic susceptibility by  m and the permanent magnet vector
by M 0 , we have
M =  m H + M 0 (10-69)
which can be substituted into Equation (10-68) to yield:

B =  H + B r (10-70)
in which  is the permeability, given by:

 = 0  1 + m  (10-71)
and Br is the remanence, given by:
Br = 0 M0 (10-72)
Notice that  1 +  m  is usually called the relative permeability  r .

Br
H
Figure 10-40 Nonlinear B-H Relationship
For isotropic linear material,  m and M 0 are scalar constants. If the material is orthotropic,  m and M change into
tensors. For real ferromagnetic material,  m and  are never constant. Instead, they depend on the strength of the magnetic
field. Usually this type of material nonlinearity is characterized by a so-called magnetization curve or B-H relation
specifying the magnitude of (a component of ) B as a function of (a component of) H.
In Marc, the magnetization curve can be entered via the B-H RELATION option. For isotropic material, only one set of data
points, representing the magnitude of the magnetic induction, B , as a function of H , the magnitude H, needs to be
given. For orthotropic material, multiple curves are needed with each curve relating a component of B to the corresponding
component of H.
When the table driven input is used, different ways of entering a magnetization curve are available. Then the ISOTROPIC or
ORTHOTROPIC material option contains either the permeability, the inverse permeability, the H-B relation, or the B-H
relation. For ORTHOTROPIC, this can be different for each component. For the H-B relation and the B-H relation, a table has
to be given where for the H-B relation, B is the independent variable, and for the B-H relation, H is the independent
variable. Permeability and inverse permeability can also be controlled by a table where the independent variable can be H,
B, or any other variable. A table can be either a set of data points or a function.
From Equations (10-70) and (10-72), it can be seen that for orthotropic materials, a permanent magnetization or remanence
can be entered through the B-H RELATION option, by putting in a nonzero B value for H = 0. For isotropic material, this does
not work since the direction of the remanence vector is still indeterminate. Therefore, in the isotropic case, the only
possibility is to supply the magnetization vector through the PERMANENT option. Any offset of the B  H  -curve,
implying B  0 at  H = 0  is disregarded in this case. For orthotropic material, it is not allowed to use the
PERMANENT option. In this case, the magnetization can exclusively be specified through the B-H RELATION option or using the
table driven input.
It is emphasized that the magnetization curve specified in the B-H relation or H-B relation must be monotonic and
uniquely defined.
Introducing the vector potential A such that:
A = B (10-73)
which automatically satisfies the constraint, Equation (10-67), and we then have the final form:
CHAPTER 10 847
   –1    A  = J +    –1 Br (10-74)
The vector potential ( A ) is not uniquely defined by Equation (10-73). In 2-D magnetostatic simulations, this indeterminacy
is removed by the reduction of A to a scalar quantity. In 3-D situations, Marc uses the Coulomb gauge for this purpose:
A = 0 (10-75)
Denote the virtual potential by W ; then, the variational formulation is:

–1 –1
    W      A dV =  W  JdV +  W      B r dV +  W   H  n d (10-76)
V V V 
where n is the outward normal to V at the boundary  .

In the three-dimensional case, the Coulomb gauge, Equation (10-75), is enforced with a penalty formulation. The resulting
term added to the variational formulation, Equation (10-76) reads:
–1 –1
    W      A dV =  W  JdV +  W      B r dV +  W   H  n d
V V V 
(10-77)
+  r    W     A  dV
V
The default value used for r is:

–4 –1
r = 10  (10-78)
Using the usual finite element interpolation functions N , the discrete curl operator G , and the weighting function
W = N.
G = N  y for two-dimensional problems

N  x
N  y – N  z (10-79)
G = N  z – N  x for three-dimensional problems
N  x – N  y
Leads to the resulting system of algebraic equations:

Ku = F (10-80)
where,
T –1
K = G  G dV (10-81)
V
T T –1 T
F = N J dV +  G  B r dV +  N  H  n  d (10-82)
V V 
The Coulomb gauge is based upon the principle of conservation of electric charge which in its steady state form reads:
J = 0 (10-83)
In Marc, it is up to you to specify the current distribution. When doing so, it is recommended to ensure that this
distribution satisfies Equation (10-82). Otherwise, condition Equation (10-75) could be violated. As a consequence, the
results could become less reliable.
From the third term on the right-hand side of Equation (10-82), it becomes clear that the natural boundary condition in
this magnetostatic formulation is H  n , the tangential component of the magnetic field intensity. Consequently, if no
other condition is specified by you, by default a zero tangential magnetic field intensity at the boundaries is assumed.
Besides, when there are no currents present on a  12 material interface separating two materials 1 and 2, the tangential
magnetic field intensity is assumed to be continuous:
n   H 1 – H 2  = 0 on  12 (10-84)
With H 1 and H 2 , the magnetic field intensities in material 1 and 2, respectively. A discontinuous tangential magnetic
field intensity can be modeled by assigning an appropriate distributed “shear” current density to the interface. This (surface)
current density J is related to H 1 and H 2 by:
n   H 1 – H 2  = J on  12 (10-85)
Iterative Procedures
It can be challenging to find a solution when a nonlinear B-H relation is used to obtain the permeability. In general, the
permeability can change orders of magnitude when the material starts to saturate. Two iterative procedures are available to
find convergence in such cases. One method is the secant method. During the iterative procedure, the estimate of the
permeability is based on the previous trial solution of the magnetic field. The other method is based on the Newton-
Raphson method. Here, the slope of the curve is also used to obtain a new estimate. The Newton-Raphson method is
generally faster, but can fail to find convergence. Convergence testing can be done on residual currents or on change in
magnetic energy. The magnetic energy is computed as:
H 
 B  dH dV
W =     
V 0 
where B is the magnetic induction, H is the magnetic field, V is the total domain, and W is the total stored magnetic
energy. It is advised to use the energy method.
CHAPTER 10 849
Edge Effects at Interface Between Materials of Different Magnetic Permeabilities

At the interface between materials of different permeabilities, there is an abrupt change of permeability value. In absence
of a surface current at the interface, the following holds:
B n 1 = B n2 and H t1 = H t 2 (10-86)
where,
H is the magnetic field intensity.

B is the magnetic flux density.
n corresponds to the normal component
t corresponds to the tangential component at interface
1 and 2 corresponds to either side of interface
A surface sheet current can flow at the interface only if one of the materials is a perfect conductor (    ). Many
magnetic materials are conducting too but are not perfect conductors.
For any material we have:
B = H (10-87)
where,
 is the permeability of the material.

Hence,
B t1 B t2
-------- = -------
- (10-88)
1 2
Now,
B = A (10-89)
where,
A is the magnetic vector potential.

The conditions (equations 10-86, (10-88), and (10-89) are satisfied if the normal component of the magnetic vector potential
A is allowed to be discontinuous along the interface normal. The tangential component of magnetic vector potential A is
continuous.
For 2-D problems, only the scalar Z component of the magnetic vector potential A is considered. This component is always
tangent to any material interfaces. Hence, the edge effect is automatically satisfied for these problems. The edge effects are
required only for three-dimensional magnetostatic problems.
In order to obtain accurate results for the magnetic vector potential A, the edge effect requires the imposition of the
Coulomb’s gauge. The Coulomb’s gauge is imposed using a penalty factor. The value of the correct value of the penalty
factor depends on the following factors:
1. Type of element: tetrahedral or hexahedral
2. Order of the element: first or second order.
3. Finite element mesh density and on the uniformity or non uniformity of the mesh. A uniform mesh requires a lesser
value.
The value of penalty factor is quite stable only for a linear hexahedral element and this value is 1.0E-04. For all other
elements, this value lies in different ranges and it is difficult to know the correct value for any given problem. It is, hence,
advisable to use linear hexahedral element with a default penalty value of 1.0E-04.
This interface conditions is implemented in Marc through the THERMAL CONTACT option, where materials on either side of
the interface have to be defined as electromagnetic contact bodies. There is now a choice between glued and touching
contact (CONTACT TABLE or CONTACT TABLE with TABLES). With glued contact, all components of the magnetic vector potential
are tied and this is recommended when two dissimilar meshes containing the same material properties need to be
connected. With touching contact, only the tangential components of the magnetic vector potential are tied and thus
satisfying the above interface conditions. This is recommended at interfaces with different material properties. The glued
or touching contact interface condition should be given on the CONTACT TABLE option. Note that for convenience, Mentat
has a special tool to create a disconnected mesh from a uniform mesh which is called matching boundaries.
Electrical Windings and Coil Current Boundary Condition

In many electrical problems, a magnetic circuit is excited by electrical windings. A winding is a set of multi-turn coils. A
set of multi-turn coils end at two termination points called terminals. Each coil has the same cross section with respect to
shape and size. Usually, a coil has a rectangular or circular shape. Other shapes are rarely used. Each coil has a thin insulation
coating. The multi-turns are bound together by some insulation material. The set of multi-turn itself defines a cross section.
This is the coil cross section. A coil cross section is usually rectangular, sometimes circular. A coil cross section contains the
coil cross section and required insulation. The current in each coil is multiplied by the number of turns (this is called
Ampere-turns). It is then assumed that this Ampere-turns flows through the full coil cross-section.
Electrical coils are specified in two ways:
1. Geometric based coils
Geometric based coils are defined by the coils cross-section that includes inter-turns insulation. The coil cross-
section traces the coil path and its orientation is decided by the coil orientation.
a. The coil path is defined by a contiguous set of nodes or a contiguous set of polylines.
b. Orientation of the coil cross section.
c. Coil cross-section type: rectangular or circular.
d. Length and breadth for rectangular type and radius for circular type.
e. Number of turns of the coil.
f. Coil cross-section type: rectangular or circular.
g. Length and breadth for rectangular type and radius for circular type.
h. Direction of the current with respect to the defined path.
CHAPTER 10 851
i. Associated electromagnetic contact body identification number. This is used for calculation of inductance of
the coil.
2. Element based coils
The list of elements define the coil cross-section and the trace of the coil along the coil path. The list includes inter-
turns insulation. The coil cross-section traces the coil path and its orientation is decided by element geometry. Coil
orientation is not required.
a. The coil path is defined by a contiguous set of nodes or a contiguous set of polylines.
b. The plane perpendicular to the coil path, that intersects the elements in the list, defines the coil cross-section.
The coil cross-section can be arbitrary and is defined solely by the element list.
c. The coil cross-sectional area must be specified.
d. The list of elements traces the coil cross-section along the coil path and defines the coil orientation.
e. Number of turns of the coil.
f. Coil cross section type: rectangular or circular.
g. Length and breadth for rectangular type and radius for circular type.
h. Direction of the current with respect to the defined path.
i. Associated circuit identification number. This is used for calculation of inductance of the selected circuits.
Coil current boundary condition
This defines the actual current in each coil which requires the following data:
a. The coil to which this boundary condition applies.
b. The magnitude of current in each coil. For harmonic analysis, the coil current has a real/imaginary or
magnitude/phase component.
c. The geometric type and the list of these geometric entities.
The complete coil data is entered in the coils modeling tools. The coil current and the winding association is entered in the
COIL CURRENT boundary condition. To activate this option, use the EMWINDING and COIL CURRENT model definition
options.
Inductance Matrix Computation

An inductor is a device that can store magnetic potential energy. This energy can be thought as the energy stored in the
magnetic field created by the current flowing in the windings.
Self-inductance
The self inductance of a coil is defined as:
N
L = --------- (10-90)
I
where,
 is the flux linking the coil

I is the current in the coil that produces the flux
N is the number of coil turns
Mutual-inductance
Two coils, C 1 and C 2 , are of the same dimensions and are glued to each other, but separated by an insulator. Coils C 1
and C 2 have N 1 and N 2 turns, respectively. A current I 1 flows through C 1 , while the current I 2 in coil C 2 is zero. I 1
produces a magnetic field intensity (H) and flux density (B) and magnetic energy is stored in this field. The magnetic field
intensity (H) depends on the current I 1 , the coil dimensions and the material surrounding the coil. The magnetic flux
produced by coil C 1 and also passing through the coil C 2 is obtained from
1 = 2 =   B  dS (10-91)
coil area
The self inductance of each coil is given by:

N1 1
L 11 = -------------
-
I1
(10-92)
N2 2
L 22 = -------------
-
I2
Then mutual-inductance M 21 of coil C 2 with respect to coil C 1 is given by:
N2 2
M 21 = -------------
- (10-93)
I1
If the roles of the two coils are reversed, then:

N1 1
M 12 = -------------
- (10-94)
I2
where M 12 is the mutual-inductance of coil C 1 with respect to coil C 2
In this case, the self and mutual inductance are related by:
L 11 = M 21 = L 22 = M 12
Hence,
L 11 I L 22 I
- = ---2- and ---------
--------- - = ---1- (10-95)
M 12 I1 M 21 I2
CHAPTER 10 853
and since,
M 12 = M 21
(10-96)
M 12 = L 11 L 22
If the two coils are loosely coupled; i.e., there is leakage flux, then:
N1 1 N1 k1 1
L 11 = -------------
- and M 12 = -------------------
- (10-97)
I1 I2
L 11 I2 k1  1
---------
- = ---------
- (10-98)
M 12 k1 I2
Similarly, for the second coil,

L 22 I1 k2  1
---------
- = ---------
- (10-99)
M
21 k2 I2
and since M 12 = M 21
M 12 = 
 k L 11 L 22
(10-100)
2
k = k1 k2 0k1
k is called the coupling factor of the coils.
The Inductance Matrix

For ‘n’ coils C 1 , C 2 , C 3 ,  C n , each coil has self inductance and mutual inductance with the remaining coils. For
multiple conductors, it is better to express inductance as an inductance matrix. For ‘n’ multiple conductors, the inductance
matrix is:
1 L 11 L 12 . . L n I1
2 L 21 L 22 . . L 2 n I2
. = . . . . . . (10-101)
. . . . . . .
n L n1 L n2 . . Ln n In
where,
L ij is the self inductance of conductor ‘i’, if i = j

L ij is the mutual inductance between conductor ‘i’ and ‘j’, if i  j .
Note that the inductance matrix is symmetric and L i j = L j i . The terms L i j are also written as M i j . The magnetic stored
energy due to a currents in ‘n’ coils C 1 , C 2 , C 3 ,  C n is:
1 2 2 2
W t = ---  L 11 I 1 + L 22 I 1 + L 33 I 1 +  + L n n I n2 + 2M 12 I 1 I 2 + 2M 23 I 2 I 3 + 2M 34 I 3
2 (10-102)
+ 2M 13 I 1 I 3 +  
The total magnetic stored energy can also be written in terms of B and H as,
H 
W =      B  dH dV (10-103)
 
V 0 
where B is the magnetic induction, H is the magnetic field, V is the total domain, and W is the total stored magnetic
energy
For inductance calculations the two expressions for the stored magnetic energy are equivalent, and Marc uses this to
compute the inductance matrix.
The inductance matrix computation requires the windings feature. The current flowing through each inductor has to be
defined by the EMWINDING model definition and COIL CURRENT boundary condition. To achieve this, the inductors must be
defined as THERMAL CONTACT (electromagnetic contact) bodies. Each winding has to refer to one of these contact bodies
using the EMWINDING model definition option. The inductors for which the inductance matrix will be computed are selected
in the EMINDUC history definition option.
Differential Inductance Computation

The differential inductance corresponds to the change in inductance due to perturbation in the exciting currents. For the
set of n coils described above, the flux linkage of any coil C 1 is the surface integral of the magnetic flux density multiplied
by the number of turns N 1 . The change in inductance can occur with current variation for nonlinear magnetic materials.
Then, the inductance matrix can be described by the flux linkage as follows:
   =  L   I    I  +  0 
where,
 is the vector of flux linkage in coils.

I is the coil current vector.
  0  is the flux linkage vector when the coil currents are zero and accounts for permanent magnet.
The terminal coil voltage vector is defined by:
 dL I I

 V  = ------------- =  -----------  I  +  L  ---------- =   L d   I   ----------
t  dI  t t
CHAPTER 10 855
where  L d  is called the differential inductance matrix. The equation shows that the induced voltage is directly
proportional to the differential inductance matrix and the time derivative of the coil current.
Losses in Ferromagnetic Lamination Cores

Materials consisting of multiple layers of ferromagnetic material separated by isolators (laminations) are always used in all
power industry devices; i.e., alternators, synchronous generators, DC generators and motors, transformers, induction
motors, relays and switch gear. Eddy currents and hysteresis is produced due to time varying magnetic flux density (B) in
laminations. For steady state operations, the time variation is assumed sinusoidal.
These laminations are used to drastically reduce eddy current and hysteresis losses in ferromagnetic materials. These losses
reduce the efficiency of the device, cause undesirable heating and require additional devices like cooling etc. and it is
required to predict them accurately. Accurate prediction goes a long way in reducing manufacturing and prototyping costs.
Core Losses
Core loss in a magnetic material occurs when the material is subjected to a time varying magnetic flux. Traditionally, core
loss, P c , consists of two components: hysteresis loss, P h , and eddy current loss, P e . According to the Steinmetz equation,
measurement and calculation of core losses are normally made with sinusoidal flux density of varying magnitude and
frequency. These measurements and calculations are based on the standard coil and frequency modeled by a two term
function of the form
Pe = Ph + Pe
, (10-104)
= k h fB n + k e f 2 B 2
where f is the frequency of the external magnetic field, B is the flux density, k h , k e , and n are the coefficients, which
depend on the lamination material, thickness, conductivity, as well as other factors. However, this formula is only applicable
under the assumption that the maximum magnetic flux density of 1.0 Tesla is not exceeded and the hysteresis loop is under
the static situation, which is not practical in electrical machines. When the magnetic flux density is over 1.0 Tesla or the
field frequency becomes high, there is a big discrepancy between the calculation based on (6.97) and experimental results.
Core loss plays an increasingly important role both in the improvement of the quality of electrical steels at the production
stage and in the optimization of their operating conditions. According to the statistical loss theory, the core loss under
sinusoidal flux condition is given by:
Pe = Ph + Pe + Pa
2 (10-105)
= k h fB  a + b B + cB + k e f 2 B 2 + k a f 1.5 B 1.5
where P a is referred to as the excess loss and k a is related to the material thickness, cross-sectional area, conductivity and
a parameter which describes the material microstructure. The coefficients k h , k e , and k a vary slightly with the magnetic
flux density B and frequency f and this variation must be considered to obtain better accuracy. The coefficients a, b, and
c depend on the frequency and this must be considered in the loss calculations.
In the present procedure, the loss coefficients are evaluated from the original loss data supplied by manufacturers. The
coefficients are obtained mathematically using polynomial curves.
In practice, the lamination steel manufacturers do not present the loss coefficients mentioned above. They only provide the
loss curves and tables in watts per kilogram (W/kg) or watts per pound (W/lb.) versus the flux density or frequency to
indicate the combination of hysteresis loss, eddy loss, and excess loss. The classical estimation of core loss (10-104) is
systematically lower than the measured values. Ratios of two or more have been observed. Thus, for the analysis and
calculation of core loss, the accuracy of the computed loss coefficients is essential.
To compute lamination losses the materials in question are defined as THERMAL CONTACT (electromagnetic contact) bodies.
Each such contact body is a stack of lamination held together by bolts, clamps etc. The lamination loss curve is provided
by the manufacturer. The loss curve is actually a set of curves for different operating frequencies. There must be at least
three curves in this set in order to compute lamination losses in Marc. Each curve in this set shows the variation of the loss
with the magnetic induction B. To achieve good accuracy a sufficient number of data points must be extracted from the
curves. The data points must also cover all values of magnetic induction B for the problem in question. The data points
are entered by the TABLE model definition option. For each lamination body, the loss curve table should be given on the
THERMAL CONTACT option. It is given as the table reference of the scaling factor of the loss curve. Needless to say that this
option only works with the table driven input. The default multiplying factor for the loss curve is 1.
To activate this feature and to set the operating frequency use the EMLAMIN history definition option in the history
definition section.
For static or time-invariant electric or magnetic sources, the first two Maxwell’s equations become decoupled and result in
two types of analyses: Electrostatic and Magnetostatic. If there is a time-varying electric current or magnetic source, the
first two Maxwell’s equations are coupled. This coupling is partial when the frequency of time variation is typically less than
1 to 100 MHz.or the wavelength of electromagnetic wave propagation is larger than the typical dimension of the electrical
device. In general, a frequency less than 1 to 100 MHz is considered Low frequency and then the first two Maxwell’s
equation are partially coupled. The analysis in this Low frequency range is called magnetodynamics. In magnetodynamic
analyses, the displacement current is very small and is neglected. In essence, the dielectric permittivity has a low value
similar to that of air or most commonly occurring materials.
Marc has the capability to perform both transient (dynamic) and harmonic coupled magnetodynamic analysis. This allows
Marc to calculate the electrical and magnetic fields subjected to external excitation. This can be solved for both two- or
three-dimensional fields. A vector potential for the magnetic field is augmented with a scalar potential for the electrical
field. If a transient analysis is performed, the magnetic permeability can be a function of the magnetic field; hence, a
nonlinear problem. The elements available for magnetodynamic analysis are described in Table 10-6.
CHAPTER 10 857
Table 10-6 Element Types for Magnetodynamic

Analysis
111 4-node planar
112 4-node axisymmetric
113 8-node brick
228 3-node planar
230 4-node tetrahedral
231 6-node planar
234 8-node quadrilateral
235 8-node axisymmetric quadrilateral
236 20-node brick
Marc prints the magnetic flux density ( B ), the magnetic field vector ( H ), electric flux density ( D ), and the electrical field
intensity at the integration points. In a harmonic analysis, these have real and imaginary components. The nodal point data
consists of the vector potential A , the scalar potential V , the charge Q , and current I .
To activate the magnetodynamic option, use the EL-MA parameter. The values of the isotropic permittivity (  ), permeability
(  ), and conductivity (  ) are given in the ISOTROPIC option. Orthotropic constants can be specified using the ORTHOTROPIC
option. Optionally, the UEPS, UMU, and USIGMA user subroutines can be used. A nonlinear permeability can be defined using
the B-H relation.
Specify nodal constraints using the FIXED POTENTIAL option. Input nodal currents and charge using the POINT CURRENT
option. Specify distributed currents by using the DIST CURRENT option and distributed charges by using the DIST CHARGE
option. Nonuniform distributed currents and charges can also be specified by the FORCEM user subroutine.
The magnetodynamic capability is linear, unless a nonlinear B-H relation is defined. In such problems, convergence is
reached when the residual satisfies the tolerance defined in the CONTROL option. The transient capability is only available
with a fixed time step; use the DYNAMIC CHANGE option to activate this option. The resultant quantities can be stored on the
post file for processing with Marc Mentat.
In magnetodynamic analysis, you can enter the current and/or the charge. In a harmonic analysis, you can enter both the
real and imaginary components or the magnitude and the phase if the table driven input format is used.
Table 10-7 Input Options for Magnetodynamicc Analysis

Input Options
Nodal Charge POINT CHARGE POINT CHARGE
Distributed Current DIST CURRENT DIST CURRENT FORCEM
Distributed Charge DIST CHARGE DIST CHARGE FORCEM
Technical Formulation
Magnetodynamic analysis is based upon the well-known Maxwell’s equations. This has been implemented in Marc for both
transient and harmonic analyses.
Transient Formulation
The Maxwell’s equations are:
·
E+B = 0 (10-106)
·
  H – E –  E = 0 (10-107)
·
   E + E  = 0 (10-108)
B = 0 (10-109)
where the constitutive relations are:
D = E
B = 0  H + M 
J = E
and
E is the electric field intensity

D is the electric flux density
H is the magnetic field intensity
B is the magnetic flux density
J is the current density
M is the magnetization
CHAPTER 10 859
and
 is the permittivity
 is the permeability
 is the conductivity
0 is the permeability of free space.
·
Additionally, we have the conservation of charge:  +   J = 0 where  is the charge density. We assume that the
magnetization vector is given by:
M = H + M 0 (10-110)
where M 0 is the strength of the permanent magnet and  is the susceptibility. The magnetic field can be defined as:
B =  H + B r (10-111)
in which  is the permeability, given by:
 = 0  1 + m  (10-112)
and B r is the remanence, given by:
Br = 0 M0 (10-113)
Notice that  1 +  m  is usually called the relative permeability  r .
A vector magnetic potential A and a scalar potential V are introduced, such that
B = A (10-114)
·
E = –  V + A  (10-115)
Note that since only the curl of A is required, an arbitrary specification of the divergence can be made. The Coulomb
gauge is then introduced.
A = 0 (10-116)
This is implemented using a penalty condition. It is important to note that E depends not only on the scalar potential,
but also upon the vector potential. Hence, interpretation of V as the usual voltage can lead to erroneous results.
Substituting into Maxwell’s equations results in:
·· ·
    – 1    A –  0 M 0   +   V· + A  +   V + A  = 0 (10-117)
·· ·
–      V· + A  +   V + A   = 0 (10-118)
·
It has been assumed that  = 0 ; in that, the permittivity has a zero time derivative.
Two time stepping schemes have been implemented in Marc for transient magnetodynamics. The default scheme uses the
Newmark-Beta algorithm. This discretizes the second-order hyperbolic equations given in Equations (10-117) and (10-118)
and is preferred in mid-to-high frequency situations where the permittivity-based terms in these equations have a significant
influence in the transient response. For low-frequency systems where the permittivity-based terms can be insignificant, the
Newmark-Beta scheme is sometimes found to produce spurious oscillations in the solutions for the potential, E and D.
In such situations, a second scheme that uses the Backward-Euler algorithm to discretize first-order equations obtained by
dropping the permittivity-based terms in Equations (10-117) and (10-118) is preferred.
For the Newmark-Beta scheme, the general form is given by:
· 1 ··
A n + 1 = A n + tA n +  --- –  t 2 A n + t 2 A n + 1 (10-119)
2 
· · ·· ··
A n + 1 = A n +  1 –  tA n + tA n + 1 (10-120)
The particular form of the dynamic equations corresponding to the trapezoidal rule:
1 1
 = ---  = --- (10-121)
2 4
results in a symmetric formulation, which is unconditionally stable for linear systems.
For the Backward-Euler scheme, the governing equations are rewritten as:
·
    – 1    A –  0 M 0   +   V + A  = 0 (10-122)
·
–      V + A   = 0 (10-123)
·
The derivative term A in Equations (10-122) and (10-123) is discretized as:
n
· An + 1 – A
A n + 1 = --------------------------- (10-124)
t
In the current formulation, a fixed time step procedure must be used. The time step is defined through the DYNAMIC CHANGE
option.
Harmonic Formulation
In harmonic analysis, it is assumed that the excitation is a sinusoidal function, and the resultant also has a sinusoidal
behavior. This results in the solution of a complex system of equations. In this case, Maxwell’s equations become:
  E + iB = 0 (10-125)
  H – iE – E = 0 (10-126)
D– = 0 (10-127)
CHAPTER 10 861
B = 0 (10-128)
where  is the excitation angular frequency and i = –1 .

Additionally, we have the conservation of charge:
i +   J = 0 (10-129)
where all symbols are the same as in the discussion above regarding transient behavior. Again, a vector potential A and a
scalar potential V are introduced. In this case, these are complex potentials. Substituting into the Maxwell’s equations
results in:
    – 1    A –  o M o   + ̃  V + iA  = 0 (10-130)
–    ̃  V + iA   = 0 (10-131)
where ̃ =  + i
With a little manipulation, a symmetric complex formulation may be obtained. The excitation frequency is prescribed
using the HARMONIC option.
Note that the capability to extract the natural frequencies of a complex system by modal analysis does not exist in Marc.
The harmonic formulation is assumed to be linear; therefore, you should not include the B-H RELATION option.
Coils
Coils are the main tool to apply a current loading to a model. In magnetodynamic analyses two types of coils are available;
the thin-wire coil and the thick-wire coil.
Thin-Wire Coil
The first type is called a thin-wire coil. Thin wires are typically used in multi-turn coils, where a cross section of this type
of coil generally contains many turns of this wire. The thickness of the wire is small compared to the penetration depth of
the induced currents.
 = 1
-------------
f  (10-132)
Then the individual turns of the wire do not have to be modeled but based on the number of turns a uniform current
distribution inside the coil is computed. Typical examples are coils used in magnets, transformers, or in analog circuits.
This type of coil can be used for magnetostatic and magnetodynamic analyses and is explained in more detail in
Magnetostatic Analysis: Electrical Windings and Coil Current Boundary Condition.
Note that the thin-wire coils can be used in both circuit and non-circuit based analyses:
a. In circuit analysis, induced currents are not handled on an individual coil basis but are handled as part of the
circuit equations.
b. In a non-circuit analysis, the total current is taken to be the same as the applied current and no induced currents
are processed for the coils.
Thick-Wire Coil
The second type represents a conductor shaped into a coil and is called a thick-wire coil. The cross section of this conductor
consists of homogeneous material. This type of coil is commonly used in induction heating processes. It can be hollow
inside, so that it can be actively cooled. With thick-wire coils induced currents inside the coil are considered. This type of
coil can only be used in harmonic magnetodynamic analyses.
Figure 10-41 Thick Wire Coil
The following effects need to be addressed for an accurate simulation of the thick wire coil:
Induced currents occur in conductors due to an alternating magnetic field. The decay
from the surface inwards is exponential. The penetration depth is the distance from the
Skin effect surface when the induced current drops to 1/e. This distance depends on frequency and
material properties. For conductors larger than this penetration depth induced currents
are generated near the surface or the “skin” of the conductor.
The magnetic field is larger on the inside of a current loop compared to the field outside
Ring effect the current loop. This causes the currents to flow mostly on the inner side of the current
loop.
The influence that the magnetic field created by a current loop has on a second current
Proximity effect loop which is in its proximity. The current will be pushed away or attracted to the other
current loop depending on the direction.
The magnetic field and the current distribution in a current loop will redistribute when
Slot effect
a conductor is placed close to a thick-wire coil.
This coil type is available as a thick-wire coil with terminals and as a thick-wire coil without terminals.
CHAPTER 10 863
Thick-Wire Coil with Terminals

A thick-wire coil with terminals represents a coil which has a begin face and an end face. An electric potential difference
( V) is considered between these two faces. From Equation (10-115) this gradient will cause an electric current to flow. The
two end faces are represented by a begin terminal node and an end terminal node. On these nodes a TERM CURRENT or TERM
VOLTAGE (Magnetodynamic) boundary condition must be applied. With these terminal nodes, coils can be connected in series
or in parallel. Note that on least one terminal node a TERM VOLTAGE (Magnetodynamic) boundary condition should be
prescribed. This coil type is available in 3D and axi-symmetric analyses. In a 3D analysis the coil can have an arbitrary
shape, and the cross-sectional area does not have to be fixed throughout the coil. In an axi-symmetric analysis the coil is
circular and effectively closed. However, in order to apply a potential difference in circumferential direction, one can
consider the cross section being modeled as a fictitious cut in the coil, where the begin and end face are located. Hence, the
two terminal nodes must be connected to the same list of elements, which define the coil cross section. With this coil type
all effects listed above can be addressed.
Thick-Wire Coil without Terminals

The thick-wire coil without terminals represents a coil which is closed, has a circular shape, and fixed cross-sectional area.
This type of coil is always loaded with a current which must be provided by the COIL CURRENT boundary condition. An
iterative scheme is implemented in Marc for this type of coil. This scheme can address all the effects referred to above. The
scheme is based on two principles:
a. The total current in the coil is given by
I =  J t  dA
A (10-133)
where Jt is the current density passing a certain cross-section of the coil and pointing in the flow direction and
∫AdA the cross-section area.
b. At each integration point, the total current density Jt is given by:
J t = J i + J a = – j  A + J a
(10-134)
where Ja is the applied current density and Ji is the induced current density.
The applied current density should be the current applied to the finite element mesh, but the total current is the physical
current that is measured in the device. Note that the total current (I) is specified via the coil current boundary condition,
this is the only physical quantity that is available to the user.
The iterative scheme works as follows:
1. The total current (I) is converted to a current density using the cross-sectional area of the thick coil. The first
estimate of the applied current density is equated to this value and applied to the coils.
2. After the solution, the total current is recomputed for each coil using equations (10-133) and (10-134). This is
compared with the requested total current and a scale factor and phase shift for the applied current density is
computed.
Steps 1 and 2 are repeated until the correct total currents are reached for all the coils in the model. When multiple coils
exist in the model, a scale factor and phase shift is computed for each coil.
The different kinds of current density can be visualized with the following post codes:
147-149 Real components of the total current density
971-973 Real components of the induced current density
974-976 Real components of the applied or source current density
167-169 Imaginary components of the total current density
981-983 Components of the induced current density
984-986 Imaginary components of the applied or source current density
Note that the different currents have different phases. The phase of the total current density will be the same as the phase
of the total current I which is applied via the coil current boundary condition.
Symmetry and Cyclic Symmetry

Symmetry and anti-symmetry conditions can be used for this coil type. When symmetry is used half of the coil is modeled,
so only half of the current must be applied. When anti-symmetry is used (or cyclic symmetry) the anti-symmetry plane
must cut the coil perpendicular to the flow direction of the current. Then the total current must be applied.
Circuit Approach for Magnetostatics, Harmonic, and Transient Magnetodynamic Analysis

The circuit approach is implemented for magnetic devices and covers the magnetostatic and magnetodynamic analysis in
Marc. A magnetic device can be excited by one or more input power sources, and it may drive one or more electric loads.
A magnetic device consists of windings made from sets of conductor coils, some of them acting as source of input power,
and the rest acting as some electrical load. In addition, the device may contain massive conductors like the core in a
transformer or a work piece in an induction heating analysis.
Definitions
Coil A coil is a thin conductor wire of a small and uniform cross section. It is wound in a loop and may have
a single turn or multiple turns in a loop. A coil ends in two points which are used for further connection.
In circuit analysis, the current density in the coil is assumed to be uniform, and the skin effect in coil is
neglected.
Winding A winding is formed as a series connection of one or more coils. Each of coil can have a single turn or
multi-turns. A winding ends in two connection points.
Circuit A circuit is an electrical circuit and is formed by a series connection of a single winding and external
components such as resistors, inductors and capacitors. This series connection ends in two terminal
points. The excitation is applied at these two terminals and can be a terminal voltage or a terminal
current. The above combination of winding, external components, and the excitation source forms a loop
or circuit. The external components are connected in series in order to produce a designed simulation
from a circuit.
CHAPTER 10 865
Circuit The usual finite element modeling is applied to the magnetic device in which the modeling of the
Approach external component is excluded. These components are the components of the electric circuit and are
used for analysis of the electric circuit using the methodology of circuit analysis. The finite element
analysis computes the usual distributed quantities: magnetic vector potential A, magnetic induction B,
magnetic field intensity H, and induced currents in conductors. In circuit analysis, the terminal current
or voltage, the induced coil voltage, the voltage drop on external components in each circuit are also
computed. In each circuit, the induced voltage is the sum of induced voltages in all coils in that circuit.
The apparent circuit power is the complex power available in that circuit and is due to power input to
the circuit or power transferred from another circuit. The Ohmic power lost is the power dissipated in
all circuit resistances. Reactive power is the measure of power that gets transferred to another circuit. A
negative sign for this power indicates that this circuit gets power from another circuit.
 Apparent power = Magnitude of (Terminal voltage x complex conjugate of circuit current)
 Ohmic power loss = (Magnitude of circuit current)2 x total circuit resistance.
 Reactive power = Imaginary part of (Terminal voltage x complex conjugate of circuit current)
The magnetostatic circuit formulation combines the magnetostatic formulation in equations (10-76) and
(10-77) and the circuit equations. The magnetodynamic circuit formulation combines the
magnetodynamic formulation in equations (10-116) and (10-117) and the circuit equations.
For both, all the basic capabilities are retained and circuit analysis adds some extra capabilities.
Nomenclature used in Finite Element Formulation for the Circuit Approach

The notations used are given below:
 is the magnetic flux linkage

f is the frequency
 is the angular frequency = 2**f
R is the resistance matrix, and it is a diagonal matrix.
L is the inductance matrix, and it is a diagonal matrix.
K is the regular stiffness matrix
 KA  is the augmented stiffness matrix
F is the regular right-hand side matrix
D is the winding distribution matrix
 GD  is is the elemental induced voltage matrix
A is the usual global column vector for magnetic vector potential A
I is the column vector for circuit current I
e is the column vector for circuit terminal voltage e
Magnetostatic Circuit Analysis

The magnetostatic circuit formulation combines the magnetostatic formulation in equations (10-76) and (10-77) and the
circuit equations. The magnetodynamic circuit formulation combines the magnetodynamic formulation in equations
(10-116) and (10-117) and the circuit equations. For both, all the basic capabilities are retained and circuit analysis adds some
extra capabilities.
For both, the circuit analysis requires an extra control node to be attached to each circuit. This extra control node is not
part of the regular Finite element mesh.
The capabilities of magnetostatic circuit analysis is the same as capabilities of magnetostatic analysis as listed in Table 10-5.
The circuit analysis is activated by using the EMCIRCUIT model definition option. The EMCIRCUIT option requires all coils to
be specified in the EMWINDING option. The material properties and boundary condition specification remains the same.
Additionally, the TERM VOLTAGE (Magnetostatic) is an extra boundary condition that can be specified at the extra control node
attached to each circuit.
For magnetostatic analysis, the formulation is explained from equations (10-66) through (10-85). The circuit is not included
in the formulation, and it remains the same here. The circuit terminal voltage is finally converted to a current density J,
and magnetostatic analysis is carried out in accordance with equations (10-66) through (10-85). The current density J is
obtained as shown below:
Applied Voltage
Current I = ----------------------------------------- and
Resistance
Current I
Current density J = -----------------------------------------------------
Cross-sectional area
As before, the resulting algebraic equation is:
KU = F (10-135)
Magnetodynamic Circuit Analysis

For magnetodynamic circuit analysis, Marc prints quantities as listed in the section for magnetodynamic analysis in
Table 10-6. Additionally, it also prints the following quantities for each circuit:
1. Terminal current
2. Terminal voltage
3. Induced voltage
4. Resistance for harmonic analysis only
5. Total inductance for harmonic analysis only
The material properties and boundary condition specification remains the same. The circuit analysis is activated by using
the EMCIRCUIT model definition option. The EMCIRCUIT option requires all coils to be specified in the EMWINDING option.
Additionally, the TERM VOLTAGE (Magnetostatic) and TERM CURRENT are two extra boundary condition that can be specified at
the extra control node attached to each circuit. TERM VOLTAGE (Magnetostatic) is the voltage across the two circuit terminals,
while TERM CURRENT is the current injected into any one of the two circuit terminals. The terminal current is a kinematic
boundary condition.
CHAPTER 10 867
For magnetodynamic analysis, the formulation is explained from equations (10-106) through (10-131). Each circuit
connected to the magnetic device is considered in the formulation.
Consider one circuit, say a circuit number ‘k’. The circuit current is I k , and it flows through all coils in series in this circuit.
The number of turns = n , h is the unit vector along the winding path, and A r is the cross-sectional area of the coil. Let
v k be the circuit voltage at the two terminals of this circuit and externally connected resistance, inductance and capacitance
are excluded.
Then,
n
d     -----  A  h dV
 Ar 
vk V
= R k I k + -----------------------------------------------
- (10-136)
dt
where R k is the circuit resistance and can be easily found as below:
1  length of each coil turn * n 

R k = --- ------------------------------------------------------------------------------
 Cross-section of each coil turn
In equation (10-135), the right hand side vector  F  can be expressed by:
 F  =  D  . I k  (10-137)
where  D  depends on the vector h and the finite element approximation for the magnetic vector potential A .
Consider any time instant ‘t’ and let e k  t  be the applied terminal voltage to a circuit ‘k’.
Now, for circuit ‘k’, the terminal voltage is written as:

d  k  t   di k  t 
- +  R k   i k  t   +  L k  --------------
 e  t   = -------------------- (10-138)
dt dt
where
ik is the circuit current.

Rk is the total series resistance in the circuit and includes external resistance and coil resistances of all coils in this
circuit.
Lk is the externally applied series inductance in this circuit.
The first term corresponds to magnetic flux linking this winding and includes leakage inductance and mutual inductance
The right side contains three terms or three ‘voltage drops’. The three drops are explained below:
1. The first drop is due to combined inductance of this winding and mutual inductance between this winding and
another winding.
2. The second drop is the resistive or ohmic drop.
3. The third drop is due to externally applied inductance
The above is expressed in matrix form for all circuits present in the problem.
   =  GD   A  (10-139)
From equations (10-137), (10-138), and (10-139), we can write the following matrix equation:
 K   –D  A + 0 0 A = 0 (10-140)

0 R I  GD   L  I e
For harmonic magnetodynamic problems, we have
 K   –D   A  + j  0   0  A = 0 (10-141)

0 R I  GD   L  I e
For transient magnetodynamic problems, we have
K  –D  A = 0 + 0 0 A (10-142)

 GD   R + L  I t + t e t + t  GD   L  I t + t
where,
 KA  = K  –D 
 GD   R + L 
L
 L  = --------
t
 GD 
 GD  = ----------------
t
A  A  t + t   –  A  t  
------------ = ----------------------------------------------------
t t
Note that the matrix  KA  is not symmetric. It is called the augmented matrix.
All other matrices (  R  ,  L  ,  GD  ,  D  , and  K  , are symmetric.

f
d i j = ---i (10-143)
Ij
CHAPTER 10 869
 G  = g i j
n
gd i j =      ------- A  h dV (10-144)
  Ar  
V
where
n is the number of turns

Ar is the coil cross-sectional area
The augmented matrix  KA  is made symmetric by a suitable modification of the matrix  D  .
Equations (10-141) and (10-142) are solved for the magnetodynamic harmonic and transient analysis, respectively. The
solution yields the usual magnetic vector potential A and the circuit currents  I  . The boundary conditions for this
analysis can be the usual magnetodynamic loads or the circuit loads specified by the circuit currents  I  or the circuit
voltages  e  .
The circuit load  I  is the current injected into the circuit, while the load  e  is the applied voltage across the circuit
terminals. Each circuit is associated with an extra control node which is not part of the regular finite element mesh. The
circuit loads, that is, currents and voltages are applied to these control nodes.
Nonlinear Magnetic Materials in Harmonic Magentodynamic Analysis

Transient Magnetodynamic analysis handles nonlinear materials and arbitrary time varying loads, but it requires a large
number of time increments and is, thus, costly. When the voltage or current loads in a problem are sinusoidally time varying
(AC), it would be preferable to use Harmonic Magnetodynamic analysis, because it requires only one sub-increment for a
given frequency. This analysis assumes that all materials and loads in the problem are linear. The response is also sinusoidal.
When a given problem containing at least one nonlinear magnetic material excited by sinusoidal voltage or current loads,
a transient magnetodynamic analysis needs to be performed. This section shows how to treat nonlinear material in a
harmonic analysis.
In most magnetic problems, nonlinear magnetic materials form the core in devices like motors, generators, solenoids,
plunger, inductors etc. In Induction heating problems, nonlinear magnetic materials forms the workpiece. For nonlinear
magnetic materials, the relationship between Magnetic Induction (flux density) B and magnetic field intensity H is
nonlinear. Hence in steady state AC operation, time has to appear as an explicit variable in the Maxwell’s equations. The
solution is made easy by finding an equivalent material and time effective calculations made by using phasor quantities.
The equivalent material, described by an effective permeability value, gives accurate estimates of magnetic stored energy
and eddy current losses in the nonlinear materials. But, it will not give accurate estimates of time averaged values of
Magnetic vector potential A, Magnetic induction B, induced currents, and induced voltages. The harmonic content of
these time averaged values cannot be extracted, and, hence, this approach must be used only when the harmonic content
is not relevant. Additionally, materials with residual magnetism are not allowed in harmonic analysis.
The effective permeability,  e f f , is calculated at each integration point using the value of magnetic co-energy, w 1 , and
linearized energy, w 2 . These values are obtained by using the amplitude, H m ax , of magnetic field intensity H .
Magnetic co-energy, w 1 :
H m ax
w1 =  B  dH (10-145)
0
Linearized energy, w 2 :
1
w 2 = ---   ma x *  H ma x  2  (10-146)
2
 m a x is the value of permability when H = H ma x .
Averaged energy, wf :
 w1 + w2 
wf = ------------------------- (10-147)
2
Effective permeability,  e f f :
2 * wf
 e ff = --------------------
-
 H ma x  2
The solution is based on an iterative procedure, since both H m ax and  ef f are unknowns. In each cycle of the procedure,
the solution H ma x is obtained at each integration point and this value is used in equations (10-145) through (10-148), to
compute effective permeability,  e f f , at each integration point. This value of  ef f is then used in the next cycle. The
default convergence method for this procedure is the combination of a check on the residual currents and on the effective
magnetic permeability. The procedure is converged if the ratio of residual and reaction current and the change of the
effective magnetic permeability are both smaller than the specified tolerance. The effective magnetic permeability
convergence check is also provided. This check is activated when FEATURE,15300 is used.
This method gets activated, when the nonlinear material is expressed by a B-H curve. Presently, HB curve is not supported.
At present, isotropic and othrotropic materials are allowed. The TABLE option is used to specify the nonlinear permeability.
Inductance Matrix Computation in Harmonic Magnetodynamic Circuit Analysis

The inductance concept and theory is discussed in the Magnetostatic section, equations (10-90) through (10-103). In
harmonic magnetodynamic analysis, the circuit approach and the total magnetic co-energy is used to compute the
inductance matrix.
The Marc analysis is first performed for the specified circuit loads. The currents flowing in each circuit is termed as the
operating current. The total magnetic co-energy is computed for operating point currents in each circuit. The total
magnetic co-energy is:
1
w =  --- B  H dv
2 (10-148)
Volume
The operating currents in each circuit is given by:

CHAPTER 10 871
I1
I2
nc is the number of circuits (10-149)

In c
Next, the current in each circuit is then perturbed by a small amount and the Marc analysis is repeated for each
perturbation. A single circuit (i) and a pair of circuits (i,k) is considered. The total magnetic co-energy (w) in each case is
shown as:
1. Perturbation of current I i in a single circuit (i):
Current I i is perturbed twice:
I i = I i + I and I i = I i – I i
Total magnetic co-energies are

w  I i + I  and w  I i – I i 
The self-inductance L i i is given by:
 w  I i + I i  + w  I i – I i  – 2 * w  I i  
L ii = ----------------------------------------------------------------------------------------------
-
 I i  2
2. Perturbation of current I i in a pair of circuits (i,k):
Current I i is perturbed twice in each circuit:
I i = I i + I i and I i = I i – I i
and
I k = I k + I k and I k = I k – I k
Total magnetic co-energies are:

w  I i + I i ,I k + I k  and w  I i – I i ,I k + I k 
and
w  I i + I i ,I k – I k  and w  I i – I i ,I k – I k 
The mutual-inductance L i k is given by:
 w  I i + I i ,I k + I k  – w  I i – I i ,I k + I k  – w  I i + I i ,I k – I k  + w  I i – I i ,I k – I k  
L ik = ---------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------
-
 4 * I i * I k 
3. Steps (1) and (2) are repeated for all circuits 1, 2, …, nc and the inductance and differential inductance matrices
are computed.
Note: When the number of coil turns is greater than 1:

a. It is assumed that the coils are tightly packed.
b. If they are not tightly packed, then each coil turn must be modeled as a separate coil in coil definition.
CHAPTER 10 873
Limitations
The known limitations are:
1. In magnetodynamic circuit analysis, circuit coils cannot be massive conductors, and it is assumed that there is no
skin effect in circuit coils. This means that the cross-section of a single coil must be less than the skin depth for a
given conductor material and operation frequency.
2. Inductance matrix computation is not supported for transient magnetodynamic circuit analysis.
3. There are two types of coils: geometry based and element based. For 3-D problems, the following issues must be
noted:
a. For geometry based coil, there can be issues with the identification of elements that belong to a coil. Some
integration points inside coil geometry may be missed or an integration point outside coil geometry may be
selected. Both will produce inaccuracies. Inaccuracy depends on the ratio incorrect integration points divided
by total number of integration points in the coil.
b. For geometry based coils, a non-conductor element may get included in a coil. Marc writes a warning message
for this in the .out file.
The remedy is:
a. Using element based coil will avoid the above two issues.
b. When geometry based coil is used:
• The coil path and coil orientation path segments should be similar in size as the element edges of the coil. It is
best to have the coil path and the coil orientation path not in the same plane. When there is abrupt change in
angle of coil path and coil orientation, replace both by a proper fillet. This ensures smaller volumes of search
and reduces chances of incorrect integration points being chosen.
• Ensure that the elements lie fully inside or outside the coil geometry. All elements in coil geometry must have
a single conductor material.
The piezoelectric effect is the coupling of stress and electric field in a material. An electric field in the material causes the
material to strain and vice versa. Marc has a fully coupled implementation of piezoelectric analysis, thus simultaneously
solving for the nodal displacements and electric potential. The elements available for piezoelectric analysis are described in
Table 10-8. They can be used in static, transient dynamic, harmonic, and eigenvalue analysis. The analysis can be
geometrically nonlinear but is materially linear. The piezoelectric elements have their equivalent heat transfer elements, so
that they can also be used in a coupled thermal-piezoelectric analysis. Such a coupled analysis is weakly coupled, and solved
using a staggered approach. When piezoelectric elements are used in a contact analysis with a node of the piezoelectric
element touching a segment of another piezoelectric element, then a multipoint constraint relation is set up for the nodal
displacements as well as for the electric potential.
Table 10-8 Element Types for Piezoelectric

Analysis
160 4-node plane stress
161 4-node plane strain
163 8-node brick
242 20-node brick
Marc prints the stresses (  ), the strains (  ), the electric displacement ( D ) and the electric field intensity ( E ) at the
integration points. The nodal data points consists of the displacements u , forces f , the potential  , and the charge Q .
To perform a piezoelectric analysis, use the PIEZO parameter. The values of the isotropic, orthotropic and anisotropic
mechanical properties are given in the ISOTROPIC, ORTHOTROPIC, and ANISOTROPIC model definition option, respectively. The
electric constants, and the constants defining the coupling between the mechanical and electric part can be specified with
the PIEZOELECTRIC model definition option.
Specify nodal constraints using the FIXED DISP option or FIXED POTENTIAL option. Input nodal loads using the POINT LOAD
option or nodal charges using the POINT CHARGE option. Specify distributed loads with the DIST LOADS option and distributed
charges with the DIST CHARGE option. Fixed nodal displacements and potentials, or nodal forces and charges can also be
specified by the FORCDT user subroutine, nonuniform distributed loads can be specified by the FORCEM user subroutine, and
nonuniform distributed charges can be specified by the FLUX user subroutine. The TABLE option may also be used to specify
spatially and temporarily varying boundary conditions.
CHAPTER 10 875
The mechanical equilibrium equation for the piezoelectric effect is:
 : dV =  t  u dA +  f  u dV (10-150)
V  V
and the electrostatic equilibrium equation is (see also Equation (10-54):
 D  E dV =  D  n dA +   V  dV (10-151)
V  V
where
 is the stress tensor

 is the strain tensor
t is the traction at a point on the surface
u is the displacement
f is the body force per unit volume
D is the electric displacement vector
E is the electric field vector
 is the volume charge density

E = –  is the virtual electric field corresponding to the virtual potential  .
x
The constitutive equations to govern piezoelectricity are written for the mechanical behavior as:
E
 = L : – e  E (10-152)
and for the electrostatic behavior as:
T 
D = e : +   E (10-153)
where,
L is the elastic stiffness

e is the piezoelectric matrix (stress based)
 is the permittivity
The superscripts E and  represent coefficients measured at constant electric field, and constant strain, respectively. The
term with e gives the electro-mechanical coupling in the two consitutive Equations (10-152) and (10-153).
We approximate the displacements and electrical potential within a finite element as
u = NU
and
 = N
where, N contains the shape functions and U and  contain the nodal degrees of freedom. The body forces and charges,
as well as the distributed loads and distributed charges, are interpolated in a similar manner.
The strains and the electric field are given as:
 = Bu U (10-154)
and
E = B  (10-155)
where B u and B  contain the gradients of N .
Denoting the virtual displacement by U and the virtual potential by  the variational formulations can be obtained
by substituting Equations (10-152) and (10-154) into Equation (10-150):
T T T
   dV =  U B u  dV =  tU dA +  fU dV
V V  V
T T T T (10-156)
 U B u LB u U dV +  U B u eB   dV =  tU dA +  fU dV
V V  V
T T T
U K u u U + U K u   = U F u
and similarly by substituting Equations (10-153) and (10-155) into Equation (10-151):
T T T
 E D dV =  –  B  D dV =  Dn dA +   V  dV
V V  V
T T T T (10-157)
–  B  eB u U dV +   B  B   dV =  Dn dA +   V  dV
V V  V
T T T
–  K u U +  K   =   
The final set of equations in matrix form is then:
K uu K u u = Fu (10-158)
– K  u K   
Strain Based Piezoelectric Coupling

It is also possible to apply strain based coefficients for the piezoelectric coupling matrix. Then the consitutive equations are
for the mechanical behavior:
 = C: + d  E (10-159)
CHAPTER 10 877
and for the electrostatic behavior:
D = d: +   E (10-160)
where
C is the elastic compliance

d is the piezoelectric matrix (strain based)
 is the permittivity (strain based)
T
When d and  are given, Marc converts this into stress based properties, where e = L:d and  =  – e  d .
Acoustic Analysis
Marc has the capability to perform acoustic analysis in a rigid as well as a deformable cavity. This allows the program to
calculate the fundamental frequencies of the cavity, as well as the pressure distribution in the cavity. This can be solved for
two- or three-dimensional fields.
Rigid Cavity Acoustic Analysis

The acoustic problem with rigid reflecting boundaries is a purely linear problem analogous to dynamic analysis, but using
the heat transfer elements.
The elements which are available for acoustic analysis of a rigid cavity are described in Table 10-9.
Marc computes and prints the following quantities: pressure and pressure gradient at the integration points. The nodal
point data consists of the pressure and the source.
Table 10-9 Element Types for Rigid Cavity Acoustic Analysis

37, 39, 131, 41 3-, 4-, 6-, 8-node planar
135, 133 4-, 10-node tetrahedral
137, 203 6-, 15-node pentahedral
217, 219 5-, 13-node pyramid
The wave equation in an inviscid fluid can be expressed in terms of the pressure p as:
2
 p
--------- = c 2  2 p (10-161)
2
t
where c is the sonic velocity:
c = K (10-162)
where K is the bulk modulus and  is the density.

