Aerospace Structures and Materials (Liu - 2016)

Frontiers in Aerospace Science
Volume 1
Aerospace Structures and
Materials
Edited By
Yucheng Liu
Department of Mechanical Engineering
Mississippi State University
MS 39762
USA
Frontiers in Aerospace Science
Volume # 1
Aerospace Structures and Materials
Editor: Yucheng Liu
ISSN (Online): 2468-4724
ISSN: Print: 2468-4716
ISBN (eBook): 978-1-68108-305-6
ISBN (Print): 978-1-68108-306-3
© [2016], Bentham eBooks imprint.
Published by Bentham Science Publishers – Sharjah, UAE. All Rights Reserved.

BENTHAM SCIENCE PUBLISHERS LTD.
End User License Agreement (for non-institutional, personal use)
This is an agreement between you and Bentham Science Publishers Ltd. Please read this License Agreement
carefully before using the ebook/echapter/ejournal (“Work”). Your use of the Work constitutes your
agreement to the terms and conditions set forth in this License Agreement. If you do not agree to these terms
and conditions then you should not use the Work.
Bentham Science Publishers agrees to grant you a non-exclusive, non-transferable limited license to use the
Work subject to and in accordance with the following terms and conditions. This License Agreement is for
non-library, personal use only. For a library / institutional / multi user license in respect of the Work, please
contact: permission@benthamscience.org.
Usage Rules:
1. All rights reserved: The Work is the subject of copyright and Bentham Science Publishers either owns the
Work (and the copyright in it) or is licensed to distribute the Work. You shall not copy, reproduce, modify,
remove, delete, augment, add to, publish, transmit, sell, resell, create derivative works from, or in any way
exploit the Work or make the Work available for others to do any of the same, in any form or by any
means, in whole or in part, in each case without the prior written permission of Bentham Science
Publishers, unless stated otherwise in this License Agreement.
2. You may download a copy of the Work on one occasion to one personal computer (including tablet,
laptop, desktop, or other such devices). You may make one back-up copy of the Work to avoid losing it.
The following DRM (Digital Rights Management) policy may also be applicable to the Work at Bentham
Science Publishers’ election, acting in its sole discretion:
● 25 ‘copy’ commands can be executed every 7 days in respect of the Work. The text selected for copying
cannot extend to more than a single page. Each time a text ‘copy’ command is executed, irrespective of
whether the text selection is made from within one page or from separate pages, it will be considered as a
separate / individual ‘copy’ command.
● 25 pages only from the Work can be printed every 7 days.
3. The unauthorised use or distribution of copyrighted or other proprietary content is illegal and could subject
you to liability for substantial money damages. You will be liable for any damage resulting from your misuse
of the Work or any violation of this License Agreement, including any infringement by you of copyrights or
proprietary rights.
Disclaimer:
Bentham Science Publishers does not guarantee that the information in the Work is error-free, or warrant that
it will meet your requirements or that access to the Work will be uninterrupted or error-free. The Work is
provided "as is" without warranty of any kind, either express or implied or statutory, including, without
limitation, implied warranties of merchantability and fitness for a particular purpose. The entire risk as to the
results and performance of the Work is assumed by you. No responsibility is assumed by Bentham Science
Publishers, its staff, editors and/or authors for any injury and/or damage to persons or property as a matter of
products liability, negligence or otherwise, or from any use or operation of any methods, products instruction,
advertisements or ideas contained in the Work.
Limitation of Liability:
In no event will Bentham Science Publishers, its staff, editors and/or authors, be liable for any damages,
including, without limitation, special, incidental and/or consequential damages and/or damages for lost data
and/or profits arising out of (whether directly or indirectly) the use or inability to use the Work. The entire
liability of Bentham Science Publishers shall be limited to the amount actually paid by you for the Work.
General:
1. Any dispute or claim arising out of or in connection with this License Agreement or the Work (including
non-contractual disputes or claims) will be governed by and construed in accordance with the laws of the
U.A.E. as applied in the Emirate of Dubai. Each party agrees that the courts of the Emirate of Dubai shall
have exclusive jurisdiction to settle any dispute or claim arising out of or in connection with this License
Agreement or the Work (including non-contractual disputes or claims).
2. Your rights under this License Agreement will automatically terminate without notice and without the
need for a court order if at any point you breach any terms of this License Agreement. In no event will any
delay or failure by Bentham Science Publishers in enforcing your compliance with this License Agreement
constitute a waiver of any of its rights.
3. You acknowledge that you have read this License Agreement, and agree to be bound by its terms and
conditions. To the extent that any other terms and conditions presented on any website of Bentham Science
Publishers conflict with, or are inconsistent with, the terms and conditions set out in this License
Agreement, you acknowledge that the terms and conditions set out in this License Agreement shall prevail.
Bentham Science Publishers Ltd.

Executive Suite Y - 2
PO Box 7917, Saif Zone
Sharjah, U.A.E.
Email: subscriptions@benthamscience.org
CONTENTS
PREFACE .................................................................................................................................................................... i
LIST OF CONTRIBUTORS .................................................................................................................................... ii
CHAPTER 1 ANALYSIS OF LINEAR/NON-LINEAR AEROELASTIC RESPONSE OF SUPERSONIC

THICK FINS ............................................................................................................................................................... 3
R.D. Firouz-Abadi, S.M. Alavi, M. Rahmanian and H. Haddadpour
NOMENCLATURE ............................................................................................................................................. 4
INTRODUCTION ................................................................................................................................................ 5
Linear Structural Modeling of a General Fin ................................................................................................. 9
Non-Linear Structural Modeling of 3D Fins ................................................................................................ 11
Aerodynamic Modeling of the Flexible Fins ................................................................................................. 14
Steady Supersonic Flow Over the Fin ............................................................................................................ 15
Localized Piston Theory Including the Effects of Thickness and Angle of Attack ..................................... 18
General Concepts in Nonlinear Analysis of Dynamical Systems ................................................................ 20
Bifurcation Theory ....................................................................................................................................... 20
Phase Plots .................................................................................................................................................... 21
Chaos ............................................................................................................................................................ 22
Lyapunov Exponents .................................................................................................................................... 22
Numerical Results ....................................................................................................................................... 23
Verification of Aerodynamic Model ..................................................................................................... 23
Effect of Thickness and Angle of Attack on the Aerodynamic Coefficient ................................................ 25
Verification of Aeroelastic Model ................................................................................................................ 26
Effect of Geometrical Parameters on the Aeroelastic Analysis ................................................................... 28
Aeroelastic Analysis of Elastic Fins ............................................................................................................. 32
Non-linear Aeroelastic Analysis .................................................................................................................. 35
Effects of Airspeed Mach Number ............................................................................................................... 35
Effects of Thickness ..................................................................................................................................... 40
Effect of Initial Angle of Attack .................................................................................................................. 44
Effect of Hinge Position ............................................................................................................................... 47
Effect of Sweep Angle ................................................................................................................................. 51
Effect of Hinge Frictional Torque ................................................................................................................ 55
CONFLICT OF INTEREST ............................................................................................................................. 61
ACKNOWLEDGEMENTS ............................................................................................................................... 62
NOTES ................................................................................................................................................................ 62
REFERENCES ................................................................................................................................................... 62
CHAPTER 2 AN ANALYTICAL AND EXPERIMENTAL INVESTIGATION INTO VIBRATORY FORCE

FOR AIRCRAFT WINGS .......................................................................................................................................... 65
Xueguang Bi and Yucheng Liu
INTRODUCTION .............................................................................................................................................. 67
LITERATURE REVIEW .................................................................................................................................... 67
Analytical Study ............................................................................................................................................. 67
Experimental Study ........................................................................................................................................ 70
EVALUATE AERODYNAMIC FORCE USING 2DOF SYSTEM ................................................................ 71
2DOF Airfoil System ..................................................................................................................................... 71
Force Determination Methods ...................................................................................................................... 74
Traditional Force Determination Method ............................................................................................ 74
Force Determination Using “H1”& “H2” Estimated Frequency Response Matrices ....................... 75
Direct Force Determination “F1” and “F2” Methods ....................................................................... 76
Validation of Force Determination Methods .............................................................................................. 76
EVALUATE AERODYNAMICS FORCE USING MDOF SYSTEM .......................................................... 85
MDOF Aircraft Wing Model ....................................................................................................................... 85
Validation of Force Determination Methods .............................................................................................. 91
DISCUSSIONS ................................................................................................................................................... 98
EXPERIMENT TESTINGS .............................................................................................................................. 98
Experiment Model ........................................................................................................................................ 98
Ground Vibration Test ................................................................................................................................. 100
Experiment Setup ................................................................................................................................. 100
Modal Analysis Methods ...................................................................................................................... 102
Results and Discussions ....................................................................................................................... 103
Wind Tunnel Test ......................................................................................................................................... 106
Theoretical Background ....................................................................................................................... 106
Experiment Setup ................................................................................................................................. 107
Results and discussions ........................................................................................................................ 108
CONCLUSION ................................................................................................................................................. 109
REFERENCES ................................................................................................................................................... 110
CHAPTER 3 COMPUTATIONAL AND ANALYTICAL INVESTIGATION OF LATERAL IMPACT

BEHAVIOR OF PRESSURIZED PIPELINES ....................................................................................................... 113
Yangqing Dou and Yucheng Liu
NOMENCLATURE ............................................................................................................................................. 114
I NTRODUCTION................................................................................................................................................ 114
PROBLEM DESCRIPTION ............................................................................................................................... 116
MATERIAL PROPERTIES ............................................................................................................................... 117
COMPUTER MODELING AND ANALYSIS .................................................................................................. 118
COMPUTATIONAL IMPACT BEHAVIOR OF THE PIPELINE MODELS ............................................. 121
FEA Results.................................................................................................................................................... 121
Effects of Impact Position on the Impact Response of Pipelines................................................................... 124
Effects of Internal Pressure on the Impact Response of Pipelines ................................................................. 130
DISCUSSION ON COMPUTER RESULTS ..................................................................................................... 132
ANALYTICAL ANALYSIS .............................................................................................................................. 134
Response Surface Method (RSM) .........
........................................................................................................ 134
RSM Models and Assessment ...................................................................................................................... 136
PARAMETRIC STUDY .................................................................................................................................... 141
Influence of Internal Pressure ...................................................................................................................... 141
INFLUENCE OF OUTSIDE DIAMETER ...................................................................................................... 144
CONCLUSION ................................................................................................................................................. 148
REFERENCES ................................................................................................................................................... 150
CHAPTER 4 EFFECT OF BONDLINE THICKNESS ON THE TRACTION-SEPARATION LAWS OF

ADHESIVELY BONDED JOINT ............................................................................................................................ 153
Gefu Ji, Zhenyu Ouyang and Guoqiang Li
INTRODUCTION ............................................................................................................................................ 154
Background .................................................................................................................................................... 154
Adhesive Bonding Technology ............................................................................................................... 154
Three Basic Modes of Fracture ..................................................................................................................... 155
Classical Fracture Mechanics ........................................................................................................................ 156
Nonlinear Fracture Mechanics ....................................................................................................................... 157
Cohesive Zone Method ........................................................................................................................... 157
J-Integral ........................................................................................................................................................ 157
Application of CZM in Adhesive Joint .......................................................................................................... 158
Effects of Adhesive Thickness ....................................................................................................................... 162
Research Objectives .................................................................................................................................... .163
THEORETICAL BACKGROUND .............................................................................................................. 164
Theoretical Derivations ................................................................................................................................ 165
Governing Equations .................................................................................................................................... 168
J- Integral for Interface Shear Behaviors ..................................................................................................... 169
J- Integral for Interface Peel Behaviors ........................................................................................................ 171
Determination of Interface Cohesive Laws .................................................................................................. 172
Fracture Tests by DCB, ENF and SLB Specimens ..................................................................................... 173
DCB Specimens for Mode-I Fracture Test .................................................................................................. 174
ENF Specimens for Mode-II Fracture Test .................................................................................................. 175
SLB Specimens for Mixed Mode I/II Fracture Test .................................................................................... .177
EXPERIMENTAL METHOD ...................................................................................................................... 180
Raw Material ............................................................................................................................................... 180
Adhesive Material ................................................................................................................................ 180
Adherends Material ............................................................................................................................. 181
Fabrication of Specimen ............................................................................................................................... 181
Steel Based DCB Specimen for Mode I Test ............................................................................................... 182
Steel Based ENF Specimen for Mode II Test .............................................................................................. 183
Steel Based Single Leg Bending Specimen for Mixed Mode I/II Test ........................................................ 184
Fabrication of Self-Aligned Ball Pin ............................................................................................................ 185
Instrumentation and Test Method ................................................................................................................ 186
RESULTS AND DISCUSSION OF MODE I TEST ................................................................................... 189
Results and Discussions of Mode I Test ..................................................................................................... 189
Global Test Results ............................................................................................................................... 189
Local Test Results ................................................................................................................................. 192
Effects of the Thickness of the Adhesive Layer ........................................................................................... 194
RESULTS AND DISCUSSIONS OF MODE II TEST .
............................................................................... 200
Global and Local Experimental Results ....................................................................................................... 200
RESULTS AND DISCUSSION OF MIXED MODE I AND MODE II TEST .
......................................... 209
Global and Local Test Results ..................................................................................................................... 209
Mode I Portion ..................................................................................................................................... 216
Mode II Portion .................................................................................................................................... 218
Mode Mixities ....................................................................................................................................... 219
Effects of the Initial Crack Tip Length of the Adhesive Layer .................................................................... 221
CONCLUSIONS AND FUTURE WORKS .
................................................................................................. 228
Mode I Test .................................................................................................................................................. 228
Mode II Test ................................................................................................................................................. 229
Mixed Mode I/II Test ................................................................................................................................... 230
REFERENCE ..................................................................................................................................................... 231
CHAPTER 5 OPTIMIZATION OF GEOMETRIC PARAMETERS AND REVERSING DESIGN
METHODOLOGY OF INVESTMENT CASTING DIE PROFILE OF TURBINE BLADE .................. 240
Yangqing Dou, Yangliu Dou, Kun Bu and Yiwei Dong
NOMENCLATURE ............................................................................................................................................. 241
INTRODUCTION ................................................................................................................................................ 242
OPTIMIZATION OF GEOMETRIC PARAMETERS FOR TURBINE BLADE........................................ 244
Geometric Parameter Extraction of Cross-section ......................................................................................... 244
Extraction of Bending Control Parameters .................................................................................................... 245
Extraction of Torsional Control Parameters................................................................................................... 247
Optimization of Torsional Control Parameters .............................................................................................. 247
Optimization of Bending Control Parameters ................................................................................................ 247
Results and Discussion ................................................................................................................................... 251
OPTIMUM DESIGN METHOD OF DIE PROFILE ..................................................................................... 254
ANALYSIS OF FILES GENERATED DURING NUMERICAL SIMULATION PROCESS .................. 256
FEM PROCESSING .......................................................................................................................................... 257
EXAMPLE VERIFICAITON ........................................................................................................................... 260
CONCLUSION ................................................................................................................................................... 265
REFERENCES ................................................................................................................................................... 266
CHAPTER 6 VIBRATION INDUCED FATIGUE ANALYSIS OF AEROSPACE STRUCTURES .............. 269

Murat Aykan and Mehmet Çelik
INTRODUCTION ................................................................................................................................................ 270
FLIGHT TESTS AND NUMERICAL ANALYSES OF AN ANTENNA FOR SAFETY OF FLIGHT
......................................................................................................................................................................... 273
Operational Flight Tests before Integration .................................................................................................. 273
Numerical Structural Analyses....................................................................................................................... 273
Operational Flight Tests After Integration ..................................................................................................... 277
VIBRATION INDUCED FATIGUE ANALYSIS AND TESTING OF A CHAFF/FLARE DISPENSER
BRACKET .................................................................................................................................................... 279
Vibration Fatigue Analysis Method ............................................................................................................... 281
Fatigue Analysis and Tests ........................................................................................................................... 286
Analysis of Design-1 ............................................................................................................................. 290
.
CONCLUSION ................................................................................................................................................. 297
REFERENCES ................................................................................................................................................... 299
CHAPTER 7 FATIGUE LIFE OPTIMIZATION OF LASER PEENED AIRCRAFT COMPONENTS ....... 302
Anoop Vasu and Ramana V. Grandhi
NOMENCLATURE ............................................................................................................................................. 303
INTRODUCTION ................................................................................................................................................ 303
INFLUENCING FACTORS FOR FATIGUE LIFE PREDICTION .............................................................. 308
Applied Load .................................................................................................................................................. 309
Laser Peening Induced Residual Stresses ...................................................................................................... 310
Residual Stress Relaxation ............................................................................................................................. 311
Fatigue Life .................................................................................................................................................... 312
Re-Peening ..................................................................................................................................................... 314
OPTIMIZATION PROCEDURE....................................................................................................................... 314
Design Variables ............................................................................................................................................ 315
Re-Peening Schedule .............................................................................................................................. 315
Mid-Span Duration ................................................................................................................................ 316
Peak Pressure ....................................................................................................................................... 317
Percentage Overlap .............................................................................................................................. 317
Optimization strategy ................................................................................................................................... 319
RESULTS AND DISCUSSION ......................................................................................................................... 320
Compressive Residual Stress Optimization Results ..................................................................................... 320
Optimization of the Re-Peening Schedule ................................................................................................... 322
CONCLUSION ................................................................................................................................................. 323
REFERENCES ................................................................................................................................................... 324
CHAPTER 8 PREDICTION OF RESIDUAL STRESS RELAXATION IN TI-6AL-4V SUBJECTED TO

LASER PEENING ...................................................................................................................................................... 327
Anoop Vasu and Ramana V. Grandhi
NOMENCLATURE ............................................................................................................................................. 328
INTRODUCTION ................................................................................................................................................ 329
LASER PEENING MECHANISM..................................................................................................................... 332
Pressure Pulse Generation .............................................................................................................................. 333
High Strain Rate Phenomenon ...................................................................................................................... 334
Plasticity in Laser Peening ............................................................................................................................. 335
FINITE ELEMENT MODELING OF LASER PEENING ............................................................................. 336
Pressure Pulse Input ....................................................................................................................................... 337
Simulation Procedure ..................................................................................................................................... 337
Constitutive Model ......................................................................................................................................... 339
Effect of Temperature in Laser Peening Process ........................................................................................... 339
Curvature Effect ........................................................................................................................................... 342
RESIDUAL STRESS RELAXATION ............................................................................................................. 344
Obtain Empirical Data for Calibration ......................................................................................................... 346
FE Modeling of Stress Relaxation ............................................................................................................... 347
RESULTS AND DISCUSSION ......................................................................................................................... 348
CONCLUSION ................................................................................................................................................. 350
REFERENCES ................................................................................................................................................... 351
CHAPTER 9 NON-DESTRUCTIVE EVALUATION (NDE) OF WELDED STRUCTURES FOR

AEROSPACE APPLICATIONS ........................................................................................................................... 355
Mohammad W. Dewan, Daniel J. Huggett and Muhammad A. Wahab
INTRODUCTION ................................................................................................................................................ 356
ULTRASONIC TESTING (UT) ......................................................................................................................... 358
FUNDAMENTALS OF ULTRASONIC TESTING ......................................................................................... 359
CONVENTIONAL ULTRASONIC TESTING (UT) ....................................................................................... 364
TIME-OF-FLIGHT-DIFFRACTION (TOFD) ULTRASONIC TESTING................................................... 366
PHASED ARRAY ULTRASONIC TESTING (PAUT) ................................................................................... 369
Defect Sizing Using PAUT .......................................................................................................................... 373
Case Study: Welding Defect Detection with PAUT .................................................................................... 376
RADIOGRAPHIC TESTING (RT) .
........................................................................................................... 378
FUNDAMENTALS OF RADIOGRAPHIC TESTING .
........................................................................... 379
Case study: Welding Defect Detection with RT .......................................................................................... 385
Comparison of PAUT and RT to obtain Welding Defect ............................................................................ 386
CONCLUSION .......................................................................................................................................... 387
REFERENCES ................................................................................................................................................... 388
CHAPTER 10 DEVELOPMENT OF A REMOTELY PILOTED HELICOPTER FOR CIVIL
APPLICATIONS ...................................................................................................................................................... 393
N. Goudarzi, R. M. Ziazi, F. Kermanshahi, S. Sadati, M. S. Sajedi and M. Mohaghegh
NOMENCLATURE ............................................................................................................................................. 394
INTRODUCTION ............................................................................................................................................ 396
CONCEPTUAL DESIGN ................................................................................................................................ 397
2.1. Weight Sizing .......................................................................................................................................... 398
2.2. Configuration Selection ......................................................................................................................... 399
2.3. Rotor Sizing............................................................................................................................................. 400
2.4. Aerodynamic Sizing ................................................................................................................................ 402
2.5. Power Sizing ........................................................................................................................................... 402
DETAILED DESIGN ....................................................................................................................................... 403
3.1. Engine and Fuel System ......................................................................................................................... 404
3.2. Blade Selection ...................................................................................................................................... 408
3.3. Power Transmitting System .................................................................................................................. 409
STABILITY AND CONTROL OF A REMOTELY PILOTED HELICOPTER .................................... 412
4.1. Dynamic Modeling ................................................................................................................................ 414
4.1.1. Methodology ............................................................................................................................... 414
4.1.2. Rigid Body Equations of Motion ................................................................................................ 418
4.2. Trim and Stability Analysis .................................................................................................................. 420
4.2.1. Analysis of the Trim .................................................................................................................... 420
4.2.2. Stability ....................................................................................................................................... 421
CATIA MODELING ..................................................................................................................................... 427
MANUFACTURING AND ASSEMBLING ................................................................................................ 428
TEST AND RELIABILITY ANALYSIS ..................................................................................................... 430
CONCLUSION ............................................................................................................................................... 437
REFERENCES ................................................................................................................................................... 438
CHAPTER 11 MATERIALS SELECTION IN DESIGN OF STRUCTURES OF SUBSONIC AND

SUPERSONIC AIRCRAFTS ................................................................................................................................. 442
Zainul Huda
NOMENCLATURE ............................................................................................................................................. 443
INTRODUCTION ................................................................................................................................................ 443
OPERATING CONDITIONS AND MATERIAL SELECTION .................................................................... 445
Environmental Service Conditions ................................................................................................................ 445
Loading Conditions and Design Relationships .............................................................................................. 447
Design Equation for Tension.................................................................................................................. 447
Design Equation for Compression ......................................................................................................... 448
Design Equation for Bending ................................................................................................................. 448
Design Equation for Torsion .................................................................................................................. 448
Materials Selection for Aircraft Structure ---- A Design Approach .............................................................. 449
MATERIALS SELECTION BASED ON SPEED AND TEMPERATURE ................................................ 451
Materials Selection for Subsonic Structures (for Speeds ≤ 1.0 Mach) ........................................................ 451
Materials Selection for Structures with Speeds ≤ 2.0 Mach ........................................................................ 454
Materials Selection for Structure for Speeds Between 2.0 and 4.0 Mach ................................................... 456
Material-Selection for Military Aircraft Wings ........................................................................................... 460
A NEW MATERIALS SELECTION CHART FOR SSTAS ......................................................................... 462
CONCLUSIONS AND RECOMMENDATIONS ........................................................................................... 463
REFERENCES ................................................................................................................................................... 464
CHAPTER 12 POTENTIAL AEROSPACE APPLICATIONS OF CARBON NANOTUBES ........................ 468
Omid Gohardani
NOMENCLATURE ............................................................................................................................................ 469
INTRODUCTION ................................................................................................................................................ 469
CARBON NANOTUBES IN AEROSPACE - HISTORICAL OVERVIEW ................................................. 471
CARBON NANOTUBES IN AERONAUTICS................................................................................................. 474
Recent Studies ................................................................................................................................................ 479
CARBON NANOTUBES AND UNMANNED AERIAL VEHICLES .......................................................... 480
Recent Studies .............................................................................................................................................. 481
CARBON NANOTUBES IN ASTRONAUTICS ............................................................................................ 481
Recent Studies .............................................................................................................................................. 482
CHALLENGES RELATED TO CARBON NANOTUBE USAGE IN AEROSPACE SCIENCES .......... 482
DISCUSSION AND FUTURE OUTLOOKS .................................................................................................. 485
ACKNOWLEDGMENTS ................................................................................................................................. 486
REFERENCES ................................................................................................................................................... 486
CHAPTER 13 COMBINATION OF CARBON FIBER SHEET MOLDING COMPOUND AND PRE-
IMPREGNATED, TAILORED CARBON FIBER REINFORCEMENTS .......................................................... 497
Marc Fette, Nicole Stöß, Jens Wulfsberg, Axel Herrmann, Gerhard Ziegmann and Georg Lonsdorfer
NOMENCLATURE ............................................................................................................................................. 498
INTRODUCTION ................................................................................................................................................ 498
STATE OF THE ART ......................................................................................................................................... 500
AIM AND PURPOSE OF THE PRESENT WORK ......................................................................................... 501
Experimental Procedures................................................................................................................................ 504
DISCUSSION ON RESULTS ........................................................................................................................... 508
CONCLUSION ................................................................................................................................................. 512
REFERENCES ................................................................................................................................................... 513
CHAPTER 14 DESIGN OPTIMIZATION OF VARIABLE STIFFNESS COMPOSITE STRUCTURES FOR
AEROSPACE APPLICATIONS ............................................................................................................................... 515
Mohammad Rouhi, Hossein Ghayoor, Suong V. Hoa and Mehdi Hojjati
INTRODUCTION ............................................................................................................................................ 516
MODELING AND ANALYSIS ...................................................................................................................... 519
DESIGN OPTIMIZATION ............................................................................................................................. 522
RESULTS .......................................................................................................................................................... 524
4.1. Bending-Induced Buckling of a Circular Cylinder ................................................................................. 524
4.1.1. Effect of the Aspect Ratio (L/R) .................................................................................................. 526
4.1.2. Effect of the Radius (R) .............................................................................................................. 528
4.1.3. Effect of the Steered Plies Percentage ....................................................................................... 529
4.2. Multi-Objective Design Optimization for Bending-Induced Buckling ................................................. 530
4.3. Axial Buckling of an Elliptical Cylinder ............................................................................................... 532
CONCLUSION ............................................................................................................................................. 535
REFERENCES ................................................................................................................................................... 535
CHAPTER 15 MANUFACTURING CHALLENGES ASSOCIATED WITH THE USE OF METAL
MATRIX COMPOSITES IN AEROSPACE STRUCTURES ......................................................................... 542
Tracie Prater
NOMENCLATURE ............................................................................................................................................. 543
INTRODUCTION ................................................................................................................................................ 544
PROBLEM DESCRIPTION ............................................................................................................................... 547
RELATIONSHIP BETWEEN WEAR AND PROCESS PARAMETERS .................................................... 548
CONSTRUCTING AN ANALYTICAL MODEL OF THE WEAR PROCESS............................................ 551
COMBATTING WEAR: SELECTION OF TOOL MATERIALS .............................................................. 557
IN SITU SENSING OF WEAR USING TORQUE ........................................................................................ 558
CONCLUSION ................................................................................................................................................. 559
REFERENCES ................................................................................................................................................... 561
SUBJECT INDEX .................................................................................................................................................... 564

i
PREFACE
This book covers a variety of aspects of design, analysis, and manufacturing of aerospace
structures and materials, which include design/failure criteria for aerospace structures,
materials selection, manufacturing methods, engineering design and analysis of structural
assemblies and components, concurrent engineering, and interdisciplinary design technology.
Challenges that current aerospace engineers are faced with are as follows. First, they must
understand the new and existing materials and the level of engineering already applied in their
creation. Second, they need to develop a working design by combining material behavior and
structural function to meet all the design requirements. Third, an optimum design has to be
conducted based on the original working design to address manufacturability and economical
constraints. Finally, the durability and longevity of the final structure should be continually
assessed.
This book equips aerospace engineers with the necessary knowledge and practical skills to
tackle these challenges in an industrial or R&D environment. Readers of this book will
acquire knowledge of a broad range of the entire process in the context of the design,
manufacturing, and analysis of aerospace structures and components. The readers will also be
exposed to advanced research and development in fields directly relevant to aerospace
engineering, which will improve our capability to design the next generation of aerospace
structures and materials.
Interesting topics that will be demonstrated through this book include: structural dynamics
and impact simulation, acoustic and vibration testing and analysis, fatigue analysis and life
optimization, reversing design methodology, non-destructive evaluation, remotely piloted
helicopter, surface enhancement of aerospace alloys, manufacturing of metal matrix
composites, aerospace applications of carbon nanotubes, carbon fiber reinforcements, variable
stiffness composites, aircraft material selection, etc. Furthermore, besides aerospace
engineering, the topics included in this book also impact a broad variety of engineering areas
such as mechanical engineering, materials engineering and science, civil engineering,
automotive engineering, computer engineering and science, etc.
The book is intended to serve as a reference for researchers, engineers, engineering faculty, as
well as graduate students. Finally, the editor would like to extend cordial appreciation to all
contributors for their great support and invaluable contributions.
Yucheng Liu
Department of Mechanical Engineering
Mississippi State University
MS 39762, USA
ii
List of Contributors
Department of Mechanical and Materials Engineering, Wright State University,
Anoop Vasu
Dayton, USA
Axel Herrmann Composite Technology Center GmbH (An Airbus Company), Stade, Germany
Department of Mechanical and Industrial Engineering, Louisiana State