Equation (10-161) can be rewritten as:
2
2  p
K p –  --------- = 0 (10-163)
2
t
Where the source terms are neglected, note that this is analogous to the dynamic equation of motion.
The modeling of rigid reflecting boundaries can be done as follows.
Mathematically, a reflecting boundary is described by:
p
------ = 0 (10-164)
n
p
where, ------ is the pressure gradient normal to the reflecting surface.
n
This is analogous to an insulated boundary in heat transfer. Hence, a reflecting boundary can be modeled by a set of nodes
at the outer surface of the area which are not connected to another part of the medium. A reflecting plate in the middle of
an acoustic medium can be modeled by double nodes at the same location.
Note: Where there are no boundary conditions applied, there is a zero-valued eigenvalue, corresponding to a constant
pressure mode. Hence, you need to have a nonzero initial shift point.
To activate the acoustic option, use the ACOUSTIC parameter. The number of modes to be extracted should also be included
on this parameter. The bulk modulus and the density of the fluid are given in the ISOTROPIC option.
Specify nodal constraints using the FIXED PRESSURE option. Input nodal sources using the POINT SOURCE option. Specify
distributed sources using the DIST SOURCES option. If nonuniform sources are required, apply these via the FLUX user
subroutine or use the TABLE option.
To obtain the fundamental frequencies, use the MODAL SHAPE option after the END OPTION. The nodes can be used in a
transient analysis by invoking the DYNAMIC CHANGE option. The point and distributed sources could be a function of time.
CHAPTER 10 879
Fluid Mechanics
Marc has the capability to perform fluid flow analysis. Marc solves the Navier-Stokes equations in the fluid under the
restrictions that the fluid is considered to be nonreactive, incompressible, single phases, and laminar. The capabilities in
Marc can be applied to four different problems listed below:
 Fluid behavior only
 Fluid-thermal coupled behavior
 Fluid-solid coupled behavior
 Fluid-thermal-solid coupled behavior
Mass Conservation
The principle of mass conservation for a single-phase fluid can be expressed in differential form as:
D
-------- +   v = 0 (10-165)
Dt
where  is the mass density, v is the Eulerian fluid velocity, and t is the time.
Momentum Conservation
The principle of conservation of linear momentum results in:
v
  ------ + v  v = f +    (10-166)
 t 
where  is the stress tensor and f is the body force per unit mass.
In general, the stress tensor can be written as the sum of the hydrostatic stress and the deviatoric stress as.
tr    1
------------- = ---  ij  ij (10-167)
3 3
1 tr   
 d =  ij – ---  ij  ij =  i j – -------------  i j (10-168)
3 3
In fluid mechanics, the fluid pressure is introduced as the negative hydrostatic pressure:
tr
p = – -------- (10-169)
3
Energy Conservation
For incompressible fluids, the principle of conservation of thermal energy is expressed by:
T T q
C p  ------- + v i -------- = – --------i + H (10-170)
 t x i x i
where T is the temperature, C p is the specific heat at constant pressure, q i is the thermal flux, and H is the internal heat
generated.
Equation of State
The equation of state for a homogeneous, single-phase gas is:
mRT
pV = ------------- (10-171)
M
m is the mass of the fluid, M is the molecular weight of the gas, R is the universal gas constant, V is the volume that the
gas occupies.
The gas constant is often expressed as R = R  M for each substance. The equation of state is then written as:
P
 = -------- (10-172)
RT
In Marc, it is assumed that the fluid is incompressible. In such cases, the density is constant  =  o
Constitutive Relations
The shear strain rate tensor is defined:
1 v v
 ij = ---  --------i + --------j (10-173)
2  x j x i
For viscous incompressible fluids, it is assumed that there is an expression:

d · ·
 ij = 2 ij =  ij (10-174)
where  is the dynamic viscosity.

If the viscosity is not a function of the strain rate, the material is considered Newtonian. Viscous fluid flow is usually
characterized by the Reynolds number Re :
Lv
Re = ---------- (10-175)

where v is a typical velocity and L is the characteristic length.

Viscosity for common fluids like air and water is fairly constant over a broad range of temperatures. However, many
materials have viscosity which is strongly dependent on the shear strain rate. These materials include glass, concrete, oil,
paint, and food products. There are several non-Newtonian fluid models in Marc to describe the viscosity as
described below.
CHAPTER 10 881
Piecewise Non-Newtonian Flow

· · 2· · 1  2
 is the equivalent strain rate =  =  ---  i j  i j (10-176)
3 
·
 =     , where  is a piecewise linear function.
Bingham Fluid
A Bingham fluid behaves as a “rigid” fluid if the stress is below a certain level, labeled the yield stress, and behaves in a
nonlinear manner at higher stress levels.
d · · ·
 i j =  o  ij + g ij   if   g (10-177)
·
 i j = 0 if   g (10-178)
1---
· · 1· · 2
The effective viscosity is  =  o + g   and  =  ---  i j  i j
2 
The materials that can be simulated with this model are cement, slurries, and pastes.
Power Law Fluid

The fluid is represented as:
d · n – 1·
i j = o K     ij (10-179)
where K is a nondimensional constant.

This model is useful for simulating flow of rubber solutions, adhesives, and biological fluids.
Carreau Model
This model alleviates the difficulties associated with the power law model and accounts for the lower and upper limiting
viscosities for an extreme value of the equivalent shear strain rate.
2 ·2 n – 1  2
 =  +  o –    1 +    (10-180)
where  o and   are the viscosity at time equals zero and infinity, respectively.
The thermal flux is governed by the Fourier law.

T
q i = k ij -------- (10-181)
x j
Finite Element Formulation

In the procedures developed, the fluid-thermal coupling can be treated as weak or strong. In the weak formulation, the
solution of the fluid and thermal equilibrium equations are solved in a staggered manner. In the strongly coupled approach,
a simultaneous solution is obtained.
This section cannot cover all of the details of the finite element formulation. It does discuss the final form of the linearized
set of equations and some of the consequences. Beginning with degrees of freedom in a system – namely, v, p, and T for
the mixed method – we utilize the traditional finite element interpolation functions to relate the values within the element
to the nodal values.
Because the pressure stabilizing Petrov-Galerkin (PSPG) method is employed in Marc, equal order interpolation functions
can be used for the velocity and the pressure. The method of weighted residuals is used to solve the coupled Navier-Stokes
equations. Based upon the conservation laws of momentum (mass and energy), we obtain first order differential equations
in the form:
Mv· + A  v v + K  T v v – Cp + B  T T = F  t  (10-182)
T
C v = 0 (10-183)
NT· + D  v T + L  T T = Q  v t  (10-184)
where the first equation is obtained from momentum, the second from mass, and the third from energy conservation,
respectively. This is expressed in matrix form as:
M 0 0  v·  A + K B –C  v 
 
 F 
 
0 N 0  T·  + 0 D+L 0  T  =  Q  (10-185)
     
0 0 0  p·  –C
T
0 0  p   0 
if the penalty method is employed:

–1 T
p = M P C v (10-186)
resulting in:
M 0  v·  + A + K + CM P C
–1 T  v   F 
B
 ·    =   (10-187)
0 N  T  0 D+L  T   Q 
The introduction of the streamline upwinding technique (SUPG) developed by Brooks and Hughes is a major
improvement for the stability of the fluid equations. This procedure controls the velocity oscillations induced by the
advection terms. Effectively, this procedure adds artificial viscosity to the true viscosity. The second stabilization method,
PSPG, allows equal order velocity and pressure interpolation functions to be used without inducing oscillations in the
pressure field. The PSPG term is not included when using the penalty formulation. The details of these stabilization terms
CHAPTER 10 883
are not provided here, but note that the magnitude of the contributions are dependent upon the element size, viscosity, and
the time step. Including these contributions, the semi-discrete set of equations takes the form:
M + M  0 0  v·  A + K + K B – C + C   v   F + F 
     
M N 0  T· + K D+L 0  T  =  Q + Q  (10-188)
     
0 0  p·
T
0  –C 0 C  p   0 
The terms K d is included for SUPG and C e and G e for PSPG. The other terms are neglected, leading to:
A + K + K B –C
M 0 0  v·   v 
 
 F 
 
0 N 0  T·  + 0 D+L 0  T  =  Q  (10-189)
 ·  T    
0 0 0  p  –C 0 C  p   0 
These submatrices can be interpreted as follows:
A represents advection of momentum

D represents advection of energy
K represents diffusion of momentum or viscosity matrix
L represents diffusion of energy or conductance matrix
M represents mass
N represents heat capacitance
B represents buoyancy
K represents SPUG stabilization matrix
C represents gradient matrix
CT represents divergence matrix
Ce represents PSPG stabilization matrix
F represents externally applied forces
Q represents externally applied fluxes
Note that typically D , K , B , L , Q , and F are dependent upon the temperature. If a non-Newtonian fluid is used, A ,
K , D , and G are dependent upon the rate of strain. If the fluid is subjected to large motions, F can also depend on the
total displacement.
In terms of physical parameter, excluding geometry, observe:
A = A  o 
D = D   o C p  = D   o C p T 
K = K    = K   T 
L = L  K  = L  K T 
M = M  o 
N = N   o C p 
B = B   0 g B 
where,
o is the initial density

Cp is the specific heat
K is the thermal conductivity
g is the gravity
B is the coefficient of thermal expansion
Penalty Method
The penalty method is an alternative method to satisfy the incompressibility constraints. The objective is to add another
term to the operator matrix (viscosity matrix) such that incompressibility is satisfied. Effectively, K is replaced with K c
where:
K c = K + K p (10-190)
where  is a large number and K p the penalty matrix. K p can be written as:
–1 T
K p = CM p C (10-191)
where C and M p are functions of the geometry and shape functions only. The value of  is typically between 10 5 to 10 9 .
The penalty method should not be used for three-node planar elements, or four-node tetrahedral elements.

It is possible to simplify the equations governing fluid flow by assuming that the time derivatives of the velocity and
temperature are zero. This is achieved by using the STEADY STATE option. For thermally coupled, non-Newtonian flows, you
still have a highly nonlinear system and multiple iterations are required.
Transient Analysis
The equations discussed above are still differential equations because of the presence of time derivatives of velocity and
temperature. The conversion of the equations to fully discrete linear equations requires an assumption of the behavior
CHAPTER 10 885
during the increment. The approximations inevitably result in problems with accuracy, artificial damping, and/or stability
problems.
Marc uses the first order backward Euler procedure such that:
v· n + 1 =  v n + 1 – v n   t (10-192)
This is substituted into the equations to yield:
A + K + K  + M  t B –C    F – M  t  v n 
 v   
0 D + L + N  t 0  T  =  Q + Q – N  t  T  (10-193)
 n
T    
–C 0 C  p   0 
Marc allows either fixed time step or adaptive time step procedure.
The equations of fluid flow are highly nonlinear even for Newtonian behavior because of the inclusion of the advection
terms. The Newton-Raphson and direct substitution procedures are available to solve these problems.
Solid Analysis
Solid can be modeled in two ways:
 The first is to model them as a fluid with a very large viscosity. In such cases, an Eulerian procedure is used
throughout the model. The constitutive laws are limited to the fluid relationships.
 The second is to model them as true solids. In such cases, a Lagrangian or updated Lagrangian approach is used in
the solid.
The complete Marc constitutive routines are available for representing the solid behavior.
An important consideration is the interface between the fluid and solid. Large structural deformation or fluid motion are
not supported in the coupled Fluid-Solid capability. For small changes in the structural deformation, the nodes at the
interface are updated. The pressure is transmitted between the fluid and the solid.
Solution of Coupled Problems in Fluids

The coupled fluid-thermal problem can be solved using either a tightly coupled procedure as shown above or in a staggered
manner. Problems such as free convection are inherently coupled and are best solved using the tightly coupled procedure.
For fluid-thermal-structural problems, only the tightly coupled procedure for the fluid-thermal part is available.
The fluid-solid interaction problem is solved in a weakly coupled manner. Every attempt has been made to maximize the
computational efficiency. It should be noted that, while the staggered procedure can result in more global iterations, there
are several motivating factors for solving the system in this manner.
 In solving fluid problems, especially in three dimensions, very large systems of equations are obtained. Any
procedure that reduces the number of equations (for example, excluding the solid) is potentially beneficial.
 The fluid flow solution always requires the solution of a nonsymmetric system. In a weakly staggered analysis, the
structural problem can still be solved with a symmetric solver.
 When using an iterative solver, the solution of the strongly coupled system will result in an ill-conditioned system,
which results in poor convergence. This is because the terms associated with the fluid, thermal, and structural
operator are of several different orders of magnitude.
The problem can be divided into two regions (fluid and solid) to accommodate these requirements. In each region, a
different solver can be invoked; different nodal optimizers can be used; and a different memory allocation can be
performed. Often, the fluid uses an out-of-core nonsymmetric solver while the structure uses an in-core symmetric solver.
Degrees of Freedom
The degrees of freedom in a fluid analysis are the velocities or, for the mixed method, the velocity and the pressure. When
the pressure is not included explicitly as a degree of freedom, the incompressibility constraint is imposed using a penalty
approach.
If a strongly coupled fluid-thermal problem is solved, the degrees of freedom are either the velocities and temperatures or
the velocities, pressure, and temperature. For a three dimensional problem, the number of degrees of freedom could be 5.
The degrees of freedom associated with the solid are the conventional displacements or the displacements and temperatures.
Element Types
The fluid region can be represented using the conventional displacement elements in Marc. When using the mixed method,
the pressure has the same order and interpolation functions as the velocity. In a coupled fluid-thermal analysis, the
temperature also has the same order as the velocity. Hence, the element types available are:
Planar
Type
4-node isoparametric bilinear 11
4-node isoparametric bilinear reduced integration 115
6-node isoparametric triangle 125
8-node isoparametric biquadratic 27
8-node isoparametric biquadratic reduced integration 54
Axisymmetric
Type
4-node isoparametric bilinear 10
4-node isoparametric bilinear reduced integration 116
6-node isoparametric triangle 126
8-node isoparametric biquadratic 28
8-node isoparametric biquadratic reduced integration 55
CHAPTER 10 887
Three Dimensional
Type
8-node trilinear brick 7
8-node trilinear brick with reduced integration 117
20-node brick 21
20-node brick with reduced integration 57
The three-node triangular elements or the four-node tetrahedral element give poor results when used with the penalty
formulation.
The shape functions used are identical to those used for structural analyses and can be found in any finite element textbook.
All of the mesh generation capabilities in Mentat can be used to generate the fluid mesh. Furthermore, Mentat can be used
to visualize the results in a manner consistent with structural analyses.
Marc’s implementation of the Navier-Stokes equations utilizes the natural boundary condition.
At nodal points, you can prescribe the time dependent values of the velocity and temperature. This can be achieved by using
the FIXED VELOCITY, FIXED TEMPERATURE, VELOCITY CHANGE, and/or TEMP CHANGE options. The FORCDT user subroutine can be
used for time-dependent behavior or the TABLE option.
Gravity and centrifugal loads can also be applied.
For coupled fluid-solid interaction problems, the pressure of the fluid is automatically applied to the structure. The
resultant deformation of the structure is applied to the fluid boundaries. In the current release, only small deformations of
the solid are permitted.
The viscosity, mass density, conductivity, and specific head are defined through the ISOTROPIC option. The STRAIN RATE or
TABLE option is used to define non-Newtonian viscous behavior. The penalty parameter can be entered through the
PARAMETERS option.
Note: In fluid analysis, data for POINT LOAD and DIST LOADS should be prescribed as total rather than incremental
quantity (as used in mechanical analysis). Similarly, POINT FLUX and DIST FLUXES for heat transfer analysis are also
given as total quantity. This specification is to be used consistently for the fluid and/or heat transfer portion of
analysis in coupled fluid-solid, fluid-thermal, and fluid-thermal-solid.
Nonlocal Analyses
In a structural analysis that suffers local softening due to material damage, the analysis may encounter numerical difficulty
to obtain a converged solution because the system matrix becomes ill conditioned when the mesh is refined. In the worst
case, the simulation may end prematurely. In the real material behavior, the local softening will spread over a relatively small
but finite region. For ductile materials, where the damage evolution is determined by the accumulation of the equivalent
plastic strain, the damage variable should be evaluated using the average (nonlocal) value of the plastic strain instead of
using the local one to avoid this numerical problem.
The implicit gradient damage approach (see [Ref. 22]) is used to evaluate the nonlocal  ep s field of the equivalent plastic
strain from the local  eps .
2 2
 e ps – l   eps =  e ps (10-194)
With a natural boundary condition:
 e ps  n = 0 (10-195)
In this approach, users should choose a length scale parameter, l, which depends not only on the material micro structure
but also on the characteristic geometry of the model.
The above differential equations are solved using a standard finite element procedure (similar with that of the heat transfer
analysis) resulting in a system matrix as follows:
 C e ps – K eps E eps = Q e p s
where E eps , is the nodal nonlocal equivalent plastic strain vector and Q ep s is the nodal RHS vector due to the local
equivalent plastic strain.
Coupled Analyses
The definition of coupled systems includes the multiple domains and independent or dependent variables describing
different physical systems. In the situation with multiple domains, the solution for both domains is obtained
simultaneously. Similarly, the dependent variables cannot be condensed out of the equilibrium equations explicitly.
Coupled systems can be classified into two categories:
1. Interface variables coupling: In this class of problems, the coupling occurs through the interfaces of the domain.
The domains can be physically different (for example, fluid-solid interaction) or physically the same but with
different discretization (for example, mesh partition with explicit/implicit procedures in different domains).
Figure 10-42 Fluid-structure Interaction (Physically Different Domains)
2. Field variables coupling: In this, the domain can be the same or different. The coupling occurs through the
governing differential equations describing different physical phenomenon; for example, coupled
thermo-mechanical problems.
CHAPTER 10 889
Figure 10-43 Metal Extrusion with Plastic Flow Coupled with Thermal Field
Marc can solve the following types of coupled problems: fluid-solid, fluid-thermal, fluid-solid-thermal interaction,
piezoelectric, electrostatic-structural, thermo-electrical (Joule heating), thermo-mechanical, electrical-thermal-mechanical
(Joule Mechanical), fluid-soil (pore pressure), magnetodynamic, and magnetodynamic-thermal. The type of coupling is
summarized in Table 10-10.
Table 10-10 Summary of Coupled Procedures

Interaction Category Coupling
Thermal- Mechanical Field Weak
Fluid-Solid
added mass approach Interface Weak
general approach Interface Weak
Fluid-Thermal
channel approach Interface Weak
general approach Field Strong/Weak
Fluid-Thermal-Solid Field/Interface Strong/Weak
Piezoelectric Field Strong
Electrostatic-Structural Field Weak
Magnetostatic-Structural Field Weak
Thermal-Electric (Joule) Field Weak
Electrical-Thermal- Mechanical Field Weak
Magnetostatic-Thermal Field Weak
Fluid-Pore Field Strong
Magnetodynamic Field Strong
Magnetodynamic-Thermal Field Weak
Magnetodynamic-Thermal-Structural Field Weak
Table 10-10 Summary of Coupled Procedures

Interaction Category Coupling
Diffusion- Thermal Field Weak
Thermal – Molecular Diffusion Field Weak
Molecular Diffusion - Structural Field Weak
Thermal - Molecular Diffusion -
Field Weak
Structural
There are two approaches for solving fluid-solid interaction problems. In the first approach, the fluid is assumed to be
inviscid and incompressible. The effect of the fluid is to augment the mass matrix of a structure. Modal shapes can be
obtained for a fluid/solid system; the modal superposition procedure predicts the dynamic behavior of the coupled
fluid/solid system. This prediction is based on extracted modal shapes. This method is discussed later.
The general fluid-solid capability models nonlinear transient behavior and is discussed in the fluid mechanics section.
There are two approaches for solving fluid-thermal interaction problems. The first is for a fluid constricted to move through
thin areas and is implemented using the CHANNEL option. This is discussed earlier in this chapter. The second approach
solves the coupled Navier-Stokes equations and is discussed in the fluid mechanics section.
The fluid-thermal-mechanical capability is described in more detail in the fluid mechanics section.
In the coupled thermo-electrical problem, the coupling takes place through the temperature-dependent electrical
conductivity in the electrical problem and the internal heat generation caused by electrical flow in the thermal problem.
The program solves for the voltage and temperature distribution.
CHAPTER 10 891
Similarly, the coupling between the thermal and mechanical problems takes place through the temperature-dependent
material properties in the mechanical (stress) problem and the internal heat generation in the mechanical problem caused
by plastic work, which serves as input for the heat transfer problem. The temperature distribution and displacements are
obtained.
In each of the coupled problems described above, two analyses are performed in each load/time increment. Iterations can
also be carried out within each increment to improve the convergence of the coupled thermo-electrical and
thermo-mechanical solutions.
In the coupled fluid-soil model, the fluid is assumed to be inviscid and incompressible. The effect of the fluid is to augment
the stress in the soil material to satisfy equilibrium, and to influence the soil’s material behavior.
In the magnetodynamic analysis, the partially coupled Maxwell’s equations are solved.
In the latter two analyses, the equations (fluid flow/structural or electrical/magnetic) are solved simultaneously.
Thermal Mechanically Coupled Analysis

Many operations performed in the metal forming industry (such as casting, extrusion, sheet rolling, and stamping) can
require a coupled thermo-mechanical analysis. The observed physical phenomena must be modeled by a coupled analysis
if the following conditions pertain:
 The body undergoes large deformations such that there is a change in the boundary conditions associated with the
heat transfer problem.
 Deformation converts mechanical work into heat through an irreversible process which is large relative to other
heat sources.
In either case, a change in the temperature distribution contributes to the deformation of the body through thermal strains
and influences the material properties.
Marc has a capability that allows you to perform mechanically coupled analysis. This capability is available for all stress
elements and for small displacement, total Lagrangian, updated Lagrangian, or rigid plastic analysis.
The COUPLE parameter is used to invoke this option. When defining the mesh, if you specify the element as a stress type
through the CONNECTIVITY option, Marc generates an associated heat transfer element, if possible. The region having an
associated heat transfer element has coupled behavior. If you specify the element as a heat transfer type through the
CONNECTIVITY option, that region is considered rigid. Only heat transfer is performed in that region.
Marc uses a staggered solution procedure in coupled thermo-mechanical analysis. It first performs a heat transfer analysis,
then a stress analysis. Use the CONTROL option to enter the convergence tolerances used in both the structural and heat
transfer passes.
Depending on the type of stress analysis performed, Marc can perform four different types of coupled analysis:
 Quasi-static coupled analysis: comprises of a transient heat transfer pass and a static mechanical pass. Fixed
stepping can be specified using TRANSIENT NON AUTO and adaptive stepping can use TRANSIENT or AUTO STEP.
AUTO STEP is preferred over the TRANSIENT option.
 Creep coupled analysis: comprises of a transient heat transfer pass and a creep mechanical pass. Fixed stepping can
be specified using CREEP INCREMENT and adaptive stepping can use AUTO CREEP or AUTO STEP
 Dynamic coupled analysis: This comprises of a transient heat transfer pass and a dynamic mechanical pass. Fixed
stepping can be specified using DYNAMIC CHANGE and adaptive stepping can use AUTO STEP.
 Harmonic coupled analysis: This can be performed by defining a HARMONIC (Dynamic) structural pass, immediately
followed by either a STEADY STATE, TRANSIENT or AUTO STEP loadcase. By using the NO STRUC input on either of
these options, this loadcase is a pure thermal loadcase. Moreover, these loadcase options enable the possibility to
include viscous heat generation determined in the previous harmonic structural pass. The viscous heat generation
conversion factor can be defined using the CONVERT option.
Load control and time step control can be specified in either of two ways:
 A fixed time step/load size can be specified by using the TRANSIENT NON AUTO option, the CREEP INCREMENT
option, or the DYNAMIC CHANGE option. In these cases, mechanical loads and deformations are incremental
quantities that are applied to each step. Fluxes are total quantities. Time variations for the mechanical and thermal
loads can be specified using appropriate TABLE input.
When using fixed time steps, the COUPLE CONTROL option may be used to specify that the structural pass will be
performed at a lower frequency than the heat transfer pass. This is beneficial when small time steps are required
because of temperature dependent thermal properties, phase transformations, or moving boundary conditions
such as welding analysis.
 An adaptive time step/load size can be specified by using the AUTO CREEP option or the AUTO STEP option. In this
case, mechanical loads and deformations are entered as the total quantities that are applied over the load set.
Fluxes are total quantities. By default, mechanical loads are linearly increased while thermal loads are applied
instantaneously.
 Exercise caution when applying boundary conditions. Use the FIXED DISP, FIXED TEMPERATURE, DISP CHANGE, TEMP
CHANGE, or TABLE options for mechanical or thermal boundary conditions, respectively.
The COUPLE CONTROL model definition or history definition option may be used to specify that the frequency of solving
the structural pass is less than solving the heat transfer pass. This is achieved by storing a copy of the temperature field of
the last increment where a full thermal-structural solution is obtained and using this temperature field to calculate the
thermal strains in the next full thermal-structural solution increment. This is often cost effective because the temperature
dependent properties, including phase transformations, dictate a small time step for the heat transfer pass, but not
necessarily for the (usually computationally more expensive) structural pass.
There are two primary causes of coupling. First, coupling occurs when deformations result in a change in the associated
heat transfer problem. Such a change can be due to either large deformation or contact. Large deformation effects are
coupled into the heat transfer problem only if the LARGE STRAIN or UPDATE parameter is invoked. The gap element in Marc
(Type 12) has been modified so that if no contact occurs, the gap element acts as a perfect insulator. When contact does
occur, the gap element acts as a perfect conductor.
The second cause of coupling is heat generated due to inelastic deformation. The irreversibility of plastic flow causes an
increase in the amount of entropy in the body. This can be expressed as:
· ·p
TS = fW (10-196)
where fW· is the fraction of the rate of plastic work dissipated into heat. Farren and Taylor measure f as approximately 0.9
p
for most metals. Using the mechanical equivalent of heat ( M ), the rate of specific volumetric flux is:
CHAPTER 10 893
·p  
R = MfW (10-197)
Use the CONVERT model definition option to define Mf .

Of course, all mechanical and thermal material properties can be temperature dependent. The governing matrix equations
can be expressed as:
Mu·· + Du· + K  T ,u ,t u = F (10-198)
C T  T T· + K T  T T = Q + Q I + Q F (10-199)
I F
In Equation (10-199), Q represents the amount of heat generated due to plastic work and Q represents the amount of
heat generated due to friction. The specific heat matrix C T and conductivity matrix K T can be evaluated in the current
configuration if the updated Lagrange option is used.
Note: All terms except M can be temperature dependent.
Coupled Acoustic-Structural Analysis

In a coupled acoustic-structural analysis, both the acoustic medium and the structure are modeled. In this way, the effect
of the acoustic medium on the dynamic response of the structure and of the structure on the dynamic response of the
acoustic medium can be taken into account. Such a coupled analysis is especially important when the natural frequencies
of the acoustic medium and the structure are in the same range. In Marc, only a harmonic coupled acoustic-structural
analysis can be performed. Since the interface between the acoustic medium and the structure is determined automatically
by Marc based on the CONTACT option, setting up the finite element model is relatively easy since the meshes do not need
to be identical at the interface.
The acoustic medium will be called the fluid, although it might also be a gas, and is considered to be inviscid and
compressible. The equilibrium equation is given by:
p +  f u·· = 0 (10-200)
in which p is the pressure,  f is the fluid density, u is the displacement vector and a superposed dot indicates
differentiation with respect to time.
Using the bulk modulus of the fluid K f , the constitutive behavior can be written as:
p = – K f u (10-201)
The Equations (10-200) and (10-201) can be combined to:

1 1
----- p·· – ---- 2p = 0 (10-202)
Kf f
It is also possible to add a damping term ru· to Equation (10-200), with r the resistivity or fluid drag:
p + ru· +  f u·· = 0 (10-203)
Since we restrict the discussion to harmonic analyses, we can write u· = iu , with  the excitation frequency and i the
imaginary unit, so that Equation (10-203) reduces to:
1 1
----- p·· – ---- 2p = 0 (10-204)
Kf f
in which the complex density  f is given by:
ir
 f =   f – ---- (10-205)
 
The weak form of Equation (10-204) is obtained in a standard way by introducing the variational field p and integrating
over the fluid volume V :
1 ·· 1
 p  K-----f p – ---- 2p dV = 0 (10-206)
V f
Applying the Green’s divergence theorem to this expression yields:

1 ·· 1 1 p
  p ----
Kf
- p + ---- p p dV +  ---- p ----- dA
   n
= 0 (10-207)
V f  f
in which  represents the boundary of the fluid with an inward normal n .

Along the boundary, various conditions can occur:
 The pressure p is prescribed on the boundary  p .
 The normal acceleration is prescribed to be a n on the accelerating boundary  a .
 The normal acceleration of the fluid equals the normal acceleration of the solid, u··f n = u··sn , on the fluid-
structure interface  f s .
 Nonreflecting or reactive boundary conditions are introduced using a spring-dashpot analogy on  f n as
--1- -----p = p·  c I + p
··  k . In this way, a spring and a dashpot are placed in series between the acoustic medium and
I
 n
its boundary, with k I the spring and c I the dashpot parameter, both per unit area. Some classical conditions can
be modeled by this expression:
1. Free surfaces waves in a gravity field: -----p = p··  g , with g the gravity acceleration;
n
2. The plane wave radiation boundary condition: -----p = p·  c , with c the wave velocity;
n
CHAPTER 10 895
3. Using the complex admittance 1  z    of the boundary, with z    the impedance, the normal velocity can be
related to the pressure by u· f n = p  z    =  1  c I + i  k I p , so that a given z    can be modeled by setting
1  c I and 1  k I . The frequency dependent imaginary acoustic impedance of absorbing materials can be measured
with the standing wave tube test.
Upon discretizing the fluid and the solid using finite elements, the fluid pressure and the displacements of the solid can be
approximated in a standard way, so that Equation (10-207) yields:
 T p 1 pT T p 1 pT  T p T p uT
  p N ----- N p·· + p  N  ----  N p dV –  p N a n dA –  p N N Gu··dA
Kf f 
V a f s
T p 1 pT · p 1 p T ··
(10-208)
+  p  N ---- N p + N ---- N p dA = 0
 cI kI 
 fn
where p and u contain the pressure and displacement degrees of freedom, N contains the interpolation functions and G
is a transformation matrix to relate the displacements in normal direction to the global displacement components. Since
·
Equation (10-208) must be valid for arbitrary admissible values of p , evaluating the various integrals and using p = ip ,
·· 2 ·· 2
p = –  p and u = –  u , results in:
 2  2
 K f + iC f –  M f p +  S f s u = F f (10-209)
 
In order to end up with a set of coupled equations, we need to consider the solid as well. The effect of the fluid on the solid
originates from the fluid pressure at the fluid-solid interface. This can be easily included by evaluating the virtual work
corresponding to the fluid pressure:
p T T u pT T T
W =  –  u ns p dA =  – u G N N p dA = – u S f s p (10-210)
 fs f s
Notice that the minus sign has been used to express that u ns is positive in the outward direction of the solid. Including
Equation (10-210), the solid behavior for a harmonic analysis is governed by:
 2  T
 K s + iC s –  M s u + S f s p = F s (10-211)
 
with K s the stiffness matrix, which can include stress-stiffening, C s is the damping matrix, M s the mass matrix, and F s
the external load vector, except for the fluid pressure.
–2
Premultiplying Equation (10-209) by  and combining the result with Equation (10-211) gives the desired coupled
complex equation system:
–2 –2
 Af S fs p  Ff
= (10-212)
T
S fs As u Fs
in which A f = K f + iC f and A s = K s + iC s .
The procedure to perform a coupled acoustic-structural is as follows. The acoustic medium and the structure are modeled
separately; the acoustic structure using heat transfer elements with acoustic material properties and the structure using
conventional stress elements. The elements representing the acoustic medium are assigned to an acoustic contact body and
the elements representing the solid to a deformable contact body. If the acoustic medium is not surrounded completely by a
deformable structure and one wants to, for example, model radiation boundary conditions along this part of the acoustic
medium, then one can define rigid contact bodies in this area. Like deformable bodies, rigid bodies can be used to define
the spring-dashpot analogy defined before. A harmonic analysis can be performed based on initial contact, so using the
undeformed configuration of the contact bodies, but also after a preload of the deformable contact bodies. In the latter
case, it might be necessary to remesh the acoustic and/or deformable bodies before the harmonic analysis can be performed,
since during the static loading of the deformable bodies the deformation of the acoustic bodies is not taken into account.
The elements which are available for coupled-structural acoustic analysis are described in Table 10-11.
Table 10-11 Element Types for Coupled-Structural Acoustic
Analysis
37, 39 3-, 4-node planar
38, 40 3-, 4-node axisymmetric
Marc computes and prints the following quantities: pressure and pressure gradient at the integration points. The nodal
point data consists of the pressure and the source.
The ACOUSTIC parameter is used to indicate that a coupled acoustic-structural analysis is performed. In addition to the
CONTACT option, the ACOUSTIC and REGION model definition options are used to define the material properties of the acoustic
medium and to set which elements correspond to the solid and the fluid region. Typical boundary conditions for the
acoustic medium can be entered using the FIXED PRESSURE model definition and PRESS CHANGE history definition
option. The DIST SOURCES and POINT SOURCE model definition and DIST SOURCES and POINT SOURCE history definition options
are used to define the load on the acoustic medium. The HARMONIC history definition option is used to define the excitation
frequencies. Spatially and frequency dependent boundary conditions may be defined by referencing the TABLE option.
CHAPTER 10 897
Fluid/Solid Interaction – Added Mass Approach

The fluid/solid interaction procedure investigates structures that are either immersed in, or contain a fluid. Examples of
problems that make use of this feature are vibration of dams, ship hulls, and tanks containing liquids.
Marc is capable of predicting the dynamic behavior of a structural system that is subject to the pressure loading of fluid.
The fluid is assumed to be inviscid and incompressible; for example, water.
In Marc, the fluid is modeled with heat transfer elements (potential theory) and the structure is modeled with normal stress
or displacement elements. The element choice must ensure that the interface between the structural and fluid models has
compatible interpolation; that is, both solid and fluid elements are either first order or second order. The TYING option can
be used to achieve compatibility if necessary. To identify the interfaces between the fluid and the structure, the FLUID SOLID
model definition set is necessary.
During increment zero, Marc calculates the stiffness matrix for the structure and the mass matrix for the structure
augmented by the fluid effect. Marc then extracts the eigenvalues of the coupled system using the MODAL SHAPE option. The
modal superposition procedure can then be used to predict the time response of the coupled system. DYNAMIC CHANGE can
be used to perform modal superposition.
To input properties of solids and the mass density of the fluid elements, use the ISOTROPIC model definition option.
The calculation of the structural mass augmentation requires triangularization of the fluid potential matrix: this matrix is
singular, unless the fluid pressure is fixed at least at one point with the FIXED DISP option.
In a fluid/solid interaction problem, the equations of motion can be expressed as:
1 T
M s a + Ku = ---- S p (10-213)
f
The pressure vector p can be calculated from:

– Sa = Hp (10-214)
The matrices in Equations (10-213) and (10-214) are defined as:
 
Ms =  Ni Ni  s dV (10-215)
V
 
K =   ij D ijkl  kl dV (10-216)
V
 
S =  R  f n i N i dA (10-217)
T
A
 
R R
H =  ---------
x i
----------- dV
x i
(10-218)
f
V
where,
 f and  s are mass densities of the fluid and solid, respectively

a is the acceleration vector
u is the displacement vector
i j is the strain displacement relation
D ijk l is the material constitutive relation
Ni is the displacement interpolation function
R is the pressure interpolation function
ni is the outward normal to the surface with fluid pressure p
T
A is the surface on which the fluid pressure acts
In the present case, the fluid is assumed to be incompressible and inviscid. Only infinitesimal displacements are considered
during the fluid vibration, so that the Eulerian and material coordinates coincide.
Substituting Equation (10-214) into Equation (10-213), we obtain:
 M s + S T H – 1 S   f a + Ku = 0
or (10-219)
Ma + Ku = 0
This equation now allows the modes and frequencies of the solid structure immersed in the fluid to be obtained by
conventional eigenvalue methods.
Coupled Electrostatic-Structural Analysis

In a coupled electrostatic-structural analysis, both the electrostatic field and the structure are modeled. In this analysis type,
the Coulomb force (the force between charged bodies) links the electrostatic part to the structural part, and, in turn, the
deformation influences the electrostatic field. This is different from a piezoelectric analysis where a change in the
electrostatic field causes the deformation which causes a change in the electrostatic field (movement of charge), but
Coulomb forces are not taken into account. The electrostatic-structural coupling is a weak coupling, where in the first pass
the electrostatic field is computed, then the corresponding Coulomb forces are transferred to the structural pass, and finally
the structural pass is evaluated. In a next cycle, the newly deformed state is used in the electrostatic field calculation. Since
the electrostatic solution is a steady-state solution, a time dependent problem will be considered as quasi static for the
electrostatic part. In general, in an electrostatic analysis, we can distinguish conductors and insulators. For example, in a
capacitor, the plates carrying a charge are the conductors, and the dielectric or air between the plates is the insulator. When
a potential difference is applied between the plates, the surfaces of these plates are charged. In Marc, the calculation of the
charge on the different bodies is performed using the concept of contact bodies. In the case of the capacitor, the plates and
the dielectric are separate contact bodies. The charge is then computed at the interface of two contact bodies. Note that it
is necessary that the insulating contact body is touching the conducting contact body. In other words, the contact bodies
on which a Coulomb force is acting are the ones being touched.
CHAPTER 10 899
Coulomb Force
Two methods are available to compute the Coulomb force in a coupled electrostatic-structural analysis. The default method
uses the following equation:
qE
F C ou lom b = ------- (10-220)
2
where q is the charge at the node where the Coulomb force is calculated, and E is the electric field at the insulator side of
the interface. Note that in principle the Coulomb force gets half the contribution from both sides of the interface, but it is
assumed that one side is the conductor where E  0 , so this contribution is neglected. This method works well if the
charged bodies are relatively close to each other (within 20 to 40 element edge lengths) such as the capacitor shown in
Figure 10-44. When the gradient in the electric field becomes too large, the computation of the Coulomb force becomes less
accurate. This is especially true when two charged bodies are located far away from each other. In such cases, the
contribution to the Coulomb force from other bodies is small relative to the self contribution. For this situation, the
following equation is available to compute the Coulomb force:
n
1 Qi Qj
Fi =  ------------ -----------
- (10-221)
4 0 r 2
j = 1 i  j ij
where Q i and Q j are the charge on two nodes, r i j the distance between the two charged nodes, and  0 the permittivity
of the medium between the two nodes. This means that for each node carrying a charge the interaction with all the other
nodes carrying a charge is calculated. However, the relevant nodes carrying a charge are at the surface of contact bodies,
which reduces the number of calculations. The total charge computed on the surface of a contact body will balance the
external and reaction charge on other nodes in the contact body. This procedure is activated by the ELECTRO,2 parameter.
Figure 10-44 Capacitor with Two Closely Spaced Charged Bodies
Equation (10-221) is valid in a 3-D analysis. For an axisymmetric analysis, the charge carrying nodes are extended to circles.
Equation (10-221) is used for a number of points on the circle. For a planar analysis, the nodal charge is a line charge and
the following equation then holds for the Coulomb force:
n
1 Qi Qj
Fi =  ------------ -----------
- (10-222)
2 0 r i j
j = 1 i  j
Analysis Procedure
The procedure to perform a coupled electrostatic-structural analysis is as follows. The different bodies (conductors and
insulators) that can be distinguished must be selected as contact bodies. The way the contact bodies are touching is
important for the computation of the Coulomb Force. The contact bodies on which a Coulomb force is acting are the ones
being touched, so in general a conductor is being touched, and air or a dielectric is the touching contact body. Both
electrostatic and structural element types are available. If an electrostatic element type is selected, then it is not active in the
structural pass. This means that this body does not deform, unless remeshing is carried out. When a structural element is
selected the corresponding electrostatic element will be automatically used in the electrostatic pass.
The ELECTRO parameter together with the STRUCTURAL parameter is used to indicate the coupled electrostatic-structural
analysis. In addition the to the CONTACT option, the ISOTROPIC or ORTHOTROPIC model definition option is used to define the
material properties for both the electrostatic and structural pass. Specify nodal constraints using the FIXED DISP or FIXED
POTENTIAL option. Input nodal loads using the POINT LOAD option or nodal charges using the POINT CHARGE option. Specify
distributed loads with the DIST LOADS option and distributed charges with the DIST CHARGE option. Fixed nodal
displacements and potentials, or nodal forces and charges can also be specified by the FORCDT user subroutine, nonuniform
distributed loads can be specified by the FORCEM user subroutine, and nonuniform distributed charges can be specified
by the FLUX user subroutine. For the history definition, the AUTO LOAD and AUTO STEP options can be used.
Coupled Magnetostatic-Structural Analysis

In a coupled magnetostatic-structural analysis, both the magnetostatic field and the structure are modeled. In this analysis
type, the Lorentz force links the magnetostatic part to the structural part, and in turn, the deformation influences the
magnetostatic field. Two methods to obtain the Lorentz force are available, the Virtual Work Method (VWM), and the
Maxwell Stress Tensor Method (MST). The magnetostatic-structural coupling is a weak coupling, where in the first pass
the magnetostatic field is computed, the Lorentz force is obtained and transferred to the structural pass, and finally the
structural pass is evaluated. In a next cycle, the newly deformed state is used in the magnetostatic field calculation. Since
the magnetostatic solution is a steady-state solution, a time dependent problem will be considered and quasi static for the
magnetostatic part.
Virtual Work Method

With the virtual work method the Lorentz force is derived from the variation of the magnetic energy, caused by small
displacements of a considered body. The magnetic energy is obtained from:
H 
 B  dH dV ,
W =     
(10-223)
V 0 
with W the magnetic energy, B the magnetic induction, H the magnetic field intensity and V the total volume of the
system. The Lorentz force is then,
F L  dr = – dW . (10-224)
A group of elements needs to be selected onto which the Lorentz force is calculated. This group of elements is then
displaced in the x-, y-, and for 3-D also in the z-direction. In Marc, this group of elements must be defined as a contact
body. For correct results, this body must be surrounded by air; to improve accuracy, Marc finds a number of layers of
CHAPTER 10 901
elements surrounding the body, and these elements together with the body get the delta displacements to obtain the Lorentz
force.
Two methods are available to compute the Lorentz force. The first method is based on finite differences. Equation (10-224)
is directly used. The total magnetic energy is computed for the different deformed states and the difference gives the
Lorentz force. The second method is based on local derivatives (see [Ref. 21]). Equations 10-223 and 10-224 can be combined
to the following form.
H
dH d V 
FL =    B ------
ds
- dV +     B dH ----------
ds
-
e Ve Ve 0
and
dV dJ
------------ =  J  – 1 ----------- dV
ds ds
With s the direction of the delta displacement, e the element, V volume, B magnetic induction, H magnetic field
intensity, and J the determinant. Note that with this implementation, only elements that distort due to the delta
displacements will contribute to the force calculation.
With both methods, the computed force is a total force on the particular body and is applied in the structural pass as a
global load on this body.
Maxwell Stress Tensor

The general expression for the Maxwell Stress Tensor is:
1 1 2
T = ------  BB – --- B I , (10-225)
0  2 
with B the magnetic induction, and I the identity tensor. The force equation at any regular point then becomes:
fL =   T . (10-226)
The total force is the integration of this force on an arbitrary closed surface as:
F =
  n  T dS , (10-227)
S
with S any closed surface lying entirely in vacuum bounding the region  , and n the normal to the surface. Note that,
although nodal forces are calculated, only the summation of these forces is a representation of the force on the specific body
or region. In Marc, the region must be defined by a contact body, where the surface of this body will be the surface S onto
which the nodal forces are calculated.
The accuracy of this method depends on the choice of the surface S and the accuracy of the magnetic field B . Best results
are obtained when the body onto which the force is calculated is surrounded by a number of layers of elements having
material properties of vacuum/air.
Analysis Procedure
The procedure to perform a coupled magnetostatic-structural analysis is as follows. Setup the model as usual, structural
elements are the default. For structural elements, the corresponding magnetostatic elements are automatically selected in
the magnetostatic pass. If, for certain regions, like the outer surrounding air needed for the magnetic field, magnetostatic
elements are chosen; then these elements are not active in the structural pass. This means that this region does not deform,
so elements at the interface could get heavily distorted or even go inside out. This can be overcome by remeshing those
regions. Either the VWM or the MST method must be selected. The VWM method only works for one body; for the MST,
method multiple bodies can be selected.
The MAGNETO parameter together with the STRUCTURAL parameter is used to indicate the coupled magnetostatic-structural
analysis. In addition to the CONTACT option to define the different bodies, the ISOTROPIC or ORTHOTROPIC model definition
option is used to define the material properties for both the magnetostatic and structural pass. Specify nodal constraints
using the FIXED DISP or FIXED POTENTIAL option. Input nodal loads using the POINT LOAD option or nodal currents using the
POINT CURRENT option. Specify distributed loads with the DIST LOADS option and distributed currents with the DIST CURRENT
option. To create a coil, the COIL CURRENT and EMWINDING can be used to define its geometry, the location, and the magnitude
of the current in this coil. Fixed nodal displacements and potentials or nodal forces and currents can also be specified by
the FORCDT user subroutine; nonuniform distributed loads or nonuniform distributed currents can be specified by the
FORCEM user subroutine. For the history definition, the AUTO LOAD and AUTO STEP options can be used. Use the FORCE
MAGNETOSTATIC option to specify which method for the force calculation should be used.
Coupled Thermal-Electrical Analysis (Joule Heating)

The coupled thermal-electrical analysis procedure can be used to analyze electric heating problems. The coupling between
the electrical problem and the thermal problem in a Joule heating analysis is due to the fact that the resistance in the electric
problem is dependent on temperatures, and the internal heat generation in the thermal problem is a function of the
electrical flow.
Marc analyzes coupled thermo-electrical (Joule heating) problems. Use the JOULE parameter to initiate the coupled thermo-
electrical analysis. This capability includes the analysis of the electrical problem, the associated thermal problem, and the
coupling between these two problems.
The electrical problem is a steady-state analysis and can involve current and/or voltage boundary conditions as well as
temperature-dependent resistivity.
The thermal analysis is generally a transient analysis with temperature-dependent thermal properties and
time/temperature-dependent boundary conditions.
Use the ISOTROPIC, ORTHOTROPIC, TEMPERATURE EFFECTS, ORTHO TEMP, or TABLE model definition options to input reference
values of thermal conductivity, specific heat, mass density, and electrical resistivity, as well as their variations with
temperatures. The mass density must remain constant throughout the analysis.
Use the FIXED VOLTAGE model definition option for nodal voltage boundary conditions, and the POINT CURRENT and/or DIST
CURRENT model definition options for current boundary conditions. No initial condition is required for the electrical
problem, since a steady-state solution is obtained.
In the thermal problem, you can use the INITIAL TEMP, FILMS, POINT FLUX, DIST FLUXES, and FIXED TEMPERATURE model
definition options to prescribe the initial conditions and boundary conditions. Use the FILM and FLUX user subroutines or
tables for complex convective and flux boundary conditions. To enter the unit conversion factor between the electrical and
CHAPTER 10 903
thermal problems, use the JOULE model definition option. Marc uses this conversion factor to compute heat flux generated
from the current flow in the structure.
Use the STEADY STATE, TRANSIENT, POINT CURRENT, DIST CURRENT, VOLTAGE CHANGE, POINT FLUX, DIST FLUXES, and TEMP CHANGE
history definition options for the incrementation and change of boundary conditions.
A weak coupling between the electrical and thermal problems is assumed in the coupled thermo-electrical analysis, such
that the distributions of the voltages and the temperatures of the structure can be solved separately within a time increment.
A steady-state solution of the electrical problem (in terms of nodal voltages) is calculated first within each time step.
The heat generation due to electrical flow is included in the thermal analysis as an additional heat input The temperature
distribution of the structure (obtained from the thermal analysis) is used to evaluate the temperature-dependent resistivity,
which in turn is used for the electrical analysis in the next time increment.
For output, voltage, current density, and heat generation are available as integration point values. Note that current density
must not be confused with Ohmic current. Current density is the electric current per unit area of cross section, while
Ohmic current is the current going through the total area. So the latter is a global quantity.
In the coupled thermo-electrical analysis, the matrix equation of the electrical problem can be expressed as;
E
K  T V = I (10-228)
and the governing equation of the thermal problem is:
C T  T T· + K T  T T = Q + Q
E
(10-229)
In Equations (10-228) and (10-229):
E
K is the temperature-dependent electrical conductivity matrix
I is the nodal current vector
V is the nodal voltage vector
C T  T  and K T  T  are the temperature-dependent heat capacity and thermal conductivity
matrices, respectively
T is the nodal temperature vector
T· is the time derivative of the temperature vector
Q is the heat flux vector
E
Q is the internal heat generation vector caused by the current flow.
E E
The coupling between the electrical and thermal problems is through terms K and Q in Equations (10-228) and (10-229).
The selection of the backward difference scheme for the discretization of the time variable in Equation (10-229) yields the
following expression:
1 1
-----  C T  T   + K T  T T n = Q n + Q nE + ----- C T  T T n – 1 (10-230)
t t
Equation (10-230) is used for the computation of nodal temperatures in each time increment t .
E
The internal heat generation vector Q is computed from:
E T E
Q = B q dV (10-231)
V
E 2
q = I R (10-232)
where I and R are the current and electrical resistance, respectively.

The controls for the heat transfer option allow input of parameters that govern the convergence solution and accuracy of
heat transfer analysis.
Resistance
A resistor is a device that can dissipate electric energy. For a pure resistor, this usually happens in the form of heat. This
energy per unit time (or power) depends on the current in the resistor and its resistance. Resistance is always associated with
all conductors allowing flow of electric current. A typical model can contain a number of conductors allowing flow of
electric current, which gives rise to a resistance value for each conductor.
All conductors in a model are assumed to be contact bodies and defined as such. This is done using the THERMAL CONTACT
model definition option. Two conductors can touch each other, but cannot overlap. If two conductors touch, it implies
flow of current between the two. A subset of the conductor bodies can be considered for a resistance computation. This
subset is specified using the EMRESIS history definition option, which refers to the bodies defined on the THERMAL CONTACT
model definition option. Marc performs a resistance computation only if the EMRESIS history definition option is specified
in the input file.
A Joule heating analysis is used to compute the resistance values. All insulators in the problem are neglected for modeling.
The computation of the resistance values results from the usual Joule heating analysis with some constraints on boundary
conditions: there must be at least one electric potential boundary condition on any node. If no boundary condition is
specified on a portion of the problem boundary, it is assumed to satisfy homogeneous Neumann condition.
The resistivity of each material is specified as per the usual Joule heating analysis. If the resistivity of any material is much
higher than those of other materials, then that material is to be treated as an insulator and should not be included in the
model.
Ohm’s law for conductors is:
V
R = --- (10-233)
I
V2
W E = I 2 R = ------ = VI (10-234)
R
CHAPTER 10 905
where,
I is the current through the resistor

R is the resistance
V is the voltage across it
W E is the generated power
The conductance G is given by:
1
G = --- (10-235)
R
Consider a single conducting body placed in an infinite homogenous insulating medium. The insulating medium can be
assumed to have negligible conductivity, in which case the current through it is negligible. All current flows in the
conductor. The current density J is then defined as the normal current passing through a unit cross-sectional area. This
means that:
I =   J  dS (10-236)
S
where S is the cross-sectional surface of the conductor.
Conduction current I is conserved; hence, I is same for any conductor cross section.
The electric field E is related to current density J in the conductor by:

J = E (10-237)
where  is the conductivity of the conductor.
The power W E in a conductor is also given by:
1
WE =    J  E dV =    --- J  J dV (10-238)
Vc Vc
where V c is the conductor domain.
The current IC flowing into the contact body and flowing out of the contact body can be obtained by adding all applied
currents and reaction currents from the finite element analysis. If this summation fails, the electric potential drop VC across
the conductor is obtained. The resistance is then calculated using Equation (10-234) with either IC or VC.
Coupled Electrical-Thermal-Mechanical Analysis

Coupled electrical-thermal-mechanical analysis (Joule-mechanical) basically combines electrical-thermal analysis (Joule
heating) with thermal-mechanical analysis. Coupled electrical-thermal-mechanical analysis is handled using a staggered
solution procedure. Using this approach, the electrical problem is solved first for the nodal voltages. Next, the thermal
problem is solved to obtain the nodal temperatures. The mechanical problem is solved last for the nodal displacements.
Coupling between the electrical and thermal problems is mainly because of heat generation due to electrical flow Q E (Joule
heating). The thermal and mechanical problems are coupled through thermal strain loads F T and heat generation due to
inelastic deformation Q I and friction Q F . Additional coupling may be introduced in case of temperature-dependant
electrical conductivity K E and mechanical stiffness K M . Nonlinearities may arise in the thermal problem due to
convection, radiation, and temperature-dependant thermal conductivity and specific heat. The mechanical problem may
involve geometric and material nonlinearities. Contact is another source of nonlinearity. If contact occurs between
deformable bodies or deformable and rigid bodies in the mechanical problem, boundary conditions of the electrical and
thermal problems are updated to reflect the new contact conditions.
E T M
Q F ,K
Electrical Thermal Mechanical
E I F
K Q,Q
The matrix equations governing the electrical, thermal and mechanical problems can be expressed as:
K E  T V = I
C T  T T· + K T  T T = Q + Q E + Q I + Q F
Mu·· + Du· + K M  T ,u ,t u = F + F T
where,
V is the nodal voltage vector.