Daniel J. Huggett
University, Baton Rouge, USA
F. Kermanshahi Ecole Polytechnique de Montreal, Quebec, Canada
Department of Mechanical & Industrial Engineering, Louisiana State

Gefu Ji
Georg Lonsdorfer Composite Technology Center GmbH (An Airbus Company), Stade, Germany
Institute of Polymer Materials and Plastics Engineering, TU Clausthal,

Gerhard Ziegmann
Clausthal, Germany

Guoqiang Li University, Baton Rouge, USA
Department of Mechanical Engineering, Southern University, Baton Rouge,
USA
Concordia Center for Composites, Department of Mechanical and Industrial

Hossein Ghayoor
Engineering, Concordia University, Montreal, Canada
Department of Aerospace Engineering, Sharif University of Technology,

H. Haddadpour
Tehran, Iran
Institute of Production Engineering, Helmut Schmidt University, Hamburg,

Jens Wulfsberg
Germany.
Kun Bu Department of Mechanical Engineering, Mississippi State University, USA
Institute of Production Engineering, Helmut Schmidt University, Hamburg,

Marc Fette Germany.
Composite Technology Center GmbH (An Airbus Company), Stade, Germany

Mehdi Hojjati
Mehmet Çelik Department of Mechanical Engineering, SST Division-ASELSAN Inc., Turkey

Mohammad Rouhi

Mohammad W. Dewan
iii

Muhammad A. Wahab
Department of Mechanical Engineering, MGEO Division-ASELSAN Inc.,

Murat Aykan
Turkey
M. Mohagheghi Aerospace Engineering Department, University of Tehran, Tehran, Iran

M. Rahmanian
Tehran, Iran
Aerospace Engineering Department, Amirkabir University of Technology,

M. S. Sajedi
Tehran, Iran
Nicole Stöß Polynt Composites Germany GmbH, Kieselstraße, Miehlen, Germany
N. Goudarzi Mechanical Engineering Department, University of Maryland, USA
Omid Gohardani Springs of Dreams Corporation, Tustin, USA
Department of Mechanical and Materials Engineering, Wright State University,

Ramana V. Grandhi
Dayton, USA

R.D. Firouz-abadi
Tehran, Iran
School of Mechanical, Industrial, and Manufacturing Engineering, Oregon State

R. M. Ziazi
University, USA

Suong V. Hoa

S.M. Alavi
Tehran, Iran
Sustainable Environment and Energy Systems Department, Middle East

S. Sadati
Technical University, Northern Cyprus Campus
Tracie Prater NASA Marshall SpaceFlight Center, Huntsville, USA
Xueguang Bi Department of Mechanical Engineering, Mississippi State University, USA
Yangliu Dou Department of Mechanical Engineering, Mississippi State University, USA
Yangqing Dou Department of Mechanical Engineering, Mississippi State University, USA
Yiwei Dong Department of Mechanical Engineering, Mississippi State University, USA
Yucheng Liu Department of Mechanical Engineering, Mississippi State University, USA
Department of Mechanical Engineering, King Abdulaziz University, Jeddah,

Zainul Huda
Saudi Arabia

Zhenyu Ouyang
Frontiers in Aerospace Science, 2016, Vol. 1, 3-64 3
CHAPTER 1
Analysis of Linear/Non-Linear Aeroelastic

Response of Supersonic Thick Fins
R.D. Firouz-Abadi*, S.M. Alavi, M. Rahmanian and H. Haddadpour

Tehran, P.O.Box 11115-8639, Iran
Abstract: This study introduces an aeroelastic model for the linear/non-linear

analysis of thick fins in supersonic or hypersonic regimes. In the first step a linear
aeroelastic model for the analysis of thick fins is developed. To this aim, a thick fin
with two degrees of freedom (2 DOF) as well as an elastic double-wedged fin in
supersonic/hypersonic flight regimes are considered. An unsteady aerodynamic
model is developed based on the shock/expansion theory by local aspplication of
the piston theory over the flat surfaces of the fin. The structural model is also
obtained based on the Lagrangian approach. Employing such developed model, the
effects of initial angle of attack, thickness and some other geometrical parameters
on the aeroelastic stability boundaries and unsteady aerodynamic loads are studied.
In the next step, a more sophisticated model describing the non-linear aeroelastic
behavior of fins with 3 DOF and free-plays in flapping, plunging and pitching
motions is investigated. To this aim, governing equation are obtained by a
modification of the linear model and effects of several geometrical parameters (e.g.
thickness, initial angle of attack, hinge frictional torque etc.) on the aeroelastic
behavior of fins are assessed1.
*Corresponding author R.D. Firouz-abadi: Department of Aerospace Engineering, Sharif University of

Technology, Tehran, P.O.Box 11115-8639, Iran; Tel: (+9821) 66164606; Fax: (+9821) 66022731; E-mail:
firouzabadi@sharif.edu.
Yucheng Liu (Ed.)

All right reserved-© 2016 Bentham Science Publishers
4 Frontiers in Aerospace Science, Vol. 1 Firouz-Abadi et al.
Keywords: Aeroelasticity, Free-play, Local piston theory, Shock and

expansion analysis, Thick supersonic fin.
NOMENCLATURE
Cα, Cβ, Ch Structural damping.

α
C Aerodynamic damping matrix.
Cs Structural damping matrix.
D Rayleigh dissipation function.
e Elastic displacement vector.
𝐞̅ Natural mode shapes of the elastic displacements.
f Generalized forces.
h Plunging displacement.
̅ , 𝐼 xz
𝐼 xx ̅ , 𝐼 zz̅ Mass moment of inertia
Kα, Kβ, K h Spring Stiffness.
Kα Aerodynamic stiffness matrix.
KE Kinetic energy.
Ks Structural stiffness matrix.
Lα Lift.
m Fin mass.
M Mach number.
Mα Flapping moment.
Ms Structural mass matrix.
n Outward unit normal of the undeformed surface.
p Pressure.
P Potential energy.
q ith generalized coordinate.
Qi ith generalized forces.
Linear/Non-Linear Aeroelastic Response Frontiers in Aerospace Science, Vol. 1 5
Sα, Sβ Static mass moment of the fin about the elastic axis.
T Temperature.
Tα Pitching moment.
u Flow velocity.
v Unit direction vector of the air velocity.
w Displacements along the normal vector of each surface.
x Vector of dynamic states.
𝑥̅ ,𝑧̅ Position of the center of gravity related to elastic axis.
α Pitch angle.
β Flap angle.
αf, βf, hf Free-plays of the corresponding degrees of freedom.
γ Specific heat ratio.
ν Prandtl-Meyer function.
θ Elastic rotation vector.
θ Natural mode shapes of the elastic rotations.
ρ Density.
ω Natural frequency.
ψ Oblique wave angle.
ξ Generalized modal coordinates.
INTRODUCTION
Commonly, an aeroelastic problem has two types of modeling including

aerodynamic loads and structural response calculation. Combination of
these two types of modeling yields a coupled fluid-structure system which
may be unstable in some flight conditions. In the present chapter,
aeroelastic response of fins in supersonic/hypersonic flow is aimed. This
matter leads to centralize on the unsteady aerodynamic loading models in
supersonic/hypersonic regimes. One of the most applied unsteady
aerodynamic tools in supersonic regime is the piston theory. This theory

describes the relation between the unsteady pressures over a flat plate in
terms of the local angle of attack. Due to its simplicity and good accuracy,
it has been considered as the most powerful as well as fast model for
derivation of the unsteady aerodynamic pressure over a flat plate in panel
flutter analysis.
Also, the piston theory is known as a great technique for computation of

unsteady aerodynamic loads, but its application on thick fins with nonzero
angle of attack is accompanied with considerable errors. Liu [1] showed
that the wing’s thickness has a considerable effect on the supersonic
aerodynamic characteristics, in particular, shifting the location of the center
of pressure forward and as a result reducing the flutter speed. Based on this
fact, many researchers have tried to improve the results of piston theory in
the aeroelastic and aerothermoelastic problems.
One method to increase the accuracy of unsteady pressure distribution

computation is the local usage of the piston theory over thick supersonic
fins which was firstly suggested by Zhang et al. [2, 3]. According to this
approach, first, flow properties are computed over the aerial structure
surfaces by utilizing the Euler method, subsequently, the piston theory is
applied locally to compute the unsteady aerodynamic loads. Firouz-abadi
and Alavi [4] extended this approach and proposed a fast algorithm for the
aeroelastic analysis of flexible supersonic fins with arbitrary polygonal
cross-section and incidence angle of attack. Based on the shock/expansion
wave theory, they proposed a procedure to calculate the steady
aerodynamic pressure over the wing instead of going through
complications of the CFD simulations. This procedure avoids the extensive
numerical computations and grid generation problems as well as providing
a highly reasonable accuracy.
Another important subject in aerodynamic models is considering the
aerodynamic nonlinearities through the simulation. Aerodynamic
nonlinear response are commonly observed in turbulent flows and gusts.

Flight at high angles of attack, shock waves in transonic and supersonic
flow regimes, turbulence, flow ionization and high temperature fields are
the other sources of aerodynamic nonlinearities which can affect the
aeroelastic behavior of the wing [5, 6].
As mentioned above, the other side of an aeroelastic problem is the

structural modeling. According to the literature, variety of structural
models have been developed for the aeroelastic analysis of supersonic fins.
Based on fin’s degree of freedom (DOF), these models can be classified
from simple mass and stiffness models for 2-DOF fins to modal analysis
based models for the elastic fins. As well as the aerodynamic loads,
nonlinear structural models can improve the accuracy for specific
situations. Structural nonlinearities are mainly due to large amplitude
vibrations (geometric nonlinearity), material nonlinearities, thermal
degradation and nonlinearity in the boundary conditions, which have
considerable contributions in accurate estimations of the structural
behavior. It’s customary, as a conventional modeling technique, to consider
structural nonlinearities as some equivalent nonlinear springs.
Seo et al. [7] followed the same structural modeling technique as

mentioned above and studied the aeroelasticity of a 2-D typical section
wing with multiple nonlinearities. They compared the nonlinearities in
pitch and plunge springs with a single nonlinearity in the pitch. This
comparison showed that, multiple structural nonlinearities in the 2-D
typical wing section models are influenced by the pitch to plunge
frequency ratio. Accordingly, it is deduced that the flutter speed is closely
related to the frequency ratio providing that both pitch and plunge
nonlinearities are present rather than just a single pitch nonlinearity.
Using a 2-DOF typical section model, McIntosh et al. [8] investigated the
stability and dynamic response of a supersonic wing. They considered both
hardening and softening nonlinearities of the support springs and gave
some comments for the accuracy improvements in the theoretical model.
Another conventional source of structural nonlinearities is the existence of

free-plays in the actuator or support mechanism of a control fin or wing due
to the manufacturing tolerances. This type of nonlinearity can impose limit
cycle oscillations (LCOs) or chaotic vibrations [9-11]. Woolston et al. [12]
investigated the effects of free-play in pitch, hysteresis and cubic
nonlinearities on the aeroelastic stability of a 3-DOF wing and confirmed
their results with the wind tunnel experiments. They showed that the
free-play in pitch angle causes a remarkable decrease of the flutter speed
compared to the results of the linear eigenvalue analysis of the same
configuration. Yang and Zhao [13] performed theoretical and experimental
studies on the aeroelastic instability of a two-dimensional wing model with
non-linear pitching stiffness and free-plays. They came up with two stable
LCOs of different amplitudes for some flight conditions. Hauenstein et al.
[14] assessed the aeroelastic response of a 2-DOF wing with the free-play
in pitch and plunge motions and obtained non-harmonic and chaotic
responses. Based on the analytical and experimental investigations, they
concluded that chaotic response does not occur when only one single
nonlinearity is considered in the wing support.
Abbas et al. [15] used the third-order piston theory along with a 2-DOF
wing section model to analyze the supersonic flutter and post flutter
behaviors under the effects of cubic stiffness nonlinearities and free-plays
in both plunging and pitching motions. They observed LCOs and chaotic
behaviors over a large variety of configurations of the aeroelastic system.
Tang and Dowell [16] performed theoretical and experimental aeroelastic
studies to investigate the effect of nonzero initial angle of attack on the
flutter, LCO and gust response of a typical wing section with control
surface free-play in an incompressible flow.
In this chapter, modeling the unsteady aerodynamic loads acting on thick
supersonic fins are estimated by the local application of the piston theory.
To this aim the steady state properties of the flow over the surface of the
fin are obtained using the shock/expansion wave equations and the piston
theory is applied locally to compute the aerodynamic loads in terms of the

fin’s DOF. To show the feasibility of the aerodynamic modeling, linear
aeroelastic analysis of two different fins (a 2 DOF rigid and an elastic fin)
are performed and the stability margins are determined. Next, a more
comprehensive model of 3D fins with three degrees of freedom
(constrained in plunging, pitching and flapping motions with springs) is
considered. Assuming free-plays in all degrees of freedom, governing
equations of motion are derived and the unsteady aerodynamic lift and
moments are calculated using the previously introduced procedure.
Finally, the linear and nonlinear aeroelastic stability is performed and the
effect of thickness, incidence angle of attack, hinge position, sweep angle
and hinge frictional viscous damping torque on the LCOs and chaotic
behavior of the fin is outlined.
Linear Structural Modeling of a General Fin
For the general aeroelastic problem, the structural model is usually

obtained by applying the Lagrangian dynamics and then implementing the
modal analysis technique. Following the Lagrangian dynamics’ principles,
a general representation of the equations of motion can be expressed as
[17].
d  ( KE  P )  ( KE  P ) D
    Qi  0 (1)
dt qi qi qi
where qi denotes the ith generalized coordinate and Qi is the ith generalized
force. Further, D, P and KE stand for the Rayleigh dissipation function,
potential and kinetic energies, respectively.
Based on the modal analysis technique, the elastic displacement and

rotation vectors at any point of the fin can be expanded as
N
e   en n  t  (2)
n 1
N
    n n (t ) (3)
n 1
where 𝐞̅ n and 𝜃̅ n represent the natural mode shapes of the elastic

displacements and rotations, respectively. They will satisfy the
orthogonality conditions, as well. The generalized modal coordinates are
denoted by  n (t ) . Using Eq. (1) along with Eqs. (2 and 3) for determination
of the kinetic and potential energies of the structure and applying the
virtual work principle and using the orthogonality of the natural modes,
one obtains the governing equations of the fin as
(4)
where ξ and f are vectors containing the generalized modal coordinates and
forces, respectively and Ks is the diagonal structural stiffness matrix,
defined as
Ks ii  i2 (5)
where ωi is the natural frequency of the ith mode. The generalized forces
due to the unsteady aerodynamic pressure p over the fin’s surface S are
calculated as follows
fm   p(n·em )dS (6)

S
where n denotes the outward unit normal of the undeformed surface.

Non-Linear Structural Modeling of 3D Fins

Fig. (1) shows a rigid control fin in supersonic flow with a reference
coordinate frame that is coincident with the fin’s root. The z axis coincides
with the fin’s shaft and the x axis is along the fin’s root, toward the leading
edge. The flexibility of the support is modeled by three springs
constraining the plunging (h), pitching (α) and flapping (β) motions. Also,
three viscous dampers, indicated by cα, cβ and ch support the fin in each
degrees of freedom.
Fr
ee
S tr
eam
1 K ,1 c
2 b 2 b
Tf
Kh ,ch
Ka ,ca
1K ,1c
2 b 2 b
Fig. (1). Geometry of the 3D model in supersonic flow.
It is assumed that free-play exists in all DOFs and static frictional torque is
applied on the pitch direction. The Rayleigh dissipation function (D) as
well as the potential (P) and kinetic (KE) energies of the system are
obtained as follows:
1
D  (ch h 2  c  2  c  2 ) (7)
2
1
P  ( K h h2  K   2  K  2 ) (8)
2
KE  (mh2  I xx  2  I zz 2 )   I xz  2mxz    mh x  mh z

1
(9)
2
where Kh , Ka and Kβ are the spring constants as depicted in Fig. (1). Also,
̅ , 𝐼 zz̅ and 𝐼 xz
‘m’ is the fin’s mass and 𝐼 xx ̅ denote the mass moments of inertia
about the corresponding axes in the reference coordinate. Furthermore, 𝑥̅
and 𝑧̅ show the position of the center of gravity.
Substitution of Eqs. (7-9) into Eq. (1), one obtains the following governing
equations of motion
M s x  Cx  fs  f a  0, (10)
where x = [h α β]T is the vector of dynamic states. Also, Ms and Cs are the
structural mass and damping matrices being defined as
 m  S S 
 
M s    S I zz I xz  2mxz  (11)
 S I xz  2mxz I xx 
 
ch 0 0
 
Cs   0 c 0 (12)
0 0 c 

where S  mx and S   mz are the static mass moment of the fin about
the elastic axis.
The vectors fs and fa in Eq. (10) represent the structural and aerodynamic
forces and moments acting on the fin which are defined as
fs   Fs (h) Ts ( )  Tf ( ) M s ( )
T
(13)
fa   La Ta Ma 
T (14)
In Eqs. (13-14) definitions, the aerodynamic lift, pitching and flapping

moments about the reference frame are shown by La, Ta, and Ma,
respectively. The resisting forces and moments exerted by the springs are
described as:
 K h (h  hf ) h  hf

Fs (h)   0 hf  h  hf (15)
 K (h  h ) h  hf
 h f
 K (   f )   f

Ts ( )   0  f     f (16)
 K (   )    f
  f
 K  (  f )   f

M s ( )   0 f    f (17)
 K (   )   f
  f
where hf, αf and βf are the free-plays of the corresponding degrees of

freedom. Also, Tf (α) is the constant frictional torque applied on the hinge,
which is positive for α & < 0 and negative for α & > 0. Fig. (2) shows
variation of elastic restoring torque, force and moment, schematically, as
functions of displacement and rotation components, h, α and β. This can be
clearly deduced that each of the typical curves shown in Fig. (2) cannot be
captured by just a single linear function and this is the main source of
nonlinearity in the present formulation. Actually, the amount of exerted
forces and moments are nonlinear functions of the corresponding
displacements and rotations which is the basic idea in nonlinear
oscillations.
F (h ), M (b )
T (a)
Kh Kb Ka
-ac , -bc - hc
h (t ), a(t ), b(t )
hc ac , bc
Fig. (2). Flap, plunge and pitch free-play stiffness nonlinearities.
Aerodynamic Modeling of the Flexible Fins
According to first order the piston theory, the unsteady pressure over a flat
plate is expressed in terms of its transverse deflection w as follows:
u 2 1 M2 2
p (w  w) (18)
M 2 1 u M 2 1
Where the over-dot represents the temporal derivation and the prime
symbol indicates spatial derivation, thus w is the slope of the transverse
deformation in the flow direction. Also M, ρ and u are the Mach number,
air density and airspeed over the surface, respectively.
For thin fins in zero angle of attack, weak compression and expansion
waves occur over the surface and thus the variations of flow properties (e.g.
Mach number, temperature and density) on the surface are negligible. But,
if the thickness of the fin is considerable or it is oriented in a nonzero angle
of attack, the flow properties change drastically behind the shock and
expansion waves. Having this in mind, one must obtain the flow properties
over the fin's surfaces and apply the piston theory locally on each surface.
Steady Supersonic Flow Over the Fin
The flow properties behind shock and expansion waves over the fin
surfaces can be obtained numerically using steady CFD solvers. However,
such computational tools are usually time consuming due to their
limitations for grid generation and iterative numerical schemes, especially
for 3D geometries. The shock/expansion wave theory is one of the simplest
methods which furnishes analytical relations for the solution of steady two
or three dimensional supersonic flows. Recalling that these waves are
inherently two dimensional in nature, a procedure is purposed to solve the
steady flow over the fin.
Fig. (3) shows a sketch of surfaces of a thick fin in supersonic flow. The
upper and lower surfaces of the fin are numbered individually and the
Mach number, density and temperature of the ith surface are denoted by
M i , i and, Ti respectively. The angle αi between the two following
surfaces i-1 and i is obtained as
i  cos1 ni 1·ni  (19)

where ni-1 and ni are the outward normal vectors of the corresponding
surfaces. When the angle αi is negative an expansion wave occurs at the
edge line Oi-1 Oʹi-1 and when it is positive an oblique shock wave is
expected to occur.
The upstream air velocity on the i-1th surface can be decomposed into
tangential and normal components with respect to the edge Oi-1 Oʹi-1, where
the corresponding Mach components are shown by M ti1 and M i n
1.
O'i
surface i+1 surf O'i-1

ace
i
surface i-1
t
M i-1 t
M i-1
__
M n
Oi M i-1
Oi-1
Fig. (3). Supersonic flow over the surfaces of a thick fin.
Assuming that the tangential component of the Mach number remains

unchanged across the wave, the downstream Mach number M ̅ behind an
expansion wave can be obtained by the Prandtl-Meyer theory as follows:
1 2  (  1)( M in1 )2 sin 2 ( i )

M (20)
sin( i   i ) 2 ( M in1 )2 sin 2  i    1
where αi is
 i   ( M )  ( M in1 ) (21)
and  is the Prandtl-Meyer function being defined as
  1 1   1 2
 (M )  tan ( M  1)  tan 1 M 2  1 (22)
 1  1
with γ describing the specific heat ratio. Also, ψi is the oblique wave angle
that is determined by solving the following equation.
(M in1 )2 sin 2 ( i )  1
tan i  2cot i (23)
(M in1 )2 (  cos 2 i )  2
By determining the Mach number over the ith surface, the density,
temperature and velocity direction over the surface are evaluated. Beyond
an oblique shock the air density and temperature are obtained as
i (  1) M i21 sin 2  i
 (24)
i 1 2  (  1) M i21 sin 2  i
Ti 2 M i21 sin 2  i  1   2  (  1) M i21 sin 2  i

( )( ) (25)
Ti 1  1 (  1) M i21 sin 2  i
The density and temperatures on a surface beyond an expansion wave are

also determined as
i 2  (  1) M i21  1
1
( ) (26)
i 1 2  (  1) M i2
Ti 2  (  1) M i21
 (27)
Ti 1 2  (  1) M i2
Also, the unit vector of the air velocity v i on the ith surface is defined as:
M it1t i 1  M (ni  t i 1 )
vi  (28)
( M it1 ) 2  ( M ) 2
where ti-1 is the unit vector of the edge Oi 1Oi1. The procedure continues
to any surface similarly. For this purpose, the tangential and normal
components of the Mach vector with respect to the edge Oi Oi are
obtained as:
M in  M cos( i )  M it1 sin( i ) (29)
M it  M sin(  i )  M it1 cos( i ) (30)

where i  cos1  ti 1.ti  is the angle between the edges Oi 1Oi1 and
Oi Oi. Starting from the leading edge and the free-stream surface, in which
the flow properties are known, this procedure is individually applied to the
upper and lower surfaces of the mid-plane. Finally, this procedure ends up
by determination of the flow properties at the trailing edge.
Localized Piston Theory Including the Effects of Thickness and Angle

of Attack
Assuming very small elastic deflections of the fin, the velocity and slope of
the transverse deformations on each surface of the fin is obtained as
w  (n   ).v (31)
w  n.e (32)
Substituting Eqs. (31) and (32) into Eq. (18), the localized piston theory
can be formulated over the fin’s surfaces as
u 2  1 M2  2 
p    
u M 2  1 
(n ).v n.e (33)
M2  1 
The unsteady aerodynamic pressure acting on the fin’s surfaces is obtained
in terms of the generalized modal coordinates after substitutions of the
modal expansions from Eqs. (2 and 3) into Eq. (33) which can be
summarized as:
N N
p   u 2  Cn  n (t )   u  Cen  n (t ) (34)
n 1 n 1
where the coefficients Cen and Cn are defined
M2  2 (35)
Cen  en .n
(M 2  1)1.5
(n  n ).v (36)
Cn 
M 1 2
Substituting Eq.(34) into Eq.(6), yields the following expression for the
generalized aerodynamic forces
N N
f m   K mn
a
 n   Cmn
a
n (37)
n 1 n 1
where Kamn and Camn are respectively the elements of the aerodynamic
stiffness and damping matrices which are obtained by calculating the
following integrals over the whole surfaces of the fin.
K amn   Cn (n.em )dS (38)

S
Camn   Cen (n.em )dS (39)

S
In order to obtain the aerodynamic stiffness and damping matrices, the fin’s
surface is discretized into small elements as shown in Fig. (4) and the
integrals in Eqs. (38 and 39) are evaluated numerically. To this aim, the
integrands of Eqs. (38 and 39) are approximated using their nodal values
and element shape functions within each surface element. The Gauss
quadrature technique can be used for the numerical integration.
Z X
Fig. (4). Discretization of the fin surfaces into small elements.
General Concepts in Nonlinear Analysis of Dynamical Systems
This chapter is particularly intended to review some highlights of the

complex behaviors like bifurcations, Limit Cycle Oscillations (LCOs) and
Chaos which are generally encountered in nonlinear dynamical systems as
well as some frequently implemented tools, namely, the Lyapunov
exponents and Phase plot diagrams.
Bifurcation Theory
Variation of one or more parameters in a nonlinear system can change the

number of equilibrium points both quantitatively and qualitatively. They
occur in both continuous systems (described by ODEs or PDEs), and
discrete systems (described by maps). The study of such changes
(especially qualitative and topological changes) is basically referred as
bifurcation theory. The value for which the qualitative changes happen in
motion is called the critical or bifurcation value. Detection of bifurcation
points are so vital to engineering designers because they provide severe and
rapid transitions (instabilities) as some control parameter is varied.
There are several classifications to the bifurcation types and bifurcating

behaviors. One major classification is local versus global bifurcation.
Local bifurcations are detectable by analyzing the local stability properties
or the equilibrium points (i.e. existence of positive real part for any of the
eigenvalues of the characteristic equation) or periodic orbits; whereas,
global bifurcations are not noticeable just by looking at the nature of
equilibrium points of the system while changing the bifurcating parameter.
In this chapter, the only presented bifurcations are the local ones, so we are
not going any further discussing global bifurcations. Some examples of
local bifurcations include: saddle-node (fold), transcritical, pitchfork (flip),
hopf, period-doubling and Neimark-Sacker (Secondary Hopf) bifurcations.
Dynamical systems can present an infinite sequence of period doubling
bifurcations under the variation of a single parameter culminating in a
chaotic motion. For more information regarding any of the above
mentioned bifurcations the reader is referred to any of [18] or [19].
Phase Plots
A phase plot is a geometrical representation of the evolutions of the system

states (which are generally position and velocity states) with time. Plane
comprising both system states is usually called the phase plane. These plots
are famous for the great amount of information they can convey for any
dynamical system. The information include the presence of limit cycle
oscillations as well as attractive and repulsive natures near each
equilibrium point. If the response tends asymptotically to some closed orbit
for any given initial value, then the system is said to have limit cycle
oscillations. Systems having LCOs are so practical because they guarantee
a specific response for any initial value. Such systems can also compensate
for any undesirable perturbation and noises imposed by external factors.
Other types of behavior such as periodic, quasi-periodic and non-periodic
responses can also be seen in phase plot diagrams.
Chaos
Chaotic vibrations are encountered in specific nonlinear systems.

Although, there is no unique and widely accepted definition for chaotic
dynamics, a common definition says that, any chaotic system must possess
the following conditions,
1. It must be sensitive to small changes in the initial conditions.

2. It must contain a broad frequency spectrum even for the case of
single harmonic excitations.
3. It must have dense periodic orbits.
4. The complexity of motion must increase while changing one or
more parameters, namely, there should be routes to chaos (e.g.
period doubling bifurcations).
5. Existence of fractal like behaviors in the Poincare maps.
The above points are accounted just as the main indications of chaotic
systems. However, in order to have a reliable identification of chaotic
behavior, it’s vital to introduce sophisticated diagnostic tests such as
maximum Lyapunov exponents.
Lyapunov Exponents
The Lyapunov exponents are a powerful tool for diagnosing whether or not
an oscillation is chaotic. It actually characterizes the rate of separation of
infinitesimally close trajectories. Quantitatively, two trajectories in phase
space with initial separation δZ0 diverge at a rate given by:
 Z (t ) et  Z0 (40)
where λ is the Lyapunov exponent. The rate of separation can be different
for different orientations of initial separation vector. Thus, there is a variety
of Lyapunov exponents. It is common to refer to the largest one as the

maximum Lyapunov exponent. A positive Maximum Lyapunov exponent
is usually taken as an indication that the system is chaotic (provided that
some other conditions are met, e.g., dense periodic orbits in the phase
space). The maximum Lyapunov exponent can be defined as [20].
1  Z (t )
  lim lim ln (41)
t   Z0 0 t  Z0
Numerical Results
Verification of Aerodynamic Model
In order to examine the validity of the present aerodynamic model, a

general fin with an aspect ratio of 1.5 and double-wedged cross section is
considered as shown in Fig. (4). The geometric parameters of the fin are
given in Table 1.
The lift coefficient slope C L and the pitching moment coefficient slope
C M  about the elastic axis of the fin are obtained using the present local
piston theory and compared with the 2D- and 3D-Euler CFD results.
Table 1. Geometric and mass characteristics of a 2-DOF rigid fin.
Parameter Description Value

Kh plunging spring stiffness 25 kN/m
Kα torsional spring stiffness 800 N.m
Iyy mass moment of inertia 1.5 kg.cm2
m Fin’s mass 0.15 kg

a location of the elastic axis, positive rearward 0.13
b semi-chord at the root 7.5 cm
xα = e - a static unbalance parameter -0.03
1.5 Local piston theory Local piston theory

Euler(2D) Euler(2D)
Euler(3D) Euler(3D)
0.3
1.35
CMD
CL D
0.2
1.2
1.05 0.1
0.05 0.1 0.15 0.2 0.25 0.05 0.1 0.15 0.2 0.25
t/c t/c
(a) (b)
Fig. (5). Comparison of the local piston theory and Euler method in estimating CL and CM versus the
thickness ratio at M=4 and zero angle of attack.
0.4
Local piston theory Local piston theory
1.8 Euler (2D) Euler (2D)
Euler (3D) Euler (3D)
0.3
1.6
CMD
CL D
0.2
1.4
0.1
1.2
1 0
0 5 10 15 20 0 5 10 15 20
D0(deg) D0(deg)
(a) (b)
Fig. (6). Comparison of the local piston theory and Euler method in estimations of CL and CM versus
the initial angle of attack at M=4 and t/c = 10%.
Fig. (5) shows the obtained results for different thickness to chord ratios
at Mach number 4 and zero angle of attack. The variation of the
aerodynamic lift and moment derivatives with respect to the initial angle
of attack for a fin with t/c=10% at Mach number 4 is depicted in Fig. (6).
The reasonable agreements between the present results and those of

2D/3D CFD simulations confirms that a combination of the local piston
and shock/expansion wave theories yields reliable results in predicting the
aerodynamic coefficients which are required in the aeroelastic analysis.
Effect of Thickness and Angle of Attack on the Aerodynamic

Coefficient
To clarify the effect of the fin’s thickness and angle of attack on the lift
coefficients CL , CL , CL and moment coefficients CM , CM , CM several
h h
case studies were carried out. The fin is considered to be embedded in a
flow stream with the following characteristics
M   4, p  1atm,   1.225kg/m3 , T  298K
u10-5 u10-3
0
M=12
M = 12
M=3 -1
1.5
-0.5
.
-2 .
CLD
CLD
CL h
1
-3 M=3
-1
M = 12
-4
0.5 M=3
-1.5
0 5 10 15 20 25 0 5 10 15 20 25 0 5 10 15 20 25
t/c t/c t/c
-1 -5
u10 u10 u10-4
4
0
M =12 M = 12
3 -0.5
-1
M=3
CM D
. .
CMh
CMD
-1
2
-2 M=3
M=12
-1.5
1
M=3
-2 -3
0 5 10 15 20 25 0 5 10 15 20 25 0 5 10 15 20 25
t/c t/c t/c
Fig. (7). Variation of the aerodynamic derivatives versus thickness ratio at zero angle of attack.
Figs. (7 and 8) illustrate the variations of the aerodynamic coefficients

with the thickness ratio for zero angle of attack as well as nonzero angle
of attack for t/c=10%. The obtained results show the remarkable effect of
thickness and initial angle of attack on the aerodynamic derivatives.
Results also show that the sensitivity of the aerodynamic derivatives to
the Mach number decreases at higher Mach number values.
u10-5 u10-3
0
M=12
M = 12
M=3 -1
1.5
-0.5
.
-2 .
CLD
CLD
CL h
1
-3 M=3
-1
M = 12
-4
0.5 M=3
-1.5
0 5 10 15 20 25 0 5 10 15 20 25 0 5 10 15 20 25
t/c t/c t/c
-1 -5
u10 u10 u10-4
4
0
M =12 M = 12
3 -0.5
-1
M=3
CM D
. .
CMh
CMD
-1
2
-2 M=3
M=12
-1.5
1
M=3
-2 -3
0 5 10 15 20 25 0 5 10 15 20 25 0 5 10 15 20 25
t/c t/c t/c
Fig. (8). Variation of the aerodynamic derivatives versus initial angle of attack at t/c = 10%.
Verification of Aeroelastic Model
To verify the aeroelastic model, a clamped delta wing as shown in Fig. (9)
is considered. An aeroelastic analysis is performed and the obtained
results are compared with those from the wind tunnel tests carried out at
NASA Langley wind tunnel by Tuovila and McCarty [21]. The
experiment was carried out at Mach 3 with the air density of 0.37kg/m3.
The wing is considered to be made of magnesium plate with 0.86mm
thickness, elastic modulus of 42.8GPa, density of 1766kg/m3 and Poisson
ratio of 0.35.

6"
45
Clamped
Fig. (9). Geometry of the delta wing in supersonic flow.
Using a finite element code, the first three natural modes and frequencies of
the wing are obtained and then the flutter speed and frequency of the wing
are determined using the proposed model.
300
200
Frequency (cps)
Damping
100 200
100
-100
500 1000 1500 2000 2500 500 1000 1500 2000 2500
Velocity (ft/s) Velocity (ft/s)
(a) (b)
Fig. (10). Variation of the aeroelastic damping and frequency versus the airspeed for the delta wing.
The variation of damping and frequency of the aeroelastic modes is shown

in Fig. (10). Further, Table 2 represents a good agreement between the
obtained flutter results in comparison with the experimental data which

confirms the validity of the present model.
Table 2. The obtained flutter results for the delta wing in comparison with experiment.
1 (Hz) 2 (Hz) 3 (Hz) flutter (Hz) uflutter (ft/s)
Ref. [21] 49 183 257 159 2030
Present 49 185 256.5 145 1962
Error (%) - 1.1 0.2 8.8 3.3
Effect of Geometrical Parameters on the Aeroelastic Analysis

This section is provided to evaluate the influence of the geometrical
parameters (e.g. thickness, angle of attack, sweep angle and taper ratio) on
aeroelastic behavior of the supersonic fin based on the local piston theory.
Accordingly, a structural model as displayed in Fig. (11) is considered. The
model is constrained by pitching and plunging springs. The geometrical
dimensions of the fin are as given in Table 1.

Fig. (11). Cross section and geometry of a 2-DOF rigid fin.

Fig. (12) shows the flutter and divergence speed of the fin versus the
thickness ratio. The results illustrate that there is a remarkable reduction in
the aeroelastic stability margin when the thickness increases, particularly in
the low angles of attack.
The incidence angle is another effective factor in the aeroelastic stability of

the supersonic fin. Increasing the initial angle of attack, causes a forward
shift in the location of the center of pressure and thus leads to a sharp
decrease in the divergent Mach number. Fig. (13) shows the variation of the
flutter and divergence Mach number with the initial angle of attack at
different thicknesses of the fin. The results show that increasing the initial
angle of attack causes the flutter Mach number of the fin to grow at high
thicknesses. On the other hand, according to the forward shift of the
pressure center, the divergent Mach number reduces with increasing angle
of attack. Qualitatively speaking, variation of the fin’s thickness or angle of
attack, may change the instability type from flutter to divergence.
Flutter
4 D = 0
q
Divergence
3.5
Mach Number
5q
2.5
q
10
2
0.05 0.1 0.15 0.2
t/c
Fig. (12). Variation of the flutter and divergence Mach number for a 2-DOF rigid fin versus
thickness ratio for different angles of attack.
Flutter
3.3 t/c=10% Divergence
Mach Number
15%
2.7 10%
20%
2.4
15%
2.1
20%
0 2 4 6 8 10
D(deg)
Fig. (13). Variation of flutter and divergence Mach number for a 2-DOF rigid fin with the angle of
attack at various thickness ratios.
10
Flutter
9 Divergence
8
Mach Number
t/c=10%
7
15%
6
20%
5
2
-5 0 5 10 15
Sweep angle (deg)
Fig. (14). Variation of flutter and divergence Mach number for a 2-DOF rigid fin versus sweep
angle.
The sweep angle is another fundamental parameter in the design of

supersonic fins. Increasing the sweep angle results in weaker
shock/expansion waves on the fin and thus increases the aeroelastic

stability. Fig. (14) illustrates the effect of sweep angle on the flutter and
divergence Mach numbers of the fin for different thickness ratios. These
results identify the increase of the aeroelastic stability with the growth of
the sweep angle. Further it is observed that for high sweep angles, the
instability type changes from flutter to divergence.
Flutter
Divergence
t/c=10%
5
Mach Number
15%
4 20%
10%
15%
3
20%
2
0 0.2 0.4 0.6 0.8 1
Taper ratio
Fig. (15). Variations of the flutter and divergence Mach number for a 2-DOF rigid fin versus taper
ratio.
Taper ratio is the last geometrical parameter studied in this section.

Actually taper ratio reduces the undesired effects of the fin’s tip and
improves the aerodynamic efficiency of the supersonic fins. Variations of
the flutter and divergence speeds of the considered fin versus taper ratio is
depicted in Fig. (15) for several thickness ratios. Taper ratio is considered
to be applied on both the chord and thickness of the fin. The obtained
results implies that the aeroelastic instability margins of the fin reduces as
the taper ratio increases.
Aeroelastic Analysis of Elastic Fins
In order to demonstrate the application of the proposed model in the

aeroelastic analysis of general supersonic thick fins and also to analyze the
effect elasticity on the stability, a flexible fin with t/c=10% with the
geometrical dimensions as shown in Fig. (16) is considered.
7.50mm
150mm
Rigid base
10
0 5
10
45mm
85 15mm
125mm
150mm
Fig. (16). Geometry and dimensions of the elastic supersonic fin.
Table 3. Material properties of the elastic thick fin.
Carbon-epoxy Foam
E1(GPa) 140 2.0
E2(GPa) 10 -
V12 0.3 0.3
G12(GPa) 5.2 -
G13(GPa) 5.2 -
G23(GPa) 3.4 -
1550 100
 (kg / m3 )
The fin is clamped to a rigid base and is made up of a composite shell with
uniform thickness of 0.5mm, which forms the outer aerodynamic surfaces
and is filled by foam. The fin surface is considered to be a four layered
Carbon/epoxy composite with symmetric cross-ply orientation. Table 3
shows the materials properties of the fin.
The aeroelastic model is developed based on the first six natural vibration
modes of the fin which are extracted using a finite element model and are
shown in Fig. (17).
(a) Z1 = 524 Hz (b) Z2 = 1053 Hz (c) Z3 = 2050 Hz
(d) Z4 = 2492 Hz (e) Z5 = 3253 Hz (f) Z6 = 4176 Hz
Fig. (17). The first six natural mode shapes of the elastic fin and their corresponding natural
frequencies.
Fig. (18) shows the variations of damping and frequency of the aeroelastic
modes with Mach number for flight at sea level. The obtained results show
that the flutter Mach number for the considered configuration is 5.97.
u10 2 u10 4
1 1.6
0 1.4
Frequency
Damping
-1 1.2
-2 1
-3 0.8
-4 0.6
3 4 5 6 7 8 3 4 5 6 7 8
Mach Number Mach Number
(a) (b)
Fig. (18). Plots of aeroelastic damping and frequency for the elastic fin at zero angle of attack.
Finally, for the last linear investigation, variation of the flutter Mach
number versus the initial angle of attack is shown in Fig. (19) which proves
that the stability of the fin decreases at higher angles of attack.
6
5.9
Mach Number
5.8
5.7
5.6
5.5
5.4
0 5 10 15 20 25
D(deg)
Fig. (19). Variation of the flutter Mach number of the elastic fin versus the initial angle of attack.
Non-linear Aeroelastic Analysis
This section aims at studying the effects of fin’s geometrical parameter on

non-linear response of thick supersonic fins. Hence, a fin with a similar
configuration as the elastic fin, shown in Fig. (16) is chosen and free-plays
are assumed to exist in all DOFs.
Effects of Airspeed Mach Number

To perform a nonlinear analysis, the values given in Table 4 are adopted.
For the comparison purposes, both linear and nonlinear analysis are
conducted for a 3D fin. To account for the linear aeroelastic behavior, an
eigenvalue analysis at the sea level conditions is performed and results are
shown in Fig. (20). These results show that the flutter takes place at
M   6.35.
500
400
-5
Frequency
Damping
-10
300
-15
Mode 1
Mode 2
200
-20 Mode 3
3 4.5 6 7.5 9 3 4.5 6 7.5 9

Fig. (20). Variations of the aeroelastic damping and frequency versus the Mach number for a 3D fin.
Table 4. Geometric and mass characteristics of the 3-DOF rigid fin.
Kh Plunging spring stiffness 90 kN/m
Kα Pitching spring stiffness 0.45 kN/m
Kβ Flapping spring stiffness 3 kN/m
hf Plunging free-play magnitude 0.01 mm
αf , βf Pitching and flapping free-play magnitudes 0.5deg
Plunging amplitude and time derivatives 0.06 mm, 0m/s, respectively

h, h
t 0 s
Pitching angle and time derivatives 6 105 , 0 deg /s, respectively

 ,  t 0 s
flapping angle and time derivatives 6 105 , 0 deg /s, respectively

,
t 0 s
Ixx Mass moment of inertia 120kg cm2
Izz Mass moment of inertia 120 kg cm2
Ixz Mass moment of inertia -12 kg cm2
m Fin's mass 1.8 kg
a Location of the elastic axis, positive rearward 0.13
b Semi-chord at the root 7.5 cm
xα = ex - ax Static unbalance parameter 0.0035
In the next step, results from the nonlinear dynamics of the same fin is of
interest. As the plunging, flapping and pitching responses are quite similar
in behavior, the graphical illustrations are restricted only to the pitching
time-histories. Fig. (21a) shows the dynamic behavior of the fin as a
function of Mach number through a bifurcation diagram. In the bifurcation

diagram, the vertical axis depicts the local maxima and local minima of the
torsional angle time series after converging to steady state response, and
the horizontal axis is the Mach number. This figure is divided into four
regions. In the first portion, for M   5.1 , since all local maxima and
minima of time series overlap each other, it is concluded that the solution
converges to an equilibrium condition. Bifurcation occurs at M   5.1 .
Considering this as the flutter onset and comparing with the estimated
flutter speed from the linear case, it’s concluded that the flutter Mach
number is decreased by 1.25 units. In the second and third regions, for
5.1  M   7.2 , LCOs occur. In these regions, LCOs' amplitude increases
as Mach number increases. Also, between these two regions bifurcation
happens again.
1 2 3 4
1
Pitch (Min/Max) Angle (deg)
0.8
0.6
0.4
3 4 5 6 7
3
Lyapunov Exponent
-3
-6
3 4 5 6 7
0.03
0.02
T(s)
T=0 T
0.01
0
3 4 5 6 7
Mach Number
Fig. (21). Variation of (a) pitch (min/max) angle of the fin, (b) Lyapunov exponent and (c) period
of the pitch time series versus the Mach number.
Finally, it seems that chaotic response starts at region four where M   7.2 .
The Lyapunov exponent is presented in Fig. (21b) at various Mach
numbers. The magnitude of the exponent changes to positive for
M   7.2 which confirms chaotic response in the fourth region. Fig. (21c)
depicts period of the oscillation at different Mach numbers. The periods of
oscillation are 0, 0.013, 0.027 and infinite for the consequent regions,
respectively. It means that at M   6.3 , a period doubling phenomenon
occurs. To have a better investigation, time histories and the phase-plane
diagrams for Mach numbers 3, 5.5, 6.5 and 7.5 are plotted in Figs. (22 to
25). These diagrams are obtained after the transient response, so the
responses converge to steady state behavior.
400 60
200
200 30
100
(deg/s)
h (mm/s)
(deg/s)
0 0 0
-100 -30
-200
-200
-60
-400
-4 -3 -2 -1 0 0 0.5 1 1.5 2 2.5 -0.6 -0.4 -0.2 0
h (mm) (deg) (deg)
-0.4
-2.1
1.4
-2.4 -0.5
1.6
(deg)
h (mm)
(deg)
-2.7 -0.6
1.8
-3 2 -0.7
0 5 10 15 0 5 10 15 0 5 10 15
Time (s) Time (s) Time (s)
Fig. (22). Phase plane diagram and time histories of the fin in pitch, flap and plunge direction at
Mach 3.
300 200
60
150 100
30
(deg/s)
h (mm/s)
(deg/s)
0 0 0
-30
-150 -100
-60
-300 -200
-4.5 -4.2 -3.9 -3.6 2.4 2.64 2.88 3.12 -0.72 -0.66 -0.6 -0.54 -0.48
h (mm) (deg) (deg)
2.4
-0.5
-3.6
2.6
-3.9
h (mm)
(deg)
(deg)
-0.6
2.8
-4.2
3 -0.7
-4.5
29.7 29.8 29.9 30 29.7 29.8 29.9 30 29.7 29.8 29.9 30

Mach 5.5.
800
100
400
400
50
200
(deg/s)
h (mm/s)
(deg/s)
0 0
0
-200 -50
-400
-400
-100
-800
-6.6 -5.5 -4.4 -3.3 2.5 3 3.5 4 4.5 -0.84 -0.72 -0.6 -0.48
h (mm) (deg) (deg)
2.1
-3
-0.84
2.8
-4 -0.72
(deg)
h (mm)
(deg)
-5 3.5 -0.6
-6 -0.48
4.2
29.7 29.8 29.9 30 29.7 29.8 29.9 30 29.7 29.8 29.9 30

Mach 6.5.
-160
2000 1200
-80
1000
600
(deg/s)
h (mm/s)
(deg/s)
0
0
0
-1000
-600 80
-2000
-1200 160
-12 -8 -4 0 0 2.5 5 7.5 -1 -0.8 -0.6 -0.4 -0.2

h (mm) (deg) (deg)
-1
7.5
0
-0.75
5
-4
h (mm)
(deg)
(deg)
-0.5
2.5
-8
-0.25
0
-12
29 29.25 29.5 29.75 30 29 29.25 29.5 29.75 30 29 29.25 29.5 29.75 30

Mach 7.5.
Effects of Thickness
Thickness is also another effective parameter which can lead to significant

changes in the aeroelastic behavior of the fin. As shown in the linear case,
aerodynamic loads are totally deviated as the thickness ratio changes. In
the nonlinear case, this parameter can also lead to both quantitatively and
qualitatively different behaviors compared to the linear analysis. The
following conditions for the free stream flow is considered.
The dynamic response of the fin at different thickness to chord ratios (t/c) is
depicted in Fig. (26a). According to this figure, which is drawn based on
local max/min values of the torsional angle time series, it is observed that
for t/c up to 10.2% the solution converges to an equilibrium condition and
at t/c = 10.2% bifurcation occurs. Stable LCOs was experienced at
thickness ratios ranging from 10.2 to 16.7, while chaotic response happens
at 16.7 <t/c<19. Based on this diagram, bifurcation occurs again between
the second and the third regions. Fig. (26b) shows the Lyapunov exponent
versus the thickness ratio which emphasizes chaotic response at
16.7<t/c<19. Period variation versus the thickness ratio is illustrated in Fig.