T is the nodal temperature vector.
u is the nodal displacement vector.
KE T  is the temperature-dependent electrical conductivity matrix.
I is the nodal current vector.
CT T  is the temperature-dependent heat capacity matrix.
KT T  is the temperature-dependent thermal conductivity matrix.
Q is the heat flux vector.
QE is the heat generation due to electrical flow vector.
QI is the heat generation due to inelastic deformation vector.
QF is the heat generation due to friction vector.
M is the mass matrix.
D is the damping matrix.
CHAPTER 10 907
K M  T ,u ,t  is the temperature, deformation and time-dependent stiffness matrix.

F is the externally applied force vector.
FT is the force due to thermal strain vector.
The JOULE and COUPLE parameters are needed to initiate the coupled electrical-thermal-mechanical analysis. When defining
the mesh through the CONNECTIVITY model definition option, specify the elements as stress type. Marc internally switches
to the associated heat transfer element in the electrical and thermal passes.
The FIXED VOLTAGE model definition option can be used for nodal voltage boundary conditions, and POINT CURRENT and/or
DIST CURRENT model definition options for current boundary conditions. No initial condition is required for the electrical
problem, since a steady-state solution is obtained.
In the thermal problem, INITIAL TEMP, FILMS, POINT FLUX, DIST FLUXES, and FIXED TEMPERATURE model definition options can
be used to prescribe the initial conditions and boundary conditions. The FILM and FLUX user subroutines can be used for
complex convective and flux boundary conditions. To enter the unit conversion factor between the electrical and thermal
problems, use the JOULE model definition option. Marc uses this conversion factor to compute heat flux generated from
the current flow in the structure.
In the mechanical problem, the usual structural model definition options such as FIXED DISP, POINT LOAD, and DIST LOADS are
used to prescribe the boundary conditions.
The history definition options (VOLTAGE CHANGE, POINT CURRENT, DIST CURRENT, TEMP CHANGE, POINT FLUX, DIST FLUXES, DISP
CHANGE, POINT LOAD, and DIST LOADS) for the incrementation and change of boundary conditions.
JOULE and CONVERT model definition options are used to enter the conversion factor for the electrical-thermal and the
thermal-mechanical problems.
Coupled Magnetostatic-Thermal Analysis

A coupled magnetostatic thermal analysis couples a magnetostatic analysis with a thermal analysis. With this analysis type,
heat generation in magnetostatic devices can be processed and simulated. The implementation in Marc follows a staggered
approach. First, a magnetostatic analysis is performed followed by a thermal analysis. In magnetostatic devices, the source
of heat generation is the applied currents flowing through conductors. The lamination loss computation is available with
the magnetostatic analysis and the heat generated due to these losses is also considered here. It should be noted that
lamination loss occurs due to a sinusoidally time-varying operation and the losses are expressed as RMS (root-mean-
squared) quantities. Hence, the heat generated due to lamination loss gives a correct picture of the subsequent temperature
distribution. The applied currents flowing through conductors are assumed to be steady time-invariant and will not contain
skin effect. The lamination loss is considered here, since the magnetostatic analysis can handle a non-linear B-H curve,
which is not possible in harmonic magnetodynamic analysis.
The total generated heat is then used in the thermal analysis. Temperature dependency for material data can also be taken
into account, which is especially important for the permeability of metals. The relative permeability drops to one when the
Curie temperature is reached, which can result in a significant change in the generated heat flux.
The heat generated in magnetostatic analysis is handled in the same way as Joule heating analysis.
To initiate a coupled magnetostatic thermal analysis, add the MAGNETO and HEAT parameters. The Magnetostatic pass is a
static magnetostatic pass; the thermal part can be either a steady state or a transient analysis. When defining the mesh, if
you specify the element as a magnetostatic type through the CONNECTIVITY option, Marc generates an associated heat transfer
element. The region having an associated heat transfer element has coupled behavior. If you specify the element as a heat
transfer type through the CONNECTIVITY option, that region is only active in the thermal pass. Use the ISOTROPIC or
ORTHOTROPIC model definition options to input values for the thermal conductivity, specific heat, mass density, and
emissivity, as well as permeability and electric conductivity. Temperature dependency for these material properties can be
added when the table driven input option is used.
For the magnetostatic pass, boundary conditions can be applied. Use FIXED POTENTIAL to prescribe the magnetostatic
potential on nodes. To apply nodal currents use POINT CURRENT. Distributed currents can be applied using DIST CURRENT. To
create a coil, the COIL CURRENT and EMWINDING can be used to define its geometry, the location, and the magnitude of the
current in this coil. The FORCDT and FORCEM user subroutines can be used to manually control the size and/or direction of
the different vectors.
For the thermal pass, use the INITIAL TEMP, FIXED TEMPERATURE, DIST FLUXES, and FILMS model definition options and the TEMP
CHANGE and POINT FLUX history definition options to prescribe the initial conditions and boundary conditions. The FILM and
FLUX user subroutines can be used for complex convective and flux boundary conditions.
The operational frequency for the lamination losses is given using the EMLAMIN history definition option.
The thermal pass can be STEADY STATE or time dependent using the TRANSIENT or AUTO STEP history definition options.
If noncompatible units are used in the magnetostatic and thermal analysis, a unit conversion factor between the current
density and the generated heat can be used. It has to be ensured that the same factor applies to the lamination losses. Use
the CONVERT model definition option for this.
Coupled Magnetodynamic-Thermal Analysis

A coupled magnetodynamic thermal analysis couples a harmonic magnetodynamic analysis with a thermal analysis. With
this analysis type, induction heating processes can be simulated. The implementation in Marc follows a staggered approach.
First, a harmonic magnetodynamic analysis is performed followed by a thermal analysis. The harmonic magnetodynamic
field generates induced currents in the model. Current is also induced in conductors that carry applied currents. The net
current in these conductors is the sum of the applied and induced current. The net current in the remaining model is the
induced current. These net currents generate heat and a heat flux is computed which is then used in the thermal analysis.
Temperature dependency for material data can also be taken into account, which is especially important for the
permeability of metals. The relative permeability drops to one when the Curie temperature is reached, which can result in
a significant change in the generated heat flux. The depth of the material induction currents generated is an important
factor in the heating process. This so-called skin depth is defined as the depth at which the magnitude of the induced
–1
current density drops to e of the magnitude at the surface
1
 = ------------- , (10-239)
f
where f is the frequency,  the electrical conductivity, and  the permeability. Hence, the penetration depth depends on
the frequency. Low frequencies are used for more uniform preheating, and higher frequencies are used for surface heating.
The heat generated in induction heating analysis is similar to a Joule heating analysis, however now in the harmonic
magnetodynamic analysis the computed current density is a complex vector instead of a scalar. The local heat generation
is (see [Ref. 17]):
CHAPTER 10 909
E 1 2 1
q = --- J = --- J  J , (10-240)
 
with  the electric conductivity, and J the complex current density.

To initiate a coupled magnetodynamic thermal analysis, add the EL-MA and HEAT parameters. The magnetodynamic pass is
a harmonic pass; the thermal part can be either a steady state or a transient analysis. When defining the mesh, if you specify
the element as an magnetodynamic type through the CONNECTIVITY option, Marc generates an associated heat transfer
element. The region having an associated heat transfer element has coupled behavior. If you specify the element as a heat
transfer type through the CONNECTIVITY option, that region is only active in the thermal pass. Use the ISOTROPIC or
ORTHOTROPIC model definition options to input values for the thermal conductivity, specific heat, mass density, and
emissivity, as well as permeability, permittivity, and electric conductivity. Temperature dependency for these material
properties can be added when the table driven input option is used.
For the magnetodynamic pass, boundary conditions can be applied. Use FIXED POTENTIAL to prescribe the magnetostatic or
electrostatic potential on nodes, and POTENTIAL CHANGE in the history definition option. To apply nodal currents or nodal
charges used POINT CURRENT and POINT CHARGE respectively. Distributed currents or distributed charges can be applied using
DIST CURRENT or DIST CHARGE, respectively. The FORCDF and FORCEM user subroutines can be used to manually control the
size and/or direction of the different vectors.
For the thermal pass INITIAL TEMP, FIXED TEMPERATURE, TEMP CHANGE, DIST FLUXES, POINT FLUX, and FILMS can be used to
prescribe the initial conditions and boundary conditions. The FILM and FLUX user subroutines can be used for complex
convective and flux boundary conditions.
The excitation frequency has to be given for the harmonic magnetodynamic pass using the HARMONIC history definition
option.
The thermal pass can be STEADY STATE or time dependent using the TRANSIENT or AUTO STEP history definition options.
If noncompatible units are used in the magnetodynamic and thermal analysis, a unit conversion factor between the current
density and the generated heat can be used. Use the CONVERT model definition option for this.
Dual Frequency
It is possible to perform induction heating using two different frequencies simultaneously. This will give more control on
how heat is generated in a conductor. At a lower frequency the heating occurs more deeply in the workpiece, while at a
higher frequency the heating occurs nearer to the surface.
Analyzing dual frequency induction heating is possible in combination with a COIL CURRENT boundary condition, or when
a circuit is used together with a TERM CURRENT or TERM VOLTAGE (Magnetostatic) boundary condition. Different boundary
conditions must be selected for the low and high frequency in the LOADCASE (Model Definition) option. In the dual frequency
approach, the solution is computed by first solving for the low frequency and then for the high frequency. Then, to
compute the local thermal energy obtained from the induced currents from both frequencies, these induced currents are
integrated over one full period ( ) of the low frequency excitation in the following way:
Where:
J is the complex current density.

Jlow is the magnitude of the current density of the low frequency.
low is the phase angle of the current density of the low frequency.
Jhigh is the magnitude of the current density of the high frequency.
high is the phase angle of the current density of the high frequency.
flow is the low frequency.
fhigh is the high frequency.
 is the electric conductivity.
Best results are obtained when the high frequency is a multiple of the low frequency.
When magnetically nonlinear material is used, an assumption must be made on how to derive the local effective
permeability from the superposition of the low frequency magnetic field and the high frequency magnetic field. This
effective permeability is computed in the following way (see the references 23. and 24.):
Where is evaluated from the B(H) curve, where the magnetic field intensity ( is derived from the
superposition of the complex low frequency magnetic field intensity and the complex high frequency magnetic field
intensity.
Coupled Magnetodynamic-Thermal-Structural Analysis

A coupled magnetodynamic-thermal-structural analysis couples a low-frequency harmonic magnetodynamic analysis with
a thermal and a structural analysis. The coupled magnetodynamic-thermal analysis (induction heating) is already explained
in the corresponding section. With the extra structural coupling, the induction heating process can be simulated such that
structural effects like motion of workpiece and/or coils, elastic plastic deformation, and thermal/residual strains can be
evaluated. Note that in this type of analysis, Lorentz forces are not taken into account. These forces operate at the frequency
of the applied current which is considered much higher than the time range of the simulation.
The implementation in Marc follows a staggered approach. For each increment, a harmonic magnetodynamic pass is
performed first, a thermal pass second, and a structural pass last. The flow chart shown in Figure 10-45 illustrates which
information is passed to the different physics passes.
CHAPTER 10 911
Harmonic
Electromagnetic where:
u
QE T
T – temperature
Thermal u – displacement
Q F,P T QE – heat due to Ohmic power loss
u
Q F, P – heat due to friction and plastic deformation
Structural
Figure 10-45 Flowchart for Coupled Magnetodynamic-Thermal Structural Analysis
Since this type of simulation contains a field problem not only the workpiece must be meshed, but also the surrounding
air must be meshed. Two approaches are available to do this.
Single Mesh Approach

The first approach is similar to the approach used in coupled electrostatic-structural and coupled magnetostatic-structural
and is illustrated in Figure 10-46. Here, the workpiece and the surrounding air are meshed in one model where separately
meshed entities (bodies) can be connected using contact. Bodies which contain the surrounding air can be treated as zero
stiffness bodies. This implies that the elements of this body are not active in the structural pass; the body is considered to
have no stiffness. To accommodate possible deformations due to deformation of bodies touching this body, this body
should be remeshed. Note that the mesh density of the workpiece must be dense enough to capture the induction currents.
Typically, element edge lengths must be smaller than the skin depth as discussed in the section about coupled
magnetodynamic-thermal.
Figure 10-46 Schematic View of the Single Mesh Approach; a Coarse Mesh is used for Illustration Purposes
Dual Mesh Approach

The second approach is the dual mesh approach, which is illustrated in Figure 10-47. The workpiece is meshed as in the
single mesh approach, but here the total region (the surrounding air plus the workpiece) needed for the magnetodynamic
solution has to be meshed separately. This region is called the dual mesh. In this approach, the mesh of the workpiece will
overlap the dual mesh. The mesh of the workpiece is used in the thermal and structural pass. All material properties must
be given for the workpiece mesh, while for the dual mesh only the magnetodynamic material properties of the space
surrounding the workpiece must be given. During the simulation, the material properties of the dual mesh will be replaced
by that of the workpiece when appropriate, so when a specific integration point of an element of the dual mesh is located
inside the workpiece, it will get the material properties of the corresponding structural element. The thermal energy density
which is computed in the dual mesh is mapped to the workpiece mesh. In conclusion, two independent meshes have to be
created. One mesh contains the actual model as will be used in the structural pass. This mesh contains all material
properties. Thermal and structural boundary conditions must be applied here. The second mesh covers the region of the
structural mesh plus enough space to take care of the far field. Note that the mesh density where the workpiece is or will
move to must be dense enough to capture the induction currents. Typically, element edge lengths must be smaller than the
skin depth as discussed in the section about coupled magnetodynamic-thermal. The dual mesh will have material properties
of the surrounding air/vacuum. Magnetodynamic boundary conditions like potential and coil currents must be applied
here. Note that when coil currents are used, these coils do not have to be physically modeled. Their shape is specified, and
then they are inserted in the magnetodynamic mesh. Marc will find all the elements which should get part of the load.
Figure 10-47 Schematic View of the Dual Mesh Approach; a Coarse Mesh is used for Illustration Purposes
To initiate a coupled magnetodynamic thermal structural analysis, add the EL-MA, HEAT, and STRUCTURAL parameters. For
this analysis the magnetodynamic pass must be a harmonic pass, and the thermal and structural pass are transient passes.
When you define the finite element mesh a structural element type should be specified for the workpiece through the
CONNECTIVITY option, Marc will generate the associated magnetodynamic and heat transfer element for the magnetodynamic
and thermal pass respectively. It is also possible the select a magnetodynamic element type for part of the mesh. These
elements will be inactive during the structural pass. It is advised to store a connected group of these elements in a contact
body (zero stiffness body) so that global remeshing (ADAPT GLOBAL) can be used to accommodate deformation. When the
dual mesh approach is used structural elements should be chosen for the workpiece and magnetodynamic elements for the
dual mesh.
This analysis only works with the multiphysics style input, so the VERSION and TABLE parameters must be used. This
multiphysics style input allows you to set the material properties per physics pass, so e.g. isotropic magnetodynamic
properties can be combined with anisotropic thermal properties and orthotropic structural properties. Material properties
can depend on a table, see Volume C for details. The CONTACT option can be used, and ADAPT GLOBAL can be used for
remeshing. A large variety of boundary conditions are available among which COIL CURRENT and EMWINDING for creating an
CHAPTER 10 913
electric current load. For the history definition, the AUTO LOAD and AUTO STEP options can be used. The excitation frequency
has to be given for the harmonic magnetodynamic pass using the HARMONIC history definition option.
The COUPLE CONTROL model definition or history definition option may be used to specify that the frequency of solving the
magnetodynamic pass and/or the structural pass is less than solving the heat transfer pass. The generated heat due to the
induced currents is computed in the magnetodynamic pass and is applied in the subsequent increments of the thermal pass.
A copy of the temperature field is stored of the last increment where a thermal and structural solution is obtained. This
temperature field is used to calculate the thermal strains in the next increment when a structural solution is obtained. This
is often cost effective because a small time step is not always required for each of the three passes. The magnetodynamic
pass is usually computationally the most expensive.
Dual Frequency
It is possible to perform a coupled magnetodynamic-thermal-structural analysis using two different frequencies
simultaneously. See Dual Frequency in the section Coupled Magnetodynamic-Thermal Analysis for more information.
Coupled Diffusion/Thermal Analysis

In a coupled diffusion/thermal analysis, the mass transport and the heat transport through a porous medium are analyzed
simultaneously. Different applications of this type of analysis can be performed in Marc, depending on the amount of detail
that is desired in the coupling.
Diffusion Analysis with Temperature Dependence

A first application of the coupled diffusion/thermal analysis is a diffusion problem where the gas percolates through a
material whose temperature evolves in time. The equations resolved for this application are those presented in the Heat
Transfer and Diffusion sections of this chapter.
The activation of this analysis is made through the DIFFUSION parameter.
The gas and solid material properties are entered through the ISOTROPIC, ORTHOTROPIC, or ANISOTROPIC options.
The initial states of temperature and pressure are respectively entered through the INITIAL TEMP option and the INIT PRESS
option.
The diffusion part requires a pressure boundary condition entered through the FIXED PRESSURE option or a mass flow rate
entered through the DIST MASS model definition option. If none are applied then a homogeneous Neumann condition (zero
gradient) is applied.
The usual thermal boundary conditions can be used. They are explained in detail in the Heat Transfer section under Boundary
Conditions.
The capability is available for two- or three-dimensional analysis using the standard heat transfer continuum elements.
The primary unknown variables of the model are the gas pressure and the temperature. Marc computes and prints the
following information at the element integration points: temperature, pressure, gradient of pressure, mass flow rate of gas,
gradient of temperature, and heat flux.
Darcy’s Law Pyrolysis Model

Another application of the coupled diffusion/thermal analysis in Marc is the Darcy’s law model of pyrolysis. The key feature
of this model that differentiates it from the pyrolysis streamline model is the use of the Darcy’s law implemented in the
diffusion model to represent the pyrolysis gas flow.
The Darcy flow model solves the coupled equations describing the gas and heat flow through a porous solid material during
the process of pyrolysis. Pyrolysis is the internal decomposition of a virgin material into a residual char and a gas as a result
of the extreme heat loads on the material. This composition is described by a kinetic law of an Arrhenius type. The gas
percolates through the porous material as a result of the high pressure gradients, and its flow is determined by Darcy’s law.
The ideal gas law is used as the state equation for the gas. The basic balance equations describing the process are the mass
balance and the energy balance.
The activation of the Darcy pyrolysis model is done through the combination of the PYROLYSIS, 3 parameter and the
DIFFUSION parameter.
A standard material must be defined for each phase of the pyrolysis, namely virgin material, charred material, and gas. The
required material properties for the solid phases are the mass density, the specific heat, the thermal conductivity, the
enthalpy of formation, the reference temperature, the porosity and the permeability. For the gas phase, the required material
properties are the specific heat, the enthalpy of formation, the reference temperature, the dynamic viscosity, and the
molecular weight (Perfect Gas Law Model). These inputs are entered through the ISOTROPIC, ORTHOTROPIC, or ANISOTROPIC
options.
Reference to these phases as well as the Arrhenius decomposition law coefficients is done in Mentat by creating a material
of type pyrolysis which will write the THERMO-PORE model definition option in the Marc input file.
The initial states of temperature, pyrolysis and pressure are respectively entered through the INITIAL TEMP option, the INITIAL
PYROLYSIS option and the INIT PRESS model option.
The diffusion part requires a pressure boundary condition entered through the FIXED PRESSURE option or a mass flow rate
entered through the DIST MASS model option. If none are applied then a homogeneous Neumann condition (zero gradient)
is applied.
The thermal boundary conditions can be entered through the SURFACE ENERGY, the FIXED TEMPERATURE, FILMS, or DIST FLUXES
options. Recession is supported through the RECEDING SURFACE option and radiation is also supported.
The capability is available for two- or three-dimensional analysis using the standard heat transfer continuum elements.
The primary unknown variables of the model are the gas pressure and the temperature. Marc computes and prints the
following information at the element integration points: temperature, pressure, gradient of pressure, mass flow rate of gas,
solid density and gas density, gradient of temperature and heat flux.
Darcy Flow Model

Initially, the material consists of virgin material only, which by virtue of its porosity already contains some gas. The initial
gas density follows from the initial pressure and the initial temperature through the ideal gas law. It is assumed that at every
point in the structure and at every instant in time, the virgin, char, and gas materials are in thermal equilibrium. Because
of the high heat load, the virgin material decomposes into a char and a gas. The primary driving force of the process is given
by a kinetic decomposition law, describing the decomposition rate as a function of temperature and other material
characteristics. In Marc, there are two forms of this decomposition law, which are explained further in this section and in
more details in the section Heat Transfer – Pyrolysis.
CHAPTER 10 915
The material density of a solid material is understood as the effective density, ̂ , of the porous material; i.e.,
̂ =  1 –   , where  is the “real” density of the non-porous material and  is the porosity of the material. The
dV p o re s
porosity is defined as the ratio of the volume taken by the pores over the total volume of a material point:  = -------------------
-.
dV
The effective gas density, ̂ g , is understood to represent the total mass of the gas inside the pores with respect to the total
volume dV ; i.e., ̂ g =  g , where  g is the real gas density.
The balance relations are derived by considering a fixed control volume of the material. Over the external boundaries of
this control volume mass transport, due to gas flow, and energy transport, due to gas flow and heat conduction, take place.
These balance relations are derived in the next two sections followed by their weak form that is used in the Finite Element
formulation.
List of Used Symbols

Notation convention: Underlined characters (e.g., n) represent vector quantities. The subscript “ x ” in the symbols below
can be either “ v ” denoting the virgin material, “ c ” denoting the char material or “S g ” denoting the pyrolysis gas.
Latin Symbols
A Area of control volume boundary [m2]
Bj pre-exponential factor of Arrhenius term j [1/s]
E a j activation energy of Arrhenius term j [J/mol]
c px specific heat at constant pressure [J/(kg K)]
H total enthalpy inside control volume [J]
·h total enthalpy flow rate over control volume boundary [W]
Hx total enthalpy per unit mass [J/kg]
H x enthalpy of formation per unit mass [J/kg]
hx sensible enthalpy per unit mass [J/kg]
H average enthalpy value of composite material [J/kg]
n outward unit normal to boundary surface of control volume []
M total mass inside control volume [kg]
·
M total mass flow rate over control volume boundary [kg/s]
m g* prescribed mass flow rate per unit area [kg/(s m2)]
Mg molar mass of gas [kg/mol]
Nd number of terms in Arrhenius law []
p pressure of gas [N/m2]
p* prescribed pressure [N/m2]
p0 initial pressure of gas [N/m2]
Q total heat flux over control volume boundary [W]

q* prescribed heat flux per unit area [W/m2]
q heat flux vector [W/m2]
R universal gas constant (  8.314 ) [J/(K mol)]
t time [s]
T temperature [K]
T* prescribed temperature [K]
T0 initial temperature [K]
T ambient or environment temperature [K]
T re f reference temperature [K]
ug real flow velocity vector of gas [m/s]
û g percolation velocity vector of gas [m/s]
V volume taken by the control volume [m3]
Vx volume taken by material [m3]
x volume fraction of char material []
Greek Symbols
 film coefficient [W/(m2 K)]
x permeability [m2]
 average permeability of composite material [m2]
x coefficient of thermal conductivity [W/(m K)]
 average coefficient of thermal conductivity of composite material [W/(m K)]
g dynamic viscosity of gas [(N s)/m2]
̂ s j density reduction of Arrhenius term j (counted as positive) [kg/m3]
̂ x effective material density [kg/m3]
x real material density of non-porous material [kg/m3]
j reduction factor of Arrhenius term j ( 0   j  1 ) []
 porosity []
x porosity []
 average porosity of composite material []
j reaction order of Arrhenius term j []
CHAPTER 10 917
Mathematical Symbols and Notations

div divergence operator
 gradient operator
. dot product operator
Balance Relations for Mass

At any instant in time, a small volume element, dV , consists of char and virgin material:
dV = dV c + dV v
When x is the volume fraction of the char, we can write:

dV c = xdV
dV  =  1 – x dV
Initially, x = 0 and when the material has completely pyrolysized, x  1 . Char and virgin material are porous materials,
so the volume element dV is also given as the sum of the volume taken by the solid, dV s , and the volume taken by the
pores filled with gas, dV g :
dV = dV s + dV g
With the definition of porosity,  , we have:
dV g = dV
dV s =  1 –  dV
The volume taken by the pores is the sum of the pore volume in the virgin material and the pore volume in the char:
dV g =  c dV c +  v dV v =   c x +  v  1 – x  dV
The current porosity is therefore given by the virgin and char states as:
 = c x + v  1 – x 
The volume taken by the solid can thus be written as:

dV s =   1 –  c x +  1 –  v   1 – x  dV
The solid mass inside the volume element, dV , can be given as:
 s dV s =   c  1 –  c x +  v  1 –  v   1 – x  dV =  ̂ c x + ̂ v  1 – x  dV =  s  1 –  dV = ̂ s dV
The mass of the gas inside the pores of the volume element, dV , can be given as:
 g dV g =  g dV = ̂ g dV
The char volume fraction can thus be expressed as:
̂ v – ̂ s
x = -----------------
-
̂ v – ̂ c
The total mass inside the control volume is:
M =   ̂ s + ̂ g  dV
V
The mass flow rate of the mass flowing out of the control volume is determined by the gas flow and is given as:
· =
M   g û g  n dA
A
The balance of mass requires that the mass change per unit time inside the control volume equals the mass flow rate of the
mass flowing over the boundaries of the control volume:
dM · , or
--------- = – M
dt
̂ s ̂ g
  --------
t
+ --------- dV
t 
=   g û g  n dA
V A
Employing the divergence theorem, we arrive at the following partial differential equation for the mass balance:
̂ ̂
--------s + --------g- + div   g û g  = 0
t t
The rate of change of the solid mass is given by a kinetic decomposition law of Arrhenius type:
Nd
̂  j
--------s =  ̂ s j -------- , where
t t
j = 1
E a j
 – ------------ 
--------j = – B j e RT  j j , with  j  t = 0  =  j 0
t
The change rate of the char volume fraction is thus given as:
x 1 ̂ s
------ = – ------------------ --------
t ̂ v – ̂ c t
The percolation (filtration) velocity of the gas through the porous material is given by Darcy’s law, which states that this
velocity is proportional to the pressure gradient:
CHAPTER 10 919

û g = – ------ p
g
The percolation velocity is not the real velocity of the gas. The two are related by:
 g û g = ̂ g u g  û g = u g
The gas density is related to the pressure and the temperature by the ideal gas law:
Mg p Mg p
- or ̂ g =  -----------
 g = ----------
RT RT
The partial differential equation resulting from the mass balance equation in terms of pressure p and temperature T can
now be written in its final form:
̂ g p ̂ g T    v –  c  ̂ s
----- ------ – ----- ------ – div   g ------ p = –  1 +  g -----------------
- --------
p t T t  g   ̂  – ̂ c t
Balance Relations For Energy

The total enthalpy (i.e., energy content) per unit mass for any of the materials is the sum of the sensible enthalpy and the
enthalpy of formation:
H x = h x + H x
The sensible enthalpy is determined with respect to some reference temperature:

T
hx  T  =  c px dT ,
T ref
where c px is the specific heat at constant pressure.
The enthalpy of the solid in volume element dV can be written as:

 s H s dV s =  c H c  1 –  c dV c +  v H v  1 –  v dV v
or
̂ s H s dV =  ̂ c H c x + ̂ v H v  1 – x  dV
The enthalpy of the gas in the pores of volume element dV can be written as:
 g H g dV g = ̂ g H g dV
The total enthalpy inside the control volume therefore is:
H =   ̂ g H g + ̂ s H s  dV
A
The enthalpy flow rate of the enthalpy flowing out of the control volume is determined by the gas flow and is given as:
·h =   g H g û g  n dA
A
The total supply of heat per unit time over the control volume boundary into the volume is given as:
Q = –  q  n dA
A
The heat flux vector q is determined by Fourier’s law of heat conduction, which states that the heat flux is proportional to
the temperature gradient:
q = – VT
The balance of energy requires that the enthalpy change per unit time inside the control volume equals the enthalpy flow
rate of the mass flowing over the boundaries of the control volume plus the rate of heat supply over these boundaries:
dH
-------- = –
dt
h + Q , or
  ̂ g H g + ̂ s H s 
 ----------------------------------------
t
- dV = –   g H g û g  n dA –  q  n dA
V A A
Employing the divergence theorem we arrive at the following partial differential equation for the energy balance:
  ̂ g H g + ̂ s H s 
----------------------------------------- + div   g H g û g  + div  q  = 0
t
By employing the mass balance equation this can be brought to:
  ̂ s H s  H g ̂ s
-------------------- + ̂ g ---------- – H g -------- +  g û g  H g + div  q  = 0
t t t
H T
With ---------x- = c px ------ (where the subscript x is c , v , or g ) and H g = c p g T , and by employing Darcy’s law and
t t
Fourier’s law and by defining:
̂ v H v – ̂ c H c
H = -------------------------------- , and  ̂c p  e f f = x̂ c c p c +  1 – x ̂ v c p v + ̂ cg c p g
̂ v – ̂ c
CHAPTER 10 921
The following final form of the partial differential equation in terms of pressure p and temperature T for the energy
balance is obtained:
T ̂ s 
 cˆp  eff ------ +  H – H g  -------- –  g c p g ------ p  T – div  T  = 0
t t g
Resulting Governing Equations and Their Weak Formulation

The partial differential equations (PDE) from the previous sections together with the appropriate boundary (BC) and
initial (IC) conditions are summarized below:
Mass balance PDE:
̂ g p ̂ g T    v –  c ̂ s
----- ------ – ----- ------ – div   g ------ p +  1 +  g -----------------
- -------- = 0
p t T t  g   ̂ v – ̂ c t
BC:
p = p* on A p : prescribed pressure
 p
 g ------ ------ = m g* on A m : prescribed gas mass flow rate
 g n
IC:
p = p 0 at t = 0 in V : initial pressure distribution
Energy balance PDE:
T  ̂ s
 ̂c p  eff ------ –  g c pg ------ p  T – div  T  +  H – H g  -------- = 0
t g t
BC:
T = T* on A T : prescribed temperature
T
 ------ = q* on A q : prescribed heat flux.
n
A prescribed heat flux can be entered directly by the POINT FLUX option or the DIST FLUXES option or can be entered by any
form present in the FILMS option or by the SURFACE ENERGY option.
IC:
T = T 0 at t = 0 in V : initial temperature distribution
The gas density in above equations can always be expressed in terms of pressure and temperature by employing the ideal
gas law. The rate of change term of the solid is defined by the Kinetic Law and serves as a driving force term in the process.
Finally, an initial condition for the Kinetic Law is required:
 j =  j0 at t = 0 in V : initial pyrolysis condition
Weak Formulation
The above equations are cast in their weak form to obtain the finite element equations of the problem. For the mass
equation, the weight functions are denoted by p and for the energy equation the weight functions are denoted by T .
Together with the short hands

m g =  g ------ p and q h = T ,
g
The weighting for the mass equation can be written as:

̂ g p  g T   v –  c ̂ s
 - -------- p dV –   m g* – m g  n p dA = 0
----- ------ – ----- ------ – div  m g  +  1 +  g -----------------
p t T t  ̂ v – ̂ c t
V A
The weighting for the energy equation can be written as:
T  ̂ s
  ̂c p  eff ------ –  g c pg ------ p  T – div  q h  +  H – H g  -------- T dV –   q* – q h  n T dA = 0
t g t
V A
By employing the divergence theorem (i.e. partial integration), the following expressions are obtained which form the basis
for the finite element equations:
ˆ ˆ  v –  c ̂ s
  g p  g T  
  ----- ------ – ----- ------ p +  g ------ p  p dV = –   1 +  g ------------------ -------- p dV +  m g* p dA
 p t T t   g  ̂ v – ̂ c t
V A
T  ̂ s
  ̂c p  eff ------ T + T  T –  g c p g ------ p  TT dV =   H g – H  -------- T dV +  q*T dA
t g t
V V A
This is a coupled set of transient equations for the unknown pressure and temperature fields. By choosing appropriate
spatial and temporal interpolations for these fields, we obtain a coupled system of nonlinear algebraic equations. This set
of equations is solved in a staggered way, where for each time step the pressure equation (from the mass balance) is solved
first, and subsequently the temperature equation (from the energy balance) is solved. On the right-hand side, the loading
terms due to the material decomposition are recognized next to the regular loading terms from prescribed heat fluxes. In
ˆg T v – c
the staggered approach, the term ----- ------ moves to the right hand side of the pressure equation. The factor  g ------------------ is
T t ̂ v – ̂ c
expected to be small against unity and is disregarded. The system of equations resulting from the pressure equation is always
symmetric. The terms related to p   in the energy equation are the convective heat transport terms as a result of the
gas flow. These render nonsymmetric matrix contributions to the system of temperature equations. If the heat capacity of
the gas can be disregarded, the convective contributions drop out and the temperature equations remain symmetric.
Symmetry is also retained if these terms are moved to the right-hand side.
CHAPTER 10 923
Average Material Properties

Since the solid material is a composite of different constituents and the mass fractions of these constituents vary spatially
and over time, it is necessary to define average values for the different material properties that appear in the balance
equations. The energy balance clearly shows how the average heat capacity for the solid must be defined:
 ̂c p  s eff =  1 – x ̂ v c p v + x̂ c c p c
A similar averaging approach is taken for the thermal conductivity:
̂ =  1 – x  v + x c
For the permeability, this averaging approach is applied to the logarithmic values of the constituents:
log    =  1 – x  log   v  + x log   c 
The last two averaging approaches do not follow from the balance equations and are, therefore, purely heuristic.
Molecular Diffusion Coupled Analyses

The molecular diffusion pass can be coupled to the thermal pass and the structural pass through:
 A thermal/molecular diffusion coupled analysis.
 A thermal/ molecular diffusion/structural coupled analysis.
 A molecular diffusion/structural coupled analysis.
The thermal/ molecular diffusion/structural coupled analysis is explained here. An application of it can be the heat
treatment process called carburizing.
In the thermal/ molecular diffusion/structural coupled analysis the effect of:
 The temperature on the structural properties, through the material properties or/and the thermal expansion
 The temperature on the diffusivity of the material
 The molar concentration on the structural properties such as yield stress (hardening) and diffusion expansion can
be taken into account.
The HEAT, MOLECULAR, STRUCTURAL parameters are used to invoke this coupled analysis.
The material properties must be entered in the multi-physics format, namely (for isotropic): ISOTROPIC (with TABLE Input -
Thermal), ISOTROPIC (with TABLE Input - Molecular Diffusion), ISOTROPIC (with TABLE Input - Molecular Diffusion).
Marc uses a staggered solution procedure in coupled thermo-mechanical analysis. It first performs a heat transfer analysis,
then a molecular diffusion analysis, then a stress analysis. Use the multi-physics format for the CONTROL option (CONTROL
(Heat Transfer - Model Definition), CONTROL (Molecular Diffusion - Model Definition), CONTROL (Mechanical - Model Definition)) to enter the
convergence tolerances used in the respective passes.
The CONTACT model option (THERMAL CONTACT (2-D) model option if the structural pass is not included), the CON INTERA
model option can be used to describe the contact between bodies or/and environment.
The COUPLE CONTROL (Model Definition) model definition or history definition option may be used to specify that the frequency
of solving the structural pass or the molecular diffusion pass is less than solving the heat transfer pass. This is achieved by
storing a copy of the temperature field of the last increment where a full thermal-structural solution is obtained and using
this temperature field to calculate the thermal strains in the next full thermal-structural solution increment. The same is
done for the molar concentration to calculate the diffusion expansion. This is often cost effective because the temperature
dependent properties and the molar concentration properties dictate a small time step for the heat transfer pass and the
molecular diffusion pass, but not necessarily for the (usually computationally more expensive) structural pass.
Technical background
The coupling between the heat transfer pass and the structural pass in this coupled analysis is the same as for the
thermal/structural pass.
The coupling between the molecular diffusion pass and the structural pass adds a contribution to the total strains through
the diffusion strain defined as:
diff 0
 = C – C 
where
diff is the diffusion strain (Please note that this strain is not available as post variable)
 is the diffusion expansion instantaneous coefficient
C is the current molar concentration
C0 is the initial molar concentration
Please note that the diffusion strain is not available as post variable. The effect of the diffusion strain is only visible through
the displacement post variable.
CHAPTER 10 925
References
1. T. J. R. Hughes, L. P. France, and M. Becestra, “A New Finite Element Formulation for Computational Fluid
Dynamics, V. Circumventing the Babuska-Brezzi Conduction, A S Petrov-Galerkin Formulation of the Stokes
Problem Accommodating Equal-Order Interpolations”, Comp. Meth. in Applied Mech. and Eng., p. 85-90, Vol.
59, 1986.
2. G. Hauke and T. J. R. Hughes, “A unified approach to compressible and incompressible flow”, Comput. Methods
Appl. Mech. Eng., p. 389-395, 113, (1994).
3. T. E. Tezduyar, S. Mittal, S. E. Ray, and R. Shih, “Incompressible Flow computations with stabilized bilinear and
linear equal order-interpolation velocity-pressure elements”, Comput. Methods. Appl. Mech. Eng., p. 221-242, 95,
(1992).
4. B. Ramaswaymy, T. C. Jue, “Some Recent Trends and Developments in Finite Element Analysis for
Incompressible Thermal Flows”, Int. J. Num. Meth. Eng., p. 671-707, 3, (1992).
5. A. N. Brooks and T. J. R. Hughes, “Streamline upwind/Petrov-Galerkin formulations for convection dominated
flows with particular emphasis on the incompressible Navier-Stokes equations”, Comp. Meth. Appl. Eng., 30,
(1982).
6. Codina, R. “Finite element formulation for the numerical solution of the convection-diffusion equation” 1993.
7. Cornfield, G. C., and Johnson, R. H. “Theoretical Predictions of Plastic Flow in Hot Rolling Including the Effect
of Various Temperature Distributions.” Journal of Iron and Steel Institute 211, pp. 567-573, 1973.
8. Hsu, M. B. “Modeling of Coupled Thermo-Electrical Problems by the Finite Element Method.” Third
International Symposium on Numerical Methods for Engineering, Paris, March, 1983.
9. Peeters, F. J. H. “Finite Element Analysis of Elasto-Hydrodynamic Lubrication Problems.” in Proceedings of the
XIth Int. Finite Element Kongress, edited by IKOSS GmbH. Baden- Baden, Germany, Nov. 15-16, 1982.
10. Yu, C C. and Heinrich, J. C. “Petrov-Galerkin methods for the time-dependent convective transport equation.”
Int. J. Numer. Meth. Engrg., Vol. 23 (1986), 883-901.
11. Yu, C C. and Heinrich, J. C. “Petrov-Galerkin methods for multidimensional time-dependent convective transport
equation.” Int. J. Numer. Meth. Engrg., Vol. 24 (1987), 2201-2215.
12. Zienkiewicz, O. C. The Finite Element Method in Engineering Science. Third Ed. London: McGraw-Hill, 1978.
13. Zienkiewicz, O. C., and Godbole, P. N. “A Penalty Function Approach to Problems of Plastic Flow of Metals with
Large Surface Deformations.” Journal of Strain Analysis 10, 180-183, 1975.
14. Zienkiewicz, O. C., and Godbole, P. N. “Flow of Plastic and viscoPlastic Solids with Special Reference to Extrusion
and Forming Processes.” Int. Num. Methods in Eng. 8, 1974.
15. Zienkiewicz, O. C., Loehner, R., Morgan, K., and Nakazawa, S. Finite Elements in Fluid Mechanics – A Decade of
Progress, John Wiley & Sons Limited, 1984.
16. J. Goldak, A. Chakravarti, and M. Bibby, “A New Finite Element Model for Welding Heat Sources”, Metallurgical
Transactions B., Volume 15B, June 1984, pp. 299 - 305
17. S. Clain, J. Rappaz, M. Swierkosz, and R. Touzani, “Numerical Modelling of Induction Heating for two-
Dimensional Geometries”,Math. Models Methods Appl. Sci., Vol 3 no 6, 805-822, 1993
18. C. Chaboudez, S. Clain, R. Glardon, J. Rappaz, M Swierkosz, and R. Touzani, “Numerical Modelling of Induction
Heating of Long Workpieces”, IEEE Trans. Magn.,Vol 30, 5026-5037, 1994
19. C. Chaboudez, S. Clain, R. Glardon, D.Mari, J. Rappaz, and M Swierkosz, “Numerical Modeling of Induction
Heating of Axisymmetric Geometries”, IEEE Trans. Magn.,Vol 33, 739-745, 1997
20. R.A. Rindal, “An analysis of the coupled chemically reacting boundary layer and charring ablator, Part IV: An
approach for characterizing charring ablator response with in-depth coking reactions.”, NASA CR-1065, June
1968.
21. J.L. Coulomb, G. Meunier, “Finite element implementation of virtual work principle for magnetic or electric force
and torque computation”, IEEE Trans. Magn. Vol. Mag-20, No. 5, September 1984.
22. S.H.A. Boers, P.J.G. Schreurs, M.G.D. Geers, “Operator-split damage-plasticity applied to groove forming in food
can lids”, International Journal of Solids and Structures 42 (2005) 4154-4178.
23. Clain, S.; Rappaz, J.;Swierkosz, M.; Touzani, R., “Numerical Modeling of Induction Heating for Two-
Dimensional Geometries”, Mathematical Models and Methods in Applied Science 3 (1993), Nr. 6, S. 805-822.
24. Labridis, D.; Dokopoulos, P., “Calculation of eddy current losses in nonlinear ferromagnetic materials”, IEEE
Trans. Mag. 25 (1989) 2665-2669.
Chapter 11: Solution Procedures for Nonlinear Systems
11 Solution Procedures for

Nonlinear Systems
 Overview 922

Considerations for Nonlinear Analysis 922

Newton-Raphson Method 936
 Modified Newton-Raphson Method 937
 Direct Substitution 939

Arc-length Methods 939

Convergence Controls 946
 Equilibrium per node 949

Singularity Ratio 950
 Solution of Linear Equations 951
 Flow Diagram 957

Remarks 958

References 958
Overview
This chapter discusses the solution schemes in Marc for nonlinear problems. Issues of convergence controls, singularity
ratio, and available solvers for linearized system of equations are also discussed. In a nonlinear problem, a set of equations
must be solved incrementally. The governing equation of the linearized system can be expressed, in an incremental form, as
Ku = r (11-1)
where u and r are the correction to the incremental displacements and residual force vectors, respectively.
There are several solution procedures available in Marc for the solution of nonlinear equations:
 Newton-Raphson Method
 Modified Newton-Raphson Method
 Direct Substitution
 Arc-length Methods
Considerations for Nonlinear Analysis

Nonlinear analysis is usually more complex and expensive than linear analysis. Also, a nonlinear problem can never be
formulated as a set of linear equations. In general, the solutions of nonlinear problems always require incremental solution
schemes and sometimes require iterations (or recycles) within each load/time increment to ensure that equilibrium is
satisfied at the end of each step. Superposition cannot be applied in nonlinear problems.
The iterative procedures supported in Marc are: Newton-Raphson, Modified Newton-Raphson, Newton-Raphson with
strain correction, and direct substitution. If the R-P flow contribution model is chosen, a direction substitution is used.
A nonlinear problem does not always have a unique solution. Sometimes a nonlinear problem does not have any solution,
although the problem can seem to be defined correctly.
Nonlinear analysis requires good judgment and uses considerable computing time. Several runs are often required. The first
run should extract the maximum information with the minimum amount of computing time. Some design considerations
for a preliminary analysis are:
 Minimize degrees of freedom whenever possible.
 Halve the number of load increments by doubling the size of each load increment.
 Impose a coarse tolerance on convergence to reduce the number of iterations. A coarse run determines the area of
most rapid change where additional load increments might be required. Plan the increment size in the final run by
the following rule of thumb: there should be as many load increments as required to fit the nonlinear results by the
same number of straight lines.
Marc solves nonlinear static problems according to one of the following two methods: tangent modulus or initial strain.
Examples of the tangent modulus method are elastic-plastic analysis, nonlinear springs, nonlinear foundations, large
displacement analysis and gaps. This method requires at least the following three controls:
 A tolerance on convergence
 A limit to the maximum allowable number of recycles
 Specification of a minimum number of recycles
CHAPTER 11 923
Solution Procedures for Nonlinear Systems
An example of the initial strain method is creep or viscoelastic analysis. Creep analysis requires the following tolerance
controls:
 Maximum relative creep strain increment control
 Maximum relative stress change control
 A limit to the maximum allowable number of recycles
To input control tolerances, use the model definition option CONTROL for all simulations. These values can be reset upon
restart or through the CONTROL history definition option. For creep simulations, addition controls are provided in the AUTO
CREEP or AUTO STEP options. See Convergence Controls in this chapter for further discussion on tolerance controls.
Behavior of Nonlinear Materials

Nonlinear behavior can be time- (rate-) independent, or time- (rate-) dependent. For example, plasticity is time-
independent and creep is time-dependent. Both viscoelastic and viscoplastic materials are also time-dependent. Nonlinear
constitutive relations must be modeled correctly to analyze nonlinear material problems. A comprehensive discussion of
constitutive relations is given in Chapter 7.
Scaling the Elastic Solution

The SCALE parameter causes scaling of the linear-elastic solution to reach the yield stress in the highest stressed element.
Scaling takes place for small displacement elastic-plastic analysis, where element properties do not depend on temperature.
The SCALE parameter causes all aspects of the initial solution to be scaled, including displacements, strains,
stresses, temperature changes, and loads. Subsequent incrementation is then based on the scaled solution.
Load Incrementation
Several history definition options are available in Marc to input mechanical and thermal load increments (see Table 11-1).
The choice is between a fixed and an automatic load stepping scheme.
Table 11-1 History Definition Options for Load Incrementation

Load Type Fixed Adaptive
Mechanical AUTO LOAD AUTO STEP
DYNAMIC CHANGE AUTO INCREMENT
CREEP INCREMENT AUTO THERM
AUTO CREEP
AUTO THERM CREEP*
Thermal TRANSIENT NON AUTO AUTO STEP
TRANSIENT*
*The option is obsolete; use AUTO STEP instead.
For a fixed scheme, the load step size remains constant during a loadcase. The fixed schemes are AUTO LOAD for static
mechanical, CREEP INCREMENT for creep, DYNAMIC CHANGE for dynamic mechanical, and TRANSIENT NON AUTO for
thermal/thermo-mechanically coupled.
In many nonlinear analyses, it is useful to have Marc automatically determine the appropriate load step size. For an adaptive
scheme, the load step size changes from one increment to the other and also within an increment depending on convergence
criteria and/or user-defined physical criteria. The adaptive schemes are AUTO STEP and AUTO INCREMENT for static mechanical,
AUTO STEP and AUTO CREEP for creep, AUTO STEP, AUTO THERM, and AUTO THERM CREEP for thermally driven mechanical
problems, and AUTO STEP for dynamic mechanical, and AUTO STEP and TRANSIENT for thermal/thermo-mechanically coupled.
The adaptive stepping scheme of choice is AUTO STEP. AUTO STEP has been designed as a unified load stepping scheme and
many of the capabilities of the other stepping schemes can now be handled by AUTO STEP. AUTO STEP can be used for
mechanical (static, creep, dynamic), thermal, and thermo-mechanically coupled analysis problems.
 In the AUTO STEP scheme, either a recycle based convergence criterion is used to automatically determine the time
step based upon a comparison of the actual number of recycles needed in an increment to a user-specified desired
number of recycles or a damping energy procedure is used. In addition to this, user-defined or program-
determined physical criteria based upon strain, stress, displacement, or temperature increments can be used to
control the time step. Reductions in the time step through cut-backs are used to satisfy both convergence criteria
and physical criteria.
More details on the AUTO STEP option are provided in the next section.
 Post buckling or snap-through analyses require the so-called arclength method which is available through the AUTO
INCREMENT option. This option can only be used in static mechanical analyses and the applied load is automatically
increased or decreased in order to maintain a certain arclength.
Note: AUTO INCREMENT can also be used for general situations without instabilities, but, in general, the AUTO STEP
option is preferred for these situations.
 For creep analysis, the available adaptive options are AUTO CREEP and AUTO STEP. The AUTO CREEP option determines
the time step in an explicit creep analysis based on the creep strain change and the stress change (see Chapter 3,
Volume C: Program Input). These checks are not available by default in AUTO STEP. These may however be
incorporated by using an absolute or relative creep strain increment criterion and an absolute or relative stress
increment criterion as additional user-defined physical criteria in combination with the default convergence
criterion. Addition of these user-defined criteria for AUTO STEP is quite simple. If an appropriate input flag is set,
two physical criteria are automatically added by the program at run-time for explicit creep problems: creep strain
increment/elastic strain = 0.5, and stress increment/stress at beginning of increment = 0.5. It should also be noted
that AUTO STEP is usually more reliable in cases involving creep and contact.
 For the case of automatic load stepping for a thermally loaded elastic-creep/elastic-plastic-creep stress analysis, the
available adaptive schemes are AUTO THERM CREEP and AUTO STEP. Allowable increments for normalized creep strain,
normalized stress and state variables can be optionally prescribed for AUTO STEP either through user-defined criteria
or program determined automatic physical criteria.
CHAPTER 11 925
 For thermally driven mechanical problems, the available options are AUTO THERM and AUTO STEP. The thermal loads
derived from a thermal analysis are applied using the CHANGE STATE option in a mechanical analysis. In the AUTO
THERM scheme, the load step of the mechanical analysis is automatically adjusted based upon user-specified
(allowed) changes in temperature from the thermal analysis per increment. For example, if there is a change of 50°
in the thermal analysis in one increment but only a change of 10° per increment is allowed in the mechanical
analysis, AUTO THERM splits up the thermal increment into five mechanical increments. Allowable state variable
increments can be optionally prescribed for AUTO STEP either through a user-defined criterion or program
determined automatic physical criterion. If these criteria are violated in an increment, AUTO STEP cuts the time-step
back and repeats the increment with a smaller time step.
Selecting Load Increment Size

Selecting a proper load step increment is an important aspect of a nonlinear solution scheme. Large steps often lead to many
recycles per increment and, if the step is too large, it can lead to inaccuracies and nonconvergence. On the other hand, using
too small steps is inefficient.
Fixed Load Incrementation

When a fixed load stepping scheme is used, it is important to select an appropriate load step size that captures the loading
history and allows for convergence within a reasonable number of recycles. For complex load histories, it is necessary to
prescribe the loading through time tables while setting up the run.
For fixed stepping, there is an option to have the load step automatically cut back in case of failure to obtain convergence.
When an increment diverges, the intermediate deformations after each recycle can show large fluctuations and the final
cause of program exit can be any of the following: maximum number of recycles reached (exit 3002), elements going inside
out (exit 1005 or 1009) or, in a contact analysis, nodes sliding off a rigid contact body (exit 2400), and nodes not being
projected properly onto 3-D NURBS (exit 2401). These deformations are normally not visible as post results (there is a
feature to allow for the intermediate results to be available on the post file, see the POST option). If the cutback feature is
activated and one of these problems occur, the state of the analysis at the end of the previous increment is restored and the
increment is subdivided into a number of subincrements. The step size is halved until convergence is obtained or the
user-specified number of cutbacks has been performed. Once a subincrement is converged, the analysis continues to
complete the remainder of the original increment. No results are written to the post file during subincrementation. When
the original increment is finished, the calculation continues to the next increment with the original increment count and
time step maintained. If the global remeshing option is activated in conjunction with the cutback feature, then, for exit
1005 or 1009, the chosen contact body is remeshed and the analysis is repeated with the original time step before the
first cutback.
User Defined Time Incrementation