(26c). Magnitude of the period in the first and fourth regions are zero and
infinite, respectively. At the beginning of the second region, it jumps to
0.013 and it doubles (period doubling) in the third region. To have a better
investigation, time histories and the phase-plane plots for various
thicknesses are shown in Figs. (27 to 30). These diagrams are plotted after
converging to the steady state behavior.
M   5, p  1atm,   1.225kg/m3 , T  298K

0.92
1 2 3 4
0.69
0.46
0.23
6 9 12 15 18
Lyapunov Exponent
-2
-4
-6
6 9 12 15 18
0.03
0.02
T(s)
T=0 T
0.01
0
6 9 12 15 18
t/c(%)
Fig. (26). Variations of (a) pitch (min/max) angle of the fin, (b) Lyapunov exponent and (c) period of
the pitch time series versus the thickness at Mach 5.
80
400
400
40
200
(deg/s)
h (mm/s)
(deg/s)
0
0 0
-400
-200 -40
-800 -400
-80
0 2 4 6 -4 -3 -2 -1 0 0 0.2 0.4 0.6
h (mm) (deg) (deg)
2.4 -3
0.72
-2.7
3
0.64
(deg)
h (mm)
(deg)
-2.4
3.6 0.56
-2.1
0.48
4.2
0 10 20 30 0 10 20 30 0 10 20 30
t/c=8%.
400
4
80
200 2
40
h (mm/s)
(deg/s)
(deg/s)
0 0
0
-200 -2
-40
-400 -4
-80
-5 -4 -3 -2 -1 1.75 2 2.25 2.5 -0.59 -0.58 -0.57 -0.56 -0.55
h (mm) (deg) (deg)
-0.54
1.6
-1.5
-0.56
2
(deg)
h (mm)
(deg)
-3
2.4 -0.58
-4.5
29.7 29.8 29.9 30 29.7 29.8 29.9 30 29.7 29.8 29.9 30

Fig. (28). Phase plane diagram and time histories of the fin in pitch, flap and plunge direction at t/c
=14%.
400 6
60
200 3
30
h (mm/s)
(deg/s)
(deg/s)
0 0 0
-200 -30 -3
-400 -60 -6
-4 -3 -2 -1 1.5 1.75 2 2.25 -0.57 -0.555 -0.54 -0.525

h (mm) (deg) (deg)
-0.52
-1
1.6
-2 -0.54
1.8
(deg)
h (mm)
(deg)
-3
-0.56
2
-4
2.2
-0.58
29.7 29.8 29.9 30 29.7 29.8 29.9 30 29.7 29.8 29.9 30

=16%.
1000 -100
600
500 -50
300
(deg/s)
h (mm/s)
(deg/s)
0 0
0
-500 50
-300
-1000 100
-600
-7.5 -6 -4.5 -3 -1.5 1 2 3 4 -0.75 -0.6 -0.45 -0.3

h (mm) (deg) (deg)
-1.7
4 -0.75
-3.4
3
(deg)
h (mm)
(deg)
-0.6
-5.1
2 -0.45
-6.8
1 -0.3
29 29.25 29.5 29.75 30 29 29.25 29.5 29.75 30 29 29.25 29.5 29.75 30

=18%.
1 2

0.6
0 3 6 9 12 15
0.03
T(s)
0.02
T=0
0.01
0
3 6 9 12 15
(deg)
Fig. (31). Variations of (a) pitch (min/max) angle of the fin and (b) period of the pitch time series
versus α0 at Mach 5.
Effect of Initial Angle of Attack
The influence of initial angle of attack on the dynamic response of fins with
t/c = 10% loaded at a flow stream just as the previous case is depicted in
Fig. (31a). As can be seen, the fin remains stable in the first region. At α0 =
10.2° bifurcation occurs and after that stable LCOs are encountered. Fig.
(31b) shows the period of the oscillations at different angles of attack. It

remains zero in the first region and jumps to 0.036s at the second part.
Also, Figs. (32 to 34) depict time histories and the phase-plane diagrams at
 0  7, 10,15 .
80
400
400
40
200
(deg/s)
h (mm/s)
(deg/s)
0
0 0
-400
-200 -40
-800 -400
-80
0 2 4 6 -4 -3 -2 -1 0 0 0.2 0.4 0.6
h (mm) (deg) (deg)
2.4 -3
0.72
-2.7
3
0.64
(deg)
h (mm)
(deg)
-2.4
3.6 0.56
-2.1
0.48
4.2
0 10 20 30 0 10 20 30 0 10 20 30
Fig. (32). Phase plane diagram and time histories of the fin in pitch, flap and plunge direction at α0
= 7°.
400
4
80
200 2
40
h (mm/s)
(deg/s)
(deg/s)
0 0
0
-200 -2
-40
-400 -4
-80
-5 -4 -3 -2 -1 1.75 2 2.25 2.5 -0.59 -0.58 -0.57 -0.56 -0.55
h (mm) (deg) (deg)
-0.54
1.6
-1.5
-0.56
2
(deg)
h (mm)
(deg)
-3
2.4 -0.58
-4.5
29.7 29.8 29.9 30 29.7 29.8 29.9 30 29.7 29.8 29.9 30

= 12°.
400 6
60
200 3
30
h (mm/s)
(deg/s)
(deg/s)
0 0 0
-200 -30 -3
-400 -60 -6
-4 -3 -2 -1 1.5 1.75 2 2.25 -0.57 -0.555 -0.54 -0.525

h (mm) (deg) (deg)
-0.52
-1
1.6
-2 -0.54
1.8
(deg)
h (mm)
(deg)
-3
-0.56
2
-4
2.2
-0.58
29.7 29.8 29.9 30 29.7 29.8 29.9 30 29.7 29.8 29.9 30

= 15°.
Effect of Hinge Position
Hinge position is another indispensable parameter which affects the

nonlinear dynamic response of the fin. Variation of this parameter changes
thoroughly the moment distribution on the fin which in return changes the
instability boundaries as well as the dynamic behavior of the fin. Using an
eigenvalue analysis, Figs. (35 and 36) are plotted to show changes in the
damping and frequency at x/c = 0.55 and x/c = 0.57, where x/c is the hinge
position to chord ratio. Comparison of these figures depicts that receding
the hinge from the leading edge changes the flutter mode from the plunging
to pitching. To dig further into the problem, effect of x/c on the dynamic
behavior of the fin is investigated via Fig. (37a). Based on the dynamic
response, this figure is divided into the three regions. In the first region,
solution converges to an equilibrium condition but in the second one LCOs
happen. Based on this diagram bifurcation occurs at x/c = 0.57.
500
0.5
400
0
Frequency
Damping
-0.5 300
-1 Mode 1
Mode 2 200
Mode 3
3 4.5 6 7.5 9 3 4.5 6 7.5 9

Fig. (35). Variation of the aeroelastic damping and frequency versus Mach number at x/c=0.55.
0.5
500
0
400
Frequency
Damping
-0.5 300
Mode 1
Mode 2 200
-1 Mode 3
3 4.5 6 7.5 9 3 4.5 6 7.5 9

Fig. (36). Variation of the aeroelastic damping and frequency versus Mach number at x/c=0.57.
of the pitch time series versus the hinge position at Mach 5.
800 400 80
200 40
400
h (mm/s)
(deg/s)
(deg/s)
0 0
0
-200 -40
-400
-400 -80
-6 -4 -2 0 0 1 2 3 4 -0.6 -0.4 -0.2 0

h (mm) (deg) (deg)
1.8
-0.4
-2.8
2.25 -0.5
h (mm)
(deg)
(deg)
-3.5
-0.6
2.7
-4.2
-0.7
3.15
0 5 10 15 0 5 10 15 0 5 10 15
x/c=0.55.
200 80
300
100 40
150
(deg/s)
h (mm/s)
(deg/s)
0 0 0
-150 -100 -40
-300 -200 -80

3 3.5 4 4.5 -3 -2.7 -2.4 -2.1 0.5 0.6 0.7 0.8
h (mm) (deg) (deg)
-3
3
0.7
-2.7
3.5
h (mm)
(deg)
(deg)
0.6
-2.4
4
0.5
-2.1
29.7 29.8 29.9 30 29.7 29.8 29.9 30 29.7 29.8 29.9 30

x/c=0.58.
800 100
600
400 50
300
(deg/s)
(deg/s)
h (mm/s)
0
0
0
-50
-400
-300
-100
-6 -4.5 -3 -1.5 1 2 3 4 -0.9 -0.75 -0.6 -0.45
h (mm) (deg) (deg)
0
-6.4 -0.4
1.5
-4.8
-0.6
h (mm)
(deg)
(deg)
3
-3.2
-0.8
4.5
-1.6
29.7 29.8 29.9 30 29.7 29.8 29.9 30 29.7 29.8 29.9 30

x/c=0.628.
2000 1400
100
1000 700
h (mm/s)
(deg/s)
(deg/s)
0 0 0
-1000 -700
-100
-2000 -1400
-10 -5 0 5 10 -4 0 4 8 -1 -0.5 0 0.5 1
h (mm) (deg) (deg)
-10
-6
0.6
-5 -3
h (mm)
(deg)
(deg)
0 0
0
3
5 -0.6
29 29.25 29.5 29.75 30 29 29.25 29.5 29.75 30 29 29.25 29.5 29.75 30

x/c=0.65.
Also, it seems at this point that the unstable mode shifts from plunging to
pitching Figs. (35 and 36). In the third region, 0.58  x / c  0.67 different
types of behavior (periodic and chaotic) occur. Fig. (37b) shows variations
of the Lyapunov exponent versus x/c. Based on this figure, periodic
response is considered at 0.627  x / c  0.63 (position A). Also, a chaotic
behavior is predicted for the fin in the remaining positions of the third
region. Similar to other investigated parameters, in Fig. (37c) the period of
oscillation as a function of the hinge position is shown. Finally, time
histories and the phase-plane diagrams at different hinge positions are
illustrated in Figs. (38-41).
Effect of Sweep Angle
Sweep angle is one of the crucial parameters in the design of supersonic

fins. Increasing the sweep angle results in weaker shock/expansion waves
on the fin and thus increases the aeroelastic stability. Fig. (42) shows
positive and negative direction of sweep angles. Dynamic behavior of the
fin is traced in Fig. (43a) as a function of sweep angle. Based on the
recorded dynamic response, the presented figure is divided into four
regions. In the first one, the dynamic behavior converges into a stable state.
Limit cycle oscillations are observed in the next two regions and finally
chaotic motion happens in the last part. Further, it can be seen that
bifurcation occurs twice at sweep angles of 1.7° and 0.9°. According to Fig.
(43b), the Lyapunov exponent turns positive in region 4 which confirms
chaotic response in that region. Fig. (43c) depicts period of oscillation at
different sweep angles. Also, time histories and the phase-plane diagrams
for sweep angles -10°, 0°, 3° and 15° are shown through Figs. (44-47).
FreeStream
Fig. (42). Positive/negative direction of sweep angle.
of the pitch time series versus the sweep angle at Mach 5.
60
300
270
150 30
(deg/s)
h (mm/s)
(deg/s)
0
0 0
-270
-150
-30
-540
-300
0 1.5 3 4.5 -3 -2 -1 0 0 0.2 0.4 0.6
h (mm) (deg) (deg)
3.9 0.48
-2.6
3.6 0.51
-2.4
h (mm)
(deg)
(deg)
0.54
3.3
-2.2
0.57
3
-2
0.6
0 5 10 15 0 5 10 15 0 5 10 15
sweep angle -10°.
60
200
100
100 30
50
(deg/s)
h (mm/s)
(deg/s)
0 0 0
-50
-100 -30
-100
-200 -60
-4 -3.8 -3.6 -3.4 2.34 2.47 2.6 2.73 2.86 -0.7 -0.65 -0.6 -0.55 -0.5
h (mm) (deg) (deg)
-0.48
-4
2.4 -0.54
-3.8
h (mm)
(deg)
(deg)
2.56 -0.6
-3.6
-0.66
2.72
-3.4
-0.72
29.7 29.8 29.9 30 29.7 29.8 29.9 30 29.7 29.8 29.9 30

sweep angle 0°.
500 300
80
250 150
40
h (mm/s)
(deg/s)
(deg/s)
0 0 0
-40
-150
-250
-80
-300
-500
3 3.5 4 4.5 5 -3.2 -2.8 -2.4 -2 0.4 0.5 0.6 0.7 0.8
h (mm) (deg) (deg)
-3.3 0.8
3
-3
0.7
3.5
(deg)
h (mm)
(deg)
-2.7
0.6
4
-2.4
0.5
4.5
-2.1
0.4
29.7 29.8 29.9 30 29.7 29.8 29.9 30 29.7 29.8 29.9 30
sweep angle 3°.
1600
2000
200
1000 800 100

(deg/s)
h (mm/s)
(deg/s)
0 0
0
-100
-1000
-800
-200
-2000
-10 -5 0 5 10 -4 0 4 8 -1 -0.5 0 0.5 1

h (mm) (deg) (deg)
-12
-5 1
-6 0.5
h (mm)
(deg)
(deg)
0
0
0
-0.5
5
6
-1
29 29.25 29.5 29.75 30 29 29.25 29.5 29.75 30 29 29.25 29.5 29.75 30

sweep angle 15°.
Effect of Hinge Frictional Torque

A 3D fin with the structural characteristics given in Table 5 is considered.
Using this configuration, the sea level aeroelastic behavior of the fin is
investigated by an eigenvalue analysis. The U-g diagrams obtained from
the analysis are presented in Fig. (48). These results predict that the flutter
takes place at M   8.94 .
Table 5. Geometric and mass characteristics of a 3-DOF rigid fin.

Kh Plunging spring stiffness 300 N/m
Kα Pitching spring stiffness 800 N/m
Kβ Flapping spring stiffness 1000 N/m
ζ Damping ratio 8 × 10-4
hf Plunging free-play magnitude 1 mm
αf , βf Pitching and flapping free-play 0.5deg

magnitudes
Plunging amplitude and time 102 mm, 102 mm / s, respectively
h, h
t 0 s
derivatives
Pitching angle and time derivatives 6 104 , 6 104 deg /s, respectively
 ,  t 0 s
Flapping angle and time derivatives 6 104 , 6 104 deg /s, respectively
, t 0 s
̅
𝐼 xx Mass moment of inertia 400kg cm2
𝐼 zz̅ Mass moment of inertia 100 kg cm2
̅
𝐼 xz Mass moment of inertia -20 kg cm2
M Fin's mass 1.5kg
A Location of the elastic axis, positive 0.08
rearward
B Semi-chord at the root 7.5 cm
x α = e x - αx Static unbalance parameter 0.0027
10
0 200 Zh
Frequency(Hz)
ZE
Damping ZD
-10
100
-20
-30 0
6 7 8 9 6 7 8 9
Fig. (48). Variation of the aeroelastic damping and frequency versus the Mach number for the
linear model.
To study the effects of frictional torque on non-linear response of thick

supersonic fins, it is assumed that the fin’s structure involves free-plays at
all degree of freedom and frictional torque is applied on pitch DOF.
Fig. (49) illustrates the dynamic behavior of the fin under consideration as
a function of Mach number through bifurcation diagrams. As shown, the
vertical axis depicts the local maxima and local minima of the
corresponding motion after converging to steady state response, and the
horizontal axis shows the Mach number. According to the fin's dynamic
response, this figure is divided to four regions. It can be seen that for the
Mach numbers of less than 4.2, since all local maxima and minima of the
time series are overlapped, the solution converges to an equilibrium
condition. It is obvious that increasing the Mach number causes LCOs
begin at Mach number 4.1. Further augmentations of the Mach number
leads to larger vibration amplitudes. Finally, it can be seen that the
amplitude of LCOs tends to infinity for M∞ ≥ 8.63. For a more
comprehensive investigation, time histories and phase-plane diagrams for
Mach values of 2.5, 4.5, 7.65 and 8 are plotted in all DOFs in Figs. (50 to
53). These diagrams are obtained after in the steady state response
conditions.
40
Flap (Max/Min) Angle (deg)

20
-20
-40
Plung (Max/Min) displacement (m)
0.4
0.2
-0.2
-0.4
6
Pitch (Max/Min) Angle (deg)
-2
-4 (I) (II) (III) (IV)
-6
2 3 4 5 6 7 8
Mach Number
Fig. (49). Variation of 1: pitch (min/max) angle, 2: plunge(min/max) displacement, and 3:

flap(min/max) angle of the fin versus the Mach number.
Fig. (50). Phase plane diagram and time histories of the fin with Tf ( ̇ ) = 0.5 N.mm in pitch
direction at Mach 2.5.
direction at Mach 8.
CONFLICT OF INTEREST
The authors confirm that they have no conflict of interest to declare for
this publication.
ACKNOWLEDGEMENTS
Declared none.
NOTES
1
Part of this chapter has been previously published in Journal of
Fluids and Structures. Volume 40, July 2013, Pages 163–184.
doi:10.1016/ j.jfluidstructs.2013.03.019.
REFERENCES
[1] D. D. Liu, Z. X. Yao, D. Sarhaddi and F. R. Chavez, "From piston theory to a
unified hypersonic/supersonic lifting surface method.," Journal of Aircraft, vol.
34, pp. 304-312, 1997.
[http://dx.doi.org/10.2514/2.2199]
[2] W. Zhang, Z. Ye and C. Zhang, "Analysis of supersonic aeroelastic problem
based on local piston theory method," 45th AIAA Aerospace Sciences Meeting
and Exhibit, Vols. AIAA Paper 2007--1072., 2007.
[http://dx.doi.org/10.2514/6.2007-1072]
[3] W. Zhang, Z. Ye, C. Zhang and F. Liu, "Supersonic flutter analysis based on a
local piston theory," AIAA Journal, vol. 47, pp. 2321-2328, 2009.
[http://dx.doi.org/10.2514/1.37750]
[4] R. D. Firouz-Abadi and S. M. Alavi, "Effect of thickness and angle of attack on
the aeroelastic stability of supersonic fins," The Aeronautical Journal , vol. 116,
pp. 777-792, 2012.
[http://dx.doi.org/10.1017/S0001924000007272]
[5] L. Librescu, P. Marzocca and W. A. Silva, "Linear/non-linear supersonic panel
flutter in a high-temperature field.," Journal of Aircraft, vol. 41, pp. 918-924,
2004.
[http://dx.doi.org/10.2514/1.679]
[6] L. Tang, R. E. Bartels, P. -C. Chen and D. D. Liu, "Numerical investigation of
transonic limit cycle oscillations of a two-dimensional supercritical wing,"

Journal of Fluids and Structures, vol. 17, pp. 29-41, 2003.
[http://dx.doi.org/10.1016/S0889-9746(02)00114-7]
[7] Y. J. Seo, S. J. Lee, J. S. Bae and I. Lee, "Effects of multiple structural
nonlinearities on limit cycle oscillation of missile control fin," Journal of Fluids
and Structures, vol. 27, pp. 623-635, 2011.
[http://dx.doi.org/10.1016/j.jfluidstructs.2011.02.009]
[8] S. C. McIntosh Jr., R. E. Reed Jr. and W. P. Rodden, "Experimental and
theoretical study of nonlinear flutter," Journal of Aircraft, vol. 18, pp.
1057-1063, 1981.
[http://dx.doi.org/10.2514/3.57600]
[9] R. Vasconcellos, A. Abdelkefi, F. D. Marques and M. R. Hajj, "Representation
and analysis of control surface freeplay nonlinearity," Journal of Fluids and
Structures, vol. 31, pp. 79-91, 2012.
[10] W. H. Shin, S. J. Lee, I. Lee and J. S. Bae, "Effects of actuator nonlinearity on
aeroelastic characteristics of a control fin," Journal of Fluids and Structures, vol.
23, pp. 1093-1105, 2007.
[11] D. Li, S. Guo and J. Xiang, "Study of the conditions that cause chaotic motion in
a two-dimensional airfoil with structural nonlinearities in subsonic flow,"
Journal of Fluids and Structures, vol. 33, pp. 109-126, 2012.
[12] D. S. Woolston, H. L. Runyan and R. E. Andrews, "An investigation of effects
of certain types of structural nonlinearities on wing and control surface flutter,"
Journal of Aeronautic Sciences, vol. 24, pp. 57-63, 1957.
[http://dx.doi.org/10.2514/8.3764]
[13] Z. C. Yang and L. C. Zhao, "Analysis of limit cycle flutter of a wing section in
incompressible flow," Journal of Sound and Vibration, vol. 123, pp. 1-13, 1988.
[http://dx.doi.org/10.1016/S0022-460X(88)80073-7]
[14] A. J. Hauenstein, J. A. Zara, W. Eversman and I. K. Qumei, "Chaotic and

nonlinear dynamic response of aerosurfaces with structural nonlinearities,"
AIAA Paper 1992-2547, pp. 2367-2375, 1992.
[http://dx.doi.org/10.2514/6.1992-2547]
[15] L. K. Abbas, Q. Chen, K. O'Donnell, D. Valentine and P. Marzocca, "Numerical
studies of an on-linear aeroelastic system with plunging and pitching," Journal
of Aerospace Science and Technology, vol. 11, pp. 405-418, 2007.
[http://dx.doi.org/10.1016/j.ast.2007.02.007]
[16] D. Tang and E. H. Dowell, "Aeroelastic airfoil with free play at angle of attack
with gust excitation," AIAA Journal, vol. 48, pp. 427-442, 2010.
[http://dx.doi.org/10.2514/1.44538]
[17] E. H. Dowell, A Modern Course in Aeroelasticity Fifth Revised and Enlarged
Edition, Springer, 2014.
[18] S. H. Strogatz, Nonlinear Dynamics and Chaos, with applications to physics,
biology, chemistry and engineering. Perseus Books, 1994.
[19] D. J. a. P. Smith, Nonlinear Ordinary Differential Equations, and introduction
for scientists and engineers, Oxford University Press, 2007.
[20] F. C. A. V. Massimo Cencini, Chaos From Simple Models to Complex Systems,
vol. 17, 2010.
[21] W. a. M. J. Tuovila, "Experimental Flutter Results for Cantilever Wing Models
at Mach Numbers up to 3.0," 1955.
CHAPTER 2
An Analytical and Experimental Investigation

into Vibratory Force for Aircraft Wings
Xueguang Bi and Yucheng Liu*
Department of Mechanical Engineering, Mississippi State University, MS

39762, USA
Abstract: This paper focus on the assessment of aerodynamic forces applied on an

aircraft. First, numerical analysis is performed on a simplified linear
two-degree-of-freedom (2DOF) airfoil system model. Methods of aerodynamic
force determination are established at different airspeed conditions and the
calculated dynamic forces correlated well to the actual forces. Next, a finite
element model of the airfoil is developed to represent its structural dynamics, and
the established force determination methods are applied to determine the
aerodynamic forces acting on such model. After that, a physical downsizing model
of the airfoil is manufactured and its natural frequencies, damping ratios, and mode
shapes are calculated and compared to those obtained from the analytical models.
Vibration experiments are also conducted to measure the physical model’s
structural response in the wind tunnel, and its aerodynamic forces are calculated
based on the data recorded by the attached accelerometers.
Keywords: Aircraft wing model, Experimental measurement, Force

determination methods, Frequency response function, Multiple-degree-of-
freedom system, Vibratory force and moment.
*Corresponding author Yucheng Liu: Department of Mechanical Engineering, Mississippi State University,
MS, 39762, USA; Tel/Fax: (001)662-325-1536/662-325-7223; E-mail: yliu@me.msstate.edu.
Yucheng Liu (Ed.)

66 Frontiers in Aerospace Science, Vol. 1 Bi and Liu
INTRODUCTION
During flight, aircraft wings always present varying structural responses

because subject to varying aerodynamic loadings. The applied
aerodynamic loadings and the aircraft wing’s structural responses are
coupled together to create a complicated vibration effect known as
structural coupling dynamics. This effect can reduce the service life of the
aircraft significantly. Sometimes, sustained vibration at natural
frequencies of airfoils may lead to catastrophic structural failure.
Structural coupling dynamics, first developed in the early 1920’s, has

become an important concern in design of aircrafts. To date, a number of
researchers have dedicated tremendous effort and time to lighten the
materials of aircrafts so as to obtain an effective loading capacity.
However, as the aircraft speed keeps on increasing, the structural
flexibility of the aircrafts receives more design concerns because it may
cause serious flutter in the airfoil and other aircraft components.
Considering a typical cross-sectional plane of an airfoil, the gravity center

G is located at the 42% ~ 45% of the chord line, and the elasticity center
is located at the 38% ~ 40% of the chord line. Therefore, the gravity
center is not coincident with the elastic center, which results in a coupling
effect between the translational and rotational degree-of-freedom of the
wing. This coupling effect and the structure – air coupling effect make it
difficult to solve the aerodynamic force of aircraft wings.
The aerodynamic forces, which are extremely difficult to obtain from a

flying aircraft, can usually be measured from experimental simulations.
As a direct, effective, and easy-to-use approach in the aerodynamic study,
experimental methods have been extensively employed to simulate the
aircraft’s flying condition, so as to measure the aerodynamic forces and
the structural responses.
In this chapter, methods are developed for determining the aerodynamic

forces acting on the aircraft wings during flight. The methods are initially
Vibratory Force for Aircraft Wings Frontiers in Aerospace Science, Vol. 1 67
developed for a 2DOF linear system, and then extended for a MDOF
system developed as a finite element model. Wind tunnel test is then
conducted to obtain the physical dynamic parameters. The finite element
model and the force determination method are then validated by
comparing the analytical results with the experimental results.
Fig. (1). The schematic diagram for the centers of an airfoil.
LITERATURE REVIEW
Dynamic response of aircraft model during flight has received a lot of

interests, and a number of analytical and experimental methods have been
developed about the simulating and analyzing the aircraft structures. The
literature search focus on two areas: the analytical study of airfoil system
and the experimental study of that system.
Analytical Study
Yosibash and Kirby [1] constructed a high order simulation model of

fluid-structure for the airfoil under flying conditions. The authors utilized
spectral/hp solver for fluid (air) and hp-FEM solver for the airfoil to
handle the coupling problem generated by aerodynamic and structure
interaction. The two solvers somehow can minimize the modeling errors
and the discretization errors. The ongoing verification and validation of
fluid-structure interaction are also presented. In their study the airfoil was
treated as a flexible thin plate in the flow and the one-way coupling and
two-way coupling were discussed separately. The coupled fluid-structure
method was also used by Liu et al. [2] in calculation of wing flutter.
Fabunmi [3] presented a method of using the pseudo-inverse technique to
determine the operating vibratory loads on a structural system. In his
work, two types of uniform beam were studied: the cantilever beam and
the free end beam. Results of numerical calculations showed that the
accuracy of the determined forces correlates well with a parameter that is
related to the number of modes contributing to the response of the
structure at a given frequency. Shyy and Kamakoti [4] studied the
interactions between rigid and flexible structures and fluid based on an
airfoil system model. A loosely coupled approach was used to perform the
combined fluid and structure computations. Two different airfoil
configurations were employed to obtain the displacement of the airfoil
and the pressure on it. A suitable interfacing technique was incorporated
to couple and synchronize the flow and structure solver. Liu and Shepard
[5] addressed an approach of dynamic force identification based on
enhanced least squares and total least-squares schemes in the frequency
domain. This approach is effective in reducing the random errors that
occur in structural response signals. Two regularization filters, named as
the truncated singular value decomposition (TSVD) filter and the
Tikhonov filter, were employed in conjunction with the conventional
least-squares scheme at specific frequencies. A new least-squares form of
the Morozov’s discrepancy principle was formulated to aid in selecting
the optimum regularization parameter for these filters at each frequency.
The accuracy in using conventional least-squares, TSVD-based
least-squares, and Tikhonov filter-based least-squares schemes were
compared analytically and numerically in their study. E. Parloo et al. [6]
demonstrated a method of identifying dynamic force by means of
in-operation modal models. The applicability of the sensitivity-based
normalization approach for force identification on the basis of output-only
data was evaluated. The quality of the reconstructed FRF data and the
advantages of using an iterative weighted pseudo-inverse approach over a
classic pseudo-inverse one were discussed through experiments
performed on a beam structure.
Aerodynamics of airfoil models remains an important topic in aircraft

research, and most recent work in the area involves some aspect of airfoil
modeling and its nonlinear aeroelastic analysis. Kim and Lee [7, 8]
created a two-dimensional flexible airfoil with a freeplay nonlinearity in
pitch and analyzed this model in the subsonic flow range. Structurally, the
airfoil was modeled as finite beam elements and two spring elements in
pitch and plunge. A doublet lattice method was used for the
two-dimensional unsteady aerodynamics to include the camber deflection
effect. The fictitious mass modal approach was adopted in order to use the
consistent modal coordinates for the structures with nonlinearity.
Nonlinear aeroelastic analyses for both the frequency domain and time
domain were performed for rigid and flexible airfoil models to investigate
the flexibility effect. Dynamic response of this airfoil model such as limit
cycle oscillation and chaotic motion were observed and they were highly
influenced by the pitch-to-plunge frequency ratio. Lee et al. [9] derived
the equations of motion of a two-dimensional airfoil oscillating in pitch
and plunge for a structural nonlinearity using subsonic aerodynamic
theory. In their works, three classical aerodynamic nonlinearities
involving cubic, freeplay and hysteresis were investigated in detail.
Besides the simplified two-dimensional models, more complicated

aircraft wing models were created for purpose of analytical analysis,
which usually are multi-degree-of-freedom models. Roy and Eversman
[10, 11] created a multi-degree-of-freedom finite element model for the
flexile wing structure, with beam elements for bending and rod elements
for torsion. This model was then used to investigate the potential of an
adaptive feed-forward controller for active flutter suppression of a
flexible wing. Dimitriadis and Cooper [12] presented a method for
identification of non-linear multi-degree-of-freedom systems and used
this method to model aeroelastic systems for tracking the stability of
aircraft during flight flutter testing.
Experimental Study
Conner, et al. [13] created an experimental model which closely

approximated a 3D aeroelastic typical section in 2D incompressible flow
and used that model to determine the system response of the aeroelastic
section with control surface freeplay. Schewe, et al. [14, 15] performed
flutter and forced oscillation experiments in a transonic wind tunnel using
an aeroelastic supercritical 2D airfoil model. From those experiments,
typical transonic phenomena were studied and small-amplitude limit
cycle oscillations were observed under different flow conditions. It was
found that very small control forces were sufficient to excite or suppress
flutter oscillations of the airfoil system. Chen [16] performed an
experimental study including wind tunnel tests to determine the flutter
derivatives of section models under smooth and turbulent flows applying
a neural-network-based method. In his study, the dynamic response
observed from the experiments and the measured mean velocities of wind
flow were used to validate an appropriate neural network. Tang and
Dowell [17] constructed an experimental high-aspect-ratio wing
aeroelastic model with a slender body at the tip and measured its response
due to flutter and limit-cycle oscillations in a wind tunnel test. Important
experimental results such as the flutter velocity, oscillation frequency, and
static deformations were correlated with numerical results. Potter and
Lind [18] developed a method to test an uncertainty model for modal
analysis and robust flutter analysis using ground vibration testing data. In
their tests the aircraft wing has a NACA-65A004 shape and is composed
of a body, an aileron, and a boom. A measurement and an excitation
system were installed on the wing model. The measure system consisted
of 16 strain gages and 2 accelerometers and the excitation system
employed a piezoelectric material. Pickrel [19] demonstrated a method to
improve and validate the structural dynamic model for ground vibration
testing. The structural dynamic model was validated by comparing its
predicted modes with experimental modes. Because the author assumed
linearity on this structural model, the errors would be too large in a high
frequency domain, in which nonlinear structure response becomes
dominant. Lee [20], Joshi [21], Sun [22], and Nakade [23] developed
different experimental methods to measure the unsteady aerodynamic
loadings and investigate the interaction between the loadings and the
object’s dynamic structural response.
EVALUATE AERODYNAMIC FORCE USING 2DOF SYSTEM
2DOF Airfoil System
The aircraft wing, which is modeled as a two degree-of-freedoms

pitch-plunge system (Fig. 2), is governed by the following equations of
motion:
Mx  (C  UL )x  (K  U 2H)x  bf a (1)
Here
 1   k  ke  0 1 
x m 0  c dC L 
x   , b   , M  0 I , C   K , K   ke k  ke 2 , H   1 c0  ,
  c a       2 d  0 c  4  c  (2)
  
 3 c  
c  0 
 c dC L  
1
L  4 c  
2 d   1 c 0  2 1 c 0  3 c 0   8   dC L 
c 
 4 c  c     
   
/ 
    4 c  4 c      d   
where m and I are mass and mass moment of inertia about the gravity
center, respectively; fa is the excitation force from the aileron; α is the
proportional damping constant; k and kθ are translational and rotational
structural stiffness, respectively, e is the distance between the gravity and
the elastic center; U denotes air speed; ρ denotes air density; c denotes
chord length; c0 is the distance between the leading edge and the elastic
center; and CL is the local lift coefficient. The derivative dCL/dθ is
assumed to be constant, with a theoretical value of 2π for incompressible
flow.
k x
 fa
U
c
k
co e ca
c
Fig. (2). The two degree freedom system for airfoil.
Rewrite Eq. (1) into the state space form and we have:
  AY  Bf
Y (3)
a
Here
 0 I 
Y  [ x, x , ,  ] T , A   1 1 ,B  [0,M 1 ] T (4)
 M (K  U H)  M (C  UL)
2
where A and B are the system and input matrices, respectively. The
system states then can be determined from Eq. (3) in the frequency
domain as:
Y(ω) = (jωI – A)-1BFa(ω) (5)
The eigenvalues λi of the system matrix A are functions of the air speed U,
which are related to the system’s natural frequencies ni and damping
ratios i as:
 i   i ni  jni 1   i
2
(6)
As shown in Fig. (3a), when U increases, n1 increases and n2 decreases
steadily. The damping ratio of the second mode ζ2 decreases with an
increasing airspeed U and it reaches zero at the critical speed Ucr. At air
speed U > Ucr, ζ2 < 0, and hence the system becomes unstable.
In the time domain, the equations of the motion for the system (Fig. 2)
can be written as:
Mx  Cx  Kx  bfa  ULx  U2Hx; (7)

where, on the right hand side are the aerodynamic force and the excitation
force at the aileron, which are referred to actual forces on the system. The
objective of this study is to determine the actual forces using the
displacement data measured during the aircraft’s flight.
Indirect force identification is a method for estimating dynamic forces

acting on a structural or mechanical system by using the system’s
frequency response matrix and response measurements. This method is
often used to determine the unknown dynamic forces that occur when the
system is operating, at which time these forces cannot be directly
measured. This method is derived from the multi-input/multi-output
(MIMO) transfer function relationship for linear systems as:
X() = H()F() (8)
where ω is frequency; X(ω) is an output vector of the system’s structural

responses; F(ω) is an input vector of applied dynamic forces; and H(ω) is
the frequency response matrix of this system. Assuming H(ω) is known,
and the F(ω) can be determined as:
F() = H()-1X() (H() is a square matrix) (9)
or
F() = [H()T H()]-1H()TX() (H() is a non-square matrix) (10)
From above equations, it can be seen that in order to determine the force
F(ω), the H(ω) and X(ω) has to be constructed or measured through
experiments.
74 Fro
ontiers in Aerospa
ace Science, Vol. 1 Bi and Liu
The objective
o of this
t study is to identify th the distributeed aerodynammic force
and moment
m actin
ng on an airrcraft wing during the fflight. By em mploying
the in
ndirect forcee identificatiion for this problem, thhe wing’s fr frequency
response matrix H(ω) has to o be obtaineed from a gground vibraation test
(GVTT), where bo oth the excitaation and ressponse are m measurable. NNext, the
outpu
ut structural response X(ω)
X is meassured and thhe F(ω) cann then be
calcullated from Eq.
E (9) or (10).
( Differeent force dettermination methods
used in
i this study are demonstrated in folllowing sections.
80
(a)
70
Natural Frequency
mode 1
60
50 mode 2
40
30
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1
U/Ucr
0.2
(b)
0.15 mode 1
Damping Ratio
0.1
mode 2
0.05
0
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1
U/Ucr
Fig. (3)). (a) Natural frequencies, (b) damping ratios as a function of airrspeed.
Forcee Determina
ation Metho
ods
Tradiitional Forcee Determina

ation Methodd
Insteaad of measu uring the H((ω) experimmentally, the frequency response

matrixx can also bee found from
m the equatioon:
H(ω) = (-ω2M + jωC + K)-1 (11)

In Eq. (11), noise can be directly added to H(ω) in order to represent an
experimentally measured frequency response matrix which involves the
noise effects. According to Fabunmi [3], we have:
p
 H ()
i, j 1
ij
H(ω) = H(ω) + Enoise, and E noise    (12)

p
Here γ is the percentage error level and ρ is a randomly generated number

which belongs to [-1, 1]. The frequency response matrix H(ω) then can be
inverted and used in Eq. (8) to identify the external forces F(ω).
Force Determination Using “H1” & “H2” Estimated Frequency

Response Matrices
As an alternative approach, consider that H(ω) can be determined from

experimental data measured through conducting a ground vibration test
on an aircraft. In order to simulate the experiment, random measurement
noise is added to the force and displacement signals.
100
F(ω) = F(ω) + Nf, and Nf   F() 
i1
(13)
100
X(ω) = X(ω) + Nx, and N x   X() 
i1
(14)
where γ is the noise level. The frequency response matrix then can be
estimated from the measured data by [6]:
H1 = GFXGFF-1 (15)
H2 = GXXGXF-1 (16)
Here “H1” and “H2” are the estimated frequency response matrices, GFF
and GFX are auto-power spectrum of force (F) and cross-power spectrum
of force (F) and displacement (X), respectively. Likewise, GXX and GXF
are auto-power spectrum of displacement (X) and cross-power spectrum
of displacement (X) and force (F). The frequency response matrices
determined by Eq. (15) or (16) are transposed and inverted, and then are
substituted into Eq. (9) or (10) to calculate the forces F(ω). Such method
is called “H1” and “H2” force determination method.
Direct Force Determination “F1” and “F2” Methods
The “H1” and “H2” force estimation methods can be improved by

following transformation, from Eq. (15) we have:
H1-1 = GFFGFX-1, H1T = GFF-TGFXT, and H1-T = GFX-TGFFT (17)
Similarly, from Eq. (16) we have:
H2-1 = GXFGXX-1, H1T = GXF-TGXXT, and H1-T = GXX-TGXFT (18)
Eqs. (17 and 18) calculate the inverted frequency response matrix directly,
which can reduce the error introduced during the inversion of the
frequency response matrix. The forces then can be calculated by
substituting Eqs. (17), (18) into Eq. (9) or (10).
Validation of Force Determination Methods

To verify the aforementioned force determination approaches, these
methods are employed to simulate the two degree-of-freedoms airfoil
system model (Figs. 1 and 2). At first, the traditional pseudo-inverse
technique is employed to determine the aerodynamic forces and it is
assumed that the frequency response function matrix for force
determination is measured from experimental test. Fig. (4) plots the
estimated Frequency Response Function H(ω) (FRF) using the traditional
Vibratorry Force for Aircrraft Wings Frontiers in A
Aerospace Sciencee, Vol. 1 77
pseuddo-inverse method,
m wherre H11 is thee very first response fuunction in
the fo
orce responsee matrix [H]. The actual and estimateed force andd moment,
as weell as the perrcentage erro
ors are also displayed inn Figs. (5 annd 6). As
shown n from these figures, by using th the pseudo-iinverse metthod, the
percentage errors of the estimmated force and momennt are very low; with
average values at 0.085% and d 0.047%, resspectively.
-4
x 10
2
Frequency Response Function H(w)
0
2 4 6 8 10 12 14 16 18 20
Frequency (Hz)
Fig. (4)). FRF H11 at U/Ucr

U = 0 with 5% noise usingg traditional invverse method, baased on two
degreess-of-freedom mo
odel.
Seconndly, the freequency resp ponse functiion, actual aand estimateed forces
and moments,
m as well as percentage errorrs are re-calcculated by ussing “H1”
and “H
“ 2” force determination
d n methods. The sample rate to calcculate H1
and H2 is 20 Hz, and the perriod is 0.05 seconds. Beecause the eexcitation
force and noise added in this study are the ranndom numbbers, the
simullation resultss will be sliightly differrent each tim
me the calcuulation is
perforrmed. Thus, we have to redo the caalculation forr a certain am mount of
times to obtain th he average values
v of thee results andd use those vvalues to
comppare and validate the diffferent force ddeterminatioons. In this sstudy, the
calcullation is repeated 100 tiimes so that the obtained average reesults are
78 Fro
ontiers in Aerospa
stablee and predicttable.
Actual Translational Force

0.2
0.15
0.1
0.05
0
2 4 6 8 10 12 14 16 18 20
Frequency (Hz)
Estimated Translational Force

0.2
0.15
0.1
0.05
0
2 4 6 8 10 12 14 16 18 20
Frequency (Hz)
% Force Determination Error
0.4
0.2
0
2 4 6 8 10 12 14 16 18 20
Frequency (Hz)
Fig. (5)). Actual, estimaated, translation

nal force and perrcentage error, U
U/Ucr = 2/3 witth 5% noise
using trraditional inversee technique, baseed on two degreees-of-freedom m
model.
0.01
Actual Moment
0.005
0
0 2 4 6 8 10 12 14 16 18 20
Frequency (Hz)
0.01
Estimated Moment
0.005
0
0 2 4 6 8 10 12 14 16 18 20
Frequency (Hz)
% Moment Determination Error
0.2
0.1
0
2 4 6 8 10 12 14 16 18 20
Frequency (Hz)
Fig. (6)). Actual, estimaated rotational moment

m and perrcentage error, U
using trraditional inversee technique, baseed on two degreees-of-freedom m
model.
-4
x 10

1
0
2 4 6 8 10 12 14 16 18 20
Frequency (Hz)
Fig. (7)). FRF H11 at U//Ucr = 0 with 5%

% noise using H 1 method, basedd on two degreess-of-freedom
model.
0.4
0.3
0.2
0.1
0
0 2 4 6 8 10 12 14 16 18 20
Frequency (Hz)
Estimated Translational Force
0.4
0.3
0.2
0.1
0
0 2 4 6 8 10 12 14 16 18 20
Frequency (Hz)
% Force Determination Error
15
10
0
0 2 4 6 8 10 12 14 16 18 20
Frequency (Hz)
Fig. (8)). Actual, estimated translationaal force and perrcentage error, U

using H1 method, based
d on two degreess-of-freedom moodel.
80 Fro
ontiers in Aerospa
0.01
Actual Moment
0.005
0
0 2 4 6 8 10 12 14 16 18 20
Frequency (Hz)
0.01
Estimated Moment
0.005
0
0 2 4 6 8 10 12 14 16 18 20
Frequency (Hz)
30
20
10
0
0 2 4 6 8 10 12 14 16 18 20
Frequency (Hz)
Fig. (9)). Actual, estimaated rotational moment

m and perrcentage error, U
-4
X 10
2
0
2 4 6 8 10 12 14 16 18 20
Frequency (Hz)
0). FRF H11 at U/Ucr

Fig. (10 U = 0 with 5%
% noise using H 2 method, basedd on two degreess-of-freedom
model.
0.4

0.3
0.2
0.1
0
0 2 4 6 8 10 12 14 16 18 20
Frequency (Hz)
Estimated Translational Force Estimated Translational Force

0.4
0.3
0.2
0.1
0
0 2 4 6 8 10 12 14 16 18 20
Frequency (Hz)
40
30
20
10
0
0 2 4 6 8 10 12 14 16 18 20
Frequency (Hz)
1). Actual, estim

Fig. (11 mated translation
nal force and per
ercentage error, U
Figs. (7 to 9) display the FRF, acttual and eestimated foorce and

displaacement, and d percentagee errors obtaained from tthe H1 methhod. Figs.
(10 to
o 12) depict the same results obtaineed from the H2 method. Here the
noise is added th hrough the method
m desccribed in Eqqs. (13 and 14). The
resultts are formu ulated at air speed of 7.6666 m/s andd 5% noise llevel. As
shown n in Figs. (8
8 and 9), thee average perrcentage errrors for the fforce and
momeent obtained d from the H1 method aree 6.24% andd 7.61%, resppectively.
Neverrtheless, from m Figs. (11 and
a 12), the average perccentage erroors for the
force and momen nt yielded fro
om the H2 meethod are 31.41% and 255.35%.
Finallly, actual, esstimated forcces and momments, and tthe percentaage errors
are reecalculated using
u the “F1” and “F2” m
methods andd plotted in (Figs. 13
to 16)). The displaayed results are
a obtainedd at the airspeed of 7.6666 m/s and
82 Fro
ontiers in Aerospa
noise level of 5% %. Based on n Figs. (13 aand 14), thee average peercentage
errorss for the forcce and moment yielded ffrom the “F1” method arre 2.69%
and 0.47%,
0 whilee these valu ues are 8.90% % and 0.47%% given by the “F2”
metho od, according g to the Figss. (15 and 166).
Actual Moment 0.015
0.01
0.005
0
0 2 4 6 8 10 12 14 16 18 20
Frequency (Hz)
0.01
Estimated Moment
0.005
0
0 2 4 6 8 10 12 14 16 18 20
Frequency (Hz)
% Moment D e termination E rror
40
30
20
10
0
0 2 4 6 8 10 12 14 16 18 20
Frequency (Hz)
2). Actual, estim

Fig. (12 mated rotational moment and peercentage error, U
0.4

0.3
0.2
0.1
0
0 2 4 6 8 10 12 14 16 18 20
Frequency (Hz)
% Force Determination Error Estimated Translational Force

0.4
0.3
0.2
0.1
0
0 2 4 6 8 10 12 14 16 18 20
Frequency (Hz)
15
10
0
0 2 4 6 8 10 12 14 16 18 20
Frequency (Hz)
3). Actual, estim

nal force and peercentage error, U
using F1 method, based
d on two degrees--of-freedom moddel.
0.01
Actual Moment
0.005
0
0 2 4 6 8 10 12 14 16 18 20
Frequency (Hz)
0.01
Estimated Moment
0.005
0
0 2 4 6 8 10 12 14 16 18 20
Frequency (Hz)
30
20
10
0
0 2 4 6 8 10 12 14 16 18 20
Frequency (Hz)
4). Actual, estim

84 Fro
ontiers in Aerospa
0.4

0.3
0.2
0.1
0
0 2 4 6 8 10 12 14 16 18 20
Frequency (Hz)
Estimated Translational Force Estimated Translational Force

0.4
0.3
0.2
0.1
0
0 2 4 6 8 10 12 14 16 18 20
Frequency (Hz)
40
30
20
10
0
0 2 4 6 8 10 12 14 16 18 20
Frequency (Hz)
5). Actual, estim

0.015
Actual Moment
0.01
0.005
0
0 2 4 6 8 10 12 14 16 18 20
Frequency (Hz)
0.001
Estimated Moment
0.005
0
0 2 4 6 8 10 12 14 16 18 20
Frequency (Hz)
40
30
20
10
0
0 2 4 6 8 10 12 14 16 18 20
Frequency (Hz)
6). Actual, estim

EVALUATE AERODYNAMICS FORCE USING MDOF SYSTEM
MDOF Aircraft Wing Model
The presented force determination methods (Eqs. 9 to 18) are then applied
for multi-degree-of-freedom wing models.
Wn fa
qn
vn
Elasticity center w3
line fa
q3
U
v3
w2 Gravity center
line f a
q2
v2
w1
fa l
q1
v1
co yq ca
Fig. (17). Finite element aeroelastic aircraft wing model.
As shown in Fig. (17), an aircraft wing is modeled as a multi-degree-

of-freedom pitch-plunge system with torsional and bending beam
elements. This system is also governed by Eq. (1), where M and K are
mass and stiffness matrices for the beam elements. For each beam
element “i”, its mass and stiffness matrices are:
 2a a 156by  0 54by  0 
 a 2 a 54by  0 156by  0 
  (19)
156by  54by  156b 22lb 54b  13lb 
Mi   
 0 0 22lb 4l 2 b 13lb  3l 2 b 
 54by  156by  54b 13lb 156b  22lb
 
 0 0  13lb  3l 2 b  22lb 4l 2 b 
c c 0 0 0  0
 c c 0 0 0 0

0 0 12 d 6ld  12 d 6ld  (20)
Ki   
0 0 6ld 4l 2 d  6ld 2l 2 d 
0 0  12 d  6lb 12 d  6ld
 
 0 0 6ld 2l 2 d  6lb 4l 2 d 
Here a = (ρIpl)/6, b = (ρAl)/420, c = GJ/l, d = EI/l3, yθ is the Y-distance
between the wing’s gravity and elastic center; ρ is the material’s density;
Ip and I are the polar and regular moment of inertia of the wing model’s
cross section, respectively; A is the cross-sectional area; l denotes the length
of single element; G denotes the shear modulus of elasticity; J represents
the polar moment of inertia; and E is the Young’s modulus (see Fig. 17).
The input vector b and the proportional damping matrix C are given as:
b = [-ca, -ca, …, -ca, 1, 1, …, 1]; C = αM + βK (21)
where ca is the Y-distance between the trailing edge and the gravity center,
and α and β are the proportional damping constant.
The displacement vector x (assume this model has n elements):
x = [θ, x]T = [θ1, …, θn, x1, …, xn]T (22)
th
where θi is the i element’s rotational displacement and xi is its
translational displacement. The system’s mass and stiffness matrices can
be written as:
 
 2a a 0  156 by  0 54by  0 0 0 
 
 a 4a a  54by  0 312 by  0 54by  0  (23)
 0 a 4a  0 0 54by  0 312 by  0 
 
           
 
 
M  156 by  54by  0  156 b 22lb 54b  13b 0 0 
 
 0 0 0  22lb 4l2b 13lb  3l2b 0 0 
 54by  312 by  54by   54lb 13lb 312 b 0 54b  13lb 
 
 0 0 0   13lb  3l2b 0 8l2b 13lb  3l2b 
 0 54by  312 by   0 0 54b 13lb 312 b 0 
 
 0 0 0  0 0  13lb  3l2b 0 8l2b 
 
           
 
 c  c 0  0 0 0 0 0 0 
 
 c 2c  c  0 0 0 0 0 0  (24)
 0  c 2c  0 0 0 0 0 0 
 
            
 
 
K   0 0 0  12 d 6 ld  12 d 6 ld 0 0 
 
 0 0 0  6 lb 4l2 d  6 ld 2l2 d 0 0 
 0 0 0   12 d  6 ld 24 d 0  12 d 6 ld 
 
 0 0 0  6 ld 2l2 d 0 8l2 d  6 ld 2l2 d 
 0 0 0  0 0  12 d  6 ld 24 b 0 
 
 0 0 0  0 0 6 lb 2l2 d 0 8l2 d 
 
           
H and L in Eq. (1) for the multi-degree-of-freedom model are

aerodynamic stiffness matrix and aerodynamic damping matrix,
respectively, which are defined as:
0 1 0 
 1 c0 
0 c(  ) 1 
 0 c dCL  4 c (25)
H  
2 d 0 0
1 c
c(  0 ) 
 4 c 
     
 3 c 
 1 c(  0 ) 0 
4 c
 1 c 1 c 3 c 8 dCL 3 c 
 c dCL c(  0 ) c2[(  0 )(  0 )  1 c(  0 ) 
(26)
L 0  
2 d  
4 c 4 c 4 c d 4 c
 1 c0 2 1 c0 3 c0 8 dCL 
 0 c(  ) c [(  )(  )  1 
4 c 4 c 4 c  d
     

where ρ0 is the air density; c is the chord length, c0 is the Y-distance

between the leading edge and the elastic center (Fig. 17); and CL is the
local lift coefficient. The derivative dCL/dθ is assumed to be constant,
with a theoretical value of 2π for incompressible flow.
Boundary conditions are added on the global system, which assume that
the initial bending, torsional, and translational displacements are zero.
Similarly, for this multi-degree-of-freedom system, the governing
88 Fro
ontiers in Aerospa
equation (1) is rew

written in to the sate spacce form as E
Eqs. (3 and 44). At this
time the
t “Y” in Eq. (3) is reprresented as:
, x n ,  1,...,, n , x 1,..., x n ] T
Y  [1,...,n , x1,..., (27)
Eqs. (5 and 6) are

a still app plicable in ccalculating tthis multi-degree-of-
freedo
om system’ss states in th
he frequencyy domain annd the eigenvvalues of
the sy
ystem matrix
x A.
Fig. (18). depicts the relationship betweenn the naturaal frequencies and the
airspeeed, and Figg. (19) displays the rellationship beetween the damping
ratioss and the airrspeed. From
m those figuures, it can bbe found thhat as the
airspeeed U increaases, the dam
mping ratio oof the first m
mode ζ2 decreeases and
becom mes zero at the
t critical speed
s Ucr. A
At air speed U > Ucr, ζ2 < 0, and
hencee the system becomes un nstable. The same phenomenon was observed
from the two degrrees-of-freed dom system.
The first
f and second mode shapes of the aircraft wing model at the
airspeeed U = 0 aree plotted in Figs.
F (20 andd 21), wheree the first moode shape
referss to bending and the secoond mode shhape refers too torsion. Tabble 1 and
2 list the relation
nships betweeen natural ffrequencies and dampinng rations
and thhe airspeed (including
( th
he first six m
modes).
40
35 (a)
30
Natunal Frequency (Hz)
25
20
15
10
0
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1
U/Ucr
8). Natural frequ

Fig. (18 uencies vs. airspeeed.
(b)
0.4
0.35
0.3
0.25
Damping Ratio
0.2
0.15
0.1
0.05
0
0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9
U/Ucr
9). Damping ratiios vs. airspeed.