The user can have complete control over the time step used at every increment by using the User-defined time stepping
option. In this case the user should provide the time stations to be reached by means of a table with Independent Variable
Time or Normalized Time. The dependent variable can be the Increment Number but it is not used by the analysis.
If a user has information about the time step requirements for his particular job, either due to the nature of the job or based
on some previous runs, this information can be used in the subsequent runs to reduce the number of cut-backs that can be
triggered if the convergence or user criterion is not met with the given time step. Avoiding time step cut-backs leads to lesser
number of increments in the analysis and it can lead to considerable savings in total analysis time.
Cutbacks are supported with this option and are triggered if convergence is not achieved with the given time step at any
increment. Note that for creep or dynamic or thermal or coupled analysis, the fixed stepping procedure with constant time
steps does not offer the cutback capability. In such cases, using the user-defined option for time stepping offers a valuable
cut-back capability for the load case. Only tables with monotonically increasing time are allowed with this option.
Note:  Non-positive definite is not activated by default for User-Defined time stepping
option whereas it is the default for a Multi-Criteria Load case.
 While changing a Loadcase from Multi-Criteria to User-Defined Time Step, care
should be taken to activate non-positive definite if required from the Solution
Control menu.
 Similarly, artificial damping may be used for stabilization of the solution during
Multi-Criteria load case, however it is not taken into account while using User-
defined time stepping scheme. This may lead to differences in behavior between a
Multi-Criteria run and a User-defined time stepping run with the same time steps.
Arc Length Method

The arc-length procedures assume that the control of the nonlinear behavior and possible instabilities is due to mechanical
loads, and that the objective is to obtain an equilibrium position at the end of the loadcase. Hence, while the program may
increase or decrease the load, the load can always be considered to be F = F b +   F e – F b  , where F b and F e are the
loads at the beginning and end of the loadcase. The scale factor does not necessarily vary linearly from 0 to 1 over the
increments, and may, in fact, become negative. This would result in negative time steps as well; hence, the AUTO INCREMENT
history definition option cannot be used with dynamics and should not be used with table driven input where the load is
a function of time.
Mechanical loads, as shown above, are applied in a proportional manner and thermal loads are applied instantaneously.
This means that any automatic load incrementation method is limited to mechanical input histories that only have linear
variations in load or displacement and thermal input histories that have immediate change in temperature. For example,
one may not use a rigid body with a linearly changing velocity, since the resulting displacement of the rigid body would
give parabolically changing displacements. In this case, one would need to use a constant velocity for the arc length method
to work properly.
For the arc length method, care must be taken to appropriately define the loading history in each loadcase. The loadcase
should be defined between appropriate break points in the load history curve. For example, in Figure 11-1, correct results
would be obtained upon defining three distinct loadcases between times 0 – t 1 , t 1 – t 2 , and t 2 – t 3 during the model
preparation. However, if only one loadcase is defined for the entire load history between 0 – t 3 , the total applied load for
the loadcase is zero. This is also true when the AUTO STEP history definition option is used when the table driven input is
not used.
CHAPTER 11 927
P2 P2
P (Load)
P (Load)
P1 P1
t1 t2 t3 t3
0 0
t (Time) t (Time)
a. Three Defined Loadcases b. One Defined Loadcase
Figure 11-1 Defining Loadcases for Automatic Load Incrementation
When the table driven input method is used with the AUTO INCREMENT history definition option and behavior as shown in
Figure 11-1 is desired, the independent variable should be the loadcase number and not time. A quasi-static total mechanical
load can, hence, be given which is both a function of position and the loadcase number.
AUTO STEP
The scheme appropriate for most applications is AUTO STEP. The primary control of the load step is based upon the number
of recycles needed to obtain convergence. There are a number of optional user-specified physical criteria that can be used
to additionally control the load step. The user inputs needed to define the AUTO STEP scheme are described in Marc Volume
C: Program Input.
Recycling Criterion
The default recycle based criterion works as follows: The user specifies a desired number of recycles. For most problems, it
is sufficient to provide a value in the range of three to five. For problems with severe nonlinearities or for problems with
very small convergence tolerances, it may be necessary to increase this number. This number is used as a target value for
the load stepping scheme. If the number of recycles required in the current increment is less than the desired number, the
load step for the next increment is increased. The time step increase is based on a factor, S u , that can also be specified by
the user. Typical values for S u are in the range of 1.2 to 1.5. While the time step increase is obviously more aggressive with
larger scale factors, it should be noted that there may be excessive recycling and cutbacks if sudden nonlinearities are
encountered. In order to avoid this, Marc uses the following logic for higher scale factors. If the actual number of recycles
in an increment is greater than 60 % of the desired number of recycles (i.e., the current increment did not converge easily),
the increase scale factor for the next increment is limited to 1.25 for scale factor values between 1.25 and 1.5625, and to
80 % of the value for scale factors above 1.5625.
See Damping Energy for alternative approach.
Time Step Cutback Scheme

The load step is never increased during an increment. If the number of recycles needed to obtain convergence exceeds the
desired number, the load step size is scaled back, the recycling cutback number N r is incremented by 1 and the increment
Nr
is performed again with the new load step. The scaleback factor for the N r th cutback is taken as s , where the factor s
2   Nr m  Nr m + 1  
is calculated from the expression s =  T s  T m  ; where N r m is the maximum number of recycling
related cutbacks for the increment and is calculated from N r m = log 10  10 5 * T s  T m  , T s is the time increment before
any recycling related cutbacks occur for the increment and T m is the minimum possible time step for the increment. T m
is equal to the value set by the user ( 10 – 5 by default) if there is no quasi-static artificial damping and is equal to 10 – 3
times the value set by the user ( 10 – 8 by default) if there is quasi-static artificial damping. The scaleback factor for any
Nr
cutback is the smaller of ( s , 1  S u ). This scheme guarantees that no matter what the starting time step for an increment,
the minimum time step is reached in a reasonable number of cutbacks if the increment consistently fails to converge. For
special cutbacks such as maximum number of recycles reached (exit 3002), elements going inside out (exit 1005 or 1009)
or, in a contact analysis, nodes sliding off a rigid contact body (exit 2400) nodes not being projected properly onto a 3-D
Nr
NURB (exit 2401), the scaleback factor is the smaller of ( s , 0.5). If the minimum time step is reached and the analysis
still fails to converge, it is terminated with exit 3015. If the ‘proceed when not converged’ option is used, then the analysis
proceeds to the next increment if and when the maximum number of recycles are reached.
Exceptions
There are some exceptions to the basic scheme outlined above. If an increment is consistently converging with the current
load step and the number of recycles exceeds the desired number, the number of recycles is allowed to go beyond the desired
number until convergence or up to the user specified maximum number. The time step is then decreased for the next
increment by 1  S u . An increment is determined to be converging if the convergence ratio was decreasing in three previous
recycles.
Special rules also apply in a contact analysis. For quasi-static problems, the AUTO STEP option is designed to only use the
automated penetration check option (see CONTACT option, 7th field of 2nd data block; option 3 is always used). During
the recycles, the contact status can keep changing (new nodes come in contact, nodes slide to new segments, separate etc.).
Whenever the contact status changes during an increment, a new set of contact constraints are incorporated into the
equilibrium equations and more recycles are necessary in order to find equilibrium. These extra recycles, due to contact
changes, are not counted when the recycle number is checked against the desired number for determining if the load step
needs to be decreased within the increment. Thus, only true Newton-Raphson iterations are taken into account. For the
load step of the next increment, the accumulated number of recycles during the previous increment is used. This ensures
that the time step is not increased when there are many changes in contact during the previous increment.
Thermal Analysis
For the most part, the recycling criterion works in a similar fashion for thermal analysis (heat transfer analysis, or thermal
part of a coupled thermo-mechanical analysis). The recycling criterion is used to satisfy the tolerance value provided for the
temperature error in estimate on the CONTROL option; that is, the analysis cuts back and starts the increment over if the
temperature error in estimate is not consistently converging within the desired number of recycles. If the temperature
estimate consistently converges in three previous recycles, the analysis continues recycling. Once the temperature estimate
tolerance is satisfied, the actual incremental temperature change T a is calculated and checked against the corresponding
CHAPTER 11 929
tolerance T m provided on the CONTROL option. It should be noted that if a user criterion on temperature is also provided,
the latter parameters over-ride the one provided on the CONTROL option. More details on how user criteria are handled are
described in the next section. If the incremental temperature change is not satisfied, the time step is scaled-back using a
factor = 0.8  T m  T a  . In this case, the thermal pass is simply continued with the smaller time step without a cut-back
because if there was a cutback, the analysis would have redo the temperature error in estimate convergence from scratch.
However, if the maximum number of allowed recycles is reached before thermal convergence is achieved, a cutback with a
Nr
scale-back factor of min( s , 0.5) is made and the increment is repeated.
User-defined Physical Criteria

In addition to allowing Marc to use the number of recycles for automatically controlling the step size for AUTO STEP, user-
specified physical criteria can optionally be used for controlling the step size. The user-specified physical criteria work as
follows. The user can specify the maximum allowed incremental change within certain ranges for specific quantities during
an increment. The quantities available are displacements, rotations, stresses, strains, strain energy, temperature (in thermal
or thermomechanically coupled analyses), and state variables (in mechanical or thermomechanically coupled analysis).
These criteria can be utilized in one of two ways.
By default, they are used as limits, which means that the load step is immediately decreased if a criterion is violated during
any iteration of the current increment, but they do not influence the decision to change the load step for the next
increment; that is, only the actual number of recycles versus desired number of recycles controls the load step for the next
increment.
The criteria can also be used as targets; in which case, they are used to control the time step for the current and next
increments. If the calculated values of the criteria are higher than the user-specified values in any iteration, the time step is
scaled down and the current increment is repeated. If the calculated values of the criteria for the current increment are
consistently smaller than the user-specified values prior to convergence, the time step for the next increment is scaled up.
The scale factor used for reduction or increase is the ratio between the actual value and the target value and this factor is
limited by user-specified minimum and maximum factors (defaults to 0.1 and 10 respectively). If this type of load step
control is used together with the recycle based control, the time step can be reduced in the current increment due to
whichever criterion that is violated. The decision to increase the step size for the next increment is solely based upon the
physical criteria.
Specification of user-defined physical criteria can be further simplified by setting a special flag in the AUTO STEP option that
allows for physical criteria to be automatically added by the solver at run-time. These automatic criteria serve as
upper-bound controls to prevent run-away Newton-Raphson iterations that ultimately cause the program to abort.
Currently, four mechanical criteria are automatically added depending on the kind of analysis that is being run: a total strain
criterion is added for any large displacement analysis and the maximum allowable equivalent total strain increment at any
point in the model is set to 50%; a plastic strain criterion is added for any large displacement, finite strain analysis and the
maximum allowable equivalent plastic strain increment at any point in the model is set to10%; a relative creep strain
criterion and a relative stress change criterion are added for any explicit creep analysis wherein the maximum allowable
creep strain change/elastic strain and the maximum allowable equivalent stress change/equivalent stress are each set to 0.5;
a state variable criterion is added for any large displacement analysis wherein the maximum allowable temperature
increment is such that the equivalent stress increment associated with the change in thermal properties of the materials does
not exceed 50% of the total equivalent stress. These criteria are only added in the analysis if there are no competing
explicitly defined user-criteria found. It should also be noted that these automatic criteria are only used as limits; i.e. they
are used to control the time step within an increment but not for the next increment.
Failure to satisfy user-defined physical criteria can occur due to two reasons — the maximum number of cutbacks allowed
by the user can be exceeded, or the minimum time step can be reached. In this case, the analysis terminates with exit 3002
and exit 3015, respectively. These premature terminations can be avoided by using the option to continue the analysis even
if physical criteria are not satisfied. If this flag is set on the AUTO STEP option, and either the maximum number of user-
allowed cutbacks or the minimum time step is reached, a mechanical analysis moves on to the next increment if it is
otherwise converged (see Convergence Controls in this chapter) or continues to recycle and scales back based solely on the
recycling criterion. Setting this flag for a thermal analysis simply allows it to move on to the next increment.
User Programmed Time Steps Selection

In addition to controlling the time step through the recycle based criterion and the physical criteria, direct control of the
time step is possible through the use of the UTIMESTEP user subroutine. In this case, the new time step that is determined by
the auto step algorithm enters the program as input and the modified time step by the user is returned as output. More
details are provided in Marc Volume D: User Subroutines and Special Routines, Chapter 2.
Post Files Output

In many analyses it is convenient to obtain post file results at specified time intervals. This is naturally obtained with a fixed
load stepping scheme but not with an automatic scheme. Traditionally, the post output frequency is given as every nth
increment. With the AUTO STEP history definition option, you can request post output to be obtained at equally spaced time
intervals. In this case, the time step is temporarily modified to exactly reach the time for output. The time step is then
restored in the following increment.
Table Considerations
When tables are used in the new input format to specify loads with complex time variations, in most cases, it is important
that the exact peaks and valleys of the loading history are not missed due to the adaptive time stepping. By default, AUTO
STEP adjusts the time step temporarily so that the peaks and valleys of the loading history are reached exactly. The time step
is then restored in the following increment. In analyses where the boundary conditions are obtained from experimental data
containing many points, it is often useful to turn this behavior off, in which case large time steps will be achieved and the
load will be smoothed out.
Quasi-Static Damping Scheme

The AUTO STEP option also has an optional artificial damping feature available for mechanical statics analyses.
Damping Energy
A comparison between the incremental strain energy and the estimated damping energy is used as a criteria for time step
control when either a 4 or 5 is given on the 10th field of the 2nd data block (idampit flag).The criteria based upon the
number of cycles is bypassed. Decisions to decrease and increase the time step are based on the damping energy rate of the
system. Furthermore, if a 4 is entered, then artificial damping is added to the system. The details of the time stepping
procedure and the damping procedure are as follows:
CHAPTER 11 931
A damping factor, F d , is introduced, which at the start of the loadcase, is set to 0. The time step for the first increment is
set equal to the user given initial time step. During the first cycle of the loadcase, a “small” strain energy value E sm a l l is
calculated as:
E small =  0.5D  1 1   1  2 dV
V
where D  1 1  is the first component of the stress-strain relation at the integration point and   1  is 10-6. This small
energy value is used later to distinguish between real deformations of the structure and zero-energy modes.
During the assembly of the stiffness matrix K and the right-hand side vector F , the following contributions are added to
K and F , respectively.
K d am p = F d M  t
F da mp = F d Mdu  t
where M is the lumped mass matrix, du is the incremental displacement and timinc is the time increment. Note that
when F d = 0 , there is no change to the equations. During the recovery phase, the incremental damping energy for the
nth iteration is calculated as:
n
E d am p = du T Mdu  t
when F d = 0 and as:
n
E d am p = du T F d Mdu  t
when F d is not 0 .
The check to determine if the time-step should be reduced through a cut-back is made as follows: If F d = 0 , perform a
cutback if:
n 1
E  E s mall and E d m a p  4E d am p
where E is the incremental strain energy. The time step reduction factor in this case is given by:
1 n
f re du c = E damp  E d a m p
Such a cutback is triggered in the first increment of the loadcase in the case of severe non-linearities, as expressed by the
fact that the damping energy for the nth iteration deviates considerably from that of the first iteration.
If F d is not 0 , perform a cutback if:
n n
E da mp  4E predict e d and E d am p  E m a x and tol  2
where E predicted is the predicted damping energy for the current increment, E m ax is the maximum damping energy
in any previously converged increment and tol is the global user-defined convergence tolerance. The reduction factor in
this case is set to 0.5. This cut-back is triggered in the case where the solution is clearly diverging and avoids unnecessary
recycles. The same reduction factor (0.5) is also used for cases when the maximum number of iterations are reached or when
user criteria are violated.
For the first increment of the loadcase, the calculation of F d and predicted energy is as follows:
 Estimated strain energy for the loadcase E e s t i m a t ed = E timinc 2
 E max = max  E e s t i m a t ed ,E  , where E = total strain energy in system .
n
 Estimated damping energy for the loadcase E d am p _ es t i m at e d = E d am p totinc  timinc , where
totinc = total time period .
 F d = E max  E d am p _ e s t i m at e d where  is a user-defined factor. Recommended default value is 10-4.
n
 Predicted damping energy for next increment E p r ed i c t ed = F d E d am p timincnew  timincold
where timincnew is the time step for the next increment and timincold is the time increment for the previous
increment.
n n
 Set E prev = F d E d am p  timinc and E m ax = F d E d am p
For subsequent increments of the loadcase, the modification of F d and the time step is as follows:
 If the total strain energy E is larger than twice the estimated strain energy E es t i m at e d , then
Fd = Fd E  Ee s t i m a t e d .
CHAPTER 11 933
n
 If E damp  timinc  1.1E p r e v , then increase the time step by a factor of 1.5.
Avoiding Exit 3015

When the minimum time step t m i n is reached, Marc normally exits with 3015. If the quasi-static damping schemes (10th
field, 2nd data block) is a 0 or 1, a quasi-static damping option is triggered in an attempt to avoid this premature exit:
When exit 3015 is encountered for the first time, the increment is repeated with a new time step given by
t ne w = fclrg t min , where fclrg is the maximum ratio between steps (4th field of the 2nd data block of the AUTO STEP
option = 10 by default). The increment is further stabilized by added the factored lumped mass matrix to the stiffness
matrix and modifying the force vector consistently. The damping factor associated with the lumped mass matrix is again
based on the ratio of the estimated strain energy to the estimated damping energy in a manner similar to the idampit = 4
scheme. If exit 3015 is again encountered, a second attempt is made with t n ew = fclrg2 t m i n and again stabilizing the
incremental solution by adding the quasi-static damping in the solution. It should be noted that this procedure is not
available for the first increment of the auto step loadcase since the damping factor is only calculated after the first increment
is completed. It should also be noted that once exit 3015 is encountered and damping is turned on, it remains on for the
rest of the loadcase.
Auto Step for Transient Dynamics

The AUTO STEP algorithm is further modified for transient dynamics problems:
 When penetration is detected in dynamic contact problems, instead of using the default iterative penetration
procedure, a time cutback is made and the increment is repeated with a smaller time step that avoids the
penetration. This scheme allows for momentum conservation without spurious penetration induced oscillations in
the response. If no other cutbacks are made, the time step is restored in the following increment.
 Additional, optimal checks are available in transient dynamic problems to control the time step since larger time
steps that may have been assessed based on the recycle based criterion or physical criteria can give rise to
unacceptable time integration errors. The time integration error check is only turned on if the 13th field of the 3rd
data block of the AUTO STEP option is set to 1. If this field is not present or set to 0, the time integration error check
is skipped and the time stepping algorithm for dynamic problems is then similar to that of static problems. If the
check is turned on, an additional check is made at the end of each increment to see if the time step needs to be
reduced for the following increment. This check is only made for the Newmark-Beta (NB) and the Single Step
Houbold (SSH) methods. The scheme that is followed is a modified version of the scheme outlined by Bergan and
Mollestad [Ref. 1]. The time step for increment n + 1 should satisfy the inequality
u nT Mu n
 t  n + 1  2 ---------------------------------
u nT K T u n
where u n is the incremental displacement vector at increment n ; M is the mass matrix; K T is the tangent
stiffness matrix, and  is a user-customizable scaling factor. The default value of  is as follows:
for the SSH operator  = 0.375 * fcsml . and
for the NB operator  = 0.75 * fcsml ,
where fcsml is the smallest ratio between time steps (3rd field of the 2nd data block of the AUTO STEP option =
0.1 by default). The smaller default value of  = 0.0375 for SSH problems compared to  = 0.075 for NB
problems is to better control the artificial damping inherent in the SSH method. The user can control the
accuracy of the solution by changing fcsml . The time step is only reduced if the value predicted by the above
equation is less than 67% of the current time step. The check is bypassed if t n + 1 is already at t m i n , if the strain
energy is negligible (for example, rigid body motion). When there are multiple loadcases in a transient dynamic
run and the time integration error check is flagged for all of them, the initial time step of a loadcase that follows a
transient dynamic loadcase is also adjusted if it is too high compared to the time step predicted by the error check.
This avoids accuracy problems associated with using an initial user-prescribed time step that is too large.
The defaults of the AUTO STEP option are carefully chosen to be adequate in a wide variety of applications. There are cases,
however, when the settings may need to be modified. Assume that the default settings are used, which means that the recycle
based control is active with an initial load of one per cent of the total. If the structure is weakly nonlinear, convergence is
obtained in just a few recycles and the time steps for successive increments get progressively larger. This can lead to
problems if the initially weakly nonlinear structure suddenly exhibits stronger nonlinearities; for instance, occurrence of
plasticity or parts coming into contact. Possible remedies to this problem include:
a. decrease the time step scale factor to a smaller number so the step size does not grow so rapidly;
b. use a physical criterion like maximum increment of displacements to limit the load step;
c. use the maximum time step to limit large steps;
d. decrease the desired and maximum number of recycles to decrease the load step if more recycles are needed.
Another situation is if the structure is highly nonlinear and convergence is slow. In this case, it may be necessary to increase
the desired number and maximum number of recycles. In general, there is a close connection between the convergence
tolerances used and the desired number and maximum number of recycles. In many cases, it may be beneficial to use one
or more physical criteria; for example, the increment of plastic strain as targets for controlling the load step. This can easily
be achieved by allowing the program to add automatic physical criteria where appropriate. This is especially a good idea if
the ‘proceed if not converged’ option is used or if the ‘non-positive definite flag’ is set since the added physical criteria then
serve as controls to limit the time step and produce a realistic numerical solution in each increment rather than letting the
solution proceed unchecked with unrealistic results.
Residual Load Correction

The residual load is applied as a correcting force to ensure that equilibrium is maintained and, hence, that an accurate
solution is obtained for nonlinear problems.
The residual load correction enforces global equilibrium at the start of each new increment. This prevents the accumulation
of out-of-equilibrium forces from increment to increment and makes the solution less sensitive to the step size. Figure 11-2
shows how stiffness is based on the state at the start of a step. The variables are defined below for increments i = 1,2,3:
 F i applied forces for i = 1,2,3
 u i calculated displacements for i = 1,2,3
 R i residual loads for i = 1,2,3

CHAPTER 11 935
F3
F2
R3
F1 R2
R1
U1 U2 U3
Figure 11-2 Stiffness Based on State at Start of Step
The residual load correction is the difference between the internal forces and the externally applied loads. The residual load
correction is expressed as
R = P –   T dV (11-2)
where  is the differential operator which transforms displacements to strains,  is the current generalized stresses, P is
the total applied load vector, and R is the residual load correction.
In order to evaluate the residual load correction accurately, evaluate the integral by summing the contributions from all
integration points. The residual load correction feature requires that stresses be stored at all the integration points. Data
storage at all integration points is the default in Marc, but can be overridden in linear analysis by use of the CENTROID
parameter.
Residual load correction should always be used unless the ELASTIC parameter is invoked; this is the default.
Restarting the Analysis

The RESTART model definition option creates a restart file for the current analysis which can be used in subsequent analyses.
It can also be used to read in a previously generated file to continue the analysis. The RESTART option is very important for
any multi-increment analysis because it allows you to continue the analysis at a later time. The default situation writes the
restart information to unit 8 and reads a previously generated file from unit 9. For postprocessing, the RESTART option can
be used to plot or combine loadcases (see CASE COMBIN). Upon restart, you can use the model definition REAUTO option to
redefine parameters associated with an automatic load sequence.
To save storage space, it is not necessary to store each increment of analysis. The frequency can be set using the RESTART
option, and subsequently modified using the RESTART INCREMENT option. It is also possible to store only the last converged
solution by using the RESTART LAST option. The input data describing the problem is not saved, and therefore must be read
in with each restart. This should include the model geometry including CONNECTIVITY and the COORDINATES material
properties contact data and boundary conditions. Additional tables may be added, or they can have additional data added.
The RESTART option specifies restart parameters; for example, input/output files, restart increment, and intervals at which
restarts are to be written.
Newton-Raphson Method
The Newton-Raphson method can be used to solve the nonlinear equilibrium equations in structural analysis by
considering the following set of equations:
K  u u = F – R  u  (11-3)
where u is the nodal-displacement vector, F is the external nodal-load vector, R is the internal nodal-load vector
(following from the internal stresses), and K is the tangent-stiffness matrix. The internal nodal-load vector is obtained from
the internal stresses as
T
R =    dv (11-4)
elem V
In this set of equations, both R and K are functions of u . In many cases, F is also a function of u (for example, if F
follows from pressure loads, the nodal load vector is a function of the orientation of the structure). The equations suggest
that use of the Newton-Raphson method is appropriate. Suppose that the last obtained approximate solution is termed
i i
u , where indicates the iteration number. Equation (11-3) can then be written as
i–1 i–1
K  u n + 1 u = F – R  u n + 1  (11-5)
i
This equation is solved for u and the next appropriate solution is obtained by
i i–1 i i i
u = u + u and u n + 1 = u n + u (11-6)
Solution of this equation completes one iteration, and the process can be repeated. The subscript n denotes the increment
number representing the state t = n . Unless stated otherwise, the subscript n + 1 is dropped with all quantities referring
to the current state.
The Newton-Raphson method is the default in Marc (see Figure 11-3). The Newton-Raphson method provides good results
for most nonlinear problems, but is expensive for large, three-dimensional problems, when the direct solver is used. The
computational problem is less significant when the iterative solvers are used. Figure 11-3 illustrates the graphical
interpretation of the Newton-Raphson iteration technique in one dimension to find the roots of the function
Fu – 1 = u – 1 = 0 , starting from increment 1 where F  u 0  = 0.2 to increment 2 where.
F  u la st  = 1.0 .The iteration process stops when the convergence criteria are satisfied. Similarly, Figure 11-4 graphically
demonstrates the modified Newton-Raphson iteration technique for the same one dimensional problem.
CHAPTER 11 937
Newton-Raphson Method
1.2
F = u1/2 Convergence Criteria:
1 Inc 2
Residual Checking
Residual 1.001
|| Fr esidua l ||
0.8 1
---------------------------- < TOL 1
8
1
0.999
4 || Fr ea ction ||
Force
8
0.998
0.997
0.996
0.6 0.995
3
0.992 0.993 0.994 0.995 0.996 0.997 0.998 0.999 1 1.001
Displacement Checking
0.4 || δu ||
--------------- < TO L1
8
Δu1 δu2 δu3
|| Δu ||
8
Recycle 1 2 3
0.2 Inc 1
Displacement (u)
0
0 0.5 1 1.5
Figure 11-3 Newton-Raphson
Modified Newton-Raphson Method

The modified Newton-Raphson method is similar to the Newton-Raphson method, but does not reassemble the stiffness
matrix in each iteration.
0 i i–1
K  u u = F – R  u  (11-7)
Modified Newton-Raphson Method

1.2
1/2
F=u Convergence Criteria:
1 Inc 2
Residual Checking
0.8
Residual
1
|| Fr esidua l ||
---------------------------- < TOL 1
1.02
8
1.01
|| Fr ea ction ||
Force
8
1
0.99
0.6 0.98
0.97
0.97 0.98 0.99 1 1.01 1.02 1.03 Displacement Checking
0.4 3 || δu ||
δu
--------------- < TO L1
8
Δu1 δu2
|| Δu ||
8
Recycle 1 2 3
0.2 Inc 1
u0
Displacement (u)
0
0 0.5 1 1.5
Figure 11-4 Modified Newton-Raphson
The process is computationally inexpensive because the tangent stiffness matrix is formed and decomposed once. From
then on, each iteration requires only forming the right-hand side and a backward substitution in the solution process.
However, the convergence is only linear, and the potential for a very large number of iterations, or even nonconvergence,
is quite high.
If contact or sudden material nonlinearities occur, reassembly cannot be avoided. The modified Newton-Raphson method
is effective for large-scale, only mildly nonlinear problems. When the iterative solver is employed, simple back substitution
is not possible, making this process ineffective. In such cases, the Newton-Raphson method should be used instead.
If the load is applied incrementally, Marc recalculates the stiffness matrix at the start of each increment or at selected
increments, as specified.
Strain Correction Method

The strain correction method is a variant of the Newton method. This method uses a linearized strain calculation, with the
nonlinear portion of the strain increment applied as an initial strain increment in subsequent iterations and recycles. This
method is appropriate for shell and beam problems in which rotations are large, but membrane stresses are small.
In such cases, rotation increments are usually much larger than the strain increments, and, hence, the nonlinear terms can
i+1 i i
dominate the linear terms. After each displacement update, the new strains E   are calculated from u and u  =  u 
which yield
i+1 i 1 i
E  = E  + ---   u   + u    + u   u   + u   u i  + u   u   (11-8)
2
This expression is linear except for the last term. Since the iteration procedures start with a fully linearized calculation of
the displacement increments, the nonlinear contributions yield strain increments inconsistent with the calculated
displacement increments in the first iteration. These errors give rise to either incorrect plasticity calculations (when using
small strain plasticity method), or, in the case of elastic material behavior, yields erroneous stresses. These stresses, in their
turn, have a dominant effect on the stiffness matrix for subsequent iterations or increments, which then causes the relatively
poor performance.
The remedy to this problem is simple and effective. The linear and nonlinear part of the strain increments are calculated
separately and only the linear part of
l i 1 i i
 E   = E  + ---  u   + u    + u   u   + u   u   (11-9)
2
is used for calculation of the stresses. The nonlinear part
nl i + 1 1
 E   = --- u   u   (11-10)
2
is used as an “initial strain” in the next iteration or increment, which contributes to the residual load vector defined by
C    nl
R =     X   L E  dV (11-11)
V0
This “strain correction” term is defined by

i i C
K  u n + 1 u = F – R  u n + 1  – R (11-12)
Since the displacement and strain increments are now calculated in a consistent way, the plasticity and/or equilibrium errors
are greatly reduced. The performance of the strain correction method is not as good if the displacement increments are
(almost) completely prescribed, which is not usually the case. Finally, note that the strain correction method can be
considered as a Newton method in which a different stiffness matrix is used.
CHAPTER 11 939
Direct Substitution
In the Eulerian formulation (R-P FLOW parameter), the governing equation of the system can be expressed as:
Kv = F (11-13)
where v is a velocity vector, and F is a force vector. This equation is very nonlinear because K is a nonlinear function of
i
v . By default, a direct substitution method is used to solve the problem. If v is the velocity at iteration i , the result of
iteration i + 1 is
i i+1
Kv  v = F (11-14)
If this method does not converge in 10 iterations, it is possible to switch into a Newton-Raphson method.
Arc-length Methods
The solution methods described above involve an iterative process to achieve equilibrium for a fixed increment of load.
Besides, none of them have the ability to deal with problems involving snap-through and snap-back behavior. An
equilibrium path as shown in Figure 11-5 displays the features possibly involved.
Snap-through Behavior
Load (F)
um Path
uilibri
r : Eq
6
2
5 Displacement (u)
Figure 11-5 Snap-through Behavior
The issue at hand is the existence of multiple displacement vectors, u , for a given applied force vector, F . The arc-length
methods provide the means to ensure that the correct displacement vector is found by Marc. If you have a load controlled
problem, the solution tends to jump from point 2 to 6 whenever the load increment after 2 is applied. If you have a
displacement controlled problem, the solution tends to jump from 3 to 5 whenever the displacement increment after 3 is
applied. Note that these problems appear essentially in quasi-static analyses. In dynamic analyses, the inertia forces help
determine equilibrium in a snap-through problem.
Thus, in a quasi-static analysis sometimes it is impossible to find a converged solution for a particular load (or displacement
increment):
 n + 1 F –  n F = F
This is illustrated in Figure 11-5 where both the phenomenon of snap-through (going from point 2 to 3) and snap-back
(going from point 3 to 4) require a solution procedure which can handle these problems without going back along the same
equilibrium curve.
As shown in Figure 11-6, assume that the solution is known at point A for load level  n F . For arriving at point B on the
equilibrium curve, you either reduce the step size or adapt the load level in the iteration process. To achieve this end, the
equilibrium equations are augmented with a constraint equation expressed typically as the norm of incremental
displacements. Hence, this allows the load level to change from iteration to iteration until equilibrium is found.
g : Co Arc- Length Procedure

nstr
ain
λn+1 F t Su
r fa
λn+2 F ce r : Equilibri
λn+3 F um
Path
B
Load (F)
th
rc leng
l: a
λn F
A
Displacement (u)
Figure 11-6 Intersection of Equilibrium Path with Constraining Spherical Surface
The augmented equation, c  u   , describes the intersection of the equilibrium curve with an auxiliary surface g for a
particular size of the path parameter  :
r  u   = F – R  u  = 0 c  u   = g  u   –  = 0 (11-15)
Variations of the parameter 
moves the surface whose intersection with the equilibrium curve r generates a sequence of points along the curve. The
distance between two intersection points, denoted with  0 and  , denoted by l is the so-called arc-length.
Linearization of Equation (11-15) around point A in Figure 11-6 yields:
K P  u   –r 
T   =  –r  (11-16)
n n 0     0
CHAPTER 11 941
where:
r r
K = ------ : P = ------ (11-17)
u 
T c c
n = ------ : n 0 = ------ (11-18)
u 
r = F – R (11-19)
r 0 = g  u   –  (11-20)
It can be noted that a standard Newton-Raphson solution procedure is obtained if the constraint condition is not imposed.
The use of the constraint equation causes a loss of the banded system of equations which would have been obtained if only
the K matrix was used. Instead of solving the N + 1 set of equations iteratively, the block elimination process is applied.
i–1
Consider the residual at iteration i to which the fraction of load level  corresponds:
i i–1 i–1 i i–1
r   =  F – R u  (11-21)
i
The residual for some variation of load level,  , becomes:
i i–1 i i i i–1
r  +   =  F + r    (11-22)
which can be written as:
i i–1 i i i–1 i i
u   +   = u    +  u * (11-23)
i i–1 i –1
where u    = K  r (11-24)
i i –1
and u * =  K  F (11-25)
i
Notice that u * does not depend on the load level. The equation above essentially establishes the influence of a change
i
in the load level  during one iteration on the change in displacement increment for that iteration. After one iteration
is solved, this equation is used to determine the change in the load level such that the constraint is followed. There are
several arc-length methods corresponding to different constraints.
Among them, the most well-known arc-length method is one proposed by Crisfield, in which the iterative solution in
displacement space follows a spherical path centered around the beginning of the increment. This requirement is translated
in the formula:
2 i i
c = l = u u (11-26)
where l is the arc length. The above equation with the help of Equations (11-25) and (11-26) is applied as:
i T i i 2 i–1 i i–1 T i i
  u *  u *     +  2  u + u     u *     +
(11-27)
i–1 i i–1 T i–1 i i–1 2
  u + u      u + u    – l  = 0
1
The equation above is interpreted with i = 1 and u = 0 in the prediction phase while retaining the full form of
Equation (11-27) in the correction phase. Two solutions for  are available. There are several methods to determine which
root to select. By default, we choose the one that maintains a positive angle of the displacement increment from one
iteration to the next.
i i
The two roots of this scalar equation are    1 and    2 . To avoid going back on the original load-deflection curve,
i–1 i
the angle between the incremental displacement vectors, u and u (before and after the current iteration,
i i i i
respectively) should be positive. Two alternative values of u (namely,  u  1 and  u  2 corresponding to    1 and
i
   2 are obtained and the cosine of two corresponding angles (  1 and  2 ) are given by
i T i–1
  u n + 1  1  u n + 1
cos  1 = ----------------------------------------------------
- (11-28)
l
i T i–1
  u n + 1  2  u n + 1
and cos  2 = ----------------------------------------------------
- (11-29)
l
0
Once again, the prediction phase is interpreted with i = 1 and u n + 1 = u n , while Equations (11-28) and (11-29)
retain their full form in the correction phase.
i i
As mentioned earlier, the appropriate root,    1 or    2 is that which gives a positive cos  . In case both the angles
are positive, the appropriate root is the one closest to the linear solution given as:
i–1 i i–1 i 2
i  u + u   u + u  – l
 = -------------------------------------------------------------------------------------- (11-30)
i–1 i i
2  u + u u *
Crisfield’s solution procedure, generalized to an automatic load incrementation process, has been implemented in Marc as
one of the options under the AUTO INCREMENT history definition option. Various components of this process are shown in
Figure 11-7.
CHAPTER 11 943
λ1 F
Load (F)
F2 = λ2 F
r1
B
δu2(λ1) = K2-1(λ1F - r1)
Displacement (u)
Figure 11-7 Crisfield’s Constant Arc Length
The constraints in Equations (11-26) and (11-27) are imposed at every iteration. Disadvantage of the quadratic equation
suggested by Crisfield is the introduction of an equation with two roots and thus the need for an extra equation to solve
1 1
the system for the calculated roots if two real roots exists. This situation arises when the contribution u (or u ) is very
large in comparison to the arc-length. This can be avoided in most cases by setting sufficiently small values of the error
tolerance on the residual force. In case the above situation still persists despite the reduction of error tolerance, Marc has
two options to proceed:
a. To attempt to continue the analysis with the load increment used in the initial step of auto increment process.
b. Use the increment resulting from the linear constraint for the load.
There are two alternate methods using the Crisfield method for selecting the roots:
 Singularity ratio method; where if the system of equations is positive definite, the largest root is used; while if the
system was negative, the smallest root is used.
 Falzon’s method described in Reference 7 is applicable for problems which have multiple nearly equal buckling
modes.
This is circumvented in Ramm’s procedure due to the linearization.
Another approach to impose the constraint is due to Ramm, who also makes use of a quadratic equation to impose the
constraint giving rise to the Riks-Ramm method. The difference is that while Crisfield imposes the constraint as a quadratic
equation, Ramm linearized the constraint.
Geometrically, the difference between the two methods is that the Crisfield method enforces the correction on the curve
of the augmented equation introducing no residual for the augmented equation. Ramm takes the intersection between the
linearizations of the curves which gives a residual of the augmented equation for the next step. Both methods converge to
the same solution, the intersection of the two curves, unless approximations are made.
The Riks-Ramm constraint is linear, in that:
2
c = l = u n u n + 1
which results in a linear equation for  :

T i i 2
u n  u +  u *  = l
Thus, the load parameter predictor is calculated as:

T i –1 i
1 u n l –  u n    K  r 
 n + 1 = ---------------------------------------------------------------------- (11-31)
T 1
u n  u * 
while during the corrector phase it is:

i T i –1 i
i  u n + 1    K  r 
 n + 1 = – ----------------------------------------------------
i
(11-32)
i T
 u n + 1   u * 
It is noted that in the definition of the constraint, the normalized displacement of the previous step is used for the normal
c
to the auxiliary surface ------ = n . Thus, problems can arise if the step size is too big. In situations with sharp curvatures in
u
the solution path, the normal to the prediction may not find intersections with the equilibrium curve. Note that the norm
of the displacement increment during the iterations is not constant in Riks-Ramm method.
In contact problems, sudden changes of the stiffness can be present (due to two bodies which are initially not in contact
suddenly make contact). Hence, a potential problem exists in the Riks-Ramm method if the inner-product of the
i
displacement due to the load vector u * and the displacement increment u n is small. This could result in a very large
value of the load increment for which convergence in the subsequent iterations is difficult to achieve. Therefore, a modified
predictor can be used resulting in a modified Riks-Ramm procedure as:
1 1 T 1
1 l n – 1 u –  u *  u
 *
= -----------------------------------------------------------------
- (11-33)
1 1
 u * u *
where,
T i
u n u *
 = -------------------------
- (11-34)
T i
u n u *
This method effectively scales the load increment to be applied in the prediction and is found to be effective for contact
problems.
Refinements and Controls

The success of the methods depend on the suitable choice of the arc-length:
2
C = l
CHAPTER 11 945
The initial value of the arc-length is calculated from the initial fraction  of the load specified by you in the following
fashion:
Ku = F – R (11-35)
2
l ini = u (11-36)
In subsequent steps the arc-length can be reduced or increased at the start of a new load step depending on the number of
iterations I 0 in the previous step. This number of iterations in compared with the desired number of iterations I d which
is typically set to 3 or 5. The new arc-length is then given by:
rt = I d  I o (11-37)
2 2
l new = rt  l prev
Often, this results in too large of an arc-length growth, especially initially when the system is nearly linear and I o is small
(often one). In this case, one can gradually grow the value as:
l ne w = sf  l prev
sf = .1 if rt  .1
sf = 1 if .1  rt  1
sf = 1.25 if 1  rt  3
sf = 1.5 if 3  rt
Two control parameters exist to limit the maximum enlargement or the minimum reduction in the arc-length.
2
l
min  -------  max (11-38)
2
l ini
The min and max defined here are the 7th and 5th field of the 2nd data block. It is also possible that the maximum fraction
of the load allowed can vary during the loadcase. This is often useful if one knows that one is approaching the limit load.
In addition, the maximum fraction of the total load may be defined. In general, control on the limiting values with respect
to the arc-length multiplier is preferred in comparison with the maximum fraction of the load to be applied in the iteration
since a solution is sought for a particular value of the arc-length.
Also, attention must be paid to the following:
1. In order to tract snap-through problems, the method of allowing solution if the stiffness matrix becomes
nonpositive needs to be set.
2. The maximum number of iterations must be set larger than the desired number of iterations.
The default procedure for convergence criterion in Marc is based on the magnitude of the maximum residual load
compared to the maximum reaction force. This method is appropriate since the residuals measure the out-of-equilibrium
force, which should be minimized. This technique is also appropriate for Newton methods, where zero load iterations
reduce the residual load. The method has the additional benefit that convergence can be satisfied without iterations.
The basic procedures are outlined below.
1. RESIDUAL CHECKING
F resid ual 
-------------------------------  TOL 1 (11-39)
F reaction 
F resid ual  M r e s i du a l 
-------------------------------  TOL 1 and ---------------------------------  TOL 2 (11-40)
F reaction  M r e a c t i on 
F resid ual 
 TOL 1 (11-41)
F resid ual 
 TOL 1 and M r e s i du a l 
 TOL 2 (11-42)
Where F is the force vector, and M is the moment vector. TOL 1 and TOL 2 are control tolerances. F 
indicates the component of F with the highest absolute value.
It should be noted that reaction forces can be related to kinematic boundary conditions or to contact constraints.
If a node has a kinematic boundary condition on a displacement degree of freedom, then the corresponding
component in the force vector is considered as a candidate for the maximum reaction force and is skipped in the
residual calculation procedure. If a node has a contact constraint, then the treatment of the node for convergence
checking depends on the contact algorithm used. If the node-to-segment algorithm is used, the contact force
component is considered as a candidate for the maximum reaction force, but the node in contact is entirely skipped
in the residual calculation procedure. If the segment-to-segment algorithm is used, the contact force component is
considered as a candidate for the maximum reaction force and the unbalanced force components are also considered
as candidates for the maximum residual force.
Residual checking has two drawbacks. First, if the CENTROID parameter is used, the residuals and reactions are not
calculated accurately. Second, in some special problems, such as free thermal expansion, there are no reaction forces.
The program uses displacement checking in either of these cases.
2. DISPLACEMENT CHECKING
u 
---------------
-  TOL 1 (11-43)
u 
u  
-  TOL 1 and ----------------  TOL 2
--------------- (11-44)
u   
u   TOL 1 (11-45)
CHAPTER 11 947
u   TOL 1 and    TOL 2 (11-46)
where u is the displacement increment vector, u is the correction to incremental displacement vector,  is
the correction to incremental rotation vector, and  is the rotation iteration vector. With this method,
convergence is satisfied if the maximum displacement of the last iteration is small compared to the actual
displacement change of the increment. A disadvantage of this approach is that it results in at least one iteration,
regardless of the accuracy of the solution.
i Correction to incremental displacements of ith iteration

F un Displacements at increment n
0 i

1 ---------------- Tolerance
k i
 j
0 j = 0
k+1
u u u
n+1 n+1
Figure 11-8 Displacement Control
3. STRAIN ENERGY CHECKING

This is similar to displacement testing where a comparison is made between the strain energy of the latest iteration
and the strain energy of the increment. With this method, the entire model is checked.
E
-------  TOL 1 (11-47)
E
where E is the strain energy of the increment and E is the correction to incremental strain energy of the
iteration. These energies are the total energies, integrated over the whole volume. A disadvantage of this approach
is that it results in at least one iteration, regardless of the accuracy of the solution. The advantage of this method is
that it evaluates the global accuracy as opposed to the local accuracy associated with a single node.
Different problems require different schemes to detect the convergence efficiently and accurately. To do this, the
following combinations of residual checking and displacement checking are also available.
4. RESIDUAL OR DISPLACEMENT CHECKING
This procedure does convergence checking on both residuals (Procedure 1) and displacements (Procedure 2).
Convergence is obtained if one converges.
5. RESIDUAL AND DISPLACEMENT CHECKING
This procedure does a convergence check on both residuals and displacements (Procedure 4). Convergence is
achieved if both criteria converge simultaneously.
In several types of analyses, maximum reactions or displacements are extremely small (even close to the round-off errors of
computers). In such circumstances, not all types of relative convergence criteria may work properly. For example, in a
problem with stress-free motion, the convergence check based on relative displacement increments works correctly but not
the convergence check based on relative residual or strain energy. In this situation, it is necessary to check the convergence
with absolute values of reactions or strain energy; otherwise, the analysis may terminate prematurely. Similarly, this kind of
situation may happen for problems with springback and free thermal expansion or constraint thermal expansion. The
details for the cases where convergence checking with relative values may encounter difficulties are listed in Table 11-2. The
AUTO SWITCH option is designed for Marc to switch to the proper convergence check scheme automatically if any of the
situations mentioned above occur during the analysis. The AUTO SWITCH option allows Marc to automatically switch the
convergence check scheme to check as required on either residuals or displacements if small reactions or displacements are
detected, or to use the absolute strain energy checking if necessary.
Table 11-2 Effectiveness of various Relative Tolerance Convergence Testing Criterion

Convergence Variable
Displacement/ Residual Strain Energy
Analysis Type Rotation Force/Torque
Stress-free motion Yes No No
Springback No Yes No
Free Thermal Expansion Yes No No
Constraint Thermal Expansion No Yes Yes
Yes – relative tolerance testing works.
No – relative tolerance testing does not work.
If AUTO SWITCH is turned on, it:

1. Switches on the relative residual checking if the relative displacement criterion is used (which fails when the
maximum incremental displacement becomes very small
Max._incremental_displacement/Smallest_element_size<1.0e-6)
2. Switches on the relative displacement checking if the relative residual force (moment) criterion is used (which fails
when the maximum reaction force becomes very small <1.0e-8)
3. To switch on the absolute energy checking if the structure is free of stress and deformation (strain energy
density < 1.0e-15).
Notes: 1. If any criterion with absolute value testing is being used as convergence checking, this feature
is deactivated.
2. If the combined convergence check scheme with relative displacements or residuals is chosen
as convergence criterion, it is not necessary to turn the AUTO SWITCH on.
3. Once the extreme cases disappear in the analysis, Marc recovers to use the convergence criteria
given by user.
CHAPTER 11 955
Equilibrium per node

The default residual convergence criterion may be easily fulfilled in models where the maximum reaction force is on the
retained node of a tying, or in case of axi-symmetric problems, where the maximum reaction force can be large if the
corresponding node is at a large radius. In addition, the element size at the location of boundary condition has an impact
on the magnitude of the reaction force. The option “Equilibrium per node” has been introduced to minimize the effect of
large reaction forces and to ensure a better, converged solution.
To check the nodal equilibrium, the residual force at a node is compared with the maximum of various force contributions
to that node. Contributions to nodal forces can be due to internal element forces, external loads, foundation forces, spring
forces, contact forces, tying forces, inertia forces, damping forces, DMIG forces and reaction forces. The equilibrium check
is considered as satisfied if the following residual convergence ratio is less than the user-defined tolerance at all the nodes
in the model:
residual force at a node
residual convergence ratio = --------------------------------------------------------------------------------------------------------------------------------------------------
maximum of internal or external or contact force at a node
Nodes with very small residual force and internal/external force are excluded from the nodal equilibrium check. In case of
Relative/Absolute checking, the “Minimum reaction force cut-off” is used to exclude nodes from the equilibrium check.
Three levels of nodal equilibrium check are available:
1. Mild: In this case, a default global convergence check and a nodal equilibrium check are performed. If the global
convergence criterion is satisfied, but the nodal equilibrium check fails, then an additional Newton Raphson
iteration is performed to check if nodal equilibrium is satisfied. Even if the latter is not satisfied, Marc proceeds to
the next increment.
2. Standard: In this case, nodal convergence ratios at all nodes are checked against the user defined convergence
tolerance. If a node does not achieve equilibrium even after 3 recycles, then an average check across the vicinity of
the node is performed. The average ratio of residuals to maximum nodal forces for all the neighboring nodes is
computed and compared against the user defined convergence tolerance. Additionally, a small fraction of nodes is
allowed to fail the equilibrium check. The Mentat default is 0.5% but this allowable fraction of non-equilibrated
nodes can be reset by the user.
3. Strong: In this case, the nodal equilibrium check is stringent and the convergence ratios at all the nodes are checked
against the user defined convergence tolerance. By default, 0.1% of the nodes are allowed to fail the equilibrium
check and can be reset by the user, if needed. The increment is considered to be converged only if the global residual
check (max residual in model / max reaction in model) and the nodal equilibrium check pass.
While using this option, the following nodal post codes can be used to better understand the Newton-Raphson
convergence process:
1. Nodal Force Convergence Ratio (nodal post code 103).
2. Nodal Moment Convergence Ratio (nodal post code 104).
3. Nodal Convergence Status (nodal post code 105).
Also, Residual force (nodal post code 101) and Residual moment (nodal post code 102) which are available for a
mechanical pass, can be utilized to see the unbalanced force at a node.
Singularity Ratio
The singularity ratio, R , is a measure of the conditioning of the system of linear equations. R is related to the conditioning
number, C , which is defined as the ratio between the highest and lowest eigenvalues in the system. The singularity ratio is
an upper bound for the inverse of the matrix conditioning number.
1RC (11-48)
C and R establish the growth of errors in the solution process. If the errors on the right-hand side of the equation are less
than E prior to the solution, the errors in the solution will be less than  , with
  CE (11-49)
The singularity ratio is a measure that is computed during the Crout elimination process of Marc using the direct solver.
In this process, a recursive algorithm redefines the diagonal terms
k–1
k k – 1
K kk = K kk –  Km k Km k 1 ik–1 (11-50)
m = i
where i is a function of the matrix profile. K k k is a diagonal of the kth degree of freedom. The singularity ratio is defined as
k k – 1
R = min K kk  K kk (11-51)
k k – 1
If all K kk and K kk are positive, the singularity ratio indicates loss of accuracy during the Crout elimination process.
This loss of accuracy occurs for all positive definite matrices. The number of digits lost during the elimination process is
approximately equal to
n l o st = – log 10 R (11-52)
The singularity ratio also indicates the presence of rigid body modes in the structure. In that case, the elimination process
k
produces zeros on the diagonal K k k  0 . Exact zeros never appear because of numerical error; therefore, the singularity
ratio is of the order
– n d igit
R = O  10  (11-53)
 
where n digit is the accuracy of floating-point numbers used in the calculation. For most versions of Marc, n digit  12 . If
k
rigid body modes are present, K k k is very small or negative. If either a zero or a negative diagonal is encountered, execution
of Marc is terminated because the matrix is diagnosed as being singular.
CHAPTER 11 957
You can force the solution of a nonpositive definite or singular matrix. In this case, Marc does not stop when it encounters
k
a negative or small term K k k on the diagonal. If you use Lagrangian multiplier elements, the matrix becomes nonpositive
k
definite and Marc automatically disables the test on the sign of K k k . However, it still tests on singular behavior.
Note: The correctness of a solution obtained for a linearized set of equations in a nonpositive definite system is not
guaranteed.
AutoSPC
The autospc capability, which can be activated either on the AUTOSPC parameter or on the SOLVER option, attempts to
eliminate rigid body motion by constraining degrees of freedom. This procedure occurs if the singularity ratio is smaller
than 1.e-11. This option is only available with the direct solution procedure. This is different than setting the non-positive
definite flag, which puts a small number into the stiffness matrix.
Solution of Linear Equations

The finite element formulation leads to a set of linear equations. The solution is obtained through numerically inverting
the system. Because of the wide range of problems encountered with Marc, there are several solution procedures available.
Most analyses result in a system which is real, symmetric, and positive definite. While this is true for linear structural
problems, assuming adequate boundary conditions, it is not true for all analyses. Marc has two main modes of solvers –
direct and iterative. Each of these modes has two families of solvers, based upon the storage procedure. While all of these
solvers can be used if there is adequate memory, only a subset uses spill logic for an out-of-core solution. Finally, there are
classifications based upon nonsymmetric and complex systems. This is summarized in the following table:
Table 11-3 Solver Behavior