Fig. (19
0
(a)
-0.005
Displacement for the wing
-0.01
-0.015
-0.02
-0.025
-0.03
-0.035
20
10 20
15
10
5
0 0
Chord direction
Span direction
0). First mode sh

Fig. (20 hape at ω1 = 16.2
244 Hz and U/Uccr = 0.
0.03 (b)
0.02
Displacement for the wing
0.01
-0.01
-0.02
-0.03
20
10 20
15
10
5
0 0
Chord direction Span direction
1). Second modee shape at ω2 = 36.473 Hz and U//Ucr = 0.

Fig. (21
Table 1. Natural frequencies (Hz) vs. airspeed (U/Ucr).
U/Ucr Mode 1 Mode 2 Mode 3 Mode 4 Mode 5 Mode 6

0 16.244 36.473 160.55 500.55 1033.2 1759.5
0.1 10.382 33.6 166.24 251.95 458.47 973.09
0.2 6.8443 34.064 75.398 76.655 77.911 79.168
0.3 21.562 21.812 63.354 75.398 76.655 77.911
0.4 16.952 20.377 71.539 75.398 76.655 77.911
0.5 4.906 31.179 81.372 133.09 284.56 722.06
0.6 4.7606 28.554 79.008 122.5 253.82 625.86
0.7 4.7012 27.346 80.797 105.5 229.52 549.58
0.8 4.6794 26.908 82.433 91.309 210.22 489.76
0.9 4.676 26.879 68.724 98.201 193.98 442.13
1 4.6819 26.788 62.333 100.17 179.66 403.43
Table 2. Damping ratios vs. airspeed (U/Ucr).
U/Ucr Mode 1 Mode 2 Mode 3 Mode 4 Mode 5 Mode 6

0 0.19876 0.23049 0.24918 0.30178 0.40083 0.47871
0.1 0.10773 0.11821 0.17493 0.1767 0.26788 0.39319
0.2 0.32367 0.32681 0.38013 0.40906 0.48424 0.58842
0.3 0.40844 0.41564 0.43326 0.4727 0.52556 0.59122
0.4 0.40034 0.43584 0.45864 0.50383 0.54092 0.6478
0.5 0.37023 0.37241 0.46516 0.48108 0.62537 0.731
0.6 0.30117 0.34676 0.3704 0.51534 0.53147 0.75383
0.7 0.23902 0.26696 0.34486 0.41992 0.56749 0.65837
0.8 0.16434 0.19748 0.29465 0.36687 0.54666 0.63189
0.9 0.071974 0.1587 0.18241 0.40808 0.41975 0.70132
1 0.00014979 0.014406 0.19345 0.27838 0.46157 0.65677

Validation of Force Determination Methods
In this section, Eqs. (9 to 18) are applied to determine the aerodynamic

forces acting on this multi-degree-freedom airfoil system and the
estimated forces and moments are compared to the actual values.
First, the traditional pseudo-inverse technique is employed to determine

the aerodynamic forces and the frequency response function matrix H(ω)
is calculated from Eqs. (11 and 12). Fig. (22) displays the first response
function in the force response matrix, H11. Figs. (23 and 24) compare the
actual and estimated forces and moments. From both figures, it can be
found that by using the traditional pseudo-inverse method, the average
percentages for the forces and moments are 0.2027% and 0.5962%,
respectively.
-4
x 10
1
0.9
0.8
Frequency respones function H(w)
0.7
0.6
0.5
0.4
0.3
0.2
0.1
0
0 10 20 30 40 50 60 70 80 90 100
Frequency(Hz)
Fig. (22). FRF H11 at U/Ucr = 0 with 5% noise using traditional inverse method, based on
multi-degree-of-freedom model.
92 Fro
ontiers in Aerospa
0.2
0.15
Actual Force
0.1
0.05
0
0 10 20 30 40 50 60 70 80 90 100
Frequecy(Hz)
0.2
0.15
Estimated Force
0.1
0.05
0
0 10 20 30 40 50 60 70 80 90 100
Frequecy(Hz)
2
%Force determination error
1.5
0.5
0
0 10 20 30 40 50 60 70 80 90 100
Frequecy(Hz)
3). Actual, estim

Fig. (23 mated, translation
using trraditional inversee technique, baseed on multi-degrree-of-freedom m
model.
0.1
0.08
Actual moment
0.06
0.04
0.02
0
0 10 20 30 40 50 60 70 80 90 100
Frequency (Hz)
0.1
Estimated moment
0.08
0.06
0.04
0.02
0
0 10 20 30 40 50 60 70 80 90 100
Frequency (Hz)
% Moment determination error
0
0 10 20 30 40 50 60 70 80 90 100
Frequency (Hz)
4). Actual, estim

using trraditional inversee technique, baseed on multi-degrree-of-freedom m
model.
Next, the frequency response function H11, actual, estimate forces and
moments, as well as percentage errors are calculated by “H1” and “H2”
methods. The noise is added through Eqs. (13 and 14). The results plotted
in following figures are formulated at air speed of 9.34 m/s and noise
level of 5%. As shown in Figs. (25 and 26), the average percentage errors
for force and moment calculated from H1 method are 17.36% and 15.80%.
Nevertheless, the percentage errors given by H2 method are 22.42% and
20.74%, as displayed in Figs. (27 and 28).
0.1
0.08
Actual Force
0.06
0.04
0.02
0
0 10 20 30 40 50 60 70 80 90 100
Frequency(Hz)
0.1
0.08
Estimated Force
0.06
0.04
0.02
0
0 10 20 30 40 50 60 70 80 90 100
Frequency(Hz)
%Force determination error
100
50
0
0 10 20 30 40 50 60 70 80 90 100
Frequency(Hz)
Fig. (25). Actual, estimated translational force and percentage error, U/Ucr = 2/3 with 5% noise
using H1 method, based on multi-degree-of-freedom model.
0.01
0.008
oment
0.006
Actual M
0.004
0.002
0
0 10 20 30 40 50 60 70 80 90 100
Frequency(Hz)
0.01
ent
0.008
atedMom
0.006
0.004
Estim
0.002
0
0 10 20 30 40 50 60 70 80 90 100
Frequency(Hz)
inationerror
150
ent determ
100
50
%Mom
10 20 30 40 50 60 70 80 90
Frequency(Hz)
Fig. (26). Actual, estimated rotational moment and percentage error, U/Ucr = 2/3 with 5% noise
0.08
Actual Force
0.06
0.04
0.02
10 20 30 40 50 60 70 80 90 100
Frequency(Hz)
0.1
0.08
atedForce
0.06
Estim
0.04
0.02
0
0 10 20 30 40 50 60 70 80 90 100
Frequency(Hz)
inationerror
150
Forcedeterm
100
50
%
0
10 20 30 40 50 60 70 80 90 100
Frequency(Hz)
0.01
0.008
Actual Moment 0.006
0.004
0.002
0
0 10 20 30 40 50 60 70 80 90 100
Frequency(Hz)
0.01
Estimated Moment
0.008
0.006
0.004
0.002
0
0 10 20 30 40 50 60 70 80 90 100
Frequency(Hz)
%Moment determination error
250
200
150
100
50
0
0 10 20 30 40 50 60 70 80 90
Frequency(Hz)
Finally, actual, estimated forces and moments, and the percentage errors
are recalculated using the “F1” and “F2” methods and plotted in Figs. (31
to 34). The displayed results are obtained at the airspeed of 9.34 m/s and
noise level of 5%. According to Figs. (29 and 30), the average percentage
errors for the force and moment yielded from the “F1” method are 8.77%
and 10.62%, while these values are 14.62% and 15.72% yielded from the
“F2” method, as shown in Figs. (31 and 32).
0.1
0.08
Actual Force
0.06
0.04
0.02
0
0 10 20 30 40 50 60 70 80 90 100
Frequency(Hz)
0.1
0.08
atedForce
0.06
Estim
0.04
0.02
0
0 10 20 30 40 50 60 70 80 90 100
Frequency(Hz)
150
inationerror
100
Forcedeterm
50
%
0
0 10 20 30 40 50 60 70 80 90 100
Frequency(Hz)
using F1 method, based on multi-degree-of-freedom model.
0.01
0.008
oment
0.006
ctual M
0.004
A
0.002
0
0 10 20 30 40 50 60 70 80 90 100
Frequency(Hz)
0.01
0.008
atedMoment
0.006
stim
0.004
E
0.002
0
0 10 20 30 40 50 60 70 80 90 100
Frequency(Hz)
120
inationerror
100
80
ent determ
60
40
Mom
20
%
10 20 30 40 50 60 70 80 90 100
Frequency(Hz)
0.1
0.08
Actual Force
0.06
0.04
0.02
0
0 10 20 30 40 50 60 70 80 90 100
Frequency(Hz)
0.1
0.08
EstimatedForce
0.06
0.04
0.02
0
0 10 20 30 40 50 60 70 80 90 100
Frequency(Hz)
150
%Forcedeterminationerror
100
50
0
0 10 20 30 40 50 60 70 80 90 100
Frequency(Hz)
0.01
0.008
oment
0.006
Actual M
0.004
0.002
0
0 10 20 30 40 50 60 70 80 90 100
Frequency(Hz)
0.015
oment
0.01
EstimatedM
0.005
0
0 10 20 30 40 50 60 70 80 90 100
Frequency(Hz)
inationerror
100
80
ent determ
60
40
om
20
%M
0
10 20 30 40 50 60 70 80
Frequency(Hz)
DISCUSSIONS
In above sections, the forces and moments estimated from the traditional
pseudo-inverse method, “H1” and “H2” methods, as well as “F1” and “F2”
methods are compared to the actual values. From these comparisons, it is
found that the errors caused by the pseudo-inverse method are much
lower than those errors yielded from other methods, which is because in
that method, the mass, stiffness and damping matrices are directly
obtained from the two degrees-of-freedom and multi-degree-of-freedom
models. However, in reality, those matrices are difficult to be obtained
from a real complicated model. In the “H1”, “H2”, “F1”, “F2” methods
which use cross-spectrum and auto-spectrum, the frequency response
function is estimated from the input force and the corresponding structural
response. Therefore, such methods are easier to apply for force
determination in real problems.
Compared to the “H1” and “H2” methods, the “F1” and “F2” methods yield
lower errors in identifying the force and moment values. This is because
the direct determination methods directly calculate the inverted frequency
response matrix, therefore eliminate the error introduced during the
inversion of the frequency response matrix. It is also found from our
study that when the airspeed and noise level increase, the force
determination errors also rise.
EXPERIMENT TESTINGS
Experiment Model
A 3D wing model with dimensions fitting the wind tunnel test was
designed using the commercial CAD package Solid Edge (Fig. 33). The
CAD model was then manufactured in the rapid prototype laboratory in
University of Louisville (Fig. 34). The airfoil shape of this model is
NACA-65A004 and its dimensions are listed in Table 3.
In order to install accelerometers in to this model for vibration test, two

rectangular cutouts were created at the side surfaces of both ends of the
boom. The cables that connect the accelerometers and signal amplifier
were fed through the interior of this model in order to eliminate
aerodynamic influence. Thus, both the wing and boom are hollow, the
thickness of the wing model is 1mm, and an 8.2 mm × 8.5 mm space is
available inside the boom.
At the root of the wing, a plate is attached to fix the wing to the wind
tunnel wall. In this plate, a helicoids hole is created to connect the model
with the wind tunnel. The material of the wing model is Dura Form
polyamide (PA), with Young’s modulus 1600 MPa, Poisson’s ratio 0.4,
and density 590 kg/m3.
Fig. (33). Airfoil model for the experiment.

100 Frrontiers in Aerosp
pace Science, Vol. 1 Bi and Liu
4). 3-D wing mo

Fig. (34 odel in Solid Edg
ge.
Table 3.
3 Dimensions of wing model.
Half Span 1 60.36 (mm)

Chord 1 61.01 (mm)
Diameter of bo
oom 1 7.27 (mm)
Total length 2 25.00 (mm)
Total width 1 87.64 (mm)
Grou
und Vibratio
on Test
Experriment Setup
p
The purpose
p of th
he ground vibration
v testt is to obtainn the airfoill model’s
frequeency respon nse functionss (FRFs) byy measuring the excitatiion force
and itts behavior of
o response. Fig. (35) shoows the flow w chart of thiis test. In
this test,
t a calibrrated impactt hammer ((with a rubbber tip) wass used to
excitee the model and two acccelerometerrs were employed to meeasure its
responses (Fig. 36 6). In that figure, a sensoor was instaalled at the end of the
impacct hammer to o measure th he excitationn force, and its sensitivitty is 2.25
mv/NN. During thee test, the wiing model w was rigidly m mounted by cclamping
it onto a steel tubbe (Fig. 33). Next, 22 tesst points weere excited separately
by the impact haammer and the t model’s responses w were recordeed by the
acceleerometers. These
T test points
p are eiither locatedd at the leadding and
trailin
ng edges, at accelerometter locations,, or at the wiing-to-boom m joint, as
shown n in Fig. (37
7). The choosing of an am mount of tesst points enssured that
the ob btained mod de shapes weere related too the entire wwing model.. In order
to miinimize the errors, 10 tests were reepeated at eevery test pooint. The
resoluution of the data
d acquisittion system is 0.5 Hz, thhe sampling time is 2
sec, and
a the samp pling frequenncy is 5000 HHz.
5). Flow chart off ground vibratio

Fig. (35 on test (GVT).
6). Impact hamm

Fig. (36 mer for the experiment.
7). Locations of test points and sensors on the moodel.

Fig. (37
Moda
al Analysis Methods
M
Threee methods were

w used to
o perform m modal analyssis: SDOF pparameter
estimation (metho od 1), directt parameter approach (m
method 2), and least
squares complex x exponentiaal algorithm m (method 3). Analysiis results
obtain
ned from theese three metthods were ccompared to each other.
The SDOF
S metho
od is develo
oped based oon frequenccy domain m
models of
the frequency resp ponse functions (FRFs) eevaluated inn the neighboorhood of
a singgle mode, which
w is an approximatte method thhat ignores complex
conjuugate contrib bution and can solve for both m modal frequeency and
residu
ue. The direect approach h identifies the continuuous-time m model and
parammeters directly from th he dynamic response ddata of the system.
Becau use the paarameters of o the moddel can ooffer clear physical
interp
pretation, thiis method haas been incrreasingly appplied in receent years
especcially in the area of ind dustrial appllication. Leaast squares complex
exponnential (LSCCE) method is a time doomain modaal analysis m method. It
exploores the relattionship betwween the immpulse respoonse functionns (IRFs)
of a multi-degree
m e-of-freedomm system andd its compleex poles andd residues
througgh a compleex exponential. The soluution of thiss model usinng LSCE
algoriithm leads too the establiishment of a polynomiall whose rootts are the
comp plex roots of the system.
Resullts and Discu

ussions
The modal
m analyssis results caalculated by these three methods aree listed in
Table 4. Figs. (388 and 39) display the freequency respponse functiion H11,11
measu ured from thhe experimen nt and the syynthesized H11,11 by assuuming the
airfoil system a lin
near two deg grees-of-freeedom model.
H 11.11
0.12
0.1
0.08
Magnitude
0.06
0.04
0.02
0
0 10 20 30 40 50 60 70 80 90 100
Frequency(Hz)
Fig. (38). The frequency response functtion H11,11 measuured from the expperiment.
Table 4. Results of model analysis.
Natural Frequencies (Hz) Damping Ratio Modal Shapes

Test Points
Method1 Method2 Method3 Method1 Method2 Method3 Method1 Method2 Method3
r=1 14.36 14.79 10.67 0.0121 0.01 0.0189 1 1.1370 1 1.129 1 0.909
1
r=2 31.82 31.57 31.11 0.0033 0.0057 0.0043 1 -0.9898 1 -1.0267 1 -1.0166
r=1 14.34 14.76 10.68 0.0116 0.0161 0.0102 1 1.0760 1 1.1950 1 1.3187
2
r=2 31.82 31.61 31.12 0.0078 0.0045 0.0044 1 -0.9598 1 -0.9793 1 -0.9665
r=1 14.4 14.1 10.74 0.0117 0.0261 0.0095 1 1.1650 1 1.1349 1 1.3135
3
r=2 31.81 31.62 31.12 0.0077 0.0043 0.0044 1 -0.9698 1 -0.9759 1 -0.9690
r=1 14.14 14.66 13.44 0.0124 0.0396 0.0094 1 1.490 1 1.3362 1 0.9876
4
r=2 31.77 31.59 31.1 0.007 0.0043 0.0043 1 -0.9731 1 -0.9769 1 -0.9847
r=1 13.85 13.85 13.43 0.0136 0.0883 0.0095 1 1.1630 1 0.9259 1 1.1606
5
r=2 31.74 31.58 31.08 0.0068 0.0045 0.0045 1 -0.9765 1 -0.9766 1 -1.0162
r=1 13.76 14.16 13.4 0.0135 0.1011 0.0096 1 1.1532 1 0.9508 1 1.2752
6
r=2 31.74 31.57 31.05 0.0066 0.0045 0.004 1 -0.9757 1 -0.9757 1 -0.8965
r=1 13.74 13.66 13.39 0.0144 0.086 0.0085 1 1.1599 1 0.8799 1 1.2246
7
r=2 31.7 31.54 31.05 0.0067 0.0046 0.0049 1 -0.9755 1 -0.9797 1 -0.9822
r=1 13.76 13.58 13.38 0.0145 0.0155 0.0098 1 1.159 1 0.9978 1 1.4757
8
r=2 31.83 31.57 31.07 0.0068 0.005 0.0049 1 -0.9761 1 -0.9779 1 -0.9862
r=1 13.74 13.59 13.35 0.0165 0.0103 0.0099 1 1.1606 1 1.1091 1 1.2032
9
r=2 31.74 31.59 31.08 0.0071 0.005 0.0051 1-0.9768 1 -0.9784 1 -0.9854
r=1 13.71 13.49 13.36 0.0174 0.0313 0.018 1 1.1667 1 1.120 1 1.0956
10
r=2 31.72 31.74 31.23 0.0048 0.0048 0.005 1 -0.9763 1 -0.9795 -0.9787
(Table 4) contd.....
r=1 13.72 13.67 13.36 0.0167 0.0451 0.012 1 1.1568 1 1.0871 1 1.1234
11
r=2 31.77 31.69 31.18 0.0054 0.0048 0.0052 1 -0.9782 1 -0.9787 1 -0.9838
r=1 13.72 13.46 13.27 0.0265 0.0317 0.0122 1 1.1739 1 1.2358 1 1.1591
12
r=2 31.72 31.75 31.24 0.0059 0.0053 0.0054 1 -1.1035 1 -0.9815 1 -0.9830
r=1 13.71 13.61 13.33 0.0191 0.0309 0.0193 1 1.1656 1 1.1811 1 1.1681
13
r=2 31.72 31.82 31.31 0.0059 0.0048 0.005 1 -0.8974 1 -0.9806 1 -0.9766
r=1 13.69 13.68 13.36 0.0165 0.0201 0.011 1 1.1587 1 1.1021 1 1.1172
14
r=2 31.73 31.78 31.28 0.0066 0.0047 0.0051 1 -0.9210 1 -0.9802 1 -0.9730
r=1 13.69 13.6 13.37 0.0154 0.0101 0.0102 1 1.1567 1 1.0605 1 1.2017
15
r=2 31.72 31.82 31.32 0.0055 0.0045 0.0048 1 -0.8032 1 -0.9792 1 -0.9700
r=1 13.7 13.8 13.35 0.0156 0.0153 0.0138 1 1.1515 1 1.0459 1 1.1893
16
r=2 31.71 31.83 31.33 0.0043 0.0033 0.0048 1 -0.9073 1 -0.9769 1 -0.9611
r=1 13.71 13.93 13.37 0.0149 0.2111 0.0102 1 1.1471 1 1.0505 1 1.1230
17
r=2 31.71 31.79 31.35 0.005 0.0042 0.0047 1 -0.8324 1 -1.2415 1 -0.9723
r=1 13.72 13.92 13.36 0.0155 0.124 0.0103 1 1.1489 1 1.1298 1 1.1156
18
r=2 31.72 31.89 31.4 0.0046 0.0045 0.0049 1 -0.8012 1 -0.9772 1 -1.0046
r=1 13.74 13.92 13.63 0.0149 0.0254 0.0228 1 1.1404 1 1.1888 1 1.0942
19
r=2 31.73 31.88 31.38 0.0074 0.0044 0.0047 1 -0.7548 1 -0.9765 1 -0.9957
r=1 14.34 12.8 13.45 0.0119 0.4737 0.0132 1 1.1683 1 1.5013 1 1.3084
20
r=2 31.75 31.62 31.13 0.0082 0.0038 0.0043 1 -0.7887 1 -0.9716 1 -0.9663
r=1 13.95 16.35 13.43 0.0151 0.03 0.0101 1 1.1648 1 1.3731 1 0.8575
21
r=2 31.72 31.86 31.35 0.0081 0.0037 0.0047 1 -0.7630 1 -0.9759 1 -0.9980
r=1 13.85 16.78 15.45 0.0139 0.008 0.0103 1 1.1099 1 1.2207 1 1.5981
22
r=2 31.72 31.72 31.89 0.0081 0.0028 0.0019 1 -0.9406 1 -0.9401 1 -0.6932
H 11.11
0.14
0.12
0.1
0.08
Magnitude
0.06
0.04
0.02
0
0 10 20 30 40 50 60 70 80 90 100
Frequency(Hz)
9). The synthesizzed frequency reesponse function H11,11.

Fig. (39
From the Table 4, the results of naturaal frequencyy, damping rratio and
modee shape obtaiined by the three
t methodds are fairly closed to eaach other.
By comparing
c the frequeency responnse functions measureed from
experriment and obtained from f synthhesized calcculation, it can be
concluuded that the airfoil system can an be repreesented as a linear
two-d
degrees-of-frreedom systeem and solvee numericallyy.
Wind
d Tunnel Tesst
Theorretical Back
kground
The purpose
p of th
his experimeent is to idenntify the air-ddynamic forrce acting
on thee wings. Asssuming no exxcitation forcce, the equattion of motioon of this
systemm can be wrritten as Eq. (1), whose terms on thhe right handd side are
the aeerodynamic force which h are called aactual forcess on the system. The
goal of
o the experriment is to identify thee actual forcces using thee data of
displaacement meaasured during g the aircraft
ft flight periood.
The equations of motion in the frequency domain are:
X() = H()F() (28)
F() = H()-1X() (29)
where H(ω) is the frequency response function of the system and F(ω) is
the actual force in the frequency domain. H(ω) is obtained from ground
vibration testing, X(ω) is the airfoil response measured during flight,
which can be measured by the two accelerometers mounted on the wing.
Thus, the air-dynamic forces acting on the wing during flight can be
identified by the experiment results.
Experiment Setup
The experiment setup for the wind tunnel test is the same as the previous
test and the data were collected using the same way (Fig. 40). The speed
of wind was measured by a Pitot gage as shown in Fig. (41). Table 5 lists
the seven wind speeds measured by the Pitot gage in the wind tunnel.
Fig. (40). Testing devices in wind tunnel experiment.