Direct Iterative Direct Multifrontal
Profile Sparse Sparse Sparse
Solver Option 0 2 4 8
Real Symmetric Yes Yes Yes Yes
Real Nonsymmetric Yes No No Yes
Complex Symmetric Yes No No Yes
Complex nonsymmetric No No No Yes
Out-of-core Yes No Yes Yes
Possible problem with poorly conditioned systems No Yes No No
Solver Parallelization No Yes No Yes
Memory Architecture N/A Shared N/A Shared
Available with DDM Yes Yes Yes Yes
Available GPU support No Yes No Yes
Table 11-4 Solver Behavior (continued)

Mixed Pardiso
Direct Direct
CASI Iterative Iterative Sparse MUMPS Direct
Solver Option 9 10 11 12
Real Symmetric Yes Yes Yes Yes
Real Nonsymmetric No No Yes Yes
Complex Symmetric No No Yes Yes
Complex nonsymmetric No No Yes Yes
Out-of-core Yes No Yes Yes
Possible problem with poorly conditioned systems No No Yes Yes
Solver Parallelization Yes No Yes Yes
Memory Architecture Shared N/A Shared Shared or
Distributed
Available with DDM No No Yes Yes
Available GPU support No No No No
The choice of the solution procedure is made through the SOLVER option.
Direct Methods
Traditionally, the solution of a system of linear equations was accomplished using direct solution procedures, such as
Cholesky decomposition and the Crout reduction method. These methods are usually reliable, in that they give accurate
results for virtually all problems at a predictable cost. For positive definite systems, there are no computational difficulties.
For poorly conditioned systems, however, the results can degenerate but the cost remains the same. The problem with these
direct methods is that a large amount of memory (or disk space) is required, and the computational costs become very large.
The main direct solvers in Marc are the Multifrontal, Mumps and Pardiso. The Multifrontal solver supports all options in
Marc and is the default solver in Mentat. In general, the Mumps solver is the recommended solver as it usually gives the
best performance. However, the memory usage is typically higher compared to the other two. The Pardiso solver is also an
efficient solver for many cases, with performance often superior to Multifrontal.
Iterative Methods
Marc offers two iterative solvers as a viable alternative for the solution of large systems:
 CASI Solver
 Iterative Sparse Solver
The Iterative Sparse solver is based upon a preconditioned conjugate gradient method. The single biggest advantage of these
iterative methods is that they allow the solution of very large systems at a reduced computational cost. The disadvantage
CHAPTER 11 959
of these methods is that the solution time is dependent not only upon the size of the problem, but also the numerical
conditioning of the system. A poorly conditioned system leads to slow convergence – hence increased computation costs.
When discussing iterative solvers, two related concepts are introduced and both are mathematical concepts, although the
fractal dimension is a simpler physical concept.
Fractal dimension: The fractal dimension, the range of which is between 1 and 3, is a measure of the “chunkiness” of the
system. For instance, a beam has a fractal dimension of 1, while a cube has a fractal dimension of 3.
Conditioning number: The conditioning number is related to the ratio of the lowest to the highest eigenvalues of the
system. This number is also related to the singularity ratio, which has been traditionally reported in the Marc output when
using a direct solution procedure. In problems involving beams or shells, the conditioning number is typically small,
because of the large differences between the membrane and bending stiffnesses. See also the section below on Iterative
Solvers.
Mixed Direct-Iterative Solver

The mixed direct iterative solver attempts to combine the advantages of both solution methods. In the first iteration of an
increment, the Multifrontal Sparse Solver (type 8) is used to obtain a solution while simultaneously creating the Cholesky
pre-conditioner. In subsequent iterations, the sparse iterative solver is used with the previously calculated preconditioner.
If solver convergence is not being obtained at a fast enough rate, the program switches back to the direct solver. This
procedure has been shown to be effective for contact problems where small deformation or inelastic problems occur. This
solver is not supported with DDM.
Preconditioners
The choice of preconditioner can substantially improve the conditioning of the system, which in turn reduces the number
of iterations required. While all positive definite systems with N degrees of freedom converges in N iterations, a well
conditioned system typically converges in less than the square root of N iterations.
The available preconditioners in the sparse iterative solver are diagonal, scaled diagonal, and incomplete Cholesky.
The available preconditioners in the CASI sparse iterative solver are: CASI Standard and CASI Primal. The Standard
preconditioner is based on grouping elements into super-elements and is susceptible to mesh distortions and element aspect
ratios. The (recommended) Primal preconditioner is based on grouping nodes into super-nodes and, compared to the
Standard preconditioner, is more robust and insensitive to mesh distortions and element aspect ratios.
The iterative solvers require an error criterion to determine when convergence occurs. The default is to use an error criterion
based upon the ratio between the residuals in the solution and the reaction force. After obtaining the solution of the linear
c
equations u evaluate:
C C
Ku = F (11-54)
The residual from the solution procedure is:
A C A C
Res = F – F = F – Ku (11-55)
If the system is linear ( K does not change) and exact numerics are preformed, then Res = 0 .
Because this is an iterative method the residual is nonzero, but reduces in size with further iterations. Convergence is
obtained when
Res  Reac  TOL (11-56)
The iterations continue until the maximum number of iterations is reached. For the CASI solver this is given by the user
with a default value of 1000. This number is sufficiently large for many situations but may have to be increased for very
large models or for models with an ill-conditioned system.
The iterative sparse solver allows an automatic calculation of the maximum number of iterations. It is calculated as the
square root of the number of degrees of freedom in the system times a factor. This factor is 10 for the incomplete Cholesky
preconditioner and 20 for the diagonal and scaled diagonal. The minimum value is taken as 1000. Hence, this takes the
model size into account but not other aspects like the conditioning of the system or the element types used. For ill
conditioned systems it may be necessary to set a larger maximum value. If the maximum number of iterations is reached
without convergence then the solution will stop. The exception is for the sparse iterative solver, where there is an option to
switch to the Multifrontal solver if convergence cannot be reached.
Storage Methods
2
In general, a system of linear equations with N unknowns is represented by a matrix of size N by N , or N variables.
Fortunately, in finite element or finite difference analyses, the system is “banded” and not all of the entries need to be stored.
This substantially reduces the memory (storage) requirements as well as the computational costs.
In the finite element method, additional zeroes often exist in the system, which results in a partially full bandwidth. This
profile storage method is used in Marc to store the stiffness matrix when solver type 0 is chosen. When many zeroes exist
within the bandwidth, the sparse storage methods can be quite advantageous. Such techniques do not store the zeroes, but
require additional memory to store the locations of the nonzero values. You can determine the “sparsity” of the system
(before decomposition) by examining the statements:
“Number of nodal entries excluding fill in” x
“Number of nodal entries including fill in” y
If the ratio ( x  y ) is large, then the sparse matrix storage procedure is advantageous.
Nonsymmetric Systems
The following analyses types result in nonsymmetric systems of equations:
 Inclusion of convective terms in heat transfer analysis
 Coriolis effects in transient dynamic analysis
 Fluid mechanics
 Steady state rolling
 Soil analysis
 Follower force stiffness
 Frictional contact
CHAPTER 11 961
The first four always result in a nonsymmetric system. The last three can be solved either fully using the nonsymmetric
solver, or (approximately) using a symmetric solver. The nonsymmetric problem uses approximately twice as much memory
for storing the stiffness matrix.
Complex Systems
Marc utilizes a complex operator matrix for dynamic harmonic analyses when the damping matrix is present or for
harmonic magnetodynamic, acoustic, or piezoelectric analyses. The matrix is always symmetric.
Iterative Solvers
In Marc, an iterative sparse solver is available using a sparse matrix technique. This method is advantageous for different
classes of problems.
There exist certain types of analyses for which the sparse iterative solver, CASI iterative or Mixed direct-iterative solvers are
not appropriate. These types include:
 Elastic analysis
 Explicit creep analysis
 Complex harmonic analysis
 Superelements/substructures
 Central difference techniques
 Eigenvalue analysis
 Use of gap elements
 Auto increment loadcases
If the CASI solver fails due to (i) non convergence (ii) failure to generate pre-conditioner matrix (iii) negative diagonals in
matrix or (iv) non-positive definite matrix, then the internal CASI direct solver will automatically be used. In such cases it
is recommended to use one of the other available direct solvers (PARDISO, Multifrontal, MUMPS etc.) for any future runs
of the analysis. It should be noted that in some cases (e.g. with thin-walled structures), the convergence of the iterative
process may be slow and the solver may fail because of reaching the maximum number of iterations. Such a failure to
converge can often be avoided by increasing the maximum number of iterations.
If the CASI solver is used with an option that is not supported, the program will switch to the direct solver type 8.
Elastic or explicit creep analysis involves repeated solutions using different load vectors. When a direct solver is used, this
is performed very efficiently using back substitution. However, when an iterative solver is used, the stiffness matrix is never
inverted, and the solution associated with a new load vector requires a complete re-solution.
The sparse iterative solver can exhibit poor convergence when shell elements or Herrmann incompressible elements are
present.
The CASI iterative solver should not be used with higher order Herrmann elements which includes element types 155,
156, 157 and 247.
If the mixed direct-iterative solver is used and convergence is poor, Marc switches to the multifrontal sparse direct solver.
When the CASI iterative solver is used, it creates a file called jid.PCS which contains the number of nonzero values in
the stiffness matrix, the number of calculations performed, the number of iterations, and the error in solving the system of
equations. It also indicates the amount of memory and computational time.
Parallel Solvers
The solution of the linear equations may be solved using multi-processors using the multifrontal, the CASI, the Pardiso,
or the MUMPS solver. For the case of the multifrontal, CASI, and Pardiso solver, the number of processors used is
controlled by the -nts ntx command when submitting the job. Here, ntx is the number of threads. For a shared memory
machine with a single chip but with four cores, one would use -nts 4 to reduce the amount of wall time consumed. When
using a distributed memory machine or cluster, one would use the MUMPS solver and the -nsolver nsx control line
argument. Where nsx is the number of processors/threads to be used. If network computing is used, then a host file is
required as well.
Basic Theory
A linear finite element system is expressed as:
Ku = F (11-57)
And a nonlinear system is expressed as:
T
K u = F – R = r (11-58)
T
where K is the elastic stiffness matrix, K is the tangent stiffness matrix in a nonlinear system, u is the displacement
vector, F is the applied load vector, and r is the residual.
The linearized system is converted to a minimization problem expressed as:
T T
  u  = 1  2u Ku – u F (11-59)
For linear structural problems, this process can be considered as the minimization of the potential energy. The minimum
is achieved when
–1
u = K F (11-60)
The function  decreases most rapidly in the direction of the negative gradient.
  u  = F – Ku = r (11-61)
The objective of the iterative techniques is to minimize function,  , without inverting the stiffness matrix. In the simplest
methods,
uk + 1 = uk + k rk (11-62)
where
CHAPTER 11 963
T T
 k = r k r k  r k Kr k (11-63)
The problem is that the gradient directions are too close, which results in poor convergence.
An improved method led to the conjugate gradient method, in which
uk + 1 = uk + k Pk (11-64)
T T
 k = P k r k – 1  P k KP k (11-65)
The trick is to choose P k to be K conjugate to P 1 P 2  P k – 1 . Hence, the name “conjugate gradient methods. Note
the elegance of these methods is that the solution may be obtained through a series of matrix multiplications and the
stiffness matrix never needs to be inverted.
Certain problems which are ill-conditioned can lead to poor convergence. The introduction of a preconditioner has been
shown to improve convergence. The next key step is to choose an appropriate preconditioner which is both effective as well
as computationally efficient. The easiest is to use the diagonal of the stiffness matrix. The incomplete Cholesky method
has been shown to be very effective in reducing the number of required iterations.
Flow Diagram
Figure 11-9 is a diagram showing the flow sequence of Marc. This diagram shows the input phase, equivalent nodal load
vector calculation, matrix assembly, matrix solution, stress recovery, and output phase. It also indicates load incrementation
and iteration within a load increment.
Input Phase:
Read Input Data
Space Allocation
Data Check
Incremental
Loads
Equivalent Nodal
Load Vector
Matrix Assembly
Iteration Loop
Matrix Solution
Time Step Loop
Stress Recovery
No
Convergence
Yes
Output Phase
Adapt Mesh
Yes Next
Increment
No
Stop
Figure 11-9 Marc Flow Diagram
Remarks
Which solution method to use depends very much on the problem. In some cases, one method can be advantageous over
another; in other cases, the converse might be true.
Whether a solution is obtainable or not with a given method, usually depends on the character of the system of equations
being solved, especially on the kind on nonlinearities that are involved.
As an example in problems which are linear until buckling occurs, due to a sudden development of nonlinearity, it is
necessary for you to guide the arc-length algorithm by making sure that the arc length remains sufficiently small prior to
the occurrence of buckling.
Even if a solution is obtainable, there is always the issue of efficiency. The pros and cons of each solution procedure, in
terms of matrix operations and storage requirements have been discussed in the previous sections. A very important variable
regarding overall efficiency is the size of the problem. The time required to assemble a stiffness matrix, as well as the time
required to recover stresses after a solution, vary roughly linearly with the number of degrees of freedom of the problem.
CHAPTER 11 965
On the other hand, the time required to go through the direct solver varies roughly quadratically with the bandwidth, as
well as linearly with the number of degrees of freedom.
In small problems, where the time spent in the solver is negligible, you can easily wipe out any solver gains, or even of
assembly gains, with solution procedures such as a line search which requires a double stress recovery. Also, for problems
with strong material or contact nonlinearities, gains obtained in assembly in modified Newton-Raphson can be nullified
by increased number of iterations or nonconvergence.
The development of new solution procedures is still an active field of research in the academic community.
References
1. Bergan, P. G. and E. Mollestad, “An Automatic Time-Stepping Algorithm for Dynamic Problems”, Computer
Methods in Applied Mechanics and Engineering, 49, (1985), pp. 299 - 318
2. Bathe, K. J. Finite Element Procedures, Prentice Hall, Englewood Cliffs, NJ, 1996.
3. Riks, E. “An incremental approach to the solution of solution and buckling problems”, Int. J. of Solids and
Structures, V. 15, 1979.
4. Riks, E. “Some Computational Aspects of the Stability Analysis of Nonlinear Structures”, Comp. Methods in Appl.
Mech. and Eng., 47, 1984.
5. Crisfield, M. A. “A fast incremental iterative procedure that handles snapthrough”, Comput. & Structures, V. 13,
1981.
6. Ramm, E. “Strategies for tracing the nonlinear response near limit points,” in K. J. Bathe et al (eds), Europe-US
Workshop on Nonlinear Finite Element Analysis in Structural Mechanics, Ruhr University Bochum, Germany,
Springer-Verlag, Berlin, pp/ 63-89. Berlin, 1985.
7. Cerini, M. and Falzon, B. G., “Use of Arc-Length Method for Capturing Mode Jumping in Postbuckling
Aerostructures”, AIAA Journal, V43,3,(2005), pp 681-689.
8. P. R. Amestoy, I. S. Duff and J.-Y. L'Excellent, Multifrontal parallel distributed symmetric and unsymmetric
solvers, in Comput. Methods in Appl. Mech. Eng., 184, 501-520 (2000).
9. P. R. Amestoy, I. S. Duff, J. Koster and J.-Y. L'Excellent, A fully asynchronous multifrontal solver using
distributed dynamic scheduling, SIAM Journal of Matrix Analysis and Applications, ol 23, No 1, pp 15-41 (2001).
10. P. R. Amestoy and A. Guermouche and J.-Y. L'Excellent and S. Pralet, Hybrid scheduling for the parallel solution
of linear systems. Parallel Computing Vol 32 (2), pp 136-156 (2006).
Chapter 12: Parallel Processing
12 Parallel Processing
 Overview 967

Different Types of Machines 967
 Supported and Unsupported Features 968
 Matrix Solvers 970
 Contact 971
 GPU Support 971
 Domain Decomposition 973
CHAPTER 12 967
Parallel Processing
Overview
Marc can make use of multiple processors at a time when performing an analysis. There are various types of parallelism
available and most, but not all, features of Marc which are available in the standard serial mode are also supported in parallel
mode. A detailed list of unsupported features is given later on in this chapter.
Domain Decomposition
The first type of parallelism used is based upon domain decomposition. A commonly used name for this is the Domain
Decomposition Method (DDM). The model is decomposed into domains of elements, where each element is part of one
and only one domain. The nodes which are located on domain boundaries are duplicated in all domains at the boundary.
These nodes are referred to as inter-domain nodes below. The total number of elements is thus the same as in a serial
(nonparallel) run but the total number of nodes can be larger. The computations in each domain are done by separate
processes on the machine used. At various stages of the analysis, the processes need to communicate data between each
other. This is handled by means of a communication protocol called MPI (Message Passing Interface). MPI is a standard
for how this communication is to be done and Marc makes use of different implementations of MPI on different platforms.
Marc uses MPI regardless of the type of machine used. The types of machines supported are shared memory machines,
which are single machines with multiple processors and a memory which is shared between the processors, and cluster of
separate workstations connected with some network. Each machine (node) of a cluster can also be a multiprocessor
machine.
Element Loop Parallelism
The second type of parallelism is based on shared memory parallel (SMP) programming using OpenMP. This can be used
for stiffness matrix assembly, mass matrix assembly and stress recovery phase of a solution step in an analysis. The basic
concept of the SMP approach can be summarized as follows. Each finite element model has a number of element groups,
where an element group typically consists of elements of the same type, using the same material model. For each element
group, the program determines unconnected sub-groups, in such a way that all elements in a sub-group don't share nodes.
Once the sub-groups are known, the assembly and stress recovery can be efficiently performed using multiple processes at
the same time (sometimes referred to as threads). In a DDM analysis, SMP can be used on a per domain basis, which implies
that several processors or cores on a multiprocessor machine, a blade server, or a cluster of workstations can be used to
further speed up the analysis.
Matrix Solution
The third type of parallelism refers to the matrix solution. Marc also has several matrix solvers which can perform the matrix
solution in parallel while the rest of the analysis is serial. The MUMPS solver supports both shared memory machines and
clusters. It uses MPI for the communication. The multifrontal and Pardiso solvers support shared memory machines. These
solvers can also be used with DDM.
The GPU option is supported for the Multifrontal direct solver and the Sparse Iterative solver.
Different Types of Machines

As mentioned above, Marc can run on a shared memory machine, a blade server, or a cluster of workstations. The main
reason for running a job in parallel on a shared memory machine is speed. Since all processes run on the same machine
sharing the same memory, the processes all compete for the same memory. There is an overhead in memory usage, so some
parts of the analysis need more memory for a parallel run than a serial job. This overhead is substantially larger for DDM
than for the parallel assembly and stress recovery using SMP. The matrix solver, on the other hand, needs less memory in
a parallel job using DDM when using the direct-iterative solver approach, since less memory is usually needed to store and
solve several smaller systems than a single large one.
In the case of a cluster, the picture is somewhat different. Suppose a number of workstations are used in a run and one
process is running on each workstation. The process then has full access to the memory of the workstation. If a job does
not fit into the memory of one workstation, the job could be run on, say, two workstations and the combined memory of
the machines may be sufficient.
The amount of speed-up that can be achieved depends on a number of factors including the type of analysis, the type of
machine used, the size of the problem, and the performance of communications. For instance, a shared memory machine
usually has faster communication than a cluster (for example, communicating over a standard Ethernet). On the other
hand, a shared memory machine may run slower if it is used near its memory capacity due to memory access conflicts and
cache misses etc.
Supported and Unsupported Features

Most of the main features of Marc are supported in parallel mode. For instance, a fully coupled thermomechanical contact
analysis can be performed in parallel.
Local adaptivity is fully supported. New elements and nodes that are created remain in the same domain as the parent
element.
Global adaptive meshing is supported with DDM, but each body to be remeshed cannot be in multiple domains.
There is limited support of Model Sections with DDM. As the algorithms to split in model in domains do not recognize
Model Sections, the single input file DDM mode cannot be used and multi-domain input files have to be prepared
manually (such that each Model Section will be completely included on one domain) and the analysis has to be started
using the command line (instead of launching the analysis from within Mentat).
Support with DDM of coordinate systems defined by vertices of CAD bodies via the COORD SYSTEM option is also limited.
Such coordinate systems cannot be used in DDM single input file mode. In multiple input file DDM mode, the domain
input files have to be prepared such that the nodes which are associated with the CAD body vertices that define the
coordinate system are defined and associated with these vertices (using the ATTACH NODE option) in every domain in which
the coordinate system is used. If the domain input files are created by Mentat, then this is automatically taken care of.
The list of unsupported features for DDM is:
 Ablation
 Acoustics
 Auto therm creep
 Beam-to-beam contact
 Bearing
 Buckling
 CASI iterative solver
 Convective terms in heat transfer
 Design sensitivity and optimization
 Eigenvalue extraction
 Electromagnetics
CHAPTER 12 969
Parallel Processing
 Explicit dynamics
 Fluids and its coupled analysis
 Gap elements (but the normal contact is supported)
 Harmonics
 Hydrodynamics
 HyperMesh, SDRC and Intergraph results
 Insert (only supported if the host and inserted elements are in the same domain)
 Pyrolysis
 Radiation
 Response spectrum
 Steady state rolling analysis
 Superelements
 VCCT with crack propagation
Moreover, some additional model and history definition options and command line arguments are not supported for
DDM in the current release. These are:
 ACTIVATE and DEACTIVATE (History Definition)
 ACTUATOR
 BEGIN SEQUENCE and END SEQUENCE
 CHANNEL
 CFAST
 CONRAD GAP
 CWELD
 DAMPING
 DISP CHANGE if not FIXED DISP (with TABLE Input - Mechanical) format. The DISP CHANGE option is supported only if
the 1st entry of the data block following DISP CHANGE is equal to 0.
 EXCLUDE (History Definition)
 FXORD
 GLOBALLOCAL
 GRID FORCE (History Definition)
 INCLUDE
 INSERT
 NEW (History Definition)
 POINT LOAD (Model Definition) with follower forces
 SUPERPLASTIC
 QVECT (with TABLE Input - Model Definition)
 TIME-TEMP
 TYING CHANGE
 UDUMP
 def command line option
 Auxiliary input files

The DDM single input file capability will not support any analyses types and input options that are not currently
supported by the traditional multiple input files DDM, except for the INSERT option.
The following notes apply to the use of the single post file option:
 Although general DDM jobs support post revisions 7 and higher, the ability to generate a single post file from a
DDM job is restricted to post revisions 9 and higher.
 Reading a single post file back into a Marc DDM job using the -pid command line option is only available for
the AXITO3D and PRE STATE options in this release.
 In the case of single input file jobs with continuous post files where the POST option follows the RESTART option,
both jobs should either use multiple post files or single post files. If multiple post files are used, then the number
of domains should be the same in both jobs.
The list of unsupported features for Element Loop Parallelism is:
 J-integral evaluation using the LORENZI option
 Direct input or creation of matrices via the DMIG, DMIG-OUT and SUPERELEM options
 Evaluating energy release rates using the VCCT option
 Harmonic analysis (see ACOUSTIC and HARMONIC options)
 Parallel CASI iterative solver (type 9, see SOLVER option)
 Multiple regions (see REGION option), as used in e.g. a coupled fluid-solid analysis
 Parallel mass matrix assembly with CASI solver
Matrix Solvers
All matrix solvers in Marc are supported in conjunction with DDM parallel mode with the exception of the unsymmetric
solver and the mixed direct-iterative. Only the multifrontal solver(solver type 8), PARDISO (solver type 11) and MUMPS
(solver type 12) supports out-of-core solution in parallel.
The matrix solvers work differently in parallel mode as compared to serial mode.
The first method in DDM usages two stage approach. This approach is supported with direct solvers (type 0, 4, and 8). In
a mechanical analysis, all inter-domain nodes are first given fixed displacements and each domain solves the resulting
equation system independently. The second stage operates on the inter-domain nodes to relax the extra imposed fixed
displacements. This stage is not necessarily if there are no connections between the meshes in the different domains. An
iterative procedure based upon the conjugate gradient method is performed for this stage. This procedure operates as
mentioned on the inter-domain nodes. The number of conjugate gradient iterations needed to converge to the final
solution is reported in the Marc output file as “nn inter-domain iterations”.
The second method is supported with the conjugate gradient iterative solver (solver type 2), which operates simultaneously
on the whole model. It works to a large extent like in a serial run. For each iteration cycle, there is a need to synchronize
the residuals from the different domains.
The third method in DDM is supported with direct solver (type 8, 11 and 12). For this case, the global stiffness matrix (or
the corresponding matrix for nonstructural analysis) is gathered to the parent process (where the job was started). For solver
type 8 and 11 (Multifrontal and PARDISO), the original child processes are put to sleep and the solver creates new
CHAPTER 12 971
Parallel Processing
processes (multiple threads) to solve the equation system in parallel. When the solution is finished, the child processes are
awakened and the data related to each domain are distributed. For solver 12 (MUMPS), original child processes takes part
in computation to solve the equation of system. For Multifrontal and PARDISO solver, multithreading is used to obtain
the performance improvement and may be used in conjunction with DDM on a shared memory architecture. The
MUMPS solver suitable for machine with shared memory architecture as well as for distributed memory or cluster
environments and may be used in conjunction with DDM.
The forth method in DDM is supported with MUMPS solver (type 12). For this case, stiffness matrix on each domain is
converted to the global stiffness on respective domains. The part of global stiffness matrix is supplied to distributed solver.
But global LHS is supplied only on parent and when the solution is finished, the data related to each domain are
distributed. This method is most suitable for machine with distributed memory or cluster environments and may be used
on shared memory architecture as well in conjunction with DDM.
Contact
Contact analysis poses a special challenge for DDM parallelization due to its global nature. The current version of Marc
fully supports contact analysis including thermal and electrical contact but excluding beam-to-beam contact.
In order to enable a fast contact search procedure, some contact data are duplicated across all domains. A node in one
domain must be able to find contact with a segment in a different domain. The duplicated data includes nodes on the
exterior boundary of deformable contact bodies. This data duplication is not necessary in a pure rigid contact job, so it is
important to define a contact table to indicate that no deformable contact can take place in this case. The default, if no
contact table is used, is that all deformable bodies check for contact with all other bodies including itself.
If a node touches a segment which is located in a different domain, the nodes making up the segment are imported into
the domain of the touching node. If the node separates, the nodes are deleted from the domain. This means that the
number of nodes in a domain can dynamically change. This procedure allows the creation of tyings spanning domains.
This also has implications on how the domains should be created. One should avoid having contact bodies with a large
number of nodes coming into contact across different domains. An extreme case would be two sheets of equal number of
shell elements coming into contact on top of each other. If two domains are used and each sheet is in a separate domain,
the number of nodes in one domain would double.
If only SMP is used, then the contact algorithm works the same as in a serial analysis.
GPU Support
Marc is capable of utilizing GPU (Graphics Processing Units) devices for additional floating point computing power. This
is available with the multifrontal direct solver (type 8) and the sparse iterative solver (type 2). GPGPU stands for General-
Purpose Computation on Graphics Processing Units. GPUs are high performance processors that can be used to accelerate
a wide range of applications. For more information on GPGPU, see http://gpgpu.org/.
Marc users who want to significantly reduce real symmetric sparse factorization run times for large models can utilize
NVDIA CUDA-capable cards.
The system requirements for using the GPU computing feature in Marc are:
 NVIDIA CUDA-capable GPGPU card(s) with at least 1.5 GB on-board memory.
 NVIDIA Developer Drivers for Linux 450.51.06 or later.

 NVIDIA Developer Drivers for Windows 451.82 or later.
Technical Discussion
An interface has been developed for the direct sparse solver (type 8) and iterative sparse solver (type 2) for NVIDIA GPU
devices. No new functionality is introduced. Utilization of GPU devices is for reduced run times only.
Improved performance only results for relatively large models. In particular, only matrices whose front sizes are larger than
a certain threshold benefit.
Input
There is a new run time parameter that controls the GPU execution.
gpuid -gpuid id Here id is the ID of a GPU device or list of IDs of GPU devices to be used in the
analysis. In a list, IDs are separated by a colon.
Default: User must specify the value.
For Serial run:
User must specify a valid GPGPU card number.
For DDM run:
Multiple IDs may be assigned to Marc DDM runs.
Single host run: GPGPU cards are assigned in round-robin fashion if
the number of GPUIDs is less than the number of
domains.
Multiple host run: GPUID must be specified for each domain.
-gpuid auto For Serial run:
Marc automatically detects all supported GPGPU cards and selects the best
card.
For DDM run:
All available GPGPU cards on a host are assigned in round-robin fashion.
Examples
On a machine with two GPGPU cards and four CPUs:
To select GPU ID number 0 for computation of your Marc analysis, one would use:
To select GPU ID number 1 for computation of your Marc analysis, one would use:
CHAPTER 12 973
Parallel Processing
To select GPU IDs 0 and 1 for DDM analysis two domains, one would use:
And for DDM analysis with four domains with two GPGPU cards, one would use:
or
This will be translated as follows:
Domain Decomposition
The way the model is decomposed into domains, in general, affects the performance of the parallel run. Traditionally,
domain decomposition had to be performed in the preprocessor stage using Mentat or Patran. Both Mentat and Patran use
METIS to perform the domain decomposition using the multiple file option. For a single input file case, the
domain decomposition can also be performed in Marc.
The general goal for a good decomposition is to minimize the number of interdomain nodes. Keeping this number low
minimizes the overhead in having duplicate nodes and also minimizes the size of the system solved in the interdomain part
of the direct solver. It is important to get a good balance between the workload in the different domains if, for instance,
different types of elements are used or if the nodes of a cluster have different performance.
Another aspect of deformable contact as mentioned above is to avoid having a large number of nodes touching segments
of a different domain.
Several different methods for creating the decomposition are available in Marc (single input), Mentat, and Patran. One
should note that some of them are based on element connectivity so they tend to put separate contact bodies in separate
domains which is not always optimal.
The optimal number of domains to use for a given job is pretty hard to estimate. It obviously depends on the number of
processors available. There is, in general, an optimal number of processors to use for a given job and a given machine or
cluster of machines. A general rule is that the larger the model is, the more it pays off to use more processors.
Running a Parallel Job

Running a Marc parallel job is not much different from running a serial job. The model can be decomposed into domains
by the preprocessor (Mentat or Patran) and a separate input file is created for each domain. From a user’s perspective, the
extra steps consist of activating the procedure for decomposing the model and activating the parallel run.
Marc can also be used to create the domains in case a single input file containing the whole model is used. Two methods
can be used to trigger the single input file option; the first is through the command line option as explained below and the
second is through the use of the PROCESSOR parameter.
If the command line option is used and the PROCESSOR parameter is not defined, the default domain decomposition
method, Metis Best, is used. Currently, the single input file option does not support any additional features that are not
supported by the traditional multiple input file option (except INSERT) as described by the list of unsupported features
above.
For the case of a run on a cluster, one would additionally have to define which machines make up the cluster (unless an
automatic cluster tool is used). The run is started as usual and, by default, separate results files are created for each domain.
The postprocessor automatically picks up the results from the different domains and combines them into a single model
which looks the same as in a serial run. One can also choose to look at the results of each domain separately, which is
particularly useful for large jobs. As opposed to the default multiple post file option, the user also has the option to request
a single post file for the whole model. For details on how to use this functionality, refer to Marc Volume C: Program Input
for the POST model definition option.
When running a parallel job from Mentat or Patran, one has to activate the button for parallel and start the job as for a
serial job. From the command line, one needs to use an extra argument to specify the number of domains:
run_marc -v n -b n -j jobname -np 5
for running a job with five domains using multiple input files generated by a preprocessor. To indicate that a single input
file will be used, the -np option is replaced with -nps and the command line in this case becomes:
run_marc -v n -b n -j jobname -nps 5
User subroutines are handled in the same way as in a serial run. See Marc Volume D: User Subroutines and Special Routines for a
description on special considerations for user subroutines in a parallel run.
Running on a cluster (a set of connected computers) requires that the machines communicate properly. The actual
connection could be anything from a telephone modem up to a high speed direct connected cluster. A standard 100 Mbit
Ethernet network is usually sufficient for reasonable performance, although it can become bottlenecked with fast
processors. Faster connections give better performance on the communication part.
In order to perform an analysis over a network, one needs to specify which machines are to be part of the analysis. This is
done through a so-called host file, which is generic for a Marc run. The Marc startup script converts it into a format that
the current MPI implementation expects. The host file is a simple text file that lists the computers that are to be used in
the run and some info about the run. The host file typically looks something like (using a Linux system)
host1 2
host2 1 /users/joe/run /programs/marc
host3 2 /usr/people/joe /programs/marc/marc
Here host1, host2, and host3 are the host names of the respective machines. It specifies a job with five domains where
two processes are run on host1 and host3 and one on host2. /users/joe/run is the directory where the job is run
on host2. This is where the input files are copied and where the results files are created. This directory could be a local
directory on a hard drive on host2 or it could be a shared directory (via NFS for example). This directory must be
accessible from host2 (but not necessarily from the other hosts). The fourth entry of the host file specifies where the Marc
installation directory is located. This can also be a local directory (if Marc is installed on the local machine), or it can point
to a shared directory. It is recommended that the run directory be a local directory to avoid I/O traffic over the network.
Not all MPI implementations currently supported uses the fourth entry. Specifically, the commonly used IBM MPI and
Intel MPI (used on Linux and Windows) do not support the specification of different installation directories. They need
to use the same path on all hosts. They can be shared or local but the path name must be the same.
The input files for the remote hosts are automatically copied to each host in case the run directory is local. When the run
is finished, the post files are copied back. Please note that only the input files and the post files (not output files, etc.) are
copied. This copying can be suppressed by buttons in Mentat and Patran or by a command line option if Marc is manually
started. If a user subroutine is used together with local run directories, the new executable is automatically transferred to
all hosts before the Marc job starts.
CHAPTER 12 975
Parallel Processing
One can also set up Marc to run via external cluster tools. A useful run time option in this context is the command line
option -dir. It specifies the run directory for the job and is where the scratch files and results files are created. Suppose
Marc is set up to run on a cluster using some kind of cluster tool to distribute the processes and each machine in the cluster
has a local disk called /scratch. By using the -dir /scratch option from the command line, the run uses the local
directory during the run. The input files are still read from where the job is started (which could be a shared directory on
a control workstation which is not even part of the run).
The following command line options related to parallel processing are available for the Marc run_marc start-up script:
-np nn specify nn number of processes.

-nps nn specify nn number of processes if using a single input file.
-ci y|n if y, copy input files to local run directories; if n, do not copy.
-cr y|n if y, copy post files from local run directories; if n, do not copy.
-ho hostfile specify the host file that describes the hosts to use in a network run.
-dir rundir specify that the run directory of the run is rundir.
-mpi specifies the MPI version to use if other than the default.
For further details on parallel processing over a number of machines, refer to the Marc Parallel Network Installation and
Operations Guide of Marc on Linux or Windows.
Using parallelization based on Element Loop Parallelism is driven by the command line option -nte. The following
command line causes the analysis to be performed using four element threads without using DDM:
run_marc -v n -b n -j jobname -nte 4
If both Element Loop Parallelism and DDM are being used, then the -nte option defines the total number of element
threads being used (summed over all the domains). So in an analysis with two domains (using a single input file) and four
element threads per domain, both the options -nte and -nps should be used as:
run_marc -v n -b n -j jobname -nte 8 -nps 2
The requested numbers of domains and threads are echoed by Marc, like:
Requested number of domains for DDM = 2
Requested number of element threads per domain = 4
Requested total number of element threads (all domains) = 8
Domain Decomposition Methods

Metis Best - method performs Metis element based, Metis node based, and Vector decomposition and picks the best
decomposition with respect to the number of interdomain nodes as well as the domains balance. (Figure 12-1)
Metis_element_based - performs decomposition of graph constructed from element connectivity.
Metis_node_based - performs decomposition of graph constructed from nodal connectivity.
Figure 12-1 Metis Best Decomposition
Vector - performs decomposition based on IDs of elements. The element IDs are sorted by coordinates of the element
centroids. The default direction of sorting is in the maximal dimension of the model bounding box in x-, y-, and z-
directions respectively. User may also specify direction of sorting by supplying the direction vector. See Marc Volume C:
Program Input, Chapter 2, Parameters – PROCESSOR. (Figure 12-2)
Figure 12-2 Vector Decomposition with Specified User Direction (1,1,0)
Radial - performs decomposition based on IDs of elements. The element IDs are sorted by distance between the element
centroids and axis of rotation. The default (axis of rotation) is specified by default direction and default point on the axis
of rotation.
Default point for axis of rotation is defined as the centroid of the model bounding box. User may also specify the axis of
rotation by point and direction. See Marc Volume C: Program Input, Chapter 2, Parameters – PROCESSOR. (Figure 12-3)
CHAPTER 12 977
Parallel Processing
Figure 12-3 Radial Decomposition with User Direction (.5,.5,.707) and

Point on Axis (2,2,2.141)
Angular - performs decomposition based on IDs of elements. The element IDs are sorted by angle between the element
centroids, axis of rotation and selected start vector. The default (axis of rotation) is specified by default direction and default
point on the axis of rotation. User may also specify the axis of rotation by point and direction. See Marc Volume C: Program
Input, Chapter 2, Parameters – PROCESSOR. (Figure 12-4).
Figure 12-4 Angular Decomposition with User Direction(.5,.5,.707) and Point on Axis (2,2,2.141)
Recursive Coordinate Bisection - performs recursively vector decomposition into two domains. Each domain is
subsequently subdivided into two or more domains until all domains are created (Figure 12-5).
Figure 12-5 Recursive Coordinate Bisection
User - performs decomposition based on element sets in the input file. The element sets must be named domaini where
i is the domain number. The domain sets are usually created by the user in the preprocessor program.
The Fine Graph flag in the 2nd data block on the PROCESSOR parameter tells Metis to use a fine graph for decomposition.
This results in higher memory requirements and lower speed for the decomposition. In some cases, it produces better
decomposition than the default (Coarse Graph). The Coarse Graph choice might produce more islands. It is
recommended to use the default unless there is a need for better quality of domain decomposition.
Sometimes the domain decomposition produces disconnected domains resembling islands (Figure 12-6). The best
performance usually is obtained when the domains are contiguous. Islands can be added to neighboring domains by
turning on the Island Remove flag in the 2nd data block on the PROCESSOR parameter. This might result in worse balance of
domains as shown in Figure 12-7.
An option on the PROCESSOR parameter is the Detect Contact. This attempts to reduce the number of nodes contacting
from different domains - essentially keeping contact within a domain where possible.
CHAPTER 12 979
Parallel Processing
Figure 12-6 Metis Node Based Decomposition without Island Removal
Figure 12-7 Metis Node Based Decomposition with Island Removal

Chapter 13: Output Results
13 Output Results
 Overview 981

Workspace Information 981
 Increment Information 984
 Selective Printout 986
 Restart 988
 Element Information 988
 Nodal Information 994

Post File 996
 Forming Limit Parameter (FLP) 997
 Program Messages 1000

Marc HyperMesh Results Interface 1001
 Marc SDRC I-DEAS Results Interface 1001
 Marc HDF Results Interface 1002

Marc - ADAMS Results Interface 1005

Status File 1007
 Element Group Information 1007

References 1008
CHAPTER 13 981
Output Results
Overview
This chapter summarizes the information that Marc provides in the output. In addition to reviewing your input, Marc
provides information about the procedures the program uses and the workspace allocation. All calculated results are
automatically written to the output file unless the user specifically requests otherwise.
Workspace Information
Marc reports several aspects of workspace information, specifically, the allocation of memory workspace and the size of the
work files.
For the general memory, Marc first gives the workspace needed for the input and stiffness assembly. This number tells how
much memory is needed to store the user-supplied data, the program-calculated data, and two-element stiffness matrices.
Each set of data comprises three parts: overhead, element information, and nodal information. Element information
consists of properties, geometries, strains, and stresses, and nodal information consists of coordinates, displacements, and
applied forces.
Next, Marc specifies the internal core allocation parameters. These values are useful in user subroutines and are often
provided to the user subroutines. The values provided here are:
 Number of degrees of freedom per node (ndeg)
 Maximum number of coordinates per node
 Maximum number of invariants per integration point (neqst)
 Maximum number of nodes per element (nnodmx)
 Maximum number of stress components per integration point (nstrmx)
 Number of strain components per integration point (ngens)
Note the following:
 The number of degrees of freedom per node indicates the number of boundary conditions necessary to eliminate
rigid body modes.
 The number of stresses per integration point is the number of stresses per layer multiplied by the number of layers.
 The number of invariants per integration point is the number of layers for either shell or beam elements.
The ELSTO parameter stores element information on an auxiliary storage device, rather than in main memory. You can
invoke this option, or Marc invokes it automatically. In this option, Marc:
 sets a flag for element storage (IELSTO)
 specifies how many words are used per element (NELSTO)
 specifies the number of elements per buffer (MXELS)
 specifies the total amount of space needed to store this information
The number of elements per buffer should be at least two, and you can change the buffer size in the ELSTO parameter option
to increase the number of elements which will be in the buffer.
Marc reports the memory for different parts as it is allocated.
For contact bodies:

allocated workspace of 4567 words of memory for body 1
For incremental backup:
space needed for incremental backup: 555313
For boundary conditions:
allocated 600 words of memory due to kinematic boundary conditions
We also get printouts like
total workspace needed with in-core matrix storage = 893281
This printout refers to the memory in general memory (see Marc Workspace Requirements in Chapter 2 in this volume) and
is not the total memory used.
The solution of the equations of equilibrium result in the formation of the operator matrix and the
decomposition/inversion of the matrix. In an mechanical analysis, this is the global stiffness matrix. For some solvers, there
is an intermediate stage where the matrix is restructured from a sparse format to a more optimal format for numerical
calculations. The allocation of this memory is reported in the memory summary (see below) either under solver:first
part or under the allocated separately: category. This is summarized as follows:
Operator Decomposition or
Solver Matrix Restructured Matrix Inverted Matrix
0 - Profile direct first part first part first part
2 - Sparse Iterative first part N/A N/A
4 - Sparse Direct first part first part first part
8 - Multi-front Direct Sparse first part N/A allocated separately
10 - Mixed Direct Iterative first part N/A allocated separately
11 - Pardiso Direct first part N/A allocated separately
12 - MUMPS direct first part N/A allocated separately
For solvers 0, 2, and 4, all the memory for the matrix solution is allocated within the general memory. For solvers 8, 11,
and 12, the workspace for the restructuring stage and decomposition part is allocated separately outside of the
general memory.
Solvers 0, 4, and 8 have an out-of-core option for both storing the operator matrix and for inverting the matrix. Solvers 11
(PARDISO) and 12 (MUMPS) have an out-of-core option for inverting the matrix only. For this solver, there must be
enough memory to assemble the operator matrix in-core.
When using solver 0, if the out-of-core solver is invoked, one observes:
total workspace needed with in-core matrix storage = 89686633
out-of-core matrix storage will be used
core allocation based on 50498 nodal entries per assembly buffer.
...file 14-- maximum record length= 40960
CHAPTER 13 983
Output Results
approximate no. of words on file= 3686400

If solver 4 were used, one would observe:
total workspace needed with sparse in-core matrix storage = 34288801
out-of-core matrix storage will be used
Note that until 14 contains the true sparse matrix, and, as the same problem is used, the I/O is the same for both solvers 0
and 4. For solver 0, this matrix gets expanded on unit 11 using a row blocking system, and decomposed to unit 12. For
solver 4, it gets expanded and decomposed using unit 2. Solver 4 uses substantially less I/O than solver 0.
For solver 8, we observe
solver workspace needed for out-of-core matrix storage = 6389794
solver workspace needed for in-core matrix storage = 31339042
matrix solution will be out-of-core
approximate disk space for out-of-core matrix storage = 62678084
At the end of the run, Marc prints out a summary of the memory used. This typically looks something like
memory usage: mbyte words % of total
within general memory:

element stiffness matrices: 0 64982 0.2
solver: first part 4 963934 3.6
overallocation initial allocation 49 12957302 48.1
other: 0 15713 0.1
allocated separately:
incremental backup: 5 1253317 4.7
solver 8 10 2581002 9.6
nodal vectors: 1 349427 1.3
contact: 1 261625 1.0
tyings: 0 13362 0.0
transformations: 0 25182 0.1

kinematic boundary conditions: 0 4200 0.0
element storage: 5 1324844 4.9
executable and common blocks: 27 7000000 26.0
miscellaneous 0 121763 0.5
---------------------------------------------------------------
total: 103 26936653
general memory allocated: 53 14001931

general memory used: 4 1044629
peak memory usage: 110 28936653
In a parallel analysis, Marc also prints out the memory usage summed over the domains. The memory printout, during the
execution, can also be obtained by means of the control file. Suppose the job defined in the input file jobname.dat is
running. The memory summary printout is obtained by creating a file called jobname.cnt with a single line containing
the word “memory”. The information is output at the beginning of the next increment.
If the ALLOCATE parameter is to be used for initial memory allocation, the value to look for is the one for the line
general memory used: 4 1044629
In this example, one would choose a value of 5 for ALLOCATE to insure that no dynamic reallocation of memory is performed.
The top part defines the memory in general memory. The overallocation initial memory specifies the amount of general
memory specified with the ALLOCATE option that is not used. The above example is very small and there is a large
overallocation due to the default used (although the total amount is small). For larger problems, this part should be small.
The line other specifies the part of the general memory that is not categorized.
Further down are the parts that are allocated separately from the general memory, listed for each category. Similar to other
above, miscellaneous specifies memory not categorized.
The two lines for general memory is for highlighting the memory allocated and the memory actually used. The difference
between the two is reflected in the overallocation line in the table. executable and common blocks is an estimate of the
memory needed for loading the executable and storing values in common blocks. This is added so the total better agrees
with the amount of memory the operative system reports for the program. Finally, the peak memory denotes the maximum
amount of memory allocated at any instant during the job. In some cases, memory is allocated and then released. The total:
in the summary table may then show a smaller value than the maximum during the run.
Increment Information
A variety of information is given for each increment, and often for each iteration, of the analysis. This information is useful
in determining the accuracy and the stability of the analysis.
CHAPTER 13 985
Output Results
Summary of Loads
The printout entitled “Load Increments Associated with Each Degree of Freedom” allows you to check the load input
quickly. It represents the sum, over all nodes in the mesh, of the point loads and the equivalent forces obtained after
distributed loads (pressures, body forces) are applied. For example, you can easily check a pressure because the total force
in a global coordinate direction would be the projected area normal to that direction of the surface upon which the pressure
is applied, multiplied by the pressure magnitude.
Timing Information
The amount of CPU necessary to reach the given location in the analysis is indicated by the following output:
 Start of increment
 Start of assembly
 Start of matrix solution
 End of matrix solution
 End of increment
As an example, subtract the time associated with the start of the matrix solution from the time associated with the end of
the matrix solution to determine how much time was spent in the equation solver.
The total amount of time required is printed at the end of the analysis and is divided into the following phases (note that
not all of them apply to all analyses):
 Total time for input
 Total time for stiffness assembly
 Total time for mass matrix assembly
 Total time for stress recovery
 Total time for matrix solution
 Total time for contact
 Total time for remeshing/rezoning
 Total time for output
 Total time for restart
 Total time for adaptivity
 Total time for cutter-mesh detection
 Total time for eigenvalue extraction
 Total time for view factors
 Total time for miscellaneous
The miscellaneous time is the remaining time, which is not counted in any of the major phases. More details about the
time required can be obtained using the PRINT,13 parameter option. In case of a multi-threaded analysis, details about
overhead due to multi-threading can be obtained using the PRINT,57 parameter option.
Singularity Ratio
The singularity ratio is a measure of the conditioning of the matrix (see Chapter 11 Solution Procedures for Nonlinear Systems for
further details). This ratio is printed each time there is a solution of the matrix equations. You can measure the influence
of the nonlinearities in the structure by examining the change in singularity ratios between increments. The singularity
ratio is not available when using the iterative solvers, and when using the MUMPS direct solver. When using the Pardiso
direct solver, the singularity ratio is available only for real symmetric, positive definite matrices.
Convergence
Several messages are printed that concern the convergence of the solution. These messages indicate the displacement,
velocity, or residual error and are very important because they provide information concerning the accuracy of the solution
procedure. These messages also indicate the ratio of the error and its relative quantity. This ratio must be less than that
given in the CONTROL option for convergence to occur. See Convergence Controls in Chapter 11for details on
convergence testing.
When the iterative solver is used, additional messages are printed regarding the convergence of the solution procedures.
Selective Printout
Marc gives you several options and user subroutines for the control of the Marc output.
Options
PRINT CHOICE allows you to select how much of the element and nodal information is to be printed. The possible selections
are
 Group of elements
 Group of nodes
 Which layers (form beam and shell elements)
 Which integration points
 Increment frequency between printouts
The data entered through this option remain in control until you insert a subsequent PRINT CHOICE set. You can include such
a set with either the model definition or with the history definition set. To obtain the default printout after a previous PRINT
CHOICE, invoke PRINT CHOICE using blank entries.
Note: The PRINT CHOICE option has no effect on the restart or post file.
You can use the PRINT ELEMENT capability as a replacement for, or in conjunction with, the PRINT CHOICE option. The
enhancements this option offers over the PRINT CHOICE option are the following:
1. You have a choice of integration and layer points for each element to be printed, which is especially useful when
several different element types are used in one analysis.
CHAPTER 13 987
Output Results
2. You have a choice of the type of quantity to be printed, for instance, stresses could be printed for all elements, but
strains for only a few.
3. The types of output quantities that can be selected are stresses, strains, creep strain, thermal strain, cracking strain,
Cauchy stress, state variables, strain energy, or all nonzero quantities.
4. The output can be placed in a file other than the standard output file.
5. The PRINT NODE option can be used to obtain element quantities such as stresses and strains at the nodal points of
each element. These are obtained by extrapolating the integration point values to the nodes.
The PRINT NODE option is an alternative to the PRINT CHOICE option for controlling the output of nodal quantities. The
additional capabilities of this option compared to the PRINT CHOICE are as follows:
1. You can choose which of the following nodal quantities are to be printed:
 Incremental displacements
 Total displacements velocities
 Accelerations
 Reaction forces
 Reneralized stresses
2. Different quantities can be printed for different nodal points.
3. The output can be placed in a file other than the standard output file.
The PRINT CONTACT and the NO PRINT CONTACT options control the output of the summary of the loads on the contact bodies
and their positions. They are also used to control the contact interaction output.
Note: All quantities are saved on the restart file, and you can obtain them at a later time by using the RESTART option.
Quantities that were not printed out are available for later use.
The SUMMARY option allows you to get a quick summary of the results obtained in the analysis. This option prints the
maximum and minimum quantities in tabular form. The is designed for direct placement into reports. You control the
increment frequency of summary information and the file unit to which the information is written.
The SUMMARY option reports on the physical components and the Tresca, von Mises, and mean values of stress, plastic
strain, and creep strain. It also reports on such nodal quantities as displacements, velocities, accelerations, and
reaction forces.
The sort options (ELEM SORT and NODE SORT) allow you to sort calculated quantities in either ascending or descending order.
These quantities may either be sorted by their real magnitude or their absolute magnitude. The sort options also allow you
to control the type of quantity to be sorted, for example, equivalent stress and the number of items to be sorted. The sort
options print the sorted values in tabular form. The is designed for direct placement into reports. You may control the
increment frequency of sorted values and the file unit to which these values are written.
The PRINT VMASS option allows you to selectively choose which elements and associated volumes, masses, first and second
moment of inertia, and strain energy are to be printed. In order to have correct mass computations, mass density for each
element must be given. The volumes, masses, and energies can be written on either standard output file unit 6, or user
specified unit. The center of mass is also calculated and output.
Grid Force Balance

The GRID FORCE option allows the user to output the different contributions to the total force applied on an element or a
node.
On an element level, the grid force balance is based upon the:
 Internal forces
 Distributed Loads
 Foundation Forces
 Reaction Force
On a nodal basis it is much more complete and includes:
Internal Forces Distributed + Point Forces

Foundation Forces Spring Forces
Contact Normal Forces Contact Friction Forces
Tying/MPC Forces Inertia Forces
Damping Forces DMIG Forces
Reaction Force
User Subroutines
User subroutines can be used to obtain additional output, which can be accessed at each time/load increment. The available
subroutines are the following:
 IMPD: Obtains nodal quantities, such as displacements, coordinates, reaction forces, velocities, and accelerations
 ELEVAR: Obtains element quantities such as strains, stresses, state variables, and cracking information
 ELEVEC: Outputs element quantities during harmonic subincrements
 INTCRD: Obtains the integration point coordinates used for forming the stiffness matrix
Restart
One capability of the RESTART option allows you to recover the output at increments where printout was suppressed in
previous runs. This option can be used to print time/load increments for a number of consecutive increments. Under this
option, Marc does not do any analysis.
In conjunction with the RESTART option, you can use either the PRINT CHOICE or PRINT ELEMENT, or PRINT NODE option to select
a region of the model for which you want to obtain results.
Element Information
The main output from an increment includes element information followed by nodal information. The system provides
the element data at each integration point. If you use the CENTROID parameter, the system provides the element data at the
CHAPTER 13 989
Output Results
centroidal point. You can control the amount of printed output by using either the PRINT CHOICE or PRINT ELEMENT option.
All quantities are total values at the current state (at the end of the current increment), and the physical components are
printed for each tensor quantity (stress, strain, and generalized stress and strain). The orientation of these physical
components is generally in the global coordinate system; however, their orientation depends on the element type. (Marc
Volume B: Element Library explains the output for each particular element type.) Also, the physical components may be
printed with respect to a user defined preferred system.
In addition to the physical components, certain invariants are given, as follows:
1. Tresca intensity - the maximum difference of the principal values, and the measure of intensity usually required for
ASME Code standards.
von Mises intensity - defined for stress as:
3 1
 = --- S ij S ij S i j =  i j – ---  i j  k k (13-1)
2 3
where  is a stress tensor and S is the deviatoric stress tensor. For beam and truss elements,  22 =  33 = 0 ,
and additional shears are zero; for plane stress elements, including plates and shells,  33 = 0 .
2. von Mises intensity - calculated for strain type quantities as:
2
 = ---  ij  ij (13-2)
3
Marc uses these measures in the plasticity and creep constitutive theories. For example, incompressible metal creep
and plasticity are based on the equivalent von Mises stress. For beam, truss, and plane stress elements, an
incompressibility assumption is made regarding the noncalculated strain components.
a. For beam and truss elements, this results in:
1
 22 =  33 = – ---  11 (13-3)
2
b. For plane stress elements
 33 = –   11 +  22  (13-4)
c. For beam and plane stress elements taken as a whole:

 kk = 0 (13-5)
3. Mean normal intensity - calculated as:

  11 +  22 +  33   3
or (13-6)
  11 +  22 +  33   3
Equation (13-6) represents the negative hydrostatic pressure for stress quantities. For strain quantities, the equation
gives the dilatational magnitude. This measurement is important in hydrostatically dependent theories (Mohr-
Coulomb or extended von Mises materials), and for materials susceptible to void growth.
4. The principal values are calculated from the physical components. Marc solves the eigenvalue problem for the
principal values using the Jacobi transformation method. Note that this is an iterative procedure and may give
slightly different results from those obtained by solving the cubic equation exactly.
5. State variables are given at any point where they are nonzero.
Mentat Computed Element Results

These element results are computed based on the Marc output. By default, Marc always outputs the elemental results at
the integration points. In Mentat, by default, these data are extrapolated linearly to the nodal positions. The data are then
averaged with the contributions from the neighboring elements. Principle values and equivalent values are computed based
on these nodal data in Mentat. If a stress or strain tensor is output in the post file, Mentat will compute their principal
values and the equivalent items (such as, equivalent stress and strain). Note that the equivalent items may be different from
ones directly output by Marc because of the way they are computed in Mentat. For example, the equivalent stress computed
in Mentat may show a slightly different result compared to the equivalent von Mises stress from Marc output.
It is also possible to turn off the extrapolation procedure in Mentat. This is recommended in problems of high gradients.
It can be done through the RESULTS>scalar plot SETTING>EXTRAPOLATION menu by picking TRANSLATE.
Solid (Continuum) Elements

For solid (continuum) elements, stress, strain, and state variables are the only element quantities given. Marc prints out
stresses and total strain for each integration point. In addition, it prints out thermal, plastic, creep, and cracking strains, if
they are applicable. Note that the total strains include the thermal contribution.
Solid-shell Elements
Marc prints the solid-shell element output in the output file using the integration point identification. But in the post file,
the element output is presented in the layer identification conforms to the standard shell convention.
For non-stacked element (only one element in the thickness direction), users can request element section outputs:
curvature, moment, stretch and section force to put in the post file (see Marc Volume C: Program Input for a detailed
description of the POST option). These are post codes 731, 741, 751, 761, respectively.
The section force and moment are calculated using Equations 13-9 and 13-10, respectively. The stresses and y-coordinate
(thickness direction) are referring to element coordinate system.
The stretch and curvature are calculated as follows.
top b ot
 ij +  i j
E ij = ---------------------------
- (13-7)
2
to p b ot
 ij –  i j
 ij = --------------------------
- (13-8)
2t
The curvatures and moments are calculated using the element mid-plane as reference.
CHAPTER 13 991
Output Results
Shell Elements
Marc prints generalized stresses and generalized total strains for each integration point. For thick shell (Type 22, 72, 75)
elements interlaminar shear stresses are printed at the interface of two laminated layers if the TSHEAR parameter is included.
The generalized stresses printed out for shell elements are:
+t2
1
---   i j dy
t (13-9)
–t  2
 Section Force per Unit Thickness 
+t2
1
---  y i j dy
t (13-10)
–t  2
 Section moment per Unit Thickness 
The generalized strains printed are
E  ;   = 1 2
(13-11)
 Stretch 
  ;   = 1 2 (13-12)
 Curvature 
Physical stress values are output only for the extreme layers unless you invoke either PRINT CHOICE or PRINT ELEMENT. In
addition, thermal, plastic, creep, and cracking strains are printed for values at the layers, if applicable. Although the total
strains are not output for the layers, you can calculate them using the following equations.
 11 = E 11 + h 11 (13-13)
 22 = E 22 + h 22 (13-14)
 12 = E 12 + h 12 (13-15)
where h is the directed distance from the midsurface to the layer; E i j are the stretches; and  i j are the curvatures as printed.
You can obtain these quantities by using the ELEVAR subroutine.
Beam Elements
The printout for beam elements is similar to shell elements, except that the section values are force, bending and torsion
moment, and bimoment for open section beams. For beam element type 45, interlaminar shear stresses are printed at the
interface of laminated layers if the TSHEAR parameter is used.
These values are given relative to the section axes (X, Y, Z) which are defined in the COORDINATES or the GEOMETRY option.
Before a beam member can be designed, it is necessary to understand the section forces distribution along the axial direction
of the beam. For example, if variations of shear force and moment along axial direction are plotted, the graphs are termed
shear diagram and moment diagram, respectively. To postprocess these diagrams, post code 261 needs to be specified along
with the post codes (shown in Table 13-1) which define the corresponding section forces of the beam element.
Table 13-1 Post Codes for Beam Section Forces

Post Code
267
264 265 266 (Shear xz 268
Element (Axial (Moment xx or (Moment yy or or Shear (Shear yz or 269 270
Type 2-D/3-D Force) Moment ip) Moment op) ip) Shear op) (Torque) (Bimoment)
5 2-D X X
16 2-D X X
45 2-D X X X
13 3-D X X X X X
14 3-D X X X X
25 3-D X X X X
31 3-D X X X X X X
52 3-D X X X X
76 3-D X X X X
77 3-D X X X X
78 3-D X X X X
79 3-D X X X X
98 3-D X X X X X X
The stresses at the section points in the cross section are only output if the PRINT CHOICE or PRINT ELEMENT option are
included. If any nonlinearity exists at the section points, they will be provided in the output file.
Gap Elements
Marc prints the gap contact force, friction forces, and the amount of slip.
Linear and Nonlinear Springs

Marc prints the spring forces.

The element temperatures are provided at the integration points. When requested, the temperature gradient and the flux
is also provided.
CHAPTER 13 993
Output Results
Joule Heating Analysis

In addition to the temperatures, the voltage, current, and the heat generated at the integration points is output.

The lubricant thickness, pressure, and pressure gradient are provided.
The electrical potential ( V ), electric field intensity vector ( E ), and electrical displacement vector/flux ( D ) are provided.
Piezoelectric or Electrostatic-Structural Analysis

In addition to the mechanical stresses and strains, the electrical field intensity vector ( E ) and electrical displacement/flux
( D ) are provided.
The magnetic potential ( A ), the magnetic field density ( H ), and the magnetic flux density ( B ) are provided. For 2-D
problems, only the A z component is given. For circuit analysis, the circuit current is provided in the .out file.
Magnetodynamics Analysis
The magnetic vector potential ( A ), the electric field intensity vector ( E ), the electrical displacement ( D ), the magnetic
field intensity ( H ), the magnetic flux density ( B ), and the induced current density ( J ) are provided.
For circuit analysis, additional quantities are provided in:
1. Mentat postprocessing as global quantities:
Harmonic analysis:
a. Real and imaginary parts of terminal voltage
b. Real and imaginary parts of terminal current
c. Real and imaginary parts of coil induced voltage
d. Total circuit resistance
e. Total Inductance
Transient analysis:
a. Terminal voltage
b. Terminal current
c. Coil induced voltage
d. Total circuit resistance
2. Global quantities in .out file:

Harmonic analysis:
a. Coil resistance for each coil
b. Real and imaginary parts of terminal voltage
c. Real and imaginary parts of terminal current
d. Real and imaginary parts of coil induced voltage
e. Total circuit resistance
f. Total circuit Reactance
g. Apparent power, ohmic loss, and reactive power
Transient analysis:
a. Coil resistance for each coil
b. Terminal voltage
c. Terminal current
d. Coil induced voltage
e. Total circuit resistance
Acoustic Analysis
The pressure and the gradient of the pressure is provided. Note in an Acoustic-Structural analysis, the pressure is only
provided as a nodal variable.
Nodal Information
Marc also prints out the following quantities at each nodal point.
Stress Analysis
 Incremental displacements - the amount of deformation that occurred in the last increment
 Total displacements - the summation of the incremental displacements
 Total equivalent nodal forces (distributed plus point loads) - the total force applied to the model through
distributed loads (pressures) and point loads
 Reaction forces at fixed boundary conditions
 Residual loads at nodal points that are not fixed by boundary conditions
Note: Each value listed above is given at each nodal point, unless you invoke the PRINT CHOICE or PRINT NODE option.
If you invoke the TRANSFORMATION option, Marc prints the nodal information relative to the user-defined
system, rather than the global coordinate system.
CHAPTER 13 995
Output Results
Reaction Forces
Marc computes reaction forces based on the integration of element stresses. This is the only way to compute total reaction
forces in a nonlinear analysis. Since such integration is only exact if the stresses are known at each integration point, the
reaction forces are not printed if you use the CENTROID parameter.
In a nonlinear analysis, you should check that the reaction forces are in equilibrium with the external forces. If they are not
in equilibrium, the analysis will be inaccurate, usually due to excessively large incremental steps. In most cases, the
equilibrium is automatically ensured due to the convergence testing in Marc.
Residual Loads
The residual loads are a measure of the accuracy of the equilibrium in the system during analysis. This measure is very
important in a nonlinear analysis and should be several orders of magnitude smaller than the reaction forces.
Dynamic Analysis
In a dynamic analysis, Marc prints out the total displacement, velocity, and acceleration at each time increment.

In a heat transfer analysis, Marc prints out nodal temperatures, externally applied fluxes, and calculated nodal heat fluxes.
Joule Heating Analysis

In Joule heating analysis, in addition to the thermal vectors, the nodal voltage and the applied and calculated currents are
provided.
Rigid-Plastic Analysis
In steady-state rigid-plastic analysis, Marc prints nodal velocities.

In a hydrodynamic bearing analysis, Marc gives the mass flux at the nodal points.
In an electrostatic analysis, Marc prints the scalar potential and the applied and calculated charge.
In a magnetostatic analysis, Marc prints the vector potential and the applied and calculated current.
In an magnetodynamic analysis, Marc prints out the vector and scalar potential, the applied and calculated current and
charge.
Piezoelectric or Electrostatic-Structural Analysis

In a piezoelectric or electrostatic-structural analysis, in addition to the structural quantities, such as displacement and
reaction force, the electric potential, and reaction charges.
Acoustic Analysis
In an acoustic analysis, Marc prints the pressure and the source.
Supplementary Information
Contact Analysis
The standard output includes, at the end of each increment, a summary of information regarding each body. This
information reports the increment’s rigid body velocity, the position of the center of rotation, and the total loads on the
body. These last values are obtained by adding the contact forces of all nodes in contact with the rigid body. Deformable
bodies being in equilibrium have no load reported.
Also at the end of each output is a summary of the interaction between contact bodies. These values are automatically
placed on the post file (post revision 14 and higher).
Additional information can be obtained by means of the PRINT,5 parameter. In such cases, all the contact activity is reported.
Namely, every time a new node touches a surface, or separates from a surface, a corresponding message is issued.
Contact with rigid surfaces entails an automatic transformation. Displacement increments and reactions in the transformed
coordinates - tangent and normal to the contact interface - are also reported for every node that is in contact in the usual
Marc manner.
If the EMRESIS or EMCAPAC options are invoked, a summary is output providing the electrical resistance or capacitance for
the conducting body.
Post File
The post file contains results of the analysis performed. This information can be viewed using Mentat or Patran to observe
the displaced mesh, contours of element quantities, principal quantities, time history behavior, response gradients, design
variables, etc. The post file contains the finite element mesh and any changes to the finite element mesh due to either
rezoning or adaptive mesh refinement. Basic nodal quantities (such as displacements, velocities, acceleration, applied loads,
and reaction forces) are automatically placed on the post file. It is also possible to include user-defined nodal vectors by use
CHAPTER 13 997
Output Results
of the UPOSTV user subroutine. Element quantities, such as stress, strain, plastic strain) must be explicitly requested by using
the POST model definition option. It is possible to have user-defined element quantities placed on the post file for
subsequent display by using the PLOTV user subroutine.
Forming Limit Parameter (FLP)

For continuum or shell elements, the Principal Engineering Strains can be calculated based on the true strain values.
Correspondingly, the Forming Limit Parameter can be obtained for shell/membranes according to the data of the Forming
Limit Diagram (FLD).
The principal engineering strains are calculated based on the true strain according to following relation:
e i = exp   i  – 1 (13-16)
where e i are the principal engineering strain components and  i are the principal true strain components.
The principal engineering strain values can be selected for postprocessing. In addition, if shell/membrane elements are used,
the Forming Limit Parameters (FLP) can also be selected. The FLP is defined as the ratio of the major principal engineering
strain to the maximum allowable major principal engineering strain given by the Forming Limit Diagram. Based on this
definition, it is calculated by
FLP = e 1  FLD  e 2  (13-17)
where e 1 means the major principal engineering strain, e 2 is the minor principal engineering strain. FLD  e 2  is the
forming limit corresponding to e 2 . FLD is defined by input data in model definition through material properties. By
plotting the major principal engineering strain of each integration point of an element as point ( e 2 , e 1 ) on the chart, as
shown in Figure 13-1, it is easy to find that for points below the FLD line is safe. Strain rates above the FLD line imply failure.
To consider the FLP value of these fields, one can find that in the field which is above FLD line, the value of FLP is larger
than 1. On the FLD line, FLP equals 1, and FLP is less than 1 if a point located in the field below the FLD line. Therefore,
by plotting the FLP value in Mentat, it can be seen if the sheet metal forming process is successful.
Major Principal Engineering Strain (e1)
FLD(e2)
Fail
Fail
FLP>1 FLD(e2)
Safe
Safe
FLP<1
Minor Principal Engineering Strain (e2)

Figure 13-1 The Definition of Forming Limit Parameter (FLP)
The methods employed in Marc to define the Forming Limit Diagram are:
1. Fitted function definition
2. Predicted function definition
3. Table definition (accepts piecewise linear FLD curves)
The details of the above methods are described below:
1. Fitted function:
By this method, the polynomial functions are utilized to fit the FLD curve. The functions are shown in Equations
13-18a and 13-19b:
FLD  e 2  = C 0 + D 1 e 2 + D 2 e 22 + D 3 e 23 + D 4 e 24  e2  0  (13-18)
FLD  e 2  = C 0 + C 1 e 2 + C 2 e 22 + C 3 e 23 + C 4 e 24  e2  0  (13-19)
2. Predicted function:
The predicted functions are generated based on the theories of local necking and diffuse necking. Both theories
assume that the material obeys the power-law strain hardening,  = K n .
Local necking: The critical strain for the onset of local necking is influenced by strain ratio  , where  =  2   1 ,
  0 . Equation shows the relationship of critical strain versus strain ratio (by Hill).
n
* = ------------------ 0
1 +  (13-20)
For uniaxial tension,  = – 0.5 . For plane strain,  = 0 .

CHAPTER 13 999
Output Results
Diffuse necking: Diffuse necking is the phenomenon that while necking happens, deformation continues. The
critical strain for the diffuse necking to happen is determined by Equation 13-21 (by Swift):
2n  1 +  +  2 
* = ------------------------------------------------------ 0 (13-21)
 1 +    2 2 –  + 2 
Note: For the case of   0 , the predicted critical strain by Equation 13-21 is well below the experiment
data. Therefore, for the purpose of prediction, the diffuse necking theory and the local necking
theory are chosen for cases of   0 and   0 , respectively.
Experiments (mainly with low carbon steels) showed that the predicted FLD usually is at a level below the
experimental FLD curve. The experimental forming limit diagram is characterized by the value of *
corresponding to plane strain. This value is referred at FLD 0 as shown in Figure 13-2. This value increases with the
strain-hardening exponent, n , and the strain-rate exponent, m . The value of * is observed to rise as the thickness
increases. This phenomenon is referred as thickness effect and it is characterized as thickness coefficient t c .
Experiments by Keeler tended to express this relationship as shown in Equation (13-22).
FLD 0 = Q   0.233 + t c  t  (13-22)
where Q = n  0.21 if n is less than 0.21. Otherwise, Q = 1.0 . T is the thickness of the sheet metal. The
thickness coefficient t c is set as 3.59 if the unit used to define the thickness is “Inch”. If the unit used is “mm”, t c
is set as 0.141. Similarly, if the unit “cm” is used, t c is then set as 1.41, etc.
It should be emphasized that, in most cases, friction is also a very important parameter that will affect the occurrence
of necking. On the other hand, Equation (13-22) is mainly from data on low-carbon steels that have relatively high
strain-rate dependence. In materials with low strain-rate dependence (for example, aluminum alloys), the thickness
effect should be much less.
Finally, it is suggested that for safety purposes, a safety zone, as shown as the shadow area in Figure 13-2, must be
set. For points above the safety zone, material is considered as failure. For points below the safety zone, material is
considered as safe. If a point locates in the safety zone, the material is in the marginal status. Usually, the thickness
of the safety zone is set as 0.1 or 10%.
3. Table definition:
The table function in Marc allows users to define any curves through the TABLE model definition option. For
example, if the user has FLD point of material, it is possible to define the FLD as piecewise linear curve. For details,
refer the Marc Volume C: Program Input, Chapter 3 Model Definition Options, TABLE.
Major Principal Engineering Strain e1
Experimental
forming limit
FLD0
Local necking Diffuse necking
0
Minor Principal Engineering Strain e2
Figure 13-2 Predicted FLD and Experimental FLD Curves
Program Messages
The messages provided by Marc at various points in the output show the current status of the problem solution. Several of
these messages are listed below.
 START OF INCREMENT x indicates the start of increment number x.
 START OF ASSEMBLY indicates Marc is about to enter the stiffness matrix assembly.
 START OF MATRIX SOLUTION indicates the start of the solution of the linear system.
 SINGULARITY RATIO prints yyy, where yyy is an indication of the conditioning of the matrix. This value is
typically in the range 10-6 to 1. If yyy is of the order of machine accuracy (10-6 for most machines), the equations
might be considered singular and the solution unreliable.
 END OF MATRIX SOLUTION indicates the end of matrix decomposition.
 END OF INCREMENT x indicates the completion of increment number x.
 RESTART DATA at INCREMENT x ON UNIT 8 indicates that the restart data for increment number x has been
written (saved) on Unit 8.
 POST DATA at INCREMENT x on UNIT y indicates that the post data for increment number x has been
written (saved) on Unit y.
In addition to these Marc messages, exit messages indicate normal and abnormal exists from Marc. Table 13-2 shows the
most common exit messages.
CHAPTER 13 1001
Output Results
Table 13-2 Marc Exit Messages

Message Meaning
MARC EXIT 3001 Normal exit.
MARC EXIT 3004 Normal exit.
MARC EXIT 13 Input data errors were detected by the program.
MARC EXIT 2004 Operator matrix (for example, stiffness matrix in stress analysis) has become
non-positive definite and the analysis terminated.
MARC EXIT 3002 Convergence has not occurred within the allowable number of recycles.
Marc HyperMesh Results Interface

Marc now outputs results for postprocessing with HyperMesh. The writing of the HyperMesh results file is invoked by the
HYPERMESH model definition option described in detail in Marc Volume C: Program Input. Mentat can also be used to write the
necessary commands into the Marc data file. The related button can be found one level below each of the ANALYSIS CLASS
buttons within the JOBS menu.The binary file created as a result has the title jobid.hmr where jobid is the name of the
data file for the job.
The geometry data may be imported into HyperMesh from a Marc data file. Alternatively, it can be created by HyperMesh
or read in from other formats via the HyperMesh import capability.
The types of data that may be selected for writing into the HyperMesh results file are listed in Marc Volume C: Program Input
under the HYPERMESH model definition option. It is important to note that, in case eigenvectors for buckling or
eigenfrequency analysis are to be written into the results file, the corresponding Marc file should have the BUCKLE INCREMENT
or MODAL INCREMENT model definition option, as appropriate. The BUCKLE, MODAL SHAPE, and RECOVER history definition
options are not to be used in these cases.
Since HyperMesh allows only one deformed shape plot per simulation, each eigenvector of an eigenvalue analysis is saved
as a separate simulation. Thus, when using HyperMesh, these eigenvectors can be plotted by skipping to the next
simulation rather than skipping to the next data type of a simulation. The contour plots can be obtained for all data types
including eigenvectors.
In case the number of requested eigenvalues is more than the number extracted, the data type in the HyperMesh
“deformed” screen will inform you for those modes that are not extracted. The “next” button may need to be clicked to see
the data type in the “deformed” mode of plotting.
Marc SDRC I-DEAS Results Interface

A facility exists in Marc to write out an SDRC I-DEAS universal file of the analysis results, in order for postprocessing by
the I-DEAS program. The writing of the universal file is invoked by the SDRC model definition option described in detail
in Marc Volume C: Program Input. Mentat can also be used to write the necessary commands into the Marc data file. The
related button can be found one level below each of the ANALYSIS CLASS buttons within the JOBS menu. The formatted
ASCII file created as a result has the title jobid.unv, where jobid is the name of the data file for the job.
The model data is imported into I-DEAS by way of the universal file as well.
The types of data that may be selected for writing into the I-DEAS universal file are listed in Marc Volume C: Program Input
under the SDRC model definition option.
Marc HDF Results Interface

A facility exists in Marc to write out an HDF (Hierarchical Data Format) database of the analysis results. The database
supports high precision, compression, and unlimited amount of data. It is open format and has multiple programming
language support.
The HDF database generated by Marc stores model input data and output result data as datasets. These datasets can be
post-processed through MSC products like Patran, Apex and can also be viewed through the HDF Viewer.
For more information on HDF5, please see the HDF5 Software Documentation:
https://portal.hdfgroup.org/display/support/Documentation
HDF File Creation

The binary HDF file has the title jobid.h5, where jobid is the name of the data file for the job. The writing of the HDF
file is invoked through the HDFPOST model definition option in conjunction with the HDFOUT option. HDFPOST
is used to define the frequency of output and the required elemental / nodal results. HDFOUT is used to identify the active
HDFPOST option for a particular job. These options are described in detail in Marc Volume C. The following points
should be noted about the Marc generated HDF file:
a. Nodal and Element output are supported for 2D elements (plane stress, plane strain, axisymmetric) and 3D
non-composite continuum elements. There is no support for beams, shells, solid shell and solid composite
elements.
b. If any unsupported element exists in the model, warnings are issued in the OUT file – separate warnings are
provided for elements that do not have model input support and for elements that do not have result output
support.
c. Only Mechanical Static and Dynamic Analysis are supported for HDF output.
d. Only three quantities for nodal output are currently supported – Displacements, External Forces and Reaction
Forces. All nodal output is in the global Cartesian coordinate system.
e. Only two quantities for elemental output are currently supported – Stress and Strain. For a large strain analysis,
the output stress and strain are the Cauchy Stress and Logarithmic Strain respectively. For a small strain analysis,
the output stress and strain are the engineering quantities. All element output is in the global Cartesian
coordinate system. For 2D elements, the output is based on integration point quantities and the center value
(average of integration point values). For 3D elements, the output is based on nodal point quantities
(extrapolated from integration point values) and the center value (average of integration point values). For 3D
quadratic elements, only the corner nodes and center point are considered.
f. There is no HDF file support for a distributed (DDM) run (Exit 58 will be issued in this case). HDF file is
supported for multi-threaded runs. For a restart run, a discontinuous HDF file that is only based on the
increments on the restart run is produced.
CHAPTER 13 1003
Output Results
g. Each dataset that is put on the HDF file is compressed. This provides for a significant reduction of the HDF
file size, especially when datasets have a number of zeroes in them. The default compression level used in Marc
is 3.
HDF File Post-Processing

The HDF file generated by Marc can be post-processed by Patran and Apex. It can also be viewed through the HDF viewer
– this is very useful to understand the HDF file layout and the schema that is used for each dataset. The following points
should be noted for HDF file post-processing:
a. The format followed by Marc is currently based on the Nastran HDF5 format. This allows post-processing of
the HDF file with existing post-processors like Patran and Apex.
b. Patran supports both model input and result output processing. The Marc generated HDF file can be directly
read into Patran for post-processing.
c. Apex currently does not support model input processing. It is necessary to have the model made available in
Apex through a different mechanism (eg. reading a Nastran input file of the same Marc model into Apex) before
post-processing the results.
HDF File Layout

The typical layout of the HDF file, as seen in the HDF viewer, is shown in the following Figure 13-3. The following points
should be noted about the layout:
Figure 13-3 HDF File Layout

a. There are two main groups INDEX and NASTRAN.
- The INDEX group contains the start and end rows of result output for each increment.
- The NASTRAN group contains data related to the model input (INPUT) and result output (RESULT).
Note: The name convention follows the fact that the Nastran schema is currently being used by Marc for the
HDF output.
CHAPTER 13 1005
Output Results
b. INPUT: Contains model input data. Currently, for nodes, nodal id and position data are supported in the
INPUT/NODE/GRID dataset. For elements, element id and nodal connectivity data are supported in the
INPUT/ELEMENT/Element_Type dataset where Element_Type refers to a specific element type (eg.
CHEXA, CTETRA, etc).
c. RESULT: Contains result output data. RESULT/ELEMENTAL group contains stress / strain data for
different element types. RESULT/NODAL group contains displacement / external force / reaction force data
for nodes.
Marc - ADAMS Results Interface

ADAMS/Flex allows flexible components to be included into ADAMS models to achieve more realistic simulation results.
Marc is capable of generating a Modal Neutral File (MNF) representing the flexible component to be integrated into the
ADAMS model. The flexible component can undergo linear or nonlinear loading in Marc up to the point where the MNF
is requested to be generated. No more analysis takes place in Marc after generating the MNF. Generating an MNF from
Marc is based on performing the most general method of Component Mode Synthesis (CMS) techniques, namely the
Craig-Bampton method.
Using the Craig-Bampton method, the degrees of freedom, u , of the flexible component are partitioned into two sets of
degrees of freedom:
 Boundary or Interface Degrees of Freedom, u B : These degrees of freedom can be used to apply loads in ADAMS or
to connect the flexible component to other rigid bodies.
 Interior Degrees of Freedom, u I : These internal degrees of freedom of the flexible component can be condensed out
using the superelement technique.
Two sets of modes shapes arranged in matrices,  I C and  I N , and their corresponding modal coordinates vectors, q C
and q N , are also defined:
 Constraint Modes,  I C : These modes are the static shapes obtained by giving each boundary degree of freedom a
unit displacement while holding all other boundary degrees of freedom fixed. Corresponding to the constraint
modes is the set of modal coordinates q C . The basis of constraint modes completely spans all possible motions of
the boundary degrees of freedom, with a one-to-one correspondence between the modal coordinates of the
constraint modes and the displacement in the corresponding boundary degrees of freedom, u B = q C .
 Fixed-Boundary Normal Modes,  I N : These modes are obtained by fixing the boundary degrees of freedom and
computing an eigensolution. There are as many fixed-boundary normal modes as the user desires. These modes
define the modal expansion of the interior degrees of freedom. The quality of this modal expansion is proportional
to the number of modes retained by the user. Corresponding to the fixed-boundary normal modes is the set of
modal coordinates q N .
The Craig-Bampton modes consist of both the constraint modes and the fixed-boundary normal modes. The relationship
between the physical degrees of freedom and the Craig-Bampton modes and their modal coordinates is given by
   
 uB  I BC O BN  q C 
  =   (13-23)
 uI  I C I N  qN 
   
where the subscripts B, I, C, and N denote Boundary Degrees of Freedom, Interior Degrees of Freedom, Constraint Mode
and Normal Mode, respectively. The above equation indicates that the Craig-Bampton analysis can be considered as a
transformation from the physical degrees of freedom to the Craig-Bampton modal basis q in the form:
u = q
where  is the modal matrix or transformation matrix. The finite element equation of motion is given by:
Mu·· + Ku = f
in which M is the mass matrix, K is the stiffness matrix and f is the load vector. Substituting by the transformation
T
relation into the equation of motion and pre-multiplying by  gives:
T ·· T T
 Mq +  Kq =  f
The above equation defines the generalized mass and stiffness matrices as:
 Generalized Mass Matrix, M̂ : Projection of the mass matrix onto the modal space given by:
T
M̂ =  M
 Generalized Stiffness Matrix, K̂ : Projection of the stiffness matrix onto the modal space given by:
T
K̂ =  K
Note that the ADAMS MNF interface expects lumped mass matrices to be used. Thus, the LUMP parameter must be
included in the Marc input. The MNF interface distinguishes between two types of loads:
 Modal Preload, f̂ PL : A projection onto the modal space of the internal load vector corresponding to the loads that
have already been applied in the finite element analysis up to the point when the MNF is generated.
 Modal Loads, f̂ L : A projection onto the modal space of the loads that may be defined in the Marc input using the
ASSEM LOAD history definition option but not applied in the finite element analysis. They may be scaled and
introduced in the ADAMS simulation. ASSEM LOAD loadcases, if exist, should precede any actual preloading to be
applied to the component in Marc.
Both types of loads can be computed using the relation
T
f̂ =  f
CHAPTER 13 1007
Output Results
Corresponding to each Craig-Bampton mode, a stress and a strain mode can be computed. These are the incremental nodal
stresses and strains found by considering each mode at a time to be the incremental displacement solution and performing
a stress recovery. To request the computation and export of stress and/or strain modes, the MNF parameter must be used.
The Craig-Bampton analysis is triggered when the SUPERELEM model or history definition option is reached in the input. The
SUPERELEM option allows direct definition of the boundary or interface DOFs, u B . The option also allows automatic
definition of interface degrees of freedom of the nodes that get in contact with selected rigid contact bodies. This is very
useful for some nonlinear analyses such as tire footprint analysis in which the interface DOFs are not known a priori. It
also allows the specification of interface degrees of freedom of the control nodes of selected load-controlled rigid contact
bodies. The two control nodes for load-controlled rigid bodies are consolidated into one node with six degrees of freedom
before exporting to the MNF.
A MODAL SHAPE loadcase using the Lanczos method must be present in the input immediately following the SUPERELEM
loadcase. No more analysis can be performed afterwards.
The units used in the Marc model are specified using the MNF UNITS model definition option. Successful MNF generation
produces Marc Exit 3018. The MNF generated will have the same filename as the input file and the extension .mnf.
Assuming the input file is jobname.dat, the MNF will have the name jobname.mnf.
Status File
The status file contains summarized information of the analysis. Each line within this file includes the loadcase number,
the associated increment number, and the number of cycles, contact separations, and cutbacks associated with that
increment; the accumulated total cycles, increment splittings, separations, cutbacks, and remeshing times numbers in the
analysis as well as the time step size of each increment and the overall time achieved by the analysis. Marc reports all this
information on one line upon completion of each increment. If the increment is partially completed (for example, the
completion of one part of a split increment or one part of the whole increment due to cutback), Marc also reports one line
for each part of the increment.
In the PRINT NODSTS option in used, the maximum and minimum resultant quantities are added to the status files. It is also
possible to specify a node in which case the results of this node are provided. This is an easy way to monitor the results.
Element Group Information

The finite element model is internally subdivided into element groups. The groups are based upon the element type and
material ID. The division into groups is internal to the program and entirely automatic. The memory allocation is done
separately for each element group, allowing the memory usage to be optimized for cases when different element types and
materials are used. Information about the element groups is printed to the output file. This printout occurs in two places.
The program first prints the amount of memory used for each group. Here is a sample printout for a job with two element
groups
memory usage per element group

group # elements nelsto MByte words
1 1368 1480 8 2024640
2 456 1284 2 585504
----------------------------------------------------------
total 1824 10 2610144
For each groups, it lists the number of elements in the group; nelsto, which is the amount of memory in words used for
each element and the total amount of memory used for all elements in the group, given in Megabytes and words.
A second printout is written further down in the output file after reading the model definition input data:
number of element groups used: 2
group # elements element type material formulation
1 1368 7 1 S
2 456 140 1 S
formulation:
S: small displacement
This listing shows the number of elements in the group and also the element type, the material ID, and the formulation
used in the group. If composites or rebars are used, it lists the composite or rebar group instead of the material ID. The
formulation column shows a number of characters indicating the type of formulation used. The table below shows the
characters that are used.
Character Description
S small displacement small strain analysis
T total Lagrange formulation
U updated Lagrange formulation
M multiplicative plasticity or hyperelastic formulation
A additive plasticity formulation
F finite strain plasticity for additive formulation
E enhanced/assumed strain formulation
C constant dilatation formulation
References
1. R. R. Craig and M. C. C. Bampton. Coupling of substructures for dynamics analyses. AIAA Journal, 6(7): 1313-1319.
1968.
2. R.R. Craig. Structural Dynamics: An Introduction to Computer Methods. John Wiley & Sons. 1981.
3. ADAMS User Guide: Using ADAMS/Flex. 2002.
Chapter 14: Code Coupling Interfaces
14 Code Coupling Interfaces
 Overview 1010

General Code Coupling 1010
 Coupling Regions 1012
 Marc Co-simulation Adapter 1015
 References 1021
Overview
This chapter describes the co-simulation interfaces to external solvers. A general code coupling is available through user
subroutines and will be discussed first. The interface allows code coupling libraries such as MpCCI (see [Ref. 1]) to couple
Marc with commercial CFD codes, such as STAR-CD and Fluent. The interface may also be used for developing dedicated
interfaces to external solvers, for applying complex boundaries to certain regions of the model, or for dedicated
postprocessing. MSC also provides co-simulation between Adams and Marc for models involving multi-body dynamics
and nonlinear finite element analysis.
General Code Coupling

Analyzing fluid-structure interaction problems has become increasingly important in industrial applications. Effects of the
deforming structure on the fluid flow and, vice versa, of the fluid pressure on the structure, as well as heat transfer between
fluid and structure can be of critical importance for a design, especially if the deformations of the structure are large.
The fluid-structure capabilities of Marc (see Chapter 10: Nonstructural Procedure Library in this manual) are limited to
nonreactive, incompressible, single phase, laminar flows. This implies that interaction problems such as combustion
reactions in airbag gas generators, involving complex turbulent flows with large deformations of the structure cannot be
solved. An additional drawback is that, at the fluid-structure interface, the fluid and structural meshes must share the same
nodes. This may be inconvenient, as the fluid analysis tends to require a denser mesh than the structural analysis.
Code coupling software, such as the MpCCI (Mesh-based parallel Code Coupling Interface) software from Fraunhofer
SCAI (see [Ref. 1]), allows structural codes such as Marc to be coupled with commercial CFD codes such as STAR-CD or
Fluent. Fluid-structure interaction problems can then be solved using the capabilities of both solvers in their respective
domains. The two solvers run simultaneously and exchange data regularly during the analysis on the common boundary
between the fluid and the structure.
Figure 14-1 shows a flow chart of such a coupled analysis, in which the CFD solver and the structural solver alternately solve
their respective parts of the coupled problem. The CFD solver solves the fluid flow problem for a given period of time (the
coupling time step) to provide the pressure that the fluid exerts on the structure, the temperature of the fluid at the
common boundary with the structure and the film coefficient to the structural solver. The structural solver applies the fluid
pressure and the film boundary condition to the structure and computes the deformations and temperatures of the
structure for the same period of time. Then it sends the current coordinates of the nodes and the temperature at the fluid-
structure interface to the CFD code. The CFD code uses these coordinates to update the flow domain and continues with
the next coupling time step. Note that the coupling time step may consist of multiple time steps in each of the solvers.
CHAPTER 14 1011
Code Coupling Interfaces
Code Coupling Software

CFD Analysis Structural Analysis
CFD Structural
Mesh Mesh
Initialization Initialization
CFD
Time Step(s)
Fluid Pressure, Fluid
Temperature, Film
Coefficient Coupling
Time Step
Structural Time
Step(s)
Current Coordinates,
Structural Temperature
CFD
Time Step(s)
Fluid Pressure, Fluid
Temperature, Film Coefficient
Coupling
Time Step
Structural Time
Step(s)
Current Coordinates,
Structural Temperature
Figure 14-1 Flow Chart of a Coupled Fluid-Structure Analysis Using Code Coupling Software
The code coupling software serves as a communication layer between the two codes. It handles the data transfer between
the solvers and interpolates the data from the CFD mesh to the structural mesh and vice versa, in case the meshes do not
match at the common boundary (Figure 14-2). To be able to interpolate the data from one mesh to the other, the code
coupling software determines the geometrical relationship between the meshes, using information (connectivity and
coordinates) provided by the two solvers at the start of the analysis.
CFD Mesh
Structural Mesh
Figure 14-2 Non-Matching Meshes at the Common Boundary Between Fluid and Structure
The basic stages, from a code coupling point of view, in a coupled analysis are:
1. Initialization. The relationship between the CFD mesh and the structural mesh is established. Both the CFD solver
and the structural solver must provide the code coupling software with the connectivity and the coordinates of the
patches on the common surface between the fluid and the structure.
2. Data exchange. Physical quantities such as the fluid pressure and the temperature of the fluid at the common surface
with the structure, are extracted from the CFD code, interpolated to the structural mesh, transferred to the
structural code and applied through appropriate boundary conditions to the structural mesh. Likewise, current
coordinates and temperature of the structure at the common surface with the fluid are extracted from the structural
code, interpolated to the CFD mesh and transferred to the CFD code. The CFD code uses the new coordinates to
update the flow domain.
3. Finalization.
Coupling Regions
Coupling regions are the basic concept to couple Marc with external solvers via code coupling software such as MpCCI. A
coupling region is that part of the surface or volume of the structural model where the interaction with the external solver
takes place. A surface region consists of a list of edges or geometric curves in 2-D and a list of faces or geometric surfaces
in 3-D. A volumetric region consists of a list of elements or contact bodies. Coupling regions are defined by the COUPLING
REGION model definition option and are accompanied by an extensive application programming interface (API). The basic
mechanical and thermal quantities (current coordinates, displacement, force, pressure, temperature, heat flux, film
coefficient) can be exchanged with an external solver on coupling regions via API calls. Quantities that are received from
the external solver are applied to the coupling region via boundary conditions.
CHAPTER 14 1013
The coupling region API consists of three user subroutines and a set of utility routines that can be called from these user
subroutines (see Chapter 9 Special Routines - Marc Post File Processor, Marc Volume D: User Subroutines and Special Routines).
The three user subroutines correspond to the basic three stages of a coupled analysis, as outlined in the previous section:
1. Initialization: CPLREG_INIT. This subroutine is called once at the start of the analysis and can be used to extract the
connectivity and coordinates of the coupling region and pass it to the code coupling software to determine the
geometric relationship between the coupling region mesh and the mesh of the external solver.
2. Data exchange: CPLREG_EXCHANGE. This subroutine is called twice per coupling time step, once at the start and once
at the end. The call at the start can be used to set the new values of physical quantities that have been received from
the external solver for the next coupling time step. The call at the end can be used to extract the new values of
physical quantities that must be sent to the external solver.
3. Finalization: CPLREG_FINALIZE. This subroutine is called once at the end of the analysis and can be used to inform
the external solver that the Marc analysis has ended.
The utility routines can be called from the user subroutines to obtain the connectivity and coordinates of the coupling
region, to obtain the current values of physical quantities on coupling regions, to set the new values of physical quantities
for the next coupling time step and even to specify the next coupling time step. The available utility routines are listed in
Table 14-1.
Table 14-1 Available Utility Subroutines for Coupling Regions

Available in User
Utility Subroutine Subroutines Purpose
CPLREG_FIND_NAME CPLREG_INIT Find coupling regions by name.
CPLREG_EXCHANGE
CPLREG_GET_INFO CPLREG_INIT Get general information about a coupling
CPLREG_EXCHANGE region.
CPLREG_GET_QUANTS CPLREG_INIT Get the prescribed physical quantities on a
CPLREG_EXCHANGE coupling region.
CPLREG_GET_MESH CPLREG_INIT Get the mesh of a coupling region.
CPLREG_EXCHANGE
CPLREG_GET_GLOBAL_VALUES CPLREG_EXCHANGE Get the values of a global quantity.
CPLREG_GET_NODE_VALUES CPLREG_EXCHANGE Get the values of a node-based quantity at a
CPLREG_GET_ALL_NODE_VALUES coupling region.
CPLREG_PUT_GLOBAL_VALUES CPLREG_EXCHANGE Put the values of a global quantity.
Table 14-1 Available Utility Subroutines for Coupling Regions (continued)

Available in User
Utility Subroutine Subroutines Purpose
CPLREG_PUT_NODE_VALUES CPLREG_EXCHANGE Put the values of a node-based quantity at a
CPLREG_PUT_ALL_NODE_VALUES coupling region.
CPLREG_PUT_EDGE_VALUES CPLREG_EXCHANGE Put the values of an edge-based quantity at a
CPLREG_PUT_ALL_EDGE_VALUES coupling region.
CPLREG_PUT_FACE_VALUES CPLREG_EXCHANGE Put the values of a face-based quantity at a
CPLREG_PUT_ALL_FACE_VALUES coupling region.
CPLREG_PUT_ELEM_VALUES CPLREG_EXCHANGE Put the values of an element-based quantity
CPLREG_PUT_ALL_ELEM_VALUES at a coupling region.
Time Step Control

The default coupling time step is equal to the Marc time step, so that the CPLREG_EXCHANGE user subroutine is called at the
start and at the end of each increment. If the AUTO STEP time stepping scheme is used, the coupling time step may be
prescribed by calling the CPLREG_PUT_GLOBAL_VALUES utility routine from the CPLREG_EXCHANGE user subroutine. In that
case, the AUTO STEP algorithm matches the next coupling time, but multiple increments may be needed to reach this
coupling time. If multiple increments are performed within a given coupling time step, then the CPLREG_EXCHANGE user
subroutine is called at the start of the first increment and at the end of the last increment of the coupling step.
Shell Elements
The top and bottom faces of a shell element are independent surface entities, as far as coupling regions are concerned. They
can both be part of the same coupling region or they can be part of two different coupling regions. The faces share the same
nodes, but the connectivity of the bottom face, as returned by the CPLREG_GET_MESH utility routine, is exactly the opposite
of that of the top face (which is the connectivity of the element).
In heat transfer and coupled thermo-mechanical analyses, the temperatures (and nodal heat fluxes) at the top and bottom
of a shell element are distinct. If a coupling region contains only top faces of shell elements, the CPLREG_GET_NODE_VALUES
utility routine returns the top temperatures (and heat fluxes) at the nodes of the region. Similarly, the
CPLREG_PUT_NODE_VALUES utility routine prescribes the top temperatures (and heat fluxes) at the nodes of the region. If the
coupling region contains only bottom faces of shell elements, these routines return and prescribe the bottom temperatures
(and heat fluxes) at the nodes. If both the top and the bottom faces of shell elements interact with the mesh of an external
solver, two coupling regions must be defined to properly access the temperatures and nodal heat fluxes on both sides of the
shell elements: one containing the top faces and the other the bottom faces of the shells.
Parallel Processing
In a parallel run with Marc, the finite element mesh is subdivided into domains where each element is part of one domain.
Nodes at the boundary between domains are present in all domains sharing that boundary. Coupling regions are allowed
to span multiple domains. If a physical quantity is prescribed on a such a coupling region, using the
CPLREG_PUT_NODE_VALUES utility routine, the value of the quantity for a given node must be the same in all domains in
CHAPTER 14 1015
which the node appears. Similarly, if the coupling time step is prescribed via CPLREG_PUT_GLOBAL_VALUES, the value must
be the same in all domains.
Marc Co-simulation Adapter

There are two types of Marc Co-simulation Adapter. The first one is called ACSI (Adams Co-Simulation Interface) adapter
which is called from within the Adams environment. This adapter only supports Marc/Adams co-simulation. The second
one is called CoSim adapter which is called from MSC CoSim GUI. This adapter supports Marc/Adams, Marc/SCflow
and Marc/Adams/SCflow co-simulation.
ACSI Adapter
The Adams Co-Simulation Interface (ACSI) provides a mechanism to solve problems dominated by rigid body motion
such as automotive ride and handling but require higher fidelity nonlinear analysis to accurately capture nonlinear behavior
in parts such as suspension systems. By utilizing Adams, accurate boundary conditions are transmitted to the component
or assembly which is modeled with Marc. In a reciprocal manner, by incorporating Marc to simulate the nonlinear
behavior, the strain energy is accurately captured, and the gross deformations are more accurately captured in Adams. It is
computationally beneficial to use this hybrid approach as opposed to a full finite element model when the duration of the
event is long and/or the number of elements required would be computationally prohibitive.
The definition of the model has three components to it: the standard Marc input file (jid.dat), the standard Adams input
file (kit.amd), and the configuration file. The configuration file is described in the Adams Co-Simulation Interface
document. The key is that a GFORCE is associated with an Adams marker. The configuration file associates the marker with
a Marc node. Fixed displacements on all six degrees of freedom are prescribed on the Marc interface node. The CO_SIM
parameter is required to activate this capability in Marc.
The ACSI is a powerful tool that links the two programs as idealized in the following figures. In Figure 14-3, the finite
element model solved by Marc transmits a Load (Force and Moment) to an Adams Marker, and the rigid model transmits
a Displacement (Translation and Rotation) to the Marc. One can observe that there can be multiple Marc parts that can
be solved either serially or in parallel. Here the circles represent traditional Adams markers and the dark circles represent
interaction between Adams and Marc.

Figure 14-3 Representation of MBD – FE Interaction Systems
In Figure 14-4, demonstrates an unfeasible configuration where the load path between Marc bodies is given by an interaction
node. One cannot use an interaction node between two Marc meshes. Interaction nodes are only used between an Adams
body and a Marc body.