Fig. (41). The pitot gage for the experiment.
Table 5. Measured wind speeds and average errors of measured aerodynamic forces under
different speed.
Speed Value Error in forces Error in moments
Speed 1 800 (ft/min) 3.7682% 8.8668%
Speed 2 1500 (ft/min) 4.8303% 10.1882%
Speed 3 2250 (ft/min) 7.1124% 11.3649%
Speed 4 2950 (ft/min) 11.2072% 13.6635%
Speed 5 3800 (ft/min) 13.4306% 18.3963%
Speed 6 4750 (ft/min) 17.3433% 20.4357%
Speed 7 5000 (ft/min) 20.5899% 21.1369%
Results and Discussions

The measured aerodynamic forces under different wind speed were
compared to the actual values and the average errors are listed in Table 6.
As shown in Table 6, the percentage errors are below 10% at lower air
speeds and the errors increased with the air speed. The maximum
percentage error for moments is 21.14%.
CONCLUSION
This chapter presents several force determination methods and applies

those methods to identify the aerodynamic forces and moments acting on
an aircraft wing model during the flight. Two wing models, a two
degrees-of-freedom model and a multi-degree-of-freedom model are
created and used for this study. The estimated force and moments are
compared to the actual values. From the comparisons, it is concluded that
with available mass, stiffness, and damping matrices, the traditional
pseudo-inverse method provides the best accuracy. However, considering
both accuracy and applicability, the direct determination methods “F1”
and “F2” are recommended for such problem. Also, the accuracy of the
created two degrees-of-freedom and multi-degree-of-freedom wing
models are verified through this study.
In the experimental analysis part, two tests, ground vibration test and
wind tunnel test were performed on a rapid prototyped airfoil model. The
model’s natural frequencies, damping ratios, mode shapes were calculated
based on the data measured from the ground vibration test. The
aerodynamic loadings acting on this model under different wind speeds
were measured from the wind tunnel test and validated by comparing to
the actual force values.
this publication.
ACKNOWLEDGEMENTS
Declared none.
REFERENCES
[1] R.M. Kirby, Z. Yosibash, and G.E. Karniadakis, "Towards stable coupling
methods for high-order discretization of fluid-structure interaction: Algorithms
and observations", J. Comput. Phys., vol. 223, no. 2, pp. 489-518, 2007.
[http://dx.doi.org/10.1016/j.jcp.2006.09.015]
[2] F. Liu, J. Cai, Y. Zhu, H.M. Tsai, and A.S. Wong, "Calculation of wing flutter by
a coupled fluid-structure method", J. Aircr., vol. 38, no. 2, pp. 334-342, 2001.
[http://dx.doi.org/10.2514/2.2766]
[3] J.A. Fabunmi, "Effects of structural modes on vibratory force determination by

the pseudoinverse technique", AIAA J., vol. 24, no. 3, pp. 504-509, 1986.
[http://dx.doi.org/10.2514/3.9297]
[4] R. Kamakoti, Y. Lian, S. Regisford, A. Kurdila, W. Shyy, Computational

aeroelasticity using a pressure-based solver, AIAA-2002-869, AIAA Aerospace
Sciences Meeting and Exhibit, 40th, Reno, NV, Jan. 14-17, 2002.
[5] Y. Liu, and W.S. Shepard Jr, "Dynamic force identification based on enhanced
least squares and total least-squares schemes in the frequency domain", J.
Sound Vibrat., vol. 282, no. 1/2, pp. 37-60, 2005.
[http://dx.doi.org/10.1016/j.jsv.2004.02.041]
[6] E. Parloo, P. Verboven, P. Guillaume, and M.V. Overmeire, "Force identification

by means of in-operation modal models", J. Sound Vibrat., vol. 262, no. 1, pp.
161-173, 2003.
[http://dx.doi.org/10.1016/S0022-460X(02)01052-0]
[7] I. Lee, and S-H. Kim, "Aeroelastic analysis of a flexible control surface with
structural nonlinearity", J. Aircr., vol. 32, no. 4, pp. 868-874, 1995.
[http://dx.doi.org/10.2514/3.46803]
[8] S-H. Kim, and I. Lee, "Aeroelastic analysis of a flexible airfoil with a freeplay
nonlinearity", J. Sound Vibrat., vol. 193, no. 4, pp. 823-846, 1996.
[http://dx.doi.org/10.1006/jsvi.1996.0317]
[9] B.H. Lee, S.J. Price, and Y.S. Wong, “Nonlinear aeroelastic analysis of
airfoils: bifurcation and chaos", Prog. Aerosp. Sci., vol. 35, no. 3, pp. 205-334,
1999.
[http://dx.doi.org/10.1016/S0376-0421(98)00015-3]
[10] I.D. Roy, and W. Eversman, "Adaptive flutter suppression of an unswept wing",
J. Aircr., vol. 33, no. 4, pp. 775-783, 1996.
[http://dx.doi.org/10.2514/3.47014]
[11] W. Eversman, and I.D. Roy, "Adaptive flutter suppression using

multi-input/multi-output adaptive least mean square control", J. Aircr., vol. 34,
no. 2, pp. 244-250, 1997.
[http://dx.doi.org/10.2514/2.2163]
[12] G. Dimitriadis, and J.E. Cooper, "A method for identification of non-linear
multi-degree-of-freedom systems", Proceedings of the Institution of Mechanical
Engineers, Part G: Journal of Aerospace Engineering, vol. 212, no. 4, pp.
287-298, 1998.
[http://dx.doi.org/10.1243/0954410981532441]
[13] M.D. Conner, D.M. Tang, E.H. Dowell, and L.N. Virgin, "Nonlinear behavior
of a typical airfoil section with control surface freeplay: a numerical and
experimental study", J. Fluids Structures, vol. 11, no. 1, pp. 89-109, 1997.
[http://dx.doi.org/10.1006/jfls.1996.0068]
[14] G. Schewe, H. Mai, and G. Dietz, “Nonlinear effects in transonic flutter with
emphasis on manifestations of limit cycle oscillations", J. Fluids Structures, vol.
18, no. 1, pp. 3-22, 2003.
[http://dx.doi.org/10.1016/S0889-9746(03)00085-9]
[15] G. Dietz, H. Mai, and G. Schewe, "Experiments on heave/pitch

limit-cycle oscillations of a supercritical airfoil close to the transonic dip", J.
Fluids Structures, vol. 19, no. 1, pp. 1-16, 2004.
[16] C.H. Chen, "Determination of flutter derivatives via a neural network approach",
J. Sound Vibrat., vol.263, no. 4, pp. 797-813, 2003.
[http://dx.doi.org/10.1016/S0022-460X(02)01279-8]
[17] D. Tang, and E.H. Dowell, "Experimental and theoretical study on aeroelastic
response of high-aspec--ratio wings", J. Aircr., vol. 39, no. 8, pp. 1430-1441,
2001.
[18] S. Potter, and R. Lind, "Developing uncertainty models for robust flutter
analysis using ground vibration test data", Proc. 42th Structures, Structural
Dynamics, and Materials Conference, 2001 Seattle, WA, USA
[http://dx.doi.org/10.2514/6.2001-1585]
[19] C.R. Pickrel, "Airplane ground vibration testing: nominal modal model
correlation", S.V. Sound and Vibration, vol. 36, no. 11, pp. 18-23, 2002.
[20] S. Lee, "Unsteady aerodynamic force prediction on a square cylinder using k-ε
turbulence models", J. Wind Eng. Ind. Aerodyn., vol. 67-68, pp. 79-90, 1997.
[http://dx.doi.org/10.1016/S0167-6105(97)00064-0]
[21] M.V. Joshi, and N.M. Reddy, "Aerodynamic force measurements over missile
configurations in IISc shock tunnel at M∞ = 5.5", Exp. Fluids, vol. 4, no. 6, pp.
338-340, 1986. [http://dx.doi.org/10.1007/BF00266300]
[22] M. Sun, and J-H. Wu, "Aerodynamic force generation and power requirements
in forward flight in a fruit fly with modeled wing motion", J. Exp. Biol., vol.
206, no. Pt 17, pp. 3065-3083, 2003.
[http://dx.doi.org/10.1242/jeb.00517] [PMID: 12878674]
[23] K. Nakade, M. Suzuki, and H. Fujimoto, "Interaction between vehicle vibration

and aerodynamic force on high-speed train running in tunnel", Veh. Syst. Dyn.,
vol. 41, pp. 717-723, 2004.
CHAPTER 3
Computational and Analytical Investigation of
Lateral Impact Behavior of Pressurized
Pipelines
Yangqing Dou* and Yucheng Liu

39762, USA
Abstract: This chapter provides a combined computational and analytical study to

investigate the lateral impact behavior of pressurized pipelines and inspect effects of
important parameters such as the outside diameter and internal pressure on such
behavior. A total of more than 300 numerical simulations are carried out on mild
steel pipe models with different internal pressure levels and were struck at the mid-span
and at the one quarter span positions. These numerical simulations of the impact
tests are performed using 3D dynamic nonlinear finite element analysis (FEA)
through LS-DYNA, where both geometrical and material nonlinearities are
considered. The computational results for the first time systematically reveal the
effects of internal pressure, impact position, and outside diameter on the lateral
impact behavior of the pipeline models. Quartic polynomial functions are applied to
formulate the maximum crushing force (Fmax), maximum permanent displacement
(Wf), and absorbed energy (Ep) of the pressurized pipelines during the impact
problem. The effects of the diameter and pressure on F, W, and E are therefore
illustrated through analyzing those functions. Response surfaces are also plotted
based on the generated quartic polynomial functions and the quality (accuracy) of
*Corresponding author Yangqing Dou: Department of Mechanical Engineering, Mississippi State University,
MS, 39762, USA; Tel: 337-704-8410; E-mail: yd120@msstate.edu.
Yucheng Liu (Ed.)

114 Frontiers in Aerospace Science, Vol. 1 Dou and Liu
those functions are verified through several techniques. The outcomes of this study
have potential benefits on research of safety and reliability of pressurized pipelines
in hydraulic system of aerospace and development of advanced pipeline materials.
Keywords: Collapse mechanism, Computer modeling and simulation,

Lateral impact, LS-DYNA, Pressurized pipeline.
NOMENCLATURE
2L Distance between two supports of a pipeline

D Outside diameter of a pipeline
σy Static uniaxial yield stress of pipe
σu Static ultimate tensile stress of pipe
εr Static uniaxial rupture strain of pipe
p Internal pressure
Wf The maximum permanent transverse displacement
Wl The local permanent transverse displacement
Wg The global permanent transverse displacement
Fmax Maximum impact force
Fm Average impact force
vi Initial velocity of indenter
vr Rebound velocity of indenter
Ep Absorbed impact energy
Ef Threshold failure energy
INTRODUCTION
Dynamic responses and failure modes of pipelines subjected to

low-velocity lateral impact were studied and described by Jones and other
researchers [1-9] through a series of experimental and theoretical analyses.
In Jones’ work [1-3, 9], pipelines with different sizes were fully clamped
at both ends and a rigid indenter struck transversely at the pipe center,
quarter span and near to the support at velocities ranging up to 14 m/s.
Lateral Impact Behavior of Pressurized Pipelines Frontiers in Aerospace Science, Vol. 1 115
Important experimental results (deformation mode, maximum permanent

transverse displacement, threshold failure energy, etc.) were observed and
associated with size of the pipelines, initial kinetic energy of the indenter,
its impact velocity, and its impact position. In addition, properties of
material of the pipelines were also considered and the influences of
material strain hardening, strain rate sensitivity, and elasticity on the
impact behavior of the pipelines were included in their analytical models.
Pressurized pipelines were also tested in Jones’ and Ng’s work [8, 9] but
the influence of internal pressure on the impact behavior of the pipelines
has not been explicitly demonstrated. In reality, the internal pressure is a
critical factor in the design and assessment of pipelines because in civil
applications such as aerospace hydraulic systems, most pipelines convey
gases and liquids under high pressures over long distances.
In this study, the dynamic inelastic behavior of clamped thin-walled pipes

with internal pressure impacted transversely by a rigid, knife-edge
indenter at the pipe center and quarter span is modeled and examined
computationally through more than 300 simulations. The obtained
numerical results are validated by comparing to several previously
published experimental data. Effects of internal pressure on the lateral
impact response of the pipelines as well as influences of other impact
conditions such as dimensions of the pipelines and impact location are
revealed from those simulation results. An explicit FEA solver, LS-DYNA
is employed to create the FEA models and run the numerical analyses.
After the computational analysis, numerical methods are employed to

establish analytical models to numerically show how the impact
parameters (internal pressure and outside diameter) affect the impact
response of the pipelines during low-speed lateral impact. This study
mainly focuses on finding the effects of internal pressure and outside
diameter on the pipeline’s impact force, deformation, and energy
absorption capacity.
PROBLEM DESCRIPTION
The indenter geometry is selected as a fairly severe test of a pipeline and

is an idealization of the edge of a flange which could be dropped
accidentally onto a pipeline. The pressurized pipes are made from
seamless cold drawn mild steel with outside diameters of 22, 42, 60, 80,
100, and 120 mm with a fixed ratio of 2L/D = 10, where 2L is the
distance between the two supports and D is the outside diameter. The
selected ratio of 2L/D = 10 is currently being used in most research
laboratory and industry plant as the largest unsupported pipe length ratio
[5]. The cold worked mild steel pipes have a 2mm wall thickness, and the
mechanical properties in the axial direction of the pipe are: static uniaxial
yield stress σy = 663 MPa, static ultimate tensile stress σu = 823 MPa, and
static uniaxial rupture strain εr = 6 ~ 7%. Several internal pressures will
be applied in addition to the tests without any internal pressure in order to
find out the influence of internal pressure on the lateral impact behavior
of the pipelines. In this study, the internal pressure p varies from 0 to 150
bar. Specifically, 6 different pressures, 0, 30, 60, 90, 120, 150 bar are
applied on inner surface of the pipelines separately in order to achieve a
complete understanding of how the internal pressure affects the lateral
impact behavior of the pipelines.
After simulations, important numerical data are measured or estimated in

order to describe the plasticity mechanisms and impact behavior of the
pressurized pipelines. Those data include Wf, the maximum permanent
transverse displacement, Wl and Wg, local and global permanent transvers
displacement respectively, maximum impact force Fmax and average
impact force Fm, rebound velocity of the striker when it separates from the
pipe vr, impact energy absorbed by the pipes Ep and the threshold failure
energy Ef, etc. Fig. (1) defines Wf, Wl and Wg, from which the
relationships among the three displacement components can be defined
as:
Wf = Wl + Wg (1)
Original position
of pipeline
wg
wf
wl
Final position of
deformed pipeline
Fig. (1). Wf, Wl, and Wg for the idealized deformed cross-section of a pipeline.
MATERIAL PROPERTIES
In creating FEA models used for impact analysis that involves large strain
and high strain rate problems, the material properties have to be carefully
defined to correctly capture the impact. Here, elastic plastic material law
with kinematic isotropic hardening is chosen to model the pipeline
material (cold drawn steel). Dynamic effects of strain rates are taken into
account by scaling static yield stress with a factor with the assumption of
Cowper-Symonds relation [12]:
   
1/ P

 0   0 1     (2)
 C  
 
where ε is the strain rate and Eq. (2) shows that during dynamic crushing,
the original yield strength σ0 should be replaced by the dynamic flow
stress σʹ0 due to the strain rate effects. C and P are constants of
Cowper-Symonds relation, which are used to define the influence of
material strain rate sensitivity in this study.
The pipelines are made of cold drawn steel, whose material properties are
listed in Table 1 and were implemented into the FEA models through the
card *MAT_PLASTIC_KINEMATIC. The indenter is assumed to be rigid
during the impact simulation and was defined using *MAT_RIGID.
Table1. Plastic kinematic material properties for the LS-DYNA pipeline models.
Material properties Value

Density ρ 7800 kg/m3
Young’s modulus E 200 GPa
Poisson’s ratio ν 0.29
Yield stress σy 663 MPa
Ultimate tensile stress σu 823 MPa
Tangential modulus KT 583 MPa
uniaxial rupture strain εr 6 ~ 7%
Failure strain ε 0.72
Hardening index 0.169
COMPUTER MODELING AND ANALYSIS
72 FEA models were created for the pressurized pipelines and one for the
rigid knife-edge indenter. The pipelines were modeled using 4-node
Belytschko-Tsay shell elements with five integration points through the
thickness. The indenter was modeled using rigid elements with rigid
material properties assigned and its mass is set as 17.48 kg. It has the
same shape and mass as the indenter used by Jones [9]. The contact
between the pipelines and the indenter was defined through automatic
surface to surface contact algorithm. Within the defined contact pair, the
impact face on the pipeline was chosen as master contact surface and the
impact face on the indenter the slave surface. An initial speed of 10 m/s
was applied on the impact, at which it impacted onto the pipeline.
Two loading curves were developed to define the internal pressure and
gravity, which were applied onto the pipeline model as loading conditions.
In applying the internal pressure curve, its direction was carefully selected
so that the internal pressure was applied onto the inner side of the pipeline
model along normal directions. For example, if the normal direction of

the shell elements of the pipeline model is pointing outward of the wall,
the negative value should be turned on in the LS-DYNA internal pressure
curve card; otherwise, a positive index value has to be applied. For more
details regarding definition and application of the internal pressure curve
card, please refer to [12].
The generated FEA models include 1000 to more than 25,000 elements.
More than 300 impact simulations were then performed on those models
after applying appropriate boundary conditions (constraints on both ends
of the pipelines), loading conditions (internal pressure on the pipelines),
and initial conditions (initial velocity on the indenter). Figs. (2 and 3)
present two impact scenarios, where the indenter impacted a pipeline
model with outside diameter of 60 mm and internal pressure of 60 bar at
its center and one quarter span, respectively.
Fig. (2). FEA model showing an indenter impacts on center of a pressurized pipeline.
Fig. (3). FEA model showing an indenter impacts on one quarter of a pressurized pipeline.
Fig. (4). Deformed configuration of FEA model presented in Fig. (2).

COMPUTATIONAL IMPACT BEHAVIOR OF THE PIPELINE

MODELS
FEA Results
After the impact simulations, important FEA results were collected and
compared, from which typical lateral impact response of the pressurized
pipelines would be described. (Figs. 4 and 5), for example, plot the
deformed configurations of the FEA models presented in Figs. (2 and 3),
respectively. Some significant FEA results are recorded in Tables 2 and 3
for pipeline models which were struck at middle and one quarter span
position, respectively.
Fig. (5). Deformed configuration of FEA model presented in Fig. (3).
Table 2. FEA results for the pipes struck at the middle position.
D p Wf Wl Wg Fmax Fm vi vr Ep
No.
(mm) (bar) (mm) (mm) (mm) (kN) (kN) (m/s) (m/s) (kJ)
1 22 0 34.31 25.37 8.94 47.14 34.65 9.93 2.02 0.83
2 22 30 34.26 25.28 8.98 47.42 34.74 9.93 2.02 0.83
No.
3 22 60 34.21 25.20 9.01 47.45 34.86 9.93 2.03 0.83
4 22 90 34.15 25.12 9.03 47.64 34.95 9.93 2.03 0.83
5 22 120 34.06 25.02 9.04 47.77 35.04 9.93 2.04 0.83
6 22 150 33.98 24.92 9.06 47.86 35.07 9.93 2.05 0.83
7 42 0 24.60 20.19 4.41 50.09 36.19 9.93 3.14 0.78
8 42 30 24.08 19.68 4.46 51.00 36.78 9.93 3.19 0.77
9 42 60 23.70 19.19 4.51 51.89 37.10 9.93 3.24 0.77
10 42 90 23.31 18.78 4.53 52.36 37.44 9.93 3.29 0.77
11 42 120 21.31 16.75 4.56 52.39 37.89 9.93 3.34 0.77
12 42 150 20.61 16.05 4.58 52.75 37.87 9.93 3.37 0.76
13 60 0 18.05 15.73 2.33 56.80 36.47 9.93 4.18 0.71
14 60 30 16.05 13.75 2.32 59.45 38.04 9.93 4.46 0.69
15 60 60 14.71 12.43 2.28 61.80 38.87 9.93 4.68 0.67
16 60 90 13.59 11.33 2.26 64.35 39.85 9.93 4.83 0.66
17 60 120 12.83 10.58 2.25 65.61 40.49 9.93 4.94 0.65
18 60 150 12.07 9.80 2.27 67.68 41.13 9.93 5.02 0.64
19 80 0 16.83 15.63 1.20 58.37 34.51 9.93 4.92 0.65
20 80 30 12.45 11.42 1.03 64.18 37.36 9.93 5.58 0.59
21 80 60 10.42 9.46 0.96 69.80 39.29 9.93 5.93 0.56
22 80 90 8.98 8.00 0.98 74.10 41.24 9.93 6.16 0.53
23 80 120 8.36 7.30 1.06 76.16 42.45 9.93 6.29 0.52
24 80 150 7.62 6.47 1.15 76.19 43.15 9.93 6.37 0.51
25 100 0 15.90 15.22 0.68 59.30 32.15 9.93 5.30 0.61
26 100 30 10.05 9.56 0.49 68.00 36.62 9.93 6.29 0.51
27 100 60 7.60 7.09 0.51 74.68 39.51 9.93 6.70 0.50
28 100 90 6.56 5.98 0.58 80.57 42.53 9.93 6.89 0.55
29 100 120 5.57 5.01 0.56 85.81 43.66 9.93 7.02 0.62
30 100 150 5.11 4.43 0.68 86.54 45.75 9.93 7.02 0.79
31 120 0 16.12 15.51 0.61 54.71 30.09 9.93 5.47 0.60
32 120 30 8.12 7.75 0.37 66.35 35.28 9.93 6.59 0.49
No.
33 120 60 6.25 5.87 0.38 79.10 38.41 9.93 7.00 0.43
34 120 90 5.37 4.88 0.49 86.63 41.61 9.93 7.21 0.41
35 120 120 4.83 4.22 0.61 91.95 44.72 9.93 7.26 0.40
36 120 150 4.48 4.71 0.77 89.86 49.10 9.93 7.33 0.39
Table 3. FEA results for the pipes struck at one quarter span position.
No. D p Wf Wl Wg Fmax Fm vi vr Ep
1 22 0 21.71 13.80 6.91 55.22 39.35 9.93 1.88 0.83
2 22 30 21.61 15.79 6.90 53.93 39.38 9.93 1.89 0.83
3 22 60 21.50 14.58 6.92 56.42 39.41 9.93 1.90 0.82
4 22 90 21.41 14.49 6.92 57.16 39.46 9.93 1.91 0.82
5 22 120 21.31 14.39 6.92 54.85 39.50 9.93 1.92 0.82
6 22 150 21.24 15.31 6.93 56.19 39.53 9.93 1.93 0.81
7 42 0 19.20 16.10 3.10 59.17 42.09 9.93 2.71 0.80
8 42 30 18.56 15.47 3.09 60.56 42.47 9.93 2.83 0.77
9 42 60 17.95 14.83 3.12 62.24 42.73 9.93 2.90 0.75
10 42 90 17.43 14.31 3.13 63.13 42.94 9.93 2.96 0.74
11 42 120 16.98 13.84 3.14 64.10 43.11 9.93 3.01 0.73
12 42 150 16.56 13.42 3.14 64.89 43.31 9.93 3.07 0.73
13 60 0 19.00 17.15 1.85 61.38 40.38 9.93 3.34 0.76
14 60 30 17.11 15.30 1.81 65.92 42.47 9.93 3.66 0.71
15 60 60 15.76 13.98 1.78 69.63 44.06 9.93 3.89 0.68
16 60 90 14.64 12.89 1.75 72.20 45.18 9.93 4.09 0.67
17 60 120 13.72 11.97 1.75 73.57 45.92 9.93 4.23 0.67
18 60 150 12.81 11.04 1.77 76.15 46.60 9.93 4.35 0.68
19 80 0 18.76 18.63 1.13 60.63 37.40 9.93 3.80 0.74
20 80 30 13.27 12.17 1.10 68.61 40.65 9.93 4.50 0.69
No. D p Wf Wl Wg Fmax Fm vi vr Ep
21 80 60 12.95 11.97 0.98 75.48 43.02 9.93 4.89 0.65
22 80 90 11.01 10.12 0.89 78.49 44.92 9.93 5.15 0.63
23 80 120 10.04 9.17 0.88 81.48 46.10 9.93 5.34 0.61
24 80 150 9.02 8.14 0.88 84.71 47.48 9.93 5.49 0.60
25 100 0 18.39 17.46 0.93 58.91 34.39 9.93 4.09 0.72
26 100 30 12.36 11.87 0.49 68.72 34.43 9.93 5.21 0.63
27 100 60 9.99 9.56 0.48 79.35 43.41 9.93 5.75 0.57
28 100 90 7.72 7.22 0.50 84.12 45.87 9.93 6.09 0.54
29 100 120 6.79 6.32 0.47 86.52 48.23 9.93 6.28 0.52
30 100 150 5.78 5.20 0.58 89.78 50.09 9.93 6.36 0.51
31 120 0 18.04 17.23 0.81 54.79 31.84 9.93 4.45 0.68
32 120 30 10.52 10.05 0.47 68.21 38.90 9.93 5.73 0.58
33 120 60 7.68 7.25 0.43 78.82 43.13 9.93 6.45 0.50
34 120 90 4.63 4.45 0.18 89.96 46.33 9.93 6.78 0.46
35 120 120 4.21 3.82 0.39 101.1 51.74 9.93 7.08 0.42
36 120 150 4.10 3.74 0.36 101.1 51.74 9.93 7.08 0.42
Some of the listed FEA results obtained from this study were compared
with several previously published experimental results [8, 9] and a good
agreement was achieved. Thus, the accuracy of the developed FEA
models and proposed computational approach was validated.
Effects of Impact Position on the Impact Response of Pipelines
(Figs. 6 to 8) compare the histories of impact force, transverse

deformation, and absorbed energy of the pipeline model when it was
struck by the indenter at its center and one quarter span (Figs. 2 and 3).
Similar results in impact force, deformation and absorbed energy can be
found from other pipeline models with different dimensions and internal
pressures.
Impact force
80000
70000
60000
50000
Force(N)
40000
30000 middle span
20000 quarter span
10000
0
1093
1275
1457
1639
1821
2003
2185
2367
2549
2731
2913
3095
3277
3459
3641
3823
183
365
547
729
911
1
Step
Fig. (6). Comparison of impact force yielded from models presented in Figs. (2 and 3).
2.50E-02
2.00E-02
1.50E-02
Wf(m)
1.00E-02
middle span
5.00E-03
quarter span
0.00E+00
47
93
139
185
231
277
323
369
415
461
507
553
599
645
691
737
783
829
875
921
967
1
Step
Fig. (7). Comparison of deformation yielded from models presented in Figs. (2 and 3).
J Absorbed Energy
9.00E+02
8.00E+02
7.00E+02
6.00E+02
5.00E+02
4.00E+02
3.00E+02
middle span
2.00E+02
1.00E+02 quarter span
0.00E+00
1006
1207
1408
1609
1810
2011
2212
2413
2614
2815
3016
3217
3418
3619
3820
202
403
604
805
1
Step
Fig. (8). Comparison of absorbed energy yielded from models presented in Figs. (2 and 3).
In order to better reveal the influence of impact position on the lateral

impact behavior of the pipelines, six response surfaces are plotted based
on results listed in Tables 1 and 2 and displayed in Figs. (9 to 11). In those
displayed response surfaces, outside diameter and internal pressure are set
as two input variables and the output result is maximum impact force,
maximum deformation, and maximum absorbed energy, separately. Figs.
(9a and 9b) compare the maximum impact force of the pipeline models
struck at its center and at its one quarter span. Figs. (10a and 10b)
compare the maximum deformation, and Figs. (11a and 11b) compare the
maximum absorbed energy.
80-100
60-80
100 40-60
20-40
80
Fmax(KN)
0-20
60
40
150
20 120
90
0 60
22 30
42 60 80 0
100
120
(a)
80-100
60-80
100 40-60
20-40
80
0-20
Fmax(KN)
60
40
150
20 120
90
0 60
22 30
42 60 80 0
100
120
(b)
Fig. (9). Maximum impact force for different impact position at (a) center (b) one quarter.
30-40
40
20-30
30
10-20
Wf(mm)
20
0-10
150
10 120
90
0 60
22 30
42
60 0
80
100
120
(a)
30-40
40
20-30
30
10-20
Wf(mm)
20
150 0-10
10 120
90
0 60
22 30
42 60 80 0
100
120
(b)
Fig. (10). Maximum deformation for different impact position at (a) center (b) one quarter.
0.8-1
0.6-0.8
1 0.4-0.6
0.8 0.2-0.4
0-0.2
0.6
Ep(kJ)
0.4
150
0.2 120
90
0 60
22 30
42 60 80 0
100
120
(a)
0.8-1
0.6-0.8
1 0.4-0.6
0.8 0.2-0.4
0-0.2
0.6
Ep(kJ)
0.4
150
0.2 120
90
0 60
22 30
42 60 80 0
100
120
(b)
Fig. (11). Maximum absorbed energy for different impact position at (a) center (b) one quarter.
Effects of Internal Pressure on the Impact Response of Pipelines
Next, the effects of internal pressure on lateral impact behavior of the

pipelines are inspected based on Figs. (12 to 17).
Fig. (12). Maximum impact force of pipeline models struck at center.
Fig. (13). Maximum impact force of pipeline models struck at one quarter.
Fig. (14). Maximum deformation of pipeline models struck at center.
Fig. (15). Maximum deformation of pipeline models struck at one quarter.
Fig. (16). Maximum absorbed energy of pipeline models struck at center.