Figure 14-4 Representation of MBD – FE Interaction Systems and FE-FE Interaction Systems
Figure 14-5 demonstrates an acceptable configuration where the load path between the Marc meshes is done using contact.
Contact
Figure 14-5 Representation of MBD-FE Interaction and FE-FE Interaction via Contact
In the example below, two contact bodies are shown with an Adam part. Note that the Marc bodies may come into self
contact. There are mlultiple ways of solving this model.
1. Put all the Marc elements into one input file with the retained nodes of the RBE2 interacting with the Adams part.
2. Put cbody1 with its RBE2 into one input file, cbody2 with its RBE2 into another input file, and run them either
serially or in parallel.
3. Put all of the Marc model into a single input file and use DDM.
CHAPTER 14 1017
Figure 14-6 Representation of MBD-FE Interaction and FE-FE Self Contact
This release of the co-simulation technology allows all of the capabilities of Marc structural and thermal-mechanically
coupled analysis to be incorporated into the combined system with the exception of local and global adaptive meshing.
Either the Adams, Adams Car, or Adams Machinery may be used in the co-simulation. There are no limitations on the
Adams model. Only the Adams C++ solver is supported.
Communication Architecture
The Glue code (also known as, master code) implements a simple control algorithm allowing a variable asynchronous
communication of the Adams and Marc codes (also known as, slave codes) at each integration step. There are no restrictions
imposed on the size of the time steps taken by both codes. The sole management is done to keep Adams running first and
Marc running last. In some cases, the Glue code allows Marc to go slightly ahead of Adams. All slave codes simulate using
their best settings. All slave codes exchange data with the master code at the end of every successful integration time step.
In principle, all codes may have different integration steps.
During dynamic co-simulation, the communication with the Glue code is constant at the end of every integration step.
This means the communication is more frequent if the co-simulation solution is difficult to compute. Similarly, the
communication with the Glue code will be less frequent when the codes are taking larger step sizes.
Co-simulation Setup
The Glue code controls the starting up and shutdown of all slave codes. Users need to follow the prompts to launch Adams
and Marc. The topology is defined using a simple configuration script. Adams and Marc models need not to be exported
into some other format; instead, users need to instrument the models following simple rules. Models can be modified
quickly, and a new co-simulation can be started immediately.
Interpolation/Extrapolation Algorithms
There is a set of supported interpolation and extrapolation algorithms which is core to all co-simulation codes. The Glue
code supports different interpolation/extrapolation algorithms for forces and displacements. The ACSI supports the
following algorithms:
Quadratic
This option uses the past history of a signal (force or displacement) to exactly fit a quadratic curve through the last three
available points in the history of the signal. The Glue code keeps a history of all transmitted data from all the slave codes.
To use this option, set interpolate_order_u=2 and/or interpolate_order_f=2 in the configuration script.
Linear
This option uses a linear fit using a weighted slope s=s1*w + s2*(1-w) where s1 is the slope of the latest data time interval
and s2 is the slope of the previous step. The weight w is provided by the user. For example, to use this option for
displacements set interpolate_order_u=1 and say linear_weight_u=0.788 in the configuration script. Option
linear_weight_u sets the weight w above. Similar options exist to set the interpolation/extrapolation for forces (see
interpolate_order_f and linear_weight_f).
Constant last
This option uses the latest value in the history of the signal. To use this option, set interpolation_order_u=0 and/or
interpolation_order_f=0 in the configuration script.
Linear least squares

This option uses a least square linear fit of the last three values in the history of a signal. To use this option, set
interpolation_order_u=-1 and/or interpolation_order_f=-1 in the configuration script.
Constant average
This option uses the average of the last three values. This option is equivalent to using a constant least square fit on the
data. To use this option, set interpolation_order_u=-2 and/or interpolation_order_f=-2 in the
configuration script.
Notes: 1. The above algorithms are supported for the interpolation/extrapolation of forces and translational
displacements only. For rotational displacements, a SLERP algorithm is used to interpolate rotational
displacements [Ref. 2].
2. At start up, when the history of a signal does not have enough points for the selected interpolation
algorithm, a simpler method is used. For example, a constant or linear interpolation is used.
Co-simulation Algorithm
In the co-simulation implementation Adams "goes first" using predicted forcing values from Marc. Later on, computed
displacements by Adams are enforced in Marc using interpolated values.
For clarity, assume both Adams and Marc are at simulation time t 1 and both are ready to move ahead in time and both
are going to take the same integration step h. Figure 14-7 (a) and (b) show plots of the current force values ( f c and f p ) and
displacement ( u ) values up to time t 1 at one interaction point.
Figure 14-7 (a) shows the status of the Adams computation at time t 1 . Curve u is the computed displacement done by
Adams while f p is the predicted force used by Adams. We have to emphasize the point that f p is not the force computed
by Marc but values predicted by the Glue code using the computed values f c by Marc. More details follow.
Figure 14-7 (b) shows the status of the Marc computation at time t 1 . Curve u is the enforced displacement imposed by
Adams; this curve matches the curve in Figure 14-7 (a). Curve f c is the computed force by Marc at the interaction point up
to time t 1 . Notice point P stands for the predicted value of the force at time t 1 + h . This prediction is done by the Glue
code (the Glue code keeps a history of all the force values from Marc.) The predicted value P will be used by Adams when
CHAPTER 14 1019
starting to move ahead first. Notice the predicted value is obtained by extrapolation of the computed values by Marc (see
dotted line in Figure 14-7 (b)).
Figure 14-7 Adams and Marc at Time t 1 Ready to Move Ahead in Time
Figure 14-8 (a) and (b) show the status of the co-simulation after both codes already took their respective simulation steps
advancing in time. First, Adams advances its simulation using the extrapolated value P for the force at the interaction point
(see Figure 14-8 (a).) Notice the value of P in the plot for f p . Next, Marc takes its simulation step using interpolated data
u from Adams and computes the force C at the interaction point. The difference between the predicted value P and the
computed value C is a measure of the error in the co-simulation.
Figure 14-8 Adams and Marc After Moving Ahead One Step
Error Control
Two conditions are constantly monitored for all components of forces and moments at all interaction points. When the
absolute value difference between P and C is larger than an absolute error e a and the relative difference is bigger than a
relative error e r , then a warning message is printed; nevertheless, the co-simulation does not stop. Both e a and e r can be
set by using the options force_absolute_error and/or force_relative_error in the configuration script. If
both P and C are zero, the relative comparison is skipped.
The warning message includes the interaction name (read from the configuration script) and the force/moment component
name (for example, F y ).
Diagnostic Messages
The Glue code creates a message file with extension *.g_msg showing basic co-simulation data and prompts to the user.
The Adams run creates an additional message file with extension *.a_msg. Finally, the Marc run also creates an additional
message file with extension *.m_msg. All three types of message files have the same root name as the configuration script.
CoSim Adapter
The CoSim adapter uses similar communication architecture as that of ACSI (Adams Co-Simulation Interface). Data
communication between solvers is mediated by a server program, called MSC CoSim (aka Glue). The type of data
communicated between the solver is shown in Figure 14-9. At this moment co-simulation with Adams and/or scFlow are
supported. For more details information on how to use MSC CoSim, please refer to its product documents.
The capability of CoSim adapter is the same as that of the ACSI for Adams-Marc co-simulation. For nonlinear FEA model,
Marc may require cut-back to achieve convergence. The advantage of the CoSim adapter is that the reduced stiffness
matrices are exchanged to Adams during the cu-back which would improve not only the accuracy but also the robustness
of the solution.
The co-simulation with scFlow extends the capability of Marc to solve FSI problem. Both fluid-structure and
fluid/thermal-structure simulation is supported. Large structural deformation simulation, like a sloshing stability of liquid
tank under earthquake loading, requires a special algorithm due to highly non-linear interaction between the liquid and
the thin structure. CoSim adapter uses iteration process during the data communication between Marc and scFlow. A mass
scaling is also introduced to improve the robustness of the solution. For most FSI problems, these extra features are not
necessary.
CoSim adapter supports Adams-Marc-SCflow co-simulation. A class of problems, where coupling of rigid body dynamics,
fluid/thermal flow and nonlinear flexible structure is required, can now be simulated using MSC CoSim.
CHAPTER 14 1021
Figure 14-9 Type of Data Communicated between Adams-Marc-scFlow
References
1. Fraunhofer Institute for Algorithms and Scientific Computing SCAI. MpCCI 3.0.4: Manuals and Tutorials. April
25, 2005, http://www.scai.fraunhofer.de/mpcci.
2. http://en.wikipedia.org/wiki/Slerp.
Appendix A: Finite Element Technology in Marc
A Finite Element Technology

in Marc
 Governing Equations of Various Structural Procedures 1017


System and Element Stiffness Matrices 1019

Load Vectors 1020
 References 1021
APPENDIX A 1017
Finite Element Technology in Marc
Governing Equations of Various Structural Procedures

This section describes the basic concepts of finite element technology. For more information, you are urged to refer to the
finite element text [Ref. 5]. Marc was developed on the basis of the displacement method. The stiffness methodology used
in Marc addresses force-displacement relations through the stiffness of the system. The force-displacement relation for a
linear static problem can be expressed as:
Ku = f (1-1)
where K is the system stiffness matrix, u is the nodal displacement, and f is the force vector.
Assuming that the structure has prescribed boundary conditions both in displacements and forces, the governing Equation
(1-1) can be written as:
K 11 K 12  u 1   f1 
  =   (1-2)
K 21 K 22  u 2   f2 
u 1 is the unknown displacement vector, f 1 is the prescribed force vector, u 2 is the prescribed displacement vector, and
f 2 is the reaction force. After solving for the displacement vector u , the strains in each element can be calculated from the
strain-displacement relation in terms of element nodal displacement as:
 e l = u el (1-3)
The stresses in the element are obtained from the stress-strain relations as:
 e l = L el (1-4)
where  el and  el are stresses and strains in the elements, and u el is the displacement vector associated with the element
nodal points;  and L are strain-displacement and stress-strain relations, respectively.
In a dynamic problem, the effects of mass and damping must be included in the system. The equation governing a linear
dynamic system is:
Mu·· + Du· + Ku = f (1-5)
where M is the system mass matrix, D is the damping matrix, Equation (1-6) is the acceleration vector, and u is the velocity
vector. The equation governing an undamped dynamic system is:
Mu·· + Ku = f (1-6)
The equation governing undamped free vibration is:
Mu·· + Ku = 0 (1-7)
Natural frequencies and modal shapes of the structural system are calculated using this equation.
2
K –  M = 0 (1-8)
The equations governing some other procedures are similar. For example, the governing equation of transient heat transfer
analysis is:
·
CTT + KTT = Q (1-9)
where C T is the heat capacity matrix, K T is the thermal conductivity matrix, Q is the thermal load vector (flux), T is
the nodal temperature vector, and t is the time derivative of the temperature. Equation (1-9) reduces to:
KTT = Q (1-10)
for the steady-state problem. Note that the equation governing steady-state heat transfer (Equation (1-10)) and the equation
of static stress analysis (Equation (1-1)) take the same form. Similarly, the hydrodynamic bearing problem is analogous to a
steady-state heat transfer problem. This problem is governed by an equation similar to Equation (1-10).
The matrix equation of the electrical problem in coupled thermo-electrical analysis is:
K E  T V = I (1-11)
The equation governing the thermal problem is:
· E
C T  T T + K T  T T = Q + Q (1-12)
in both Equation (1-11) and Equation (1-12), v is the voltage, K E  T  is the temperature-dependent electrical-conductivity
matrix, I is the nodal-current vector, C T  T  is the temperature-dependent, heat-capacity matrix, and K T  T  is the
E
thermal-conductivity matrix. T is the nodal temperature vector, Q is the heat-flux vector, and Q is the internal heat-
generation vector that results from the electrical current. Electrical and thermal problems are coupled through K E  T  and
E
Q .
The matrix equations for the thermal-mechanical problem are as follows:
Mu·· + Du· + K  T ,u ,t u = F (1-13)

· I
C T  T T + K T  T T = Q + Q + Q F (1-14)
In Equations (1-13) and (1-14) the damping matrix D , stiffness matrix K , heat-capacity matrix C T and thermal-
I
conductivity matrix K T are all dependent on temperature. Q is the internal heat generated due to inelastic deformation.
The coupling between the heat transfer problem and the mechanical problem is due to the temperature-dependent
mechanical properties and the internal heat generated. If an updated Lagrangian analysis is performed, K and K T are
dependent upon prior displacement.
The governing equations described above are either sets of algebraic equations (Equations (1-1), (1-10), and (1-11)), or sets of
ordinary differential equations (Equations (1-5) through (1-8) and all equations through Equation (1-4)). The time variable
is a continuous variable for the ordinary differential equations. Select an integration operator (for example, Newmark-beta,
Houbolt, or central difference for dynamic problems, and backward difference for heat transfer) to reduce the set of
APPENDIX A 1019
differential equations to a set of algebraic equations. The final form of governing equations of all analysis procedures is,
therefore, a set of algebraic equations.
System and Element Stiffness Matrices

The previous section presented system matrices assembled from element matrices in the system. For example, the system
el
stiffness matrix K is expressed in terms of the element stiffness matrix K i as:
N
el
K =  Ki (1-15)
i = 1
where N is the number of elements in the system. The system stiffness matrix is a symmetric-banded matrix. See Figure B-1
for a schematic of the assemblage of element matrices. When the element-by-element iterative solver is used, the total
stiffness matrix is never assembled.
1 2
2 3
1
(a) 3 4 5
4 5
6 8
7
1 2 3 4 5
(b) 
[K] {F}
(c)

Band
Figure B-1 Schematic of Matrix Assemblage

The element stiffness matrix can be expressed as

el T el
K =   Ldv (2-1)
el
v
el
where v is the volume of the element,  is the strain-displacement relation, and L is the stress-strain relation.
 = u (2-2)
 = L (2-3)
The mass matrix can be expressed as:
el T
M =  N Ndv (2-4)
el
v
The integration in (2-1) is carried out numerically in Marc, and is dependent on the selection of integration points. The
element stiffness matrix can be fully or under integrated. The mass matrix is always fully integrated.
Note: The  matrix is directly associated with the shape functions and geometry of each element. Shape functions
associated with different element types are explained in Marc Volume B: Element Library . Stress-strain relations L
are discussed in Chapter 5: Material Library of this volume.
Load Vectors
The nodal force vector f in (1-1) includes the contributions of various types of loading.
f = f po int + f s urface + f b od y + f (2-5)
where f point is the point load vector, f surface is the surface load vector, f body is the body (volumetric) load vector, and f*
represents all other types of load vectors (for example, thermal and creep strains, and initial stress).
The point load is associated with nodal degrees-of-freedom and can be added to the nodal force vector directly. Equivalent
nodal force vectors f surface , f body must be calculated from the distributed (surface/volumetric) load first and then added
to the nodal force vector. In Marc, the computation of equivalent nodal forces is carried out through numerical integration
of the distributed load over the surface area or volume to which the load is applied. This can be expressed as:
T
f su rfa ce = N pdA (2-6)
A
T
f bo d y = N pbdV (2-7)
V
where p is the surface load and b the volumetric load.

APPENDIX A 1021
Figure B-1 shows the assemblage of the nodal force vector.
References
1. Zienkiewicz, O. C. , R. L. Taylor and J.Z. Zhu The Finite Element Method: Its Basis and Fundamentals (6th ed.),
Elsevier, Oxford, 2005.
2. Bathe, K. J. Finite Element Procedures, Prentice-Hall, Englewood Cliffs, NJ, 1995.
3. Hughes, T. J. R. The Finite Element Method–Linear Static and Dynamic Finite Element Analysis, Dover Publications,
New York, 2000.
4. Ogden, R. W. “Large Deformation Isotropic Elasticity: On The Correlation of Theory and Experiment for
Incompressible Rubberlike Solids,” Proceedings of the Royal Society, Vol. A (326), pp. 565-584, 1972.
5. Cook, R. D., D. S. Malkus, and M. E. Plesha, Concepts and Applications of Finite Element Analysis (4th ed.), John
Wiley & Sons, New York, NY, 2001.
Appendix B: Finite Element Analysis of NC Machining Processes
B Finite Element Analysis of

NC Machining Processes
 General Description 1023


NC Files (Cutter Shape and Cutter Path Definition) 1023

Intersection Between Finite Element Mesh and Cutter 1025
 Deactivation of Elements 1025
 Adaptive Remeshing 1025

Input of Initial Stresses 1026
APPENDIX B 1023
Finite Element Analysis of NC Machining Processes
General Description
Machining (or Metal Cutting) is a type of material removal processes widely used to produce a part with the desired
geometry. After removal of the machined material, re-establishment of equilibrium within the remaining part of the
structure causes some distortion due to the relief of the residual stresses in the machined materials.
Depending on the stress level and geometry of the part, the severe distortion due to machining can result in high scrap rates
and increased manufacturing costs, especially for structures with thin wall or large thin plates.
A numerical analysis procedure has been developed in Marc to predict the distortion during the 3-D bulk machining so
that engineers can optimize their structure design and machining processes. This procedure is based on the assumption that
effects introduced by the cutting process are local and small compared to those due to pre-existing residual stresses.
The simulation procedure uses the cutter shape and path information obtained from NC machining data. Cutter motion
information is used to predict the material to be removed by automatically deleting finite elements that are located within
the cutter path. The distortion of the remaining part can be predicted by re-calculating the equilibrium. An interface has
been developed in Marc to convert cutter path data into the finite elements to be removed. The cutter path data is stored
in either the Automatically Programmed Tools (APT) source or the Cutter Location (CL) data format. The APT source is
the NC data output by CAD software CATIA. The CL data is the cutter location data provided by APT compilers.
For details on how to use this functionality, refer to Marc Volume C: Program Input for the MACHINING parameter and the
DEACTIVATE model definition option and the DEACTIVATE history definition option.
NC Files (Cutter Shape and Cutter Path Definition)

The cutter shape and the cutter path are the key parameters to determining the volume of the material to be cut. The cutter
shape is defined by the CUTTER statement in the APT source or CL files. The expanded form of the CUTTER statement is
as follows:
CUTTER  d ,r ,e ,f , , ,h
where the parameters
d The tool diameter.

r The radius of the corner circle; it can be zero or larger than d  2 .
e The radial distance from the tool axis to the center of the corner circle.
f The distance from the tool endpoint to the center of the corner circle measured parallel with the tool axis.
The angle from a radial line through the tool endpoint to the lower line segment; it is in range of zero to 90°, and

positive.
 The angle between the upper segment and the tool axis; it is in the ranges of -90° to 90°.
h The cutter height measured from the tool endpoint along the tool axis.
have the following meaning as shown in Figure 2-1.

Tool axis
Upper Line
segment
Envelope h
e
Lower line
r segment
f

Endpoint
d
Figure 2-1 Definition of Cutter Geometry
The cutter path is calculated based on the motion of the cutter and cutter shape. The motion is defined by cutter motion
statements like GOTO/, GODLTA/, CYCLE/, and DRILL/, etc. In Marc, these motion types are translated into point-to-
point, circular, or drilling motion modes, respectively. For details, refer to references by Irvin H. Kral and Dassault Systems.
Notes: 1. Marc accepts APT data files written by CATIA V4 and the corresponding CL data files by APT
compilers.
2. The cutter can be of either multi-axis or fixed axis.
3. Two cutter motion types: Point-To-Point (for example, GOTO/ and GODLTA/ statements) and
Cycle (DRILL/…) are directly supported. Circular motions (either in-plane or out-of-plane) can be
interpolated into Point-To-Point motions. The interpolation can be done inside CATIA by
choosing the proper option when writing out APT source file.
4. In addition, some statements in the APT/CL files are skipped. If the user chooses statements like
TRACUT/, COPY/, /MACRO, or /CALL to define cutter motions, they must be converted into
explicit motion statements like Point-to-Point or Cycle (DRILL/) motions before writing out the
APT data file. The cutting process is stopped if statements, such as /FINI, /STOP, or /END, or the
end of APT/CL file is reached.
5. It is assumed that the part and cutter are defined in the same coordinate system. So, if a part needs
to be flipped over between two machining stages, the user can rotate the cutter but keep the part
unchanged in space.
6. If the user wants to change the boundary conditions for some nodes during the cutting process
defined by one APT/CL file, time synchronization should be turned on when defining the
machining loadcase.
APPENDIX B 1025
7. For visualization purpose, the user can specify a contact surface as cutter surface. During
postprocessing, user will be able to visualize the cutter motion along the cutting process. In order to
do this, you MUST exclude this cutter surface ID from all kinds of contact checking during analysis.
The user can do this by creating CONTACT TABLE and turn off the contact body ID of the cutter
surface. In addition, it is also necessary to make sure that the cutter surface is initially set at the
location of the origin of the coordinate system and the orientation is initially set to be in the
direction of Z-axis.
Intersection Between Finite Element Mesh and Cutter

When the cutter moves, its path forms the volume of the material to be cut. This volume is calculated based on the cutter
geometry and its path in space. The finite element mesh is checked against this volume to decide the intersection between
the volume and the finite element nodes and elements.
Deactivation of Elements
The automatic deactivation of finite elements depends on the intersection status between the cutter path and the mesh.
The approach is designed for elements that are fully or partially located inside the cut volume. Elements that are fully
located inside the cut volume are automatically deactivated and the stress that existed before deactivation is released. For
elements that are partially intersected by the cut volume, integration points are checked. In the current version, the check
of integration points is limited to the element centroid only. If element centroid is found inside the cut volume, the
associated element is deactivated. For deactivated elements, the stresses/strains are set to zero permanently.
After the deactivation of elements along the motion of cutter, the part will distort progressively. The updated geometry of
the part is used for the next cutting step. Using this method, the whole machining process is simulated with Marc. After
the machining is finished, the final distortion of the work piece can be calculated by spring back analysis. The spring back
analysis is conducted by removing all fixturing restraints except for those needed to remove rigid body motion.
Adaptive Remeshing
Generally speaking, the accuracy of the finite element analysis of machining processes depends on, plays a key role to get
the accuracy of initial (residual) stresses. Accurate initial stress the correct deformation of the part. From the view point of
FE analysis, the initial FE mesh may affect the accuracy significantly. Finer mesh is needed for accuracy but it results in
higher analysis cost due to the increased number of finite elements and increased total number of analysis increments. Also,
for most machining processes, the cutter path is complicated enough that it is not possible to properly define the best suited
fine mesh at the beginning of the analysis. The adaptive remeshing technology has been enhanced in Marc to obtain
accurate results for the FE mesh located inside the cutter path.
An element is subdivided if the cutter only cuts part of the element in the current cutting step. For example, one 3-D brick
element generates 8 new elements after first level subdivision and 64 new elements after the second level subdivision.
Therefore, users need to be aware of the potentially large number of new elements with increased levels of subdivision.
Typical Features for Machining

1. Multiple level subdivision of element in one incremental step:
Typically, Marc subdivides only once per load step. If the 17th criterion of ADAPTIVE model definition option is used
(for NC machining), Marc automatically subdivides the element until the maximum specified level has been
reached. This process may generate a huge number of new elements and nodes. Consequently, these newly created
elements and nodes may result in excessive memory usage and CPU time, though most of them are immediately
cut off. For some problems, it is possible to cut the part with a few large elements but without sacrifice of accuracy
if only allowing remeshing to be conducted at some specific locations of the part where finer meshing is required.
In such circumstances, it might be reasonable to conduct one round of rough cutting followed by fine cutting based
on finer mesh after adaptive remeshing. Details are described in Item 3.
2. ADAPTIVE REMSHING at EACH increment:
At the loadcase of machining, Marc does adaptive remeshing at each increment based on the intersection of the
cuter path and the FE mesh.
3. ADAPTIVE REMSHING at LAST increment:
A cutting analysis in Marc is not done without adaptive remeshing until the cutting process is completed. However,
the cutting process may not be completely finished yet because some elements that are partially cut by cutter are
still retained. In order to cut those elements off, the cutter path is re-calculated and those elements that are partially
cut are subdivided in order to improve the accuracy of the cutting. In this case, extra incremental steps are added
into this machining loadcase. The added number of extra load steps equal the maximum subdivision level that is
defined by the ADAPTIVE option.
Input of Initial Stresses

There are various approaches to input the initial stresses into the FE model for machining analysis. Generally speaking,
Marc allows user to input the initial stress through the INIT STRESS model definition option. It also allows user to import
the data through the PRE STATE option. With the PRE STATE option, the user can transfer information, including strain, stress
etc., from previous analysis into the model for new analysis.
By using option of INIT STRESS, user may define the stress based on the input values or tables (if the stresses are defined as
function of, say, coordinates). Marc also allows the user to define the initial stresses with an ASCII format text file. The text
file contains the raw initial stress data of points randomly located in a space field. In Marc, the given initial stress data is
processed and interpolated into the finite element mesh according to the location of each element. For details about this
data structure and format, see INIT STRESS model definition option in Marc Volume C: Program Input.
References
1. Irvin H. Kral, Numerical Control Programming in APT, Prentice-Hall, 1986.
2. Dassault Systems, CATIA Training Guide, Version 4, Release 1.6, April, 1996.
3. Houtzeel Manufacturing systems, Inc., APT Language Reference Manual, March, 1997.
APPENDIX B 1027
MESH2D
MESH2D generates a mesh of quadrilateral or triangular elements for a two-dimensional body of any shape. The generated
mesh is written to a separate file and must be read with the CONNECTIVITY, COORDINATES, and FIXED DISP, etc., options.
Block Definition
In MESH2D, a physical object or domain is divided into quadrilateral and/or triangular parts, called blocks. Quadrilateral
blocks are created by Marc by mapping with polynomials of the third order from a unit square. These blocks can, therefore,
be used to approximate curved boundaries. The geometry of a quadrilateral block is defined by the coordinates on 12 nodes
shown in Figure 2-2. If the interior nodes on an edge of the block are equal to zero or are not specified, the edge of the block
is straight.
Triangular blocks have straight edges. The geometry of a triangular block is defined by the coordinates of the three vertices.
Merging of Nodes
Marc creates each block with a unique numbering scheme. The MERGE option fuses all nodes that lie within a small circle,
renumbers the nodes in sequence, and then removes all gaps in the numbering system. You can select which blocks are to
be merged together, or you can request that all blocks be merged. You must give the closeness distance for which nodes will
be merged.
 P3
P9
y P4 P10
4 10 9 3 P8
P11 P7
11 8 P12 P2
2  P6
P1
P5
12 7 x
1 5 6 2 12
x   ,  =  x  P i  i   , 
i = 1
2
12
y   ,  =  y  P i  i   , 
i = 1
Figure 2-2 Typical Quadrilateral Block Mapping
Block Types
MESH2D generates four types of quadrilateral blocks.
1. Block Type 1:
Is a quadrilateral block that is covered by a regular grid. The program obtains this grid by dividing the block edge
into M by N intervals. Figure 2-3 illustrates the division of block edges into intervals with M = 4, N = 3.
The P1 P4 face of the block is the 1-4 face of triangular elements and the P1 P2 face of the block becomes the 1-
2 face of quadrilateral elements.


P4 P3
16 17 18 19 20
24 16 17 18 19 20
2/3
17 12 13 14
2/3 9 10 11 12
11 15
N = 3 2/3 16 11 12 13 14 15

9 7 8 9
N = 3 2/3 5 6 7 8 
6 10
2 4 5 8 6 7 8 9 10
2/3 P2
1 3 6 7 2/3 1 2 3 4
1 2 3 4 5 1 2 3 4 5
P1
1/2 1/2 1/2 1/2 1/2 1/2 1/2 1/2
M=4 M=4
2. Block Type 2:
Is a quadrilateral block that allows the transition of a coarse mesh to a finer one. In one direction, the block is
divided into M, 2M, 4M ...; while in the other direction, the block is divided into N intervals. Figure 2-4 illustrates
the division of Block Type 2 edges into intervals.
APPENDIX B 1029
The P1 P2 face of the block becomes the 1-2 face of quadrilateral elements, and the P2 P3 face of the block is the
2-3 face of triangular elements.

P4 P3
19
18 34 
2/7
9 17 1/2 1/2 1/2 1/2
7 10 11 12 13 14 15 16
4/7
8 7 8 9 10 11
M=2 5 6 7
4 8
N=3 x 1 3 4 6
1 3 4 6
4 5 6 x
3/7 2
2 5 2 5
1 2 3
P1 1 2 3 P2
1 1
1 1

3. Block Type 3:
Is a triangular block. The program obtains the mesh for this block by dividing each side into N equal intervals.
Figure 2-5 illustrates Block Type 3 for triangular and quadrilateral elements.
P3 P3
15 10
13
16
10 6
14 8 9
P1 6 11
8
1 2 7 12 4 5
1 6 7
3 4 8 5
2
9 P1 1 2 3
3 P2
7
4 P2
1 2 3 4
N=4 5
4. Block Type 4:
Is a refine operation about a single node of a block (Figure 2-6). The values of N and M are not used.
If quadrilateral elements are used in a triangular block, the element near the P2 P3 face of the block is collapsed by
MESH2D in every row. The P1 P2 face of the block is the 1-2 face of the generated elements.
5 6 7
4 3
4
3
1 2
1 2
Symmetry, Weighting, and Constraints

MESH2D contains several features that facilitate the generation of a mesh: use of symmetries, generation of weighted
meshes, and constraints. These features are discussed below.
MESH2D can use symmetries in physical bodies during block generation. An axis of symmetry is defined by the
coordinates of one nodal point, and the component of a vector on the axis. One block can be reflected across many axes to
form the domain. Figure 2-7 illustrates the symmetry features of MESH2D.
APPENDIX B 1031
2
1 1
Original Block One Symmetry Axis
3 2 3 2
4 1 4 1
5 8
6 7
Two Symmetry Axes Three Symmetry Axes

Figure 2-7 Symmetry Option Example
A weighted mesh is generated by the program by spacing the two intermediate points along the length of a boundary. This
technique biases the mesh in a way that is similar to the weighting of the boundary points. This is performed according to
the third order isoparametric mapping function.
Note: If a weighted mesh is to be generated, be cautious not to move the interior boundary points excessively. If the
points are moved more than 1/6 of the block length from the 1/3 positions, the generated elements can turn
inside-out.
The CONSTRAINT option generates boundary condition restraints for a particular degree of freedom for all nodes on one side
of a block. The option then writes the constraints into the file after it writes the coordinate data. The FIXED DISP, etc., option
must be used to read the boundary conditions generated from the file.
Additional Options
Occasionally, you might want to position nodes at specific locations. The coordinates of these nodes are entered explicitly
and substituted for the coordinates calculated by the program. This is performed using the SPECIFIED NODES option. Some
additional options in MESH2D are as follows:
 MESH2D can be used several times within one input file.
 The START NUMBER option gives starting node and element numbers.
 The CONNECT option allows forced connections and/or disconnections with other blocks. This option is useful
when the final mesh has cracks, tying, or gaps between two parts.
 The MANY TYPES option specifies different element types.
FXORD Option
The FXORD model definition option (Volume C: Program Input) generates doubly curved shell elements of element type 4,
8, or 24 for the geometries most frequently found in shell analysis. Since the mathematical form of the surface is well-
defined, the program can generate the 11 or 14 nodal coordinates needed by element type 8, 24, or 4 to fit a doubly curved
surface from a reduced set of coordinates. For example, you can generate an axisymmetric shell by entering only four
coordinates per node. The FXORD option automatically generates the complete set of coordinates required by the elements
in the program from the mathematical form of the surface.
A rotation and translation option is available for all components of the surface to give complete generality to the surface
generation. The input to FXORD consists of the reduced set of coordinates given in a local coordinate system and a set of
coordinates which orient the local system with respect to the global system used in the analysis. The program uses these
two sets of coordinates to generate a structure made up of several shell components for analysis.
The FXORD option allows for the generation of several types of geometries. Because you may need to analyze shells with well-
defined surfaces not available in this option, you can use the UFXORD user subroutine to perform your own coordinate
generation (Volume D: User Subroutines and Special Routines). The FXORD option can also be used to convert cylindrical
coordinates or spherical coordinates to Cartesian coordinates for continuum elements.
Major Classes of the FXORD Option

The cases are considered are as follows:
 Shallow Shell (Type I)
 Axisymmetric Shell (Type 2)
 Cylindrical Shell Panel (Type 3)
 Circular Cylinder (Type 4)
 Plate (Type 5)
 Curved Circular Cylinder (Type 6)
 Convert Cylindrical to Cartesian (Type 7)
 Convert Spherical to Cartesian (Type 8)
Shallow Shell (Type I)

Type 1 is a shallow shell with:
 1 = x 1 2 = x2 (2-1)
The middle surface of Figure 2-8 (Type 1) is defined by an equation of the form:
x 3 = x 3  x 1 x 2  (2-2)
and the surface is determined when the following information is given at each node:
APPENDIX B 1033
x 3 x 3  2 x 3
x 1 x 2 x 3 --------- --------- ------------------ (2-3)
x 1 x 2 x 1 x 2
The last coordinate is only necessary for Element Type 4.

X3
X3
R
f
X2
X2
q
X1
Type 1 Type 2
X3
X2
R
X2 
3
X1
X3
X1
Type 4
X3
Type 3
Figure 2-8 Classification of Shells
Axisymmetric Shell (Type 2)

The middle surface symmetric to the x 3 axis (Figure 2-8, Type 2) is defined as:
x 1 = R    cos  cos 
x 2 = R    cos  sin 
(2-4)
x 3 = R    sin 
where  and  are the angles shown in Figure 2-8. In this case, the surface is defined by:
dR
  R ------- (2-5)
d
The angles  and  are given in degrees.
Cylindrical Shell Panel (Type 3)

The middle surface is the cylinder defined by Figure 2-8:
x1 = x1  s 
x2 = x2  s  (2-6)
x3 = x3
The nodal geometric data required is:

dx 1 dx 2
s x 3 x 1 x 2 --------- --------- (2-7)
ds ds
Circular Cylinder (Type 4)

This is the particular case of Type 3 where the curve,
x 1  s  x 2  s  (2-8)
is the circle given by Figure 2-8 (Type 4).
x 1 = R cos 
(2-9)
x 2 = R sin 
The only nodal information is now:

 x 3 R (2-10)
Note that  is given in degrees and, because R is constant, it needs to be given for the first nodal point only.
Plate (Type 5)
The shell is degenerated into the plate,
x3 = 0 (2-11)
The data is reduced to,

x 1 x 2 (2-12)
Curved Circular Cylinder (Type 6)

Figure 2-9 illustrates this type of geometry.
APPENDIX B 1035
X3
Shell Middle Surface
q1
q2 R
f
X2
q
X1
Type 6
Figure 2-9 Curved Circular Cylinder
The middle surface of the shell is defined by the equations as:

x 1 = r cos 
x 2 = r sin  cos  + R  1 – cos   (2-13)
x 3 =  R + – r sin  r sin   sin 
The Gaussian coordinates on the surface are:

 1 = r
(2-14)
 2 = R
and form an orthonormal coordinate system. The nodal point information is:
  r R (2-15)
 and  in degrees. You need to specify the radii r and R only for the first nodal point.
Convert Cylindrical to Cartesian (Type 7)

Type 7 allows you to enter the coordinates for continuum elements in cylindrical coordinates, which are converted by Marc
to Cartesian coordinates. In this way, you can enter R ,  , Z and obtain x, y, z where  is given in degrees and,
x = R cos 
y = R sin  (2-16)
z = Z
Convert Spherical to Cartesian (Type 8)

Type 8 allows you to enter the coordinates for continuum elements in spherical coordinates, which are converted by Marc
to Cartesian coordinates. In this way, you can enter R ,  ,  and obtain x, y, z where  and  are given in degrees and,
x = R cos  sin 
y = R sin  cos  (2-17)
z = R cos 
Recommendations on Use of the FXORD Option

When a continuous surface has a line of discontinuity, for example, a complete cylinder at  = 0 = 360 , you must
place two nodes at each nodal location on the line to allow the distinct coordinate to be input. You must use tying element
type 100 to join the degrees of freedom. Generally, when different surfaces come together, you must use the intersecting
shell tyings.
The FXORD option cannot precede the COORDINATES option, because it uses input from that option.
Incremental Mesh Generators

Incremental mesh generators are a collection of options available in Marc to assist you in generating the mesh. Incremental
mesh generators generate connectivity lists by repeating patterns and generate nodal coordinates by interpolation. Use these
options directly during the model definition phase of the input.
During the model definition phase, you can often divide the structure into regions, or blocks, for which a particular mesh
pattern can be easily generated. This mesh pattern is established for each region and is associated with a single element
connectivity list. Use the CONNECTIVITY option to input this element connectivity list. The incremental mesh generators then
generate the remainder of the connectivity lists.
Critical nodes define the outline of the regions to be analyzed. Use the COORDINATES option to enter the critical nodes. The
incremental mesh generators complete the rest and join the regions by merging nodes.
A special connectivity interpolator option generates midside nodes for elements where these nodes have not been specified
in the original connectivity. A separate mesh generation run is sometimes required to determine the position of these nodes.
This run can be followed by mesh display plotting.
The incremental mesh generators are listed below:
Element Connectivity Generator – The CONN GENER option repeats the pattern of the connectivity data for
previously defined master elements. One element can be removed for each series of elements, allowing the
program to generate a tapered mesh. Two elements can be removed for each series with triangular elements.
Element Connectivity Interpolator – The CONN FILL option completes the connectivity list by generating midside
nodes. You first generate the simpler quadrilateral or brick elements without the midside nodes. You can then fill
in the midside nodes with this option.
Coordinate Generator – The NODE GENER option creates a new set of nodes by copying the spacing of another
specified set of nodes.
APPENDIX B 1037
Coordinate Interpolator – The NODE FILL generates intermediate nodes on a line defined by two end nodes. The
spaces between the nodes can be varied according to a geometric progression.
Coordinate Generation for Circular Arcs – The NODE CIRCLE option generates the coordinates for a series of
nodes which lie on a circular arc.
Appendix C: Material Data Files and Data Encryption
C Material Data Files and Data

Encryption
 External Material Data Files 1028


Data Encryption 1028
External Material Data Files

It is possible to use material data properties by referring to an external file. The extension of the file is .umt. In this file each
complete material definition must be given inside the MAT BEGIN and MAT END option (see MAT BEGIN, MAT END), so
multiple material property definitions can be defined in a single file. Material property entries as discussed in Marc Volume
C can be entered here. A complete material definition must be given so including tables and material options which are
cross-referenced. When e.g. a composite or multi network material is defined all materials which are used in the layers or
in the multi network and its related properties must be defined inside the MAT BEGIN and MAT END option (see MAT
BEGIN, MAT END). They should be written in multi physics style and in extended format. The material properties described
here may contain physics quantities which are not used in the current model. The list of elements should be omitted.
Material id’s and table id’s are made unique inside Marc, so they cannot be accessed from other options in the model. The
reference to the material is done by material name using the READ MATERIAL option.
Data Encryption
A tool is available to encrypt the material data file. This provides a mechanism to protect the material data against
unauthorized use. This encryption makes it possible to allow e.g. subcontractors to use material data in a Marc analysis
without actually seeing the data. The data encryption will support an expiration date (i.e. after a given date, the data can
no longer be used) and unique data from a selected number of users (i.e. only users with a given MSC Customer
Entitlement ID can use the data).
Encrypting the material data file is done via Mentat or the standalone program run_encryption (on Linux) and
run_encryption.bat (on Windows) located in the Marc directory tools. It is a command line tool which prints the
following listing when it is called as run_encryption –help.
Marc encryption program - Version 1
usage: run_encryption key filename [OPTIONS]
key ASCII formatted encryption key, with a maximum of 64 entries.

Name of the file to be encrypted.This file must have the extension .umt and will be
encrypted into a file with the same base name, but with the extension .ume. Likewise, the
filename
key provided above will be stored in a file with the extension .umk. This file is also
encrypted.
options
Name of the file containing CEID's. If this option is used, then only users having an MSC
-ceid filename Customer Entitlement ID from the list of ID's in this file will be able to use the encrypted
file. The CEID's are stored in the encrypted key file.
Expiration date (mm/dd/yyyy). If this option is used, then the encrypted file cannot be used
beyond this date. The expiration date is stored in both the encrypted material file and the
-expdate date
encrypted key file. By default, an encrypted file can be used for a period of one year after it
has been created.
APPENDIX C 1029
Material Data Files and Data Encryption
A new key file is created without encrypting a material file. Use this to e.g. recreate a key file
-keyfile
with other CEID's.
-help Print this help and exit.
When this program is executed it will create a new encrypted material data file (extension .ume) and a file containing the
decryption key (extension .umk). The expiration date is stored in both the encrypted material data file and the key file.
The MSC Customer Entitlement ID’s are stored in the key file (which is encrypted based on a hard-coded key). This allows
one customer to create an encrypted material data file, and have another customer share this material data file to specific
(other) users without exposing the original encryption key and material data.
MSC Nastran Implicit Nonlinear (SOL 600) User’s Guide
Index
ABCDEFGHIJKLMNOPQRSTUVWXYZ
Index eigenvalue 135
electrical 676
INDEX Index
A electristatic-structural 496
electromagnetic 856
ABLATION (parameter) 781
electrostatic 495, 678, 749, 993, 995
acoustic
electrostatic-structural 996
analysis 496, 678, 749, 877, 994, 996
electro-structural 993
element types 877, 896
error 619
input options 678
fluid 496, 750, 879
ACOUSTIC (parameter) 878
fluid/solid interaction 495, 749, 897
ACTIVATE (history definition option) 596
Fourier 62
ADAMS 1005
harmonic 860
ADAPT GLOBAL (model definition option) 533, 781
heat transfer 494, 676, 758, 992, 995
ADAPTIVE (model definition option) 61, 62, 533, 622,
hydrodynamic bearing 495, 677, 749, 835, 995
756
Joule heating 995
ADAPTIVE (parameter) 756
large strain 752
adaptive mesh
linear 60
contact 548
magnetodynamic 750, 996
global 625
magnetodynamics 993
local 622
magnetostatic 495, 750, 844, 993, 995
refinement 756
nonlinear 65
adaptive time control 101
perturbation 83
ALL POINTS (parameter) 744, 827
piezoelectric 494, 678, 750, 873, 993, 996
ALLOCATE (parameter) 36
post buckling 930
analysis 495, 906
rigid cavity acoustic 877
acoustic 496, 678, 749, 877, 994, 996
rigid-plastic 161, 752, 995
contact 996
snap-through 930
coupled 540, 888
soil 165, 463, 750
coupled acoustic-structural 893
stress 994
coupled contact 540
structural zooming 188
coupled diffusion/thermal 913
thermal contact 542
coupled electromagnetic-thermal 908
thermo-electrical 495
coupled electrostatic-structural 749
analytical contact 554
coupled thermo-electrical (Joule heating) 902
ANELAS (user subroutine) 61, 202, 204, 395, 396
coupled thermo-mechanical 891
ANEXP (user subroutine) 61, 204, 447
crack 107
ANISO HYPE (model definition option) 75, 292, 297
creep 930
ANISOTROPIC (model definition option) 61, 202, 204,
design sensitivity 172
208, 238, 310, 339, 340, 446, 831, 874
dynamic 995
anisotropic hyperelasticity 295
anisotropic viscoelastic material 374 pin code 717
ANKOND (user subroutine) 494, 759 BEAM SECT (parameter) 568, 571, 572, 612, 613, 731,
ANPLAS (user subroutine) 204, 310 732, 747
application phase 130 beam-shell offsets 714
APPROACH (history definition option) 508 adaptive meshing 716
Arbitrary Eularian-Lagrangian (AEL) formulation 79 contact 716
arc-length methods 945 dynamics 716
Arrhenius 105, 793 field analysis 716
ARRUDBOYCE (model definition option) 106, 292 offset vector specification 715
ASSEM LOAD (model definition option) 1006 post file 716
ASSUMED STRAIN (parameter) 755 supported elements 716
assumed strain formulation 755 beam-to-beam intersection tying constraints 688
ATTACH EDGE (model definition option) 621, 663 BEARING (parameter) 836
ATTACH FACE (model definition option) 663 BEGIN SEQUENCE (history definition option) 672
AUTO CREEP (history definition option) 101, 103, 105, behavior
106, 370, 892, 929, 930 incompressible 85
AUTO INCREMENT (history definition option) 82, 463, nearly incompressible 88
611, 666, 929, 930, 932, 933, 948 nonlinear 929
AUTO LOAD (history definition option) 381, 611, 666, Bergan and Mollestad 939
900, 902, 929, 930 BERGBOYCE (model definition option) 106, 383
AUTO STEP (history definition option) 105, 148, 373, 381, Bergström-Boyce viscoelastic model 381
463, 554, 561, 672, 760, 764, 810, 892, 900, 902, B-H RELATION (electromagnetic, model definition option)
908, 909, 929, 930, 931, 933, 934, 935, 936, 939, 495, 861
940 B-H RELATION (magnetostatic, model definition option)
AUTO THERM (history definition option) 673, 929, 930, 495, 844, 846
931 Bingham fluid 881
AUTO THERM CREEP (history definition option) 104, 105, block definition 597
929, 930 body motion 513
automatic global remeshing 625 Bonding 537
Automatic Thermally Loaded Elastic-Creep/Elastic- Bonora damage model 493
Plastic-Creep Stress Analysis (AUTO THERM CREEP) boundary conditions
104 input options 681
Automatically Programmed Tools (APT) 1029 loading types 649
AUTOMSET (parameter) 702 time-dependent 155
AUTOSPC (parameter) 534 treatment of 88
AXITO3D (model definition option) 181 Breaking glue 537
breaking glue 533
B Brown and Smith 297
bandwidth optimization 608 BUCKLE (history definition option) 82, 83
Barlat BUCKLE (parameter) 65, 82
1991 312, 314 BUCKLE INCREMENT (model definition option) 65, 82, 83
2003 315 buckling analysis 82
Yld2004-13p 315 bushing
Yld2004-18p 314 coordinate system 704
beam elements 747, 991 bushings 703
nodal offsets for elements 704
Index 1038
properties 705 linear model 796
water drying model 797
C combined hardening 339
Cam-Clay model 464 common material characteristics 84
capacitor 841 COMPOSITE (model definition option) 179, 203, 204
Carreau model 881 composite continuum elements 204
CASE COMBIN (model definition option) 61, 63, 941 composite material 203
CASI iterative solver 161, 958, 959, 961 CON INTERA (model definition option) 506
cavity CON TABLE (model definition option) 506
liquid filled 670 configuration file 1015
pressure loading 667 CONN FILL (model definition option) 607
radiating 819 CONN GENER (model definition option) 607
rigid, acoustic analysis 877 CONNECT (model definition option) 602
surface elements 672, 755 connection methods 717
CAVITY DEFINITION (model definition option) 677, 812, CONNECTIVITY (model definition option) 597, 607, 703,
822, 823 720, 727, 732, 733, 736, 891, 907, 908, 909, 941
central difference operator 149, 152, 153 Conrad Gap 825
CENTROID (parameter) 72, 75, 82, 673, 744, 827, 941, CONRAD GAP (model definition option) 759
952, 988, 995 constant dilatation elements 753
CFAST (model definition option) 703, 719, 720, 736, 737, constitutive relations 342
738, 739, 740, 742 constraint
Chaboche 344, 366 rigid link 693
CHANGE PORE (model definition option) 170 CONSTRAINT (model definition option) 601
CHANGE STATE (history definition option) 104, 105, 931 contact
CHANGE STATE (model definition option) 38, 673, 759, adaptive meshing 548
827 analysis 996
CHANNEL (model definition option) 759, 890 analytical 554
classical lamination theory, multi-Layered shells 207 beams 545
clearance 584 bodies 502
closest point projection procedure 96 cohesive 539
CO_SIM (parameter) 1015 constraint implementation 517
code coupling coupled 540
basic stages 1012 deformable 557
software 1010 detection 513
COEFFICIENT (model definition option) 706 dynamic 547
COHESIVE (model definition option) 751 friction modeling 519
COHESIVE (with TABLE Input - Thermal) 751 neighbor relations 516
cohesive contact 539 optimize ordering 510
cohesive zone 485 quadratic 545
coil 864 release 534
current 850 segment-to-segment 563
multi-turn coils 850 separation 533
COIL CURRENT (magnetostatic, model definition option) shell 515, 547
851, 854, 908 thermal 542
coking 794 tolerance values 551
tube-tube 569 coupled electrical-thermal-mechanical analysis 906
CONTACT (model definition option) 78, 80, 81, 88, 149, coupled electrostatic-structural analysis 496
441, 503, 505, 531, 545, 547, 649, 699, 707, 760, coupled emectromagnetic-thermal analysis 908
811, 900, 902, 934 coupled thermo-electrical analysis (Joule heating) 902
CONTACT CHANGE (rezoning option) 548 coupled thermo-mechanical analysis 891
CONTACT NODE (model definition option) 503, 510 coupling region
contact penetration criteria 624 parallel processing 1014
contact penetration criteria, remeshing 631 shell elements 1014
CONTACT TABLE (history definition option) 508, 513, time step 1014
514, 534, 535 coupling regions 1012
CONTACT TABLE (model definition option) 506, 519, 529, crack
530, 531, 542, 544, 547, 560 analysis 107
CONTINUE (history definition option) 57, 58 initiation 130
continuum composite elements 754 propagation 119
interlaminar stresses 235 CRACK DATA (model definition option) 461, 462
continuum elements 746 crack growth direction 127
CONTROL (electromagnetic, model definition option) 857 crack growth increment 120
CONTROL (heat transfer, model definition option) 891 CRACK INIT (model definition option 130
CONTROL (history definition option) 105, 710, 715, 761, crack propagation
929 fatigue crack growth 119
CONTROL (magnetostatic, model definition option) 845 crack propgation
CONTROL (model definition option) 73, 83, 525, 552, direct crack growth 122
710, 715, 760, 819, 929, 934, 935, 986 mesh update methods 124
control algorithm 611 cracking
convergence controls 137 closure 463
CONVERT (model definition option) 893, 907, 908, 909 low tension 462
COORD SYSTEM (model definition option) 212, 507, 704 uniaxial 462
COORDINATES (model definition option) 597, 607, 732, Craig-Bampton 1005
941, 991 CREATE SEC (model definition option) 184
co-simulation 22, 1010 create sections
Coulomb friction model 520 beam 571
arctangent model 521 creep 101
bilinear model 525 buckling 104
limitations 528 control tolerances 102
stick-slip (modified step function) model 523 implicit formulation 365
COUPLE (parameter) 891, 907 Maxwell model 359
COUPLE CONTROL (history definition option) 892 CREEP (model definition option) 106, 359, 360, 365, 370,
coupled acoustic-structural analysis 893 447
acoustic medium 896 CREEP (parameter) 104, 105, 106, 107, 358, 365, 369
divergence theorem of Green 894 creep analysis 930
fluid pressure 895 CREEP INCREMENT (history definition option) 892, 929,
harmonic analyses 894 930
coupled analysis 540, 888 Crisfield 947, 949
coupled contact analysis 540 criteria
coupled diffusion/thermal analysis 913 contact penetration 624
coupled electrical-thermal-mechanical 906 equivalent values 624
Index 1040
mean strain energy 622 deactivate 596
node within a box 624 element method 807
nodes in contact 624 DEFINE (model definition option) 47, 48, 49
previously refined mesh 625 deformable contact, solution strategy 560
remeshing 630 deformable-deformable contact 557
temperature gradient 625 degree of freedom transformation 681
user-defined 625 DELAMINATION (model definition option) 132, 134
Zienkiewicz-Zhu 622 DENSITY EFFECTS (model definition option) 323
Zienkiewicz-Zhu creep strain 624 DESIGN DISPLACEMENT CONSTRAINTS (model definition
Cross Section 700 option) 179
CRPLAW (user subroutine) 101, 102, 103, 106, 107, 359, DESIGN FREQUENCY CONSTRAINTS (model definition
361, 364, 366, 370, 447 option) 179
CRPVIS (user subroutine) 105, 358, 370 DESIGN OBJECTIVE (model definition option) 179
crushing 463 design optimization 175
cure kinetics 192 DESIGN OPTIMIZATION (parameter) 179, 202
CURE SHRINKAGE (model definition option) 192 DESIGN SENSITIVITY (parameter) 179, 202
CURVES (model definition option) 212, 621 design sensitivity analysis 172
Cutter Location (CL) 1029 design space 176
CWELD (model definition option) 719, 720, 725, 726, DESIGN STRAIN CONSTRAINTS (model definition option)
727, 728, 729, 731, 732, 733, 735, 736, 737, 738, 179
739, 740, 742 DESIGN STRESS CONSTRAINTS (model definition option)
CWELD patch-to-patch connections 720 179
direct 720 DESIGN VARIABLES (model definition option) 179
indirect 722 diffusion 831
cyclic loading 672 DIFFUSION (parameter) 831, 914
cyclic symmetry dilatational creep 365
tying constraints 706 direct
CYCLIC SYMMETRY (model definition option) 707, 708 integration 148
CYLINDRICAL (model definition option) 596, 715 mesh definition input 590
methods 958
D direct crack growth 122
DAMAGE (model definition option) 106, 240, 482 DISP CHANGE (history definition option) 105, 697, 701,
damage models 892, 907
Bonora 493 DIST CHARGE (electromangetic, model definition option)
cohesive zone 485 857
ductile metals 475 DIST CHARGE (model definition option) 874, 900
elastomers 478 DIST CHARGES (electrostatic, model definition option)
damping 156, 157 840, 900, 909
frequency dependent 389 DIST CURRENT (electromagnetic, model definition option)
DAMPING (model definition option) 140, 147, 157 857
DAMPING COMPONENTS (history definition option) 836 DIST CURRENT (history definition option) 907
Darcy flow model 914 DIST CURRENT (Joule, history definition option) 902, 903,
Darcy’s law pyrolysis model 914 909
deact glue 533 DIST CURRENT (magnetostatic, model definition option)
DEACT GLUE (model definition option) 117, 536 844, 908
DIST CURRENT (model definition option) 907 Coulomb force 899
DIST FLUXES (history definition option) 903, 907, 909 coupled electrostatic-structural 496, 679, 749, 898
DIST FLUXES (model definition option) 166, 759, 764, element types 839
836, 902, 907, 908, 914, 921 input options 678
DIST LOADS (history definition option) 105, 907 electrostatic-structural
DIST LOADS (model definition option) 166, 664, 666, analysis 993, 996
874, 900, 902, 907 ELEM SORT (model definition option) 987
DIST LOADS (parameter) 179 element distortion criteria, remeshing 631
DIST MASS (diffusion, model definition option) 913, 914 element loop parallelism 967
DIST MASS (model definition option) 831 elements
DIST SOURCE (model definition option) 878 acoustic analysis 877, 896
Domain Decomposition Method (DDM) 967, 975 beam 747, 991
dual mesh approach 911 beams, contact 545
ductile metals 475 cavity surface 755
dynamic classes of stress-strain relations 205
analysis 995 composite continuum 204
crack propagation 119, 130 connectivity data 590
fracture methodology 129 constant dilatation 753
impact 547 contact 544
DYNAMIC (parameter) 65, 130, 135, 147, 148 continuum 746
DYNAMIC CHANGE (electromagnetic, history definition continuum composite 754
option) 857, 860 deactivate 807
DYNAMIC CHANGE (history definition option) 147, 148, electrostatic 839
154, 878, 892, 897, 929, 930 fluid 886
Fourier analysis 64
E gap-and-friction 750, 992
eddy 855 groups 1007
Edge I-J orientation 209 heat transfer 748
effects of temperature and strain rates on yield stress heat transfer convection 827
334 hydrodynamic bearing 835
eigenvalue analysis 135 incompressible 752
ELASTIC (parameter) 60, 61, 619 interface 134, 485, 751
elastic models 464 magnetodynamic 857
ELASTICITY (parameter) 61 magnetostatic 844
elastomeric material 286 membrane 746
elastomers 478 piezoelectric 874
electrical plate 748
analysis 676 quarter point 113
input options 678 rebar 751
ELECTRO (parameter) 840, 899 reduced integration 754
electromagnetic semi-infinite 754
analysis 856 shell 547, 748
input options 679 special 750
electrostatic truss 745
analysis 495, 678, 749, 993, 995 ELEMENTS (parameter) 134, 720, 733, 736, 744
ELEVAR (user subroutine) 988, 991
Index 1042
ELEVEC (user subroutine) 988 701, 874, 892, 897, 900, 902, 907
EL-MA (parameter) 857, 909 FIXED POTENTIAL (electromagnetic, model definition
ELSTO (parameter) 35, 36, 37, 38, 981 option) 857
EMCAPAC (history definition option) 543, 841 FIXED POTENTIAL (electrostatic, model definition option)
EMCIRCUIT (magnetostatic, model definition option) 866 840, 909
EMINDUC (history definition option) 854 FIXED POTENTIAL (magnetostatic, model definition
EMISSIVITY (model definition option) 822 option) 844, 908
EMLAMIN (history definition option) 856, 908 FIXED PRESSURE (diffusion, model definition option) 913,
EMRESIS (history definition option) 543, 904 914
EMWINDING (magnetostatic, model definition option) FIXED PRESSURE (model definition option) 166, 831,
851, 854, 866, 908 836, 878
Encryption 1034 FIXED TEMPERATURE (model definition option) 759, 763,
END (parameter) 57 822, 887, 892, 902, 907, 908, 909, 914
END OPTION (model definition option) 46, 57, 58, 179, FIXED VELOCITY (model definition option) 887
878 FIXED VOLTAGE (model definition option) 902, 907
END REZONE (rezoning option) 58 FLD (Forming Limit Diagram) 997
END SEQUENCE (history definition option) 672 flow
energy release rate 107, 121 diagram 963
Equivalent Values Criterion 624 rule 341
error analysis 619 FLP (Forming Limit Parameter) 997
ERROR ESTIMATE (model definition option) 62, 88, 619 fluid
Eulerian formulation 77 analysis 496, 750, 879
examples of lists input 49 elements 886
EXCLUDE (history definition option) 515 FLUID (parameter) 750
EXCLUDE (model definition option) 510 FLUID DRAG (model definition option) 666
expanded beam 569 fluid drag and wave
exponential cap 321, 329, 333 input options 666
EXPORT SEC (model definition option) 184 loads 666
EXTENDED (parameter) 47 fluid mechanics 879
Bingham fluid 881
F Carreau model 881
FAIL DATA (model definition option) 214, 235, 236, 237, element types 886
462 piecewise non-Newtonian flow 881
fatigue crack growth 119 power law fluid 881
FEATURE (parameter) 710, 715 FLUID SOLID (model definition option) 897
Feature Vertex Angle 640 fluid/solid interaction analysis 495, 749, 897
file units 37 FLUX (user subroutine) 676, 677, 759, 764, 766, 836,
FILM (user subroutine) 676, 677, 759, 764, 766, 902, 840, 844, 874, 878, 900, 902, 907, 908, 909
907, 908, 909 FOAM (model definition option) 75, 106, 381
FILMS (model definition option) 676, 759, 764, 775, 811, FOLLOW FOR (parameter) 72, 82, 664, 684, 755
902, 907, 908, 909, 914, 921 follow force stiffness contribution 755
FILMS (with TABLE input, model definition option) 677 FORCDF (user subroutine) 909
finite element technology FORCDT (user subroutine) 155, 763, 764, 766, 837, 840,
basic concepts 1023 844, 874, 887, 900, 902, 908
FIXED DISP (model definition option) 166, 597, 601, 697, force components 649
FORCEM (user subroutine) 155, 621, 857, 874, 900, 902, edges 640
908, 909 faces 641
Forming Limit Diagram (FLD) 997 nodes 640
Forming Limit Parameter (FLP) 997 harmonic
FOUNDATION (model definition option) 61, 81 analysis 860
FOURIER (model definition option) 61, 62, 63, 65 response 139
FOURIER (parameter) 62, 65 HARMONIC (electromagnetic, history definition option)
Fourier analysis 62 861, 909
coefficients 63 HARMONIC (parameter) 139, 140, 141, 398
elements used 64 Hashiguchi Plasticity Model 347
fracture mechanics 107 Hashin
frequency dependent damping 389 fabric failure criterion 223
frequency-dependent material behavior 388 failure criterion 221
friction modeling 519 tape failure criterion 224
Coulomb 520 HEAT (parameter) 749, 827, 908, 909
FXORD (model definition option) 602, 603, 607, 748 heat fluxes 540, 673
(special) conditions 676
G input options 676
gap-and-friction element 750, 992 heat transfer
gasket 241 analysis 494, 676, 758, 992, 995
GASKET (model definition option) 245 convection 827
Gasser 298 convection elements 827
generalized alpha operator 149, 152 elements 748
HHT 153 temperature effects 762
spectral radius 153 hereditary integral model 370
generation phase 129 Hill
GENT (model definition option) 75, 106, 292 failure criterion 217
geometric domains 744 yield function 310
GEOMETRY (model definition option) 179, 568, 569, 570, history data mapping technique 644
571, 613, 714, 717, 729, 731, 732, 748, 753, 755, history definition options 57
836, 991 Hockett-Sherby 337
geometry preservation 639 Hoffman failure criterion 218
GLK (model definition option) 189 HOOKLW (user subroutine) 61, 202, 204, 395, 396
global HOOKVI (user subroutine) 374, 395, 396
meshing 548 host file 41, 962, 974
ZX plane orientation 210 Houbolt operator 149
GLOBALLOCAL (model definition option) 38, 189 hydrodoynamic bearing 993
glue model, friction 533 hydrodynamic bearing
gradient analysis 495, 677, 749, 993, 995
objective or response function 172 elements 835
pressure 166 mass fluxes 677
Gurson model, modified 475 pump pressures 677
restrictors 677
HYPELA2 (user subroutine) 72, 247, 248, 464
H HYPOELASTIC (model definition option) 247
hard hysteresis 280, 386, 480, 855
Index 1044
I initial stress and initial plastic strain 674
IBOOC (parameter) 35, 36, 38 magnetodynamic 858
IMPD (user subroutine) 988 magnetostatic currents 679
IMPORT SEC (model definition option) 184 mass fluxes and restrictors 677
incompressible mechanical loads 664
Herrmann elastic formulation 87 thermal loads 673
material behavior 85 INSERT (model definition option) 702
material modeled 87 INTCRD (user subroutine) 988
incompressible elements 752 interface
rigid-plastic flow 752 ADAMS 1005
incremental mesh generators 607 hyperMesh results 1001
induction SDRC I-DEAS results 1001
heating 908, 909 interference fit 697
magnetic 53, 844 interpolation (shape) functions, type of 745
motors 855 inverse power sweep 136
inertia relief 158 ISOTROPIC (acoustic, model definition option) 496, 878
loads evaluation 159 ISOTROPIC (electromagnetic, model definition option)
Support method 158 495, 857
initial conditions 155, 508 ISOTROPIC (electrostatic, model definition option) 495,
previous analysis result data 181 840
INITIAL DISP (model definition option) 155 ISOTROPIC (fluid, model definition option) 495, 496, 897
initial gap 509 ISOTROPIC (heat transfer, model definition option) 494,
INITIAL PC (model definition option) 170 495, 673, 902, 909, 913, 914
initial plastic strain 673 ISOTROPIC (hydrodynamic, model definition option) 495
INITIAL PORE (model definition option) 170 ISOTROPIC (magnetostatic, model definition option) 495,
INITIAL POROSITY (model definition option) 166, 831 844, 846
INITIAL PRESSURE (diffusion, model definition option) ISOTROPIC (model definition option) 47, 49, 179, 202,
913, 914 204, 238, 245, 309, 310, 319, 320, 334, 336, 339,
INITIAL PYROLYSIS (model definition option) 799, 914 340, 361, 364, 365, 366, 372, 373, 446, 462, 664,
initial stress and initial plastic strain input options 674 758, 762, 793, 822, 823, 831, 836, 874, 887, 908
INITIAL TEMP (heat transfer, model definition option) 913, ISOTROPIC (with TABLE input - Stress) 335
914 isotropic hardening 338
INITIAL TEMP (model definition option) 758, 763, 902,
907, 908, 909 J
INITIAL VEL (model definition option) 155 Johnson-Cook
INITIAL VOID RATIO (model definition option) 166 damage model 494
input conventions 46 model 336
input options Joule
acoustic analysis sources 678 analysis 995
boundary conditions 681 heating 902
electrical 678 structural 906
electromagnetic 679 JOULE (model definition option) 903, 907
electrostatic 678 JOULE (parameter) 902, 907
fluid drag and wave loads 666
heat fluxes 676
K user-defined physical criteria 935
Kachanov factor 481 vectors 1026
Kelvin-Voigt model 369 LOAD COR (parameter) 664
kinematic load incrementation
constraints 679 adaptive scheme 930
hardening 339 fixed scheme 930
Lorentz 900
LORENZI (model definition option) 61, 111, 113, 130
L losses 855, 907
Lagrangian formulation 69
Lanczos method 136
large deformations 85
M
LARGE DISP (parameter) 71, 72, 75, 82, 87, 247, 334, machining (or metal cutting) 1029, 1034
695 MAGNETO (parameter) 844, 902, 908
large strain magnetodynamic
analysis 752 analysis 750, 864, 993, 996
elasticity 85 circuit approach 864
plasticity 87 element types 857
LARGE STRAIN (parameter) 71, 72, 75, 98, 149, 247, input options 858
334, 398, 466, 684, 695, 823, 892 magnetodynamic-thermal
linear analysis 60 induction heating 908
linear elastic material 201 magnetostatic
linear fracture mechanics 107 analysis 495, 679, 750, 993, 995
linear springs and elastic foundations input options 703 current input options 679
linked/unlinked 179 elements types 844
list of items, input 47 magnetostatics
load analysis 864
AUTO STEP scheme 933 MANY TYPES (model definition option) 602
automatic incrementation 932 Marc
cavity pressure 667 overview 31
fixed 931 MARLOW (model definition option) 106, 303, 304
fluid drag and wave 666 mass fluxes and restrictors 677
history definition options 929 MASSES (model definition option) 140
increment size 931 material
mass fuxes 677 instabilities 88
mechanical 663 low tension 461
mechanical and thermal increments 929 mechanical shape memory 280
piezoelectric 678 powder 321
pump pressures 677 preferred direction 208
recycling criterion 933 material behavior
residual 995 frequency-dependent 388
residual load correction 940 Material Data File 1034
restrictors 677 material dependent failure criteria 214
surface/volumetric 650 Hashin 221
thermal 673 Hashin Fabric 223
types 649 Hashin Tape 224
Index 1046
Hill 217 remeshing based growth 124
Hoffman 218 MESH2D 597
maximum strain 216 MESH2D (model definition option) 601
maximum stress 214 meshers
Puck 225 2-D solid 632
SIFT 227 3-D solid 634
strain failure 216 shell 633
Tsai-Wu 220 meshing techniques 632
user-defined 221 MICROSTRUCTURE (model definition option) 448, 810
material model mixture model 238
annealing 353 MNF (parameter) 1007
composite 203 MNF UNITS (model definition option) 1007
elastomers 286 modal
gasket 241 stresses and reactions 138
linear elastic 201 superposition 138
Mohr-Coulomb (hydrostatic yield dependence) 319 MODAL INCREMENT (model definition option) 65
nonlinear hypoelastic 247 MODAL SHAPE (history definition option) 136, 137, 142,
time-independent inelastic 306 878, 897, 1007
Mean Strain Energy Criterion 622 mode separation 112
mean-normal method 97 model connections
mechanical CWELD patch-to-patch 720
loads 663 model definition options 57
loads input options 664 model section
mechanical shape memory model define 184
conversation thermo-mechanical to mechanical export 184
SMA 285 include 184
experimental data fitting 284 model sections 183
mechanical wear 170 Mohr-Coulomb
membrane elements 746 linear material 319
MERGE (model definition option) 597 parabolic material 319
mesh MOONEY (model definition option) 72, 75, 106, 238, 290,
definition 21, 590 293, 381, 403, 446, 482
refinement tying constraints 686 Mooney-Rivlin 290, 294
severe distortion 88 MOTION (2-D, user subroutine) 507
mesh density control 634 MOTION CHANGE (history definition option) 508, 534
2-D 634 motion of bodies 506
3-D solid 634 initial conditions 508
general 635 initial gap 509
shell 634 MPC-CHECK (parameter) 702
mesh generation, remeshing techniques 632 MPI (Message Passing Interface) 967
mesh splitting 132 Mullin’s effect 478
mesh update methods 124 multiplicative decomposition 92, 305
growth along element edges or faces 127 multistage return mapping 98
growth by mesh cutting 127 MUMPS 41, 958, 967
growth by releasing tyings or glued contact 126
N HYPELA2 user subroutine 247
Narayanaswamy model 105, 384 NLELAST model definition 247
NASSOC (user subroutine) 106, 107, 366 nonlinearities, three types 69
Nastran RBE2 and RBE3 709 Nonlocal 887
Navier-Stokes 879, 882, 887, 890 non-Newtonian 880
NC machining data 1029 nonstructural materials 494
cutter path 1030 NURBS 504, 518, 554, 555
cutter shape 1029 trimmed 504
element deactivation 1031
intersection 1031, 1035 O
nearly incompressible behavior 88 Oak Ridge National Laboratory (ORNL) 101, 334, 363
neighbor relations 516 objective function 172, 173, 178, 179, 180, 181
Neo-Hookean 290 objective stress rates 90
Newmark-beta operator 148, 150 OGDEN (model definition option) 72, 75, 106, 291, 381,
Newton Cooling 460 446, 482
Newton-Raphson method Ogden-Roxburgh Damage model 482
modified 943 Ohm’s law 904
NiTi alloys 261 Ohmic power loss 910
NLELAST 247 Ohmic power lost 865
(bimodulus) elasticity model 259 OpenMP 967
(bimodulus) elasticity model with either no tension, operator
limited tension, no compression, or limited central difference 149
compression enhancements 259 generalized alpha operator 149
basic assumptions and definitions 248 Houbolt 149
converting engineering strain/stress to true strain/ Newmark-beta 148
stress 248 single step Houbolt 149
input of uniaxial stress-strain data 248 OPTHOTROPIC (model definition option) 340
Nastran like nonlinear elasticity model 249 optimization algorithm 178
nonlinear orthotropic elasticity model 261 OPTIMIZE (model definition option) 608
principal strain space model 258 Option
strain invariant model 253 ACTIVATE (history definition) 596
NLELAST (model definition option) 72, 204, 247, 250, ADAPT GLOBAL (model definition) 533, 781
251 ADAPTIVE (model definition) 61, 62, 533, 622, 756
nodal degrees of freedom tying constraints 693 ANISO HYPE (model definition) 75, 292, 297
NODE CIRCLE (model definition option) 608 ANISOTROPIC (model definition) 61, 202, 204, 208,
NODE FILL (model definition option) 608 238, 310, 339, 340, 446, 831, 874
NODE GENER (model definition option) 607 APPROACH (history definition) 508
NODE SORT (model definition option) 987 ARRUDBOYCE (model definition) 106, 292, 398,
Node Within A Box Criterion 624 406, 435, 440
Nodes In Contact Criterion 624 ASSEM LOAD (model definition) 1006
nodes, merging of 597 ATTACH EDGE (model definition) 621, 663
nonlinear ATTACH FACE (model definition) 663
analysis 65 AUTO CREEP (history definition) 101, 103, 105, 106,
fracture mechanics 110 370, 892, 929, 930
nonlinear hypoelastic material 247 AUTO INCREMENT (history definition) 82, 463, 611,
Index 1048
666, 929, 930, 932, 933, 948 760, 811, 900, 902, 934
AUTO LOAD (history definition) 381, 611, 666, 900, CONTACT CHANGE (rezoning) 548
902, 929, 930 CONTACT NODE (model definition) 503, 510
AUTO STEP (history definition) 105, 148, 373, 381, CONTACT TABLE (history definition) 508, 513, 514,
463, 554, 561, 672, 760, 764, 810, 892, 900, 534, 535
902, 908, 909, 929, 930, 931, 933, 934, 935, CONTACT TABLE (model definition) 506, 519, 529,
936, 939, 940 530, 531, 542, 544, 547, 560
AUTO THERM (history definition) 673, 929, 930, 931 CONTINUE (history definition) 57, 58
AUTO THERM CREEP (history definition) 104, 105, CONTROL (electromagnetic, model definition) 857
929, 930 CONTROL (heat transfer, model definition) 891
AXITO3D (model definition) 181 CONTROL (history definition) 105, 710, 715, 761,
BEGIN SEQUENCE (history definition) 672 929
BERGBOYCE (model definition) 106, 383 CONTROL (magnetostatic, model definition) 845
B-H RELATION (electromagnetic, model definition) CONTROL (model definition) 73, 83, 525, 552, 710,
495, 861 715, 760, 819, 929, 934, 935, 986
B-H RELATION (magnetostatic, model definition) CONVERT (model definition) 893, 907, 908, 909
495, 844, 846 COORD SYSTEM (model definition) 212, 507, 704
BUCKLE (history definition) 82, 83 COORDINATES (model definition) 597, 607, 732,
BUCKLE INCREMENT (model definition) 65, 82, 83 941, 991
CASE COMBIN (model definition) 61, 63, 941 COUPLE CONTROL (historyl definition) 892
CAVITY DEFINITION (model definition) 677, 812, CRACK DATA (model definition) 461, 462
822, 823 CRACK INIT (model definition) 130
CFAST (model definition) 703, 719, 720, 736, 737, CREATE SEC (model definition) 184
738, 739, 740, 742 CREEP (model definition) 106, 359, 360, 365, 370,
CHANGE PORE (model definition) 170 447
CHANGE STATE (history definition) 104, 105, 931 CREEP INCREMENT (history definition) 892, 929,
CHANGE STATE (model definition) 38, 673, 759, 827 930
CHANNEL (model definition) 759, 890 CURE SHRINKAGE (model definition) 192
COEFFICIENT (model definition) 706 CURVES (model definition) 212, 621
COHESIVE (model definition) 751 CWELD (model definition) 719, 720, 725, 726, 727,
COIL CURRENT (magnetostatic, model definition) 728, 729, 731, 732, 733, 735, 736, 737, 738,
851, 854, 908 739, 740, 742
COMPOSITE (model definition) 179, 203, 204, 395 CYCLIC SYMMETRY (model definition) 707, 708
CON INTERA (model definition) 506 CYLINDRICAL (model definition) 596, 715
CON TABLE (model definition) 506 DAMAGE (model definition) 106, 240, 482
CONN FILL (model definition) 607 DAMPING (model definition) 140, 147, 157
CONN GENER (model definition) 607 DAMPING COMPONENTS (history definition) 836
CONNECT (model definition) 602 DEACT GLUE (model definition) 117
CONNECTIVITY (model definition) 597, 607, 703, DEFINE (model definition) 47, 48, 49
720, 727, 732, 733, 736, 891, 907, 908, 909, DELAMINATION (model definition) 132, 134
941 DENSITY EFFECTS (model definition) 323
CONRAD GAP (model definition) 759 DESIGN DISPLACEMENT CONSTRAINTS (model
CONSTRAINT (model definition) 601 definition) 179
CONTACT (model definition) 78, 80, 81, 88, 149, DESIGN FREQUENCY CONSTRAINTS (model
441, 503, 505, 531, 545, 547, 649, 699, 707, definition) 179
DESIGN OBJECTIVE (model definition) 179 ERROR ESTIMATE (model definition) 62, 88, 619
DESIGN STRAIN CONSTRAINTS (model definition) EXCLUDE (history definition) 515
179 EXCLUDE (model definition) 510
DESIGN STRESS CONSTRAINTS (model definition) EXPORT SEC (model definition) 184
179 FAIL DATA (model definition) 214, 235, 236, 237,
DESIGN VARIABLES (model definition) 179 462
DISP CHANGE (history definition) 105, 697, 699, FILMS (model definition) 676, 759, 764, 775, 811,
701, 892, 907 902, 907, 908, 909, 914, 921
DIST CHARGE (electromangetic, model definition) FILMS (with TABLE input, model definition) 677
857 FIXED DISP (model definition) 166, 597, 601, 697,
DIST CHARGE (model definition) 874, 900 701, 874, 892, 897, 900, 902, 907
DIST CHARGES (electrostatic, model definition) 840, FIXED POTENTIAL (electromagnetic, model
900, 909 definition) 857
DIST CURRENT (electromagnetic, model definition) FIXED POTENTIAL (electrostatic, model definition)
857 840, 909
DIST CURRENT (history definition) 907 FIXED POTENTIAL (magnetostatic, model definition)
DIST CURRENT (Joule, history definition) 902, 903, 844, 908
909 FIXED PRESSURE (diffusion, model definition) 913,
DIST CURRENT (magnetostatic, model definition) 914
844, 908 FIXED PRESSURE (model definition) 166, 831, 836,
DIST CURRENT (model definition) 907 878
DIST FLUXES (history definition) 903, 907, 909 FIXED TEMPERATURE (model definition) 759, 763,
DIST FLUXES (model definition) 166, 759, 764, 836, 822, 887, 892, 902, 907, 908, 909, 914
902, 907, 908, 914, 921 FIXED VELOCITY (model definition) 887
DIST LOADS (history definition) 105, 907 FIXED VOLTAGE (model definition) 902, 907
DIST LOADS (model definition) 166, 664, 666, 874, FLUID DRAG (model definition) 666
900, 902, 907 FLUID SOLID (model definition) 897
DIST MASS (diffusion, model definition) 913, 914 FOAM (model definition) 75, 106, 381, 398, 402,
DIST MASS (model definition) 831 406, 435, 440
DIST SOURCE (model definition) 878 FOUNDATION (model definition) 61, 81
DYNAMIC CHANGE (history definition, FOURIER (model definition) 61, 62, 63, 65
electromagnetic) 857, 860 FXORD (model definition) 602, 603, 607, 748
DYNAMIC CHANGE (history definition) 147, 148, GASKET (model definition) 245
154, 878, 892, 897, 929, 930 GENT (model definition) 75, 106, 292, 398, 406,
ELEM SORT (model definition) 987 435, 440
EMCAPAC (history definition) 543, 841 GEOMETRY (model definition) 179, 568, 569, 570,
EMCIRCUIT (magnetostatic, model definition) 866 571, 613, 714, 717, 729, 731, 732, 748, 753,
EMINDUC (history definition) 854 755, 836, 991
EMISSIVITY (model definition) 822 GLK (model definition) 189
EMLAMIN (history definition) 856, 908 GLOBALLOCAL (model definition) 38, 189
EMRESIS (history definition) 543, 904 HARMONIC (electromagnetic, history definition)
EMWINDING (magnetostatic, model definition) 851, 861, 909
854, 866, 908 HYPOELASTIC (model definition) 247
END OPTION (model definition) 46, 57, 58, 179, 878 IMPORT SEC (model definition) 184
END REZONE (rezoning) 58 INITIAL DISP (model definition) 155
END SEQUENCE (history definition) 672 INITIAL PC (model definition) 170
Index 1050
INITIAL PORE (model definition) 170 MOTION CHANGE (history definition) 508, 534
INITIAL POROSITY (model definition) 166, 831 NLELAST (model definition) 72, 204, 247, 250, 251
INITIAL PRESSURE (diffusion, model definition) 913, NODE CIRCLE (model definition) 608
914 NODE FILL (model definition) 608
INITIAL PYROLYSIS (model definition) 799, 914 NODE GENER (model definition) 607
INITIAL TEMP (heat transfer, model definition) 913, NODE SORT (model definition) 987
914 OGDEN (model definition) 72, 75, 106, 291, 381,
INITIAL TEMP (model definition) 758, 763, 902, 907, 398, 406, 435, 440, 446, 482
908, 909 OPTIMIZE (model definition) 608
INITIAL VEL (model definition) 155 ORIENTATION (model definition) 202, 204, 208,
INITIAL VOID RATIO (model definition) 166 310, 374, 395, 744
INSERT (model definition) 702 ORTHO TEMP (model definition) 204, 334, 446, 758,
ISOTROPIC (acoustic, model definition) 496, 878 902
ISOTROPIC (electromagnetic, model definition) 495, ORTHOTROPIC (electrical, model definition) 495
857 ORTHOTROPIC (electromagnetic, model definition)
ISOTROPIC (electrostatic, model definition) 495, 840 495, 857
ISOTROPIC (fluid, model definition) 495, 496, 897 ORTHOTROPIC (electrostatic, model definition) 840,
ISOTROPIC (heat transfer, model definition) 494, 900, 902
495, 673, 902, 909, 913, 914 ORTHOTROPIC (magnetostatic, model definition)
ISOTROPIC (hydrodynamic, model definition) 495 495, 844, 846
ISOTROPIC (magnetostatic, model definition) 495, ORTHOTROPIC (model definition) 61, 179, 202, 204,
844, 846 238, 309, 310, 334, 336, 339, 340, 365, 366,
ISOTROPIC (model definition) 47, 49, 179, 202, 204, 372, 374, 393, 394, 395, 440, 446, 447, 664,
238, 245, 309, 310, 319, 320, 334, 336, 339, 758, 793, 822, 823, 831, 874
340, 361, 364, 365, 366, 372, 373, 392, 406, ORTHOTROPIC (thermal, model definition) 494, 495,
435, 440, 446, 462, 664, 758, 762, 793, 822, 673, 902, 909, 913, 914
823, 831, 836, 874, 887, 908 ORTHOTROPIC (with TABLE input - Stress) 335
ISOTROPIC (with TABLE input - Stress) 335 ORTHOTROPIC (with TABLE input - thermal) 908
ISOTROPIC (with TABLE input - thermal) 908 PARAMETERS (history definition) 150
JOULE (model definition) 903, 907 PARAMETERS (model definition) 161, 163, 519,
LORENZI (model definition) 61, 111, 113, 130 541, 547, 570, 572, 823, 887
MANY TYPES (model definition) 602 PAYNE (model definition) 417, 423, 426, 427, 428,
MARLOW (model definition) 106, 303, 304, 398, 434, 440, 441
406, 435, 440 PBUSH (model definition) 703, 705, 706
MASSES (model definition) 140 PERMANENT (magnetostatic, model definition) 844,
MERGE (model definition) 597 846
MESH2D (model definition) 601 PFAST (model definition) 703, 705, 720, 736
MICROSTRUCTURE (model definition) 448, 810 PHI-COEFFICIENTS (model definition) 403
MNF UNITS (model definition) 1007 PIEZOELECTRIC (model definition) 874
MODAL INCREMENT (model definition) 65 PIN CODE (model definition) 717
MODAL SHAPE (history definition) 136, 137, 142, POINT CHARGE (electrostatic, model definition) 858,
878, 897, 1007 900
MOONEY (model definition) 72, 75, 106, 238, 290, POINT CHARGE (model definition) 840, 874
293, 381, 398, 403, 406, 433, 435, 440, 446, POINT CHARGE (with TABLE Input, model definition
482 option) 909
POINT CURRENT (electromagnetic, history RECEDING SURFACE (model definition) 780, 914
definition) 857 RECOVER (history definition) 138
POINT CURRENT (history definition) 907 RELATIVE DENSITY (model definition) 323
POINT CURRENT (Joule, history definition) 903 RELEASE (history definition) 534
POINT CURRENT (Joule, model definition) 902 RESTART (model definition) 37, 61, 101, 941, 987,
POINT CURRENT (magnetostatic, model definition) 988
844, 908 RESTART INCREMENT (history definition) 941
POINT CURRENT (model definition) 907 RESTART LAST (model definition) 941
POINT CURRENT (with TABLE Input - Magnetostatic, RESTRICTOR (model definition) 836
history definition) 909 REZONE (rezoning) 58
POINT FLUX (history definition) 903, 907, 908 ROTATION A (model definition) 664
POINT FLUX (model definition) 166, 759, 764, 836, RROD (model definition) 117, 126
902, 907, 909, 921 SERVO LINK (model definition) 648, 686
POINT LOAD (electrostatic, model definition) 900, SHELL TRANSFORMATION (model definition) 682,
902 684
POINT LOAD (history definition) 105, 907 SHIFT FUNCTION (model definition) 105, 378, 417,
POINT LOAD (model definition) 47, 166, 507, 874, 427, 429, 430, 440, 441
907 SOIL (model definition) 166, 170, 464
POINT MASS (model definition) 831 SOLVER (model definition) 755, 957, 958
POINT SOURCE (model definition) 878 SPECIFIED NODES (model definition) 601
POINTS (model definition) 621 SPLINE (model definition) 503, 504, 505, 558, 567,
POROSITY CHANGE (model definition) 241 570, 576, 577, 578, 782
POST (model definition) 105, 235, 392, 395, 399, SPRINGS (model definition) 61, 81, 140
407, 444, 759, 798, 827, 931, 997 START NUMBER (model definition) 601
POTENTIAL CHANGE (history definition) 909 STEADY STATE (electrostatic, history definition) 840
POWDER (model definition) 323 STEADY STATE (history definition) 760, 884, 892,
PRE STATE (model definition) 38, 181, 182, 183 903, 908, 909
PRINT CHOICE (model definition) 142, 233, 986, STEADY STATE (magnetostatic, history definition)
987, 988, 989, 991, 994 845
PRINT ELEMENT (model definition) 233, 986, 988, STIFFNS COMPONENTS (history definition) 836
989, 991 STIFFSCALE (model definition) 612
PRINT NODE (model definition) 987, 988, 994 STRAIN RATE (model definition, fluid) 496
PRINT VMASS (model definition) 987 STRAIN RATE (model definition) 163, 204, 334, 335,
PROPORTIONAL INCREMENT (history definition) 887
105, 106, 611, 666 SUMMARY (model definition) 987
PSHELL (model definition) 208, 241 SUPERELEM (history definition) 1007
PWELD (model definition) 720, 721, 724, 729, 731, SUPERELEM (model definition) 1007
732 SURFACE ENERGY (model definition) 768, 780, 914,
RAD-CAVITY (model definition) 677, 775, 822, 823 921
RADIATING CAVITY (model definition) 812 SURFACES (model definition) 621
RBE2 (model definition) 117, 126, 710, 717, 726, SWLDPRM (model definition) 720, 732, 735, 737,
727 738
RBE3 (model definition) 711, 712, 713, 717, 719, SYNCHRONIZED (history definition) 508
722, 726, 727, 736, 742 TABLE (model definition) 52, 243, 303, 339, 340,
REAUTO (model definition) 101, 941 372, 392, 394, 398, 423, 428, 432, 434, 441,
REBAR (model definition) 461 446, 447, 540, 822, 823, 856, 870, 892
Index 1052
TEMP CHANGE (history definition) 759, 763, 887, VISCELORTH (model definition) 374, 393, 394, 395,
892, 903, 907, 908, 909 440
TEMPERATURE EFFECTS (coupled thermal-stress, VISCELPROP (model definition) 373, 379, 392, 434,
model definition) 495 440
TEMPERATURE EFFECTS (heat transfer, model VOID CHANGE (model definition) 241
definition) 494, 902 VOLTAGE CHANGE (history definition) 903, 907
TEMPERATURE EFFECTS (hydrodynamic, model WELD FILL (model definition) 801, 802, 807, 808
definition) 495 WELD FLUX (model definition) 801, 802, 805, 808,
TEMPERATURE EFFECTS (model definition) 204, 810
290, 323, 334, 335, 372, 378, 446, 447, 496, WELD PATH (model definition) 801, 802, 805
758, 762, 823, 836 WORK HARD (model definition) 163, 204, 335, 339,
TERM CURRENT (magnetostatic, model definition) 340
866 ORIENT (user subroutine) 61, 202, 310, 374
TERM VOLTAGE (magnetostatic, model definition) ORIENT2 (user subroutine) 209, 211, 212
866 ORIENTATION (model definition option) 202, 204, 208,
THERMAL CONTACT (2-D) (model definition) 543, 310, 374, 744
544, 760, 811, 831, 841, 850, 854, 856, 904 ORTHO TEMP (model definition option) 204, 334, 446,
THERMO-PORE (model definition) 793, 914 758, 902
THICKNESS (model definition) 836 ORTHOTROPIC (electrical, model definition option) 495
THICKNS CHANGE (history definition) 836 ORTHOTROPIC (electromagnetic, model definition option)
TIME STEP (history definition) 334, 373, 381 495, 857
TIME-TEMP (model definition) 448, 459, 460 ORTHOTROPIC (electrostatic, model definition option)
TRANSFORMATION (model definition) 507, 681, 840, 900, 902
689, 699, 715, 994 ORTHOTROPIC (magnetostatic, model definition option)
TRANSIENT (history definition) 759, 760, 761, 764, 495, 844, 846
810, 892, 903, 908, 909, 929, 930 ORTHOTROPIC (model definition option) 61, 179, 202,
TRANSIENT (historyl definition) 810 204, 238, 309, 310, 334, 336, 339, 340, 365, 366,
TYING (model definition) 648, 685, 686, 759, 897 372, 374, 393, 446, 447, 664, 758, 793, 822, 823,
TYING CHANGE (history definition) 686 831, 874
UFCONN (model definition) 597 ORTHOTROPIC (thermal, model definition option) 494,
UFXORD (model definition) 597 495, 673, 902, 909, 913, 914
UTRANFORM (model definition) 681 ORTHOTROPIC (with TABLE input - Stress) 335
VCCT (model definition) 107, 113 ORTHOTROPIC (with TABLE input - thermal) 908
VELOCITY (model definition) 759, 827, 836 out-of-core storage options 36
VELOCITY CHANGE (history definition) 759, 827, 887 output results
VIEW FACTOR (model definition) 677, 815 workspace information 315, 981, 1010
VISCEL EXP (model definition) 106, 388 overclosure 697
VISCELCORR (model definition) 428, 429, 432, 434, tying 696
440, 441
VISCELFOAM (model definition) 105, 381, 398, 434, P
440 parallel processing
VISCELMOON (model definition) 105, 290, 381, 398, contact analysis 971
417, 423, 426, 427, 428, 429, 434, 440, 482 domain decomposition 973
VISCELOGDEN (model definition) 105, 291, 381, GPU Support 971
398, 434, 440, 482 solvers 970
types of machines 967 MNF 1007
unsupported features 968 MPC-CHECK 702
Parameter PIEZO 874
ABLATION 781 PORE 831
ACOUSTIC 878 PRINT 83, 525, 768, 821, 996
ADAPTIVE 756 PYROLYSIS 914
ALL POINTS 744, 827 RADIATION 494, 677, 811, 815
ALLOCATE 36 RBE 727
ASSUMED STRAIN 755 RESTRICTOR 836
AUTOMSET 702 REZONING 608
AUTOSPC 534 R-P FLOW 161, 945
BEAM SECT 568, 571, 572, 612, 613, 731, 732, SCALE 929
747 SHELL SECT 748
BEARING 836 SIZING 744, 984
BUCKLE 65, 82 STATE VARS 836
CENTROID 72, 75, 82, 673, 744, 827, 941, 952, STOP 36
988, 995 STRUCTURAL 900, 902
CO_SIM 1015 TSHEAR 233, 991
COUPLE 891, 907 T-T-T 459
CREEP 104, 105, 106, 107, 358, 365, 369 UPDATE 149, 892
DESIGN OPTIMIZATION 179, 202 VISCELHARM 405, 432
DESIGN SENSITIVITY 179, 202 VISCO ELAS 105
DIFFUSION 831, 914 WELDING 810
DIST LOADS 179 parameters 57
DYNAMIC 65, 130, 135, 147, 148 PARAMETERS (history definition option) 150
ELASTIC 60, 61, 619 PARAMETERS (model definition option) 161, 163, 519,
ELASTICITY 61 541, 547, 570, 572, 823, 887
ELECTRO 840, 899 Pardiso 958, 967
ELEMENTS 134, 720, 733, 736, 744 Paris’ law 120
EL-MA 857, 909 participation factor 138
ELSTO 35, 36, 37, 38, 981 particle tracking 38
END 57 Payne 405
EXTENDED 47 PBUSH (model definition option) 703, 705, 706
FEATURE 489, 710, 715 penalty 161, 884
FLUID 750 PERMANENT (magnetostatic, model definition option)
FOLLOW FOR 72, 82, 664, 684, 755 844, 846
FOURIER 62, 65 perturbation analysis 83
HARMONIC 139, 140, 141, 392, 394, 398, 440 PFAST (model definition option) 703, 705, 720, 736
HEAT 749, 827, 908, 909 phase transformations 448
IBOOC 35, 36, 38 PHI-COEFFICIENTS (model definition option) 403
JOULE 902, 907 piecewise linear representation 446
LARGE DISP 71, 72, 75, 82, 87, 247, 334, 695 PIEZO (parameter) 874
LARGE STRAIN 71, 72, 75, 98, 149, 247, 334, 398, piezoelectric
466, 684, 695, 823, 892 analysis 494, 678, 750, 873, 993, 996
LOAD COR 664 coupling 876
MAGNETO 844, 902, 908 element types 874
Index 1054
loads 678 Power Law Fluid 881
PIEZOELECTRIC (model definition option) 874 PRE STATE (model definition option) 38, 181, 182, 183
pin code 717 compatible element types 182
PIN CODE (model definition option) 717 model sections 183
pipe bend element tying constraints 692 preconditioners, iterative solvers 959
plate elements 748 presentation of the energy equation 799
PLOTV (user subroutine) 997 Previously Refined Mesh Criterion 625
POINT CHARGE (electrostatic, model definition option) Principal Engineering Strains 997
858, 900 PRINT (parameter) 83, 525, 768, 821, 996
POINT CHARGE (model definition option) 840, 874 PRINT CHOICE (model definition option) 142, 233, 986,
POINT CHARGE (with TABLE Input - Electrostatic, model 987, 988, 989, 991, 994
definition option) 909 PRINT ELEMENT (model definition option) 233, 986, 988,
POINT CURRENT (electromagnetic, history definition 989, 991
option) 857 PRINT NODE (model definition option) 987, 988, 994
POINT CURRENT (history definition option) 907 PRINT VMASS (model definition option) 987
POINT CURRENT (Joule, history definition option) 903 program
POINT CURRENT (Joule, model definition option) 902 initiation 21, 40
POINT CURRENT (magnetostatic, model definition option) messages 1000
844, 908 progressive composite failure 235
POINT CURRENT (model definition option) 907 PROPORTIONAL INCREMENT (history definition option)
POINT CURRENT (with TABLE Input - Magnetostatic, 105, 106, 611, 666
history definition option) 909 PSHELL (model definition option) 208, 241
POINT FLUX (history definition option) 903, 907, 908 PSPG 882
POINT FLUX (model definition option) 166, 759, 764, 836, Puck failure criterion 225
902, 907, 909, 921 PWELD (model definition option) 720, 721, 724, 729,
POINT LOAD (electrostatic, model definition option) 900, 731, 732
902 PWL (piece-wise linear) 503
POINT LOAD (history definition option) 105, 907 pyrolysis 791
POINT LOAD (model definition option) 47, 166, 507, 874, alternative formulation 794
907 presentation of the mass equation 792
POINT MASS (model definition option) 831 principle of 792
POINT SOURCE (model definition option) 878 PYROLYSIS (parameter) 914
POINTS (model definition option) 621
Poisson’s 60, 201, 202, 204, 322, 445, 459, 462, 464, Q
466 Qiu and Pence 297
PORE (parameter) 831 quadratic contact 545
POROSITY CHANGE (model definition option) 241 quarter point elements 113
POST (model definition option) 105, 235, 392, 395, 759,
798, 827, 931, 997
post buckling analysis 930
R
post file 996 RAD-CAVITY (model definition option) 677, 775, 822, 823
POTENTIAL CHANGE (history definition) 909 radial return 95
powder 66 RADIATING CAVITY (model definition option) 812
POWDER (model definition option) 323 radiation 811
powder material 321 computational approaches 811
direct adaptive integration 818
Monte Carlo method 815 ROTATION A (model definition option) 664
pixel based modified Hemi-Cube method 812 R-P FLOW (parameter) 161, 945
viewfactors 812 RROD (model definition option) 117, 126
RADIATION (parameter) 494, 677, 811, 815 rubber foam model 293
radiation viewfactors 812 RVCNST 521, 522
RBE (parameter) 727
RBE2 (model definition option) 117, 126, 710, 717, 726, S
727 SCALE (parameter) 929
RBE2 and RBE3 709 scaling, linear-elastic solution 929
RBE3 (model definition option) 711, 712, 713, 717, 719, segment-to-segment 563
722, 726, 727, 736, 742 selective printout 986
reaction forces 995 grid force balance 988
REAUTO (model definition option) 101, 941 options 986
REBAR (model definition option) 461 semi-infinite elements 754
REBAR (user subroutine) 461 sensitivity 172
rebar elements 751 separation, contact 533
RECEDING SURFACE (model definition option) 780, 914 SEPFOR (user subroutine) 557
RECOVER (history definition option) 138 SEPSTR (user subroutine) 557
reduced integration elements 754 Sequential Quadratic Programming 178
RELATIVE DENSITY (model definition option) 323 SERVO LINK (model definition option) 648, 686
RELEASE (history definition option) 534 shape memory
remeshing mechanical 280
automatic global 625 thermo-mechanical material 261
criteria 630 shell
remeshing techniques contact 515
adaptive 1031 elements 547, 748
mesh generation 632 stiffener tying constraints 692
target number of elements 642 transformation 682
residual load correction 940 SHELL SECT (parameter) 748
resistance 543, 705, 902, 904, 996 shell stiffener tying constraints 692
resistor 904 shell thickness update 79
RESTART (model definition option) 37, 61, 101, 941, 987, SHELL TRANSFORMATION (model definition option) 682,
988 684
restart analysis 941 shell-to-shell intersection tying constraints 688
RESTART INCREMENT (history definition option) 941 shell-to-solid intersections tying constraints 689
RESTART LAST (model definition option) 941 SHIFT FUNCTION (model definition option) 105, 378
RESTRICTOR (model definition option) 836 single input
RESTRICTOR (parameter) 836 DDM 41, 970, 973, 975
result evaluation, contact 549 single step Houbolt operator 149
REZONE (rezoning option) 58 singularity ratio 956
rezoning 548, 608 SIZING (parameter) 744, 984
REZONING (parameter) 608 snap-through analysis 930
rigid contact, solution strategy 556 soft
rigid link constraint 694, 699 edges 641
rigid-plastic analysis 161, 752, 995 faces 641
Riks-Ramm 949, 950
Index 1056
SOIL (model definition option) 166, 170, 464 SUPERELEM (history definition option) 1007
soil analysis 165, 463, 750 SUPERELEM (model definition option) 1007
SOLVER (model definition option) 755, 957, 958 SUPG 882
special elements 750 surface energy 768
cable 751 SURFACE ENERGY (model definition option) 768, 780,
curved-pipe 751 914, 921
Gap-and-Friction 750 SURFACES (model definition option) 621
pipe-bend 751 SWLDPRM (model definition option) 720, 732, 735, 737,
rebar 751 738
shear panel 751 symmetries 600
SPECIFIED NODES (model definition option) 601 Symmetry Conditions 586
spectrum response 142 SYNCHRONIZED (history definition option) 508
SPLINE (model definition option) 503, 504, 505, 558, system and element stiffness matrices 1025
567, 570, 576, 577, 578, 782
SPRINGS (model definition option) 61, 81, 140 T
springs, linear and nonlinear 992 TABLE (model definition option) 52, 243, 303, 339, 340,
START NUMBER (model definition option) 601 372, 446, 447, 540, 822, 823, 856, 870, 892
STATE VARS (parameter) 836 TEMP CHANGE (history definition option) 759, 763, 887,
status file 1007 892, 903, 907, 908, 909
STEADY STATE (electrostatic, history definition option) TEMPERATURE EFFECTS (coupled thermal-stress, model
840 definition option) 495
STEADY STATE (history definition option) 760, 884, 892, TEMPERATURE EFFECTS (heat transfer, model definition
903, 908, 909 option) 494, 902
STEADY STATE (magnetostatic, history definition option) TEMPERATURE EFFECTS (hydrodynamic, model
845 definition option) 495
Stefan-Boltzmann 677 TEMPERATURE EFFECTS (model definition option) 204,
STIFFNS COMPONENTS (history definition option) 836 290, 323, 334, 335, 372, 378, 446, 447, 496, 758,
STIFFSCALE (model definition option) 612 762, 823, 836
STOP (parameter) 36 Temperature Gradient Criterion 625
storage methods 960 temperature-dependent
strain correction method 944 creep 447
STRAIN RATE (model definition option, fluid) 496 material properties 445
STRAIN RATE (model definition option) 163, 204, 334, tension softening 462
335, 887 TERM CURRENT (magnetostatic, model definition option)
strain rate effects on elastic-plastic constitutive relation 866
343 TERM VOLTAGE (magnetostatic, model definition option)
stress 866
analysis 994 terminal voltage 864
intensity factor defined 108 thermal
intensity factors 108 loads 673
stress intensity factor 121 loads input options 673
STRUCTURAL (parameter) 900, 902 THERMAL CONTACT (2-D) (model definition option) 543,
structural zooming analysis 188 544, 760, 811, 831, 841, 850, 854, 856, 904
substructuring 609 thermal contact analysis 542
SUMMARY (model definition option) 987 thermal expansion coefficient 447
thermo-electrical 495 shell-to-solid intersections 689
thermo-mechanical shape memory model 261
constitutive theory 265 U
phase transformation strains 265 UADAP (user subroutine) 625
transformation induced deformation 263 UADAPBOX (user subroutine) 810
THERMO-PORE (model definition option) 793, 914 UBEAR (user subroutine) 837
Thermo-Rheologically Simple material 375 UBREAKGLUE (user subroutine) 537
THICKNESS (model definition option) 836 UCOHESIVE (user subroutine) 489
thickness update 79 UCOHESIVET (user subroutine) 489
THICKNS CHANGE (history definition option) 836 UCOMPL (user subroutine) 140
TIME STEP (history definition option) 334, 373, 381 UCONORT (user subroutine)T 532
time step definition 154 UCOORD (user subroutine) 621
time-dependent UCORRFAC (user subroutine) 432
boundary conditions 155 UCRACK (user subroutine) 462
inelastic behavior 355 UCRACK_FATIGUE_LAW (user subroutine) 120, 122
time-independent, inelastic behavior 306 UCRACKGROW (user subroutine) 128
TIME-TEMP (model definition option) 448, 459, 460 UCRPLW (user subroutine) 101, 104, 365, 447
Time-Temperature-Transformation (T-T-T) 459 UDELAM (user subroutine 133
tolerance values, contact 551 UELASTOMER (user subroutine) 293
total Lagrangian procedure 72 UELASTOMER_ANISO (user subroutine) 298
TRANSFORMATION (history definition option) 699 UELDAM (user subroutine) 482
TRANSFORMATION (model definition option) 507, 681, UENERG (user subroutine) 293
689, 699, 715, 994 UEPS (user subroutine) 495, 840, 844, 857
TRANSIENT (history definition option) 759, 760, 761, UFAH (user subroutine) 770
764, 810, 892, 903, 908, 909, 929, 930 UFAIL (user subroutine) 214, 221
trimming 644 UFCONN (model definition option) 597
TRSFAC (user subroutine) 378 UFCONN (user subroutine) 597
truss elements 745 UFINITE (user subroutine) 309
Tsai-Wu failure criterion 220 UFORMSN (user subroutine) 686, 693
TSHEAR (parameter) 233, 991 UFOUR (user subroutine) 63
T-T-T (parameter) 459 UFRIC (user subroutine) 529, 530, 531
tube-tube contact 569 UFTHP (user subroutine) 774
tying UFXORD (model definition optiion) 597
overclosure 696 UFXORD (user subroutine) 597, 603
shell-to-solid 694 UGLAW (user subroutine) 774
TYING (model definition option) 648, 685, 686, 759, 897 UGLUESTIF_STS (user subroutine) 539
TYING CHANGE (history definition option) 686 UGROOV (user subroutine) 836
tying constraints UGROWRIGID (user subroutine) 508
beam-to-beam intersection 688 UHTCOE (user subroutine) 542, 543
cyclic symmetry 706 UHTCON (user subroutine) 542, 543
mesh refinement 686 UHTNRC (user subroutine) 542, 543
nodal degrees of freedom 693 UINITGAP (user subroutine) 509
pipe bend element 692 UMOONY (user subroutine) 290
rigid link 694 UMU (user subroutine) 495, 857
shell stiffener 692 UNEWTN (user subroutine) 163
shell-to-shell intersection 688
Index 1058
uniaxial cracking data 462 UBREAKGLUE 537
UNLCMD (user subroutine) 493 UCOHESIVE 489
UOGDEN (user subroutine) 291 UCOHESIVET 489
UPDATE (parameter) 149, 892 UCOMPL 140
updated Lagrange Formulation, nonlinear elasticity 305 UCONORT 532
updated Lagrangian Procedure 75 UCOORD 621
UPNOD (user subroutine) 162 UCORRFAC 432
UPOSTV (user subroutine) 997 UCRACK 462
URESTR (user subroutine) 836 UCRACK_FATIGUE_LAW 120, 122
URPFLO (user subroutine) 163 UCRACKGROW 128
User subroutine UCRPLW 101, 104, 365, 447
ANELAS 61, 202, 204, 395, 396 UDELAM 133
ANEXP 61, 204, 447 UELASTOMER 293
ANKOND 494, 759 UELASTOMER_ANISO 298
ANPLAS 204, 310 UELDAM 482
CRPLAW 101, 102, 103, 106, 107, 359, 361, 364, UENERG 293
366, 370, 447 UEPS 495, 840, 844, 857
CRPVIS 105, 358, 370 UFAH 770
ELEVAR 988, 991 UFAIL 214, 221
ELEVEC 988 UFCONN 597
FILM 676, 677, 759, 764, 766, 902, 907, 908, 909 UFINITE 309
FLUX 676, 677, 759, 764, 766, 836, 840, 844, 874, UFORMSN 686, 693
878, 900, 902, 907, 908, 909 UFOUR 63
FORCDF 909 UFRIC 529, 530, 531
FORCDT 155, 763, 764, 766, 837, 840, 844, 874, UFTHP 774
887, 900, 902, 908 UFXORD 597, 603
FORCEM 155, 621, 857, 874, 900, 902, 908, 909 UGLAW 774
HOOKLW 61, 202, 204, 395, 396 UGLUESTIF_STS 539
HOOKVI 374, 395, 396 UGROOV 836
HYPELA2 72, 247, 248, 464 UGROWRIGID 508
IMPD 988 UHTCOE 542, 543
INTCRD 988 UHTCON 542, 543
MOTION (2-D) 507 UHTNRC 542, 543
NASSOC 106, 107, 366 UINITGAP 509
ORIENT 61, 202, 310, 374 UINVAR 431, 432, 434
ORIENT2 209, 211, 212 UMOONY 290
PLOTV 997 UMU 495, 857
REBAR 461 UNEWTN 163
SEPFOR 557 UNLCDM 493
SEPSTR 557 UOGDEN 291
TRSFAC 378, 430 UPAYNE 406, 428, 429, 441
UADAP 625 UPNOD 162
UADAPTBOX 810 UPOSTV 997
UAMPLT 406 URESTR 836
UBEAR 837 URPFLO 163
USIGMA 495, 857 398, 482
USINC 155, 763 VISCELOGDEN (model definition option) 105, 291, 381,
USPLIT_MESH 132, 135 482
USPRNG 81, 703 VISCELORTH (model definition option) 374, 393, 395
UTHERDEG 798 VISCELPROP (model definition option) 373, 379
UTHICK 836, 837 VISCO ELAS (parameter) 105
UTIMESTEP 936 viscoelastic material
UTRANS 681 incompressible isotropic 374
UVELOC 759, 827, 836 isotropic 372
UVSCPL 102, 103, 107, 366 Thermo-Rheologically Simple 375
UWATERSL 798 viscoelasticity 105
UWELDFLUX 802, 803 Bergström-Boyce viscoelastic model 381
VSWELL 102, 103, 365, 370 frequency dependent damping 389
WKSLP 339, 340 viscoplasticity
YIEL 107, 366 explicit formulation 366
ZERO 107, 366 explicit method 106
user subroutine input 597 implicit method 107
user-defined Criterion 625 VOID CHANGE (model definition option) 241
user-defined failure criteria 221 VOLTAGE CHANGE (history definition option) 903, 907
USIGMA (user subroutine) 495, 857 von Mises 95, 96, 103, 175, 308, 312, 322, 338, 339,
USINC (user subroutine) 155, 763 341, 342, 343, 344, 361, 987, 989
USPLIT_MESH (user subroutine) 132, 135 VSWELL (user subroutine) 102, 103, 365, 370
USPRNG (user subroutine) 81, 703
UTHERDEG (user subroutine) 798 W
UTHICK (user subroutine) 836, 837 weighted mesh 601
UTIMESTEP (user subroutine) 936 WELD FILL (model definition option) 801, 802, 807, 808
UTRANFORM (model definition option) 681 WELD FLUX (model definition option) 801, 802, 805, 808,
UTRANS (user subroutine) 681 810
UVELOC (user subroutine) 759, 827, 836 WELD PATH (model definition option) 801, 802, 805
UVSCPL (user subroutine) 102, 103, 107, 366 welding 801
UWATERSL (user subroutine) 798 weld flux 802
UWELDFLUX (user subroutine) 802, 803 weld path 805
WELDING (parameter) 810
V Williams-Landel-Ferry (WLF) 105, 376
VCCT (model definition option) 107, 113 windings 850, 864
VELOCITY (model definition option) 759, 827, 836 WKSLP (user subroutine) 339, 340
VELOCITY CHANGE (history definition option) 759, 827, WORK HARD (model definition option) 163, 204, 335,
887 339, 340
VIEW FACTOR (model definition option) 677, 815 workhardening
virtual crack closure technique (VCCT) 107 isotropic 338
VISCEL EXP (model definition option) 106, 388 rules 338
VISCELCORR (model definition option) 432 slope 307
VISCELFOAM (model definition option) 105, 381 workspace allocation 35
VISCELHARM (parameter) 405, 432
VISCELMOON (model definition option) 105, 290, 381,
Index 1060
Y
YIEL (user subroutine) 107, 366
Young’s 204
Young’s modulus 201, 204, 295, 308, 338, 445, 459,
462, 464
Z
ZERO (user subroutine) 107, 366
Zienkiewicz-Zhu Creep Strain Criterion 624
Zienkiewicz-Zhu Criterion 622
zooming 188

Vola

Uploaded by

Document Information

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Vola

Uploaded by

Copyright:

Available Formats

Marc Volume A: Theory and User Information

Preface  About This Guide, 20

 About This Guide 20

About This Guide

Purpose of This Guide

Contents of This Guide

Chapter Title Description

Chapter Title Description

Example: Volume A: Theory and User Information

Accessing Marc Manuals

Downloading the PDF Documentation Files

Navigating the PDF Files

Printing the PDF Files

Training and Internet Resources

MAR101 - Basic Nonlinear Analysis using Marc and Mentat

MAR102 - Advanced Nonlinear Analysis using Marc and Mentat

MAR103 - Experimental Elastomer Analysis

MAR120 - Basic Nonlinear Analysis using Marc and Patran

MAR121 - Advanced Nonlinear Analysis using Marc and Patran

Help Us Help You

1 The Marc System

Preprocessing Mentat or Patran

Postprocessing Mentat or Patran

Figure 1-1 The Marc System

Marc for Analysis

Mentat or Patran for GUI

Program Function Library

Features and Benefits of Marc

Marc Workspace Requirements

General Memory stored in common/space/

Second Group of Memory stored in individual dynamically allocated vectors

Incremental Backup (4) Decomposed Stiffness Matrix (5)

5. Decomposed Stiffness matrix is for solver type 6 and 8.

Table 2-1 FORTRAN File Units Used by Marc

Table 2-1 FORTRAN File Units Used by Marc (continued)

Table 2-1 FORTRAN File Units Used by Marc (continued)

Table 2-1 FORTRAN File Units Used by Marc (continued)

Table 2-2 Keyword Descriptions*

Table 2-2 Keyword Descriptions* (continued)

Examples of Running Marc Jobs

Input of List of Items

1. Element numbers 2. Body numbers

Input File Version 13

Figure 3-1 Solid, Sheet and Wire CAD Bodies

b. Sheet bodies (surfaces):

Table Driven Input

1 time 32 current 63 3rd state variable

Independent variable types (5, 6, 7) will have the following characteristics.

Allowing Extrapolation No Extrapolation

The following mathematical functions may be used in an equation:

cos cosine (x) x in radians

dsin sine (x) x in degrees

Model Definition Options

History Definition Options

Figure 3-2 The Marc Input Data File

Table 4-1 Basic Input

More complex linear analyses require additional data blocks.

Table 4-2 Fourier Coefficients – Increment Number

Table 4-3 Elements Used for Fourier Analysis

Nodal forces are:

Modal Shapes and Buckling Load Estimations During a Fourier Analysis

Figure 4-3 Snap-Through

Figure 4-4 Contact and Friction Problem

Figure 4-5 Centrifugal Load Problem (Conservative)