Fig. (17). Maximum absorbed energy of pipeline models struck at one quarter.
DISCUSSION ON COMPUTER RESULTS
From Figs. (6 to 17) it can be observed that the different impact positions
did not evidently alter the characteristic lateral impact response of the
pressurized pipelines. The response surfaces of maximum impact force,
displacement, and absorbed energy obtained from both impact scenarios
showed similar shapes (Figs. 9 to 11). The influences of internal pressure
and diameter on the impact response of those pipelines also rendered
similar tendencies under both mid-position impact and one
quarter-position impact (Figs. 12 to 17).
However, different impact positions did affect the values of maximum

impact force, transverse displacement, and absorbed energy. Figs. (12 and
13) argue that the maximum impact forces yielded during the one
quarter-position impact were apparently higher than those generated
during the mid-position impact. This is because that the “one quarter
position” is closer to fixture and therefore is more rigid than the “middle
position”. Due to the same reason, the permanent transverse
displacements yielded from the one quarter-position impact were
significantly lower than those resulted from the mid-position impact (as
shown in Figs. (14 and 15), which can also be well explained through the
classic cantilever beam theory. However, as reflected from (Figs. 16 and
17), the impact energy absorbed during the mid-position impact was
slightly lower than that absorbed during the one quarter-position impact.
Those computational findings agree well with the phenomena observed
experimentally by Jones and his coworkers [1-3].
As for the influences of the internal pressure and outside diameter, Figs.
(12 and 13) reveal that when the internal pressure increased, the
maximum impact force also increased, even such tendency was not
apparent when the out diameter was low. It is believed that the internal
pressure will enhance impact resistance of the pipeline model, therefore
leads to higher impact forces. Meanwhile, those figures also tell us that
under the same pressure, the pipeline models with larger outside diameter
may be subjected to higher impact force.
From Figs. (14 and 15) it is found that the maximum transverse
displacement decreased with an increasing internal pressure. Similar to
the maximum impact force, such tendency was not clear when the outside
diameter was low. This phenomenon can also be explained through the
improved impact resistance of the pipelines due to the increased internal
pressure. On the other side, the total permanent transverse displacement
Wf decreased when the outside diameter D increased. However, on
inspecting the results listed in Tables 1 and 2, it can be seen that the
permanent local displacement Wl did increase as the outside diameter
increased, which can be perfectly explained using the theoretical model
produced by Oliveira et al. [10].
Figs. (16 and 17) show that when the internal pressure increased, the
impact energy absorbed by the pipelines decreased (still, such tendency
was not that evident for pipelines with a small outside diameter). Similar
finding was also reported in Ng and Shen’s work, which were obtained
through experiments [8]. Since the internal pressure enhances the rigidity
of the pipeline, that way the indenter will rebound at a comparatively high
speed after impacting the pipeline and less impact energy will be
absorbed by the pipeline. Meanwhile, from both figures it is observed that
similar to the displacement case, the absorbed impact energy decreased
when the outside diameter increased.
ANALYTICAL ANALYSIS
Response Surface Method (RSM) [11]
In modern industry, RSM is extensively applied in developing, improving,

formulating, and optimizing processes. During a design process, such
method is used to determine potential influences of several input variables
(independent variables) on the performance or quality of the entire system
so as to acquire optimized responses from that system. This method has
also been extensively applied by the author in optimum design of
thin-walled columns in order to optimize their energy absorption capacity
during crash analysis [12-17]. In this study, RSM is employed to
determine how the internal pressure (p) and outside diameter (D) of the
mild-steel pipelines affect their impact response, including impact force
(Fmax), permanent deformation (Wf), and absorbed energy (Ep).
In this study, the impact response of the pipelines (it can be the impact
force, deformation, and absorbed energy) is approximated using a series
of the basic functions in a form of
 n
y(x)  F( p, D), W ( p, D), orE ( p, D)   β i i (p, D) (3)
i 1
where n represents the number of basic functions φi(p, D). In this paper,
the polynomials are used to build up these basic functions to formulate F,
W, and E.
In Eq. (3), the βi, known as the regression coefficients, are estimated using
the method of least squares. Suppose we have m (m > n) observations
(obtained from FEA) for the yielded response yi (y1 – ym) based on the m
sampling design points (p, D)i, the least squares function is therefore
expressed as
2
m m  n 
L   εi    y i   β j j (p, D) 
2
(4)
i 1 i 1  j1 
where the design points (p, D)i are selected from the specified design
space, εi is the error between the response yi observed at these points, and
the RS approximation at that point. Afterwards, the coefficient vector b =
(β1, β2, …, βn)T can be determined by L /   0 , which is
b = (ФTФ)-1ФTy (5)
where Ф denotes the matrix consisting of basic functions evaluated using
m sampling points, which is
 1 (p, D)1   n (p, D)1 
Φ      
 (6)
1 (p, D) m   n (p, D) m 
By substituting Eq. (5) into (3), the response surface model is created and
the response functions (F(p, D), W(p, D), and E(p, D)) then can be fully
determined.
The accuracy of the developed response surface model can be verified

through several techniques. The relative error (RE) between the observed
response at those sampling points y(x) and the original response ŷ(x) is
RE = [ŷ(x) – y(x)]/y(x) (7)
Other two important properties in evaluating the model’s accuracy are the
sum of squares of the residuals (SSE) and the total sum of squares (SST),
which are
m
SS E   (y i  ŷ i ) 2 (8)
i 1
m
SS T   (y i  y i ) 2 (9)
i 1
Where y̅i is the mean value of yi.
The model’s fitness can be evaluated based on the F statistic, coefficient

of multiple determination R2, adjusted R2 statistic, and root mean square
error (RMSE) respectively, which are calculated as
(SS T  SS E )/n
F (10)
SS E /(m  n  1)
SS E
R2  1 (11)
SS T
m 1
2
R adj  1 (1  R 2 ) (12)
mn
SS E
RMSE  (13)
m  n 1
According to the classical RSM theory, the larger the values of R2 and
R2adj, and the smaller the value of RMSE, the better the model fit.
RSM Models and Assessment
As shown in Tables 2 and 3, six pressures and diameters were selected for
modeling and simulation and totally 36 combinations were presented.
Thus, in Eq. (4) the matrix Φ has 36 rows (m = 36), which corresponds to
the 36 combinations of p and D (No. 1 to 36 in Tables 1 and 2). In this
study, quartic polynomial is used as the basic function because it provides
the best fitness to the real problems [12]. Thus, the basic functions φi are
terms in a full quartic form, which are 1, p, D, p2, pD, D2, p3, p2D, pD2,
D3, p4, p3D, p2D2, pD3, D4. Substituting p and D values into Eq. (6), a 36
× 15 matrix Φ then can be created.
Next, response surface models (quartic polynomials) are created for F, W,

and E separately. Based on the FEA results listed in Tables 1 and 2, the
regression coefficients are determined using Eq. (7) and then the
corresponding quaritc polynomial functions are achieved as:
Mid-span impact
Fmax(p, D) = 32.6109 + 0.1092D – 0.0554p + 0.0003D2 + 0.0017Dp +

0.0006p2 – 2.6280×10-5D3 + 1.2623×10-5D2p – 2.3841×10-5Dp2 –
2.4003×10-8p3 + 1.2543×10-7D4 – 7.4857×10-8D3p + 7.0706×10-8D2p2 +
5.1554×10-8Dp3 – 7.5446×10-9P4 (14)
Wf(p, D) = 45.0506 – 0.4446D + 0.0670p – 0.0047D2 – 0.0030Dp +

0.0002p2 + 0.0001D3 – 7.3635D2p + 2.8224×10-5Dp2 – 1.0382×10-5p3 –
4.5878×10-7D4 + 2.8113×10-8D3p + 5.3156×10-8D2p2 – 1.1328×10-7Dp3 +
4.9168×10-8p4 (15)
Ep(p, D) = -0.0071 + 0.0676D + 0.0048p – 0.0017D2 – 0.0003Dp +

6.6842×10-6p2 + 1.6911×10-5D3 + 3.5967×10-6D2p + 4.9025×10-7Dp2 –
5.2068×10-8p3 – 5.6130×10-8D4 – 1.7108×10-8D3p – 4.5308×10-10D2p2 –
7.7971×10-10Dp3 + 2.1434×10-10P4 (16)
One quarter-span impact
Fmax(p, D) = 27.4686 – 0.3884D – 0.0101p + 0.0068D2 – 0.0004Dp +

0.0009p2 – 5.5470×10-5D3 – 3.0335×10-5D2p + 1.1911×10-5 Dp2 –
1.1831×10-5p3 + 1.6885×10-7D4 + 9.1578×10-8D3p +8.5204×10-8D2p2 –
6.3737×10-8Dp3 + 4.5053×10-8p4 (17)
Wf(p, D) = 30.7352 + 0.5217D – 0.0294p – 0.0059D2 – 0.0029Dp +

0.0003p2 – 1.3492×10-5D3+ 6.2449×10-5D2p + 9.7825×10-6Dp2 –
6.0572×10-6p3 + 2.2537×10-7D4 – 2.4567×10-7D3p – 3.1488×10-8D2p2 –
5.0011×10-8Dp3 + 2.7521×10-8p4 (18)
Ep(p, D) = 0.9443 – 0.0069D + 0.0004p + 0.0001D2 – 3.4112×10-5Dp –

7.2483×10-7p2 – 8.2700×10-7D3 – 6.4675×10-9D2p + 2.0426×10-7Dp2 –
1.0603×10-8p3 + 2.1813×10-9D4 – 1.1507 ×10-10D3p – 3.0748×10-10D2p2 –
1.5731×10-10Dp3 + 4.4457p4 (19)
The approximation of the responses obtained from the response functions

and the FEA results are then substituted into Eqs. (8) to (13) to evaluate
the fitness of the developed RS models. In those equations m = 36 and n =
15 (number of basic functions in a full quartic polynomial form). The
values of RE, R2, R2adj, and RMSE are calculated and displayed in Table 3.
From that table it can be seen that the R2 and R2adj for force (mid-span),
displacement (mid-span and one quarter span), and energy (one quarter
span) are very close to the perfect value (1) and the values for energy
(mid-span) and force (one quarter span) are still very high. The tabulated
values prove that the developed quartic polynomial functions accurately
reflect the pressurized pipeline’s responses during the lateral low speed
impact process. Meanwhile, Table 4 also validates the selection of outside
diameters and internal pressures over the entire design domain. The
generated response surfaces with respect to variables p and D are plotted
in Figs. (18 and 19).
(a)
(b)
(c)
Fig. (18). Quartic response surfaces of (a) Fmax; (b) Wf; (c) Ep under mid-span impact.
(a)
(b)
(c)
Fig. (19). Quartic response surfaces of (a) Fmax; (b) Wf; (c) Ep under one quarter span impact.
Table 4. Accuracy of quartic polynomial RS models.
RS model R2 R2adj RMSE Max. RE (%)

Fmax mid-span 0.9967 0.9945 1.4838 0.7854
Wf mid-span 0.9981 0.9969 0.8091 0.9236
Ep mid-span 0.9171 0.8618 0.3027 0.6960
Fmax one quarter span 0.9933 0.9889 1.4148 0.5874
Wf one quarter span 0.9965 0.9942 0.6151 0.4127
Ep one quarter span 0.9969 0.9948 0.0123 0.6739
PARAMETRIC STUDY
Influence of Internal Pressure
In order to reveal the effects of internal pressure on the pipelines’ lateral

impact response, as well as validate the accuracy of the developed
response surface models, we substitute D = 22, 42, 60, 80, 100, and 120
into Eqs. (14 to 19) to obtain a series of simplified analytical models with
the internal pressure p as the only variable. Curves are then plotted from
those analytical models and compared with the results obtained from FEA
simulations. (As displayed in Figs. 20 to 22).
22D
22D
100 42D
42D
60D
100 60D
80D
90 80D
100D
100D
120D 90 120D
80
80
70
Fmax (KN)
Fmax (KN)
70
60
50 60
40 50
30 40
0 30 60 90 120 150 0 30 60 90 120 150
Pressure (bar) Pressure (bar)
(a) (b)
Fig. (20). Effects of p on maximum impact force, (a) mid-span impact (b) one quarter span impact.
22D
40 42D
35 60D
30 80D
25 100D
W (mm)
20 120D
15
10
0
0 30 60 90 120 150
Pressure (bar)
(a)
22D
25 42D
60D
20
80D
100D
15
W (mm)
120D
10
0
0 30 60 90 120 150
Pressure (bar)
(b)
Fig. (21). Effects of p on maximum permanent deformation, (a) mid-span impact (b) one quarter
span impact.
22D
42D
60D
0.8
80D
100D
120D
0.7
0.6
Ep(KJ)
0.5
0.4
0.3
0 30 60 90 120 150
Pressure (bar)
(a)
22D
42D
0.9
60D
80D
100D
0.8
120D
0.7
Ep (KJ)
0.6
0.5
0.4
0.3
0 30 60 90 120 150
Pressure (bar)
(b)
Fig. (22). Effects of p on absorbed energy, (a) mid-span impact (b) one quarter span impact.
From above figures, it was found that the internal pressure affects the
behavior of pressurized pipelines under both mid-span and one quarter
span impact in similar ways. As shown in (Figs. 20a and 20b), the
maximum impact forces yielded in the one quarter span impact are
apparently higher than those generated during the mid-span impact. This
is because that the “one quarter position” is closer to the fixture and
therefore is more rigid than the “middle position”, which leads to higher
impact force values. However, as reflected in (Figs. 22a and 22b), the
change of impact positions does not evidently alter the energy absorbed
during the impacts.
Figs. (20 to 22) also reveal that when the internal pressure increases, the
maximum impact force also increases, even such tendency is not evident
when the outside diameter is small. Thus, it can be deduced that the
internal pressure can enhance the impact resistance of the pipeline and
therefore leads to higher impact forces. On the contrary, as the pressure
increases, the impact energy absorbed by the pipelines decreases, still
such tendency is not that clear when the outside diameter of the pipelines
is small. Since the internal pressure improves the overall impact
resistance of the pipeline model, therefore after impacting a pipeline with
higher internal pressure, the indenter will rebound at a higher speed and
only small portion of impact energy will be absorbed by the pipeline itself.
These findings agree very well with the results presented in ng and shen’s
work [8], who drawn similar conclusions through a series of impact tests
and experimental analysis.
INFLUENCE OF OUTSIDE DIAMETER

Similarly, in order to study the effects of outside diameter on the pipelines’
lateral impact response, p = 0, 30, 60, 90, 120, 150 are substituted into
Eqs. (14-19) separately to create a series of analytical models with the
outside diameter as the only variable. Those analytical models are then
compared with the FEA results and the comparison results are shown in
Figs. (23 to 25).
0bar
100
30bar
90
60bar
80
90bar
70
120bar
60
Fmax (kN) 150bar
50
40
30
20
10
0
0 20 40 60 80 100 120
Diameter (mm)
(a)
0bar
120
30bar
100 60bar
90bar
80 120bar
Fmax (kN)
150bar
60
40
20
0
0 20 40 60 80 100 120
Diameter (mm)
(b)
Fig. (23). Effects of D on maximum impact force, (a) mid-span impact (b) one quarter span
impact.
0bar
40
30bar
35
60bar
30 90bar
W (mm) 25 120bar
150bar
20
15
10
5
0
0 20 40 60 80 100 120
Diameter (mm)
(a)
0bar
25
30bar
60bar
20
90bar
120bar
15
W (mm)
150bar
10
0
0 20 40 60 80 100 120
Diameter (mm)
(b)
Fig. (24). Effects of D on maximum permanent deformation, (a) mid-span impact (b) one quarter
span impact.
0bar
0.9
30bar
0.8
60bar
0.7
90bar
Ep (KJ) 0.6 120bar
0.5 150bar
0.4
0.3
0.2
0.1
0
0 20 40 60 80 100 120
Diameter (mm)
(a)
0bar
0.9
30bar
0.8
60bar
0.7
90bar
0.6 120bar
Ep (KJ)
0.5 150bar
0.4
0.3
0.2
0.1
0
0 20 40 60 80 100 120
Diameter (mm)
(b)
Fig. (25). Effects of D on absorbed energy, (a) mid-span impact (b) one quarter span impact.
At first, it was observed that just like the internal pressure, the outside
diameter affects the behavior of the pipeline models impacted at two
positions in similar ways. From above figures, it was found that the
internal pressure affects the behavior of pressurized pipelines under both
mid-span and one quarter span impact in similar ways. As illustrated in
Figs. (20 and 22), it is verified in Figs. (23 and 25) that the maximum
impact forces yielded in the one quarter span impact are higher than those
generated during the mid-span impact, and the energy absorbed in both
cases are very close to each other.
From Fig. (23a and 23b), it can be seen that as the outside diameter
increases, the maximum impact forces also increases for most internal
pressures. Exceptions were however observed at low internal pressures.
For mid-space impacts, when the internal pressure is lower than 30 bar,
the maximum impact force value goes down when the outside diameter
exceeds 100 mm. For the one quarter span impacts, the increasing
tendency of the maximum force value becomes much less apparent when
the internal pressure is 30 bar and the outside diameter passes 80 mm, and
when the internal pressure is 60 bar and the outside diameter reaches 100
mm. Furthermore, for pipelines without internal pressure, the maximum
force starts decreasing when the outside diameter reaches 60 mm. As
reflected in Fig. (25a and 25b), when the outside diameter increases, the
energy absorbed by the pipelines decreases and the decreasing tendency is
much more obvious than the decreasing caused by the increasing internal
pressure (Fig. 22a and 22b). In summary, the influence of the outside
diameter on pressurized pipelines’ impact behavior is similar to that of the
internal pressure, both of them will enhance impact resistance of the
pipeline models.
CONCLUSION
In this chapter, the lateral impact behavior of pressurized pipelines is

studied through a series of numerical simulations and analytical analyses.
72 FEA models were created and more than 300 simulations were
performed to simulate that the pressurized pipelines with different

dimensions and different internal pressures applied were struck at the
mid-span and at the one quarter span positions. Effects of impact position,
outside diameter, and internal pressure on the lateral impact behavior of
the pressurized pipelines are demonstrated based on the FEA results and
RS models, which can be described as follows:
1. The different impact positions do not evidently affect the
characteristics of lateral impact response of the pressurized
pipelines.
2. The maximum impact force yielded during the mid-position
impact is lower than that produced during the one quarter-position
impact; the permanent transverse displacements caused by
mid-position impact is apparently higher than that resulted from
the mid-position impact; while the energy absorbed during the
mid-position impact was slightly lower than that absorbed during
the one quarter-position impact.
3. When the internal pressure increases, the maximum impact force
increases, the maximum transverse displacement decreases, and
the absorbed impact energy decreases also.
4. When the outside diameter increases, the maximum impact force
increases, the maximum transverse displacement decreases, and
the absorbed impact energy decreases also.
5. The internal pressure shows less influence on the lateral impact
response of the pipelines with smaller outside diameter.
Based on the outcomes of this study, it is expected that a complete

understanding of plasticity mechanisms that govern the impact of the
pressurized pipelines can be fully obtained by implementing the proposed
research plans. In addition, the present computational and analytical
approaches will propel the research in safety and reliability of low
pressure pipelines, which are used in hydraulic system of aerospace.
this publication.
ACKNOWLEDGEMENTS
Declared none.
REFERENCES
[1] N. Jones, S.E. Birch, R.S. Birch, L. Zhu, and M. Brown, "An experimental
study on the lateral impact of fully clamped steel pipes", In: Proceedings of
the Institution of Mechanical Engineers., vol. 206, pp. 111-127, 1992.
[http://dx.doi.org/10.1243/PIME_PROC_1992_206_207_02]
[2] N. Jones, and W.Q. Shen, "A theoretical study of the lateral impact of fully
clamped pipelines", In: Proceedings of the Institution of Mechanical
Engineers, vol. 206, pp. 129-146, 1992.
[http://dx.doi.org/10.1243/PIME_PROC_1992_206_208_02]
[3] N. Jones, and R.S. Birch, "Influence of internal pressure on the impact
behavior of steel pipelines", ASME Journal of Pressure Vessel Technology, vol.
118, pp. 464-471, 1996.
[http://dx.doi.org/10.1115/1.2842215]
[4] S.Y. Shin, B. Hwang, S. Lee, N.J. Kim, and S.S. Ahn, "Correlation of
microstructure and charpy impact properties in API X70 and X80 line-pipe
steels", Mater. Sci. Eng. A, vol. 458, pp. 281-289, 2007.
[http://dx.doi.org/10.1016/j.msea.2006.12.097]
[5] B.N. Leis, R.B. Francini, R. Mohan, D.L. Rudland, and R.J. Olson,
"Pressure-displacement behavior of transmission pipelines under outside
forces towards a serviceability criterion for mechanical damage", In:
Proceedings of the 8th International Offshore and Polar Engineering
Conference, Canada, 1998, pp. 60-67.
[6] K.S. Chen, and W.Q. Shen, "Further experimental study on the failure of fully
clamped steel pipes", Int. J. Impact Eng., vol. 21, no. 3, pp. 177-202, 1998.
[http://dx.doi.org/10.1016/S0734-743X(97)00083-3]
[7] W.Q. Shen, and K.S. Chen, “An investigation on the impact performance
of pipelines", Intern. J. Crashworthiness, vol. 3, no. 2, pp. 191-209, 1998.
[http://dx.doi.org/10.1533/cras.1998.0070]
[8] C.S. Ng, and W.Q. Shen, "Effect of lateral impact loads on failure of
pressurized pipelines supported by foundation", In: Proceedings of the
Institution of Mechanical Engineers, Part E: J. Process Mech. Eng., vol. 220,
pp. 193-206, 2006.
[http://dx.doi.org/10.1243/0954408JPME97]
[9] N. Jones, and R.S. Birch, "Low-velocity impact of pressurized pipelines", Int.
J. Impact Eng., vol. 37, no. 2, pp. 207-219, 2010.
[http://dx.doi.org/10.1016/j.ijimpeng.2009.05.006]
[10] J.G. de Oliveira, T. Wierzbicki, and W. Abramowicz, "Plastic behavior of
tubular members under lateral concentrated loading", Technical Report
82-0708, Det norske Veritas, Research Division, 1982.
[11] H. Oktem, T. Erzurumlu, and H. Kurtaran, "Application of response surface
methodology in the optimization of cutting conditions for surface roughness",
J. Mater. Process. Technol., vol. 170, pp. 11-16, 2005.
[http://dx.doi.org/10.1016/j.jmatprotec.2005.04.096]
[12] Y-C. Liu, "Optimum design of thin-walled box section beams for
crashworthiness analysis", Finite Elem. Anal. Des. vol. 44, no. 3, pp. 139-147,
2008.
[http://dx.doi.org/10.1016/j.finel.2007.11.003]
[13] Y-C. Liu, “Design optimization of tapered thin-walled square tubes", Intern. J.
Crashworthiness, vol. 13, no. 5, pp. 543-550, 2008.
http://dx.doi.org/10.1080/13588260802222102]
[14] Y-C. Liu, "Crashworthiness design of multicorner thin-walled columns",
Thin-walled Struct., vol. 46, no. 12, pp. 1329-1337, 2008.
[http://dx.doi.org/10.1016/j.tws.2008.04.003]
[15] Y-C. Liu, "Crashworthiness design of thin-walled curved beams with box and
channel cross sections", Intern. J. Crashworthiness, vol. 15, no. 4, pp.
413-423, 2010.
[http://dx.doi.org/10.1080/13588261003647410]
[16] Y-C. Liu, "Thin-walled curved hexagonal beams in crashes – FEA and design",
Intern. J. Crashworthiness, vol. 15, no. 2, pp. 151-159, 2010.
[http://dx.doi.org/10.1080/13588260903094400]
[17] Y-C. Liu, "“Optimization of the crushing performance of tubular structures“,
HKIE (The Hong Kong Institution of Engineers)", Transactions, vol. 17, no. 1,
pp. 16-25, 2010.
CHAPTER 4
Effect of Bondline Thickness on the Traction-

separation Laws of Adhesively Bonded Joint
Gefu Ji1, Zhenyu Ouyang1 and Guoqiang Li1,2,*
1
University, Baton Rouge, LA 70803, USA
2
Department of Mechanical Engineering, Southern University, Baton
Rouge, LA 70813, USA
Abstract: A number of existing and emerging industrial applications are dependent

on layered substrates through adhesive bonding. The interfacial fracture of
adhesively bonded structures is a critical issue for their extensive applications to a
variety of modern industries. In the recent two decades, cohesive zone models
(CZMs) have been receiving intensive attentions for fracture problems in adhesively
bonded joints due to its fairly simple and accurate predictive ability. In CZMs the
nonlinear interfacial fracture behaviors are described by the traction-separation laws
(also referred to as cohesive laws). The cohesive laws represent the local constitutive
behavior, instead of the global parameter, such as toughness. While numerous global
tests have been conducted to measure the interfacial toughness of adhesive joints,
limited local tests have been conducted to determine the interfacial traction-
separation laws or interfacial cohesive laws. Among the limited local tests in some
recent experimental studies, very few studies have considered the effects of adhesive
thickness on the local interfacial traction-separation laws. In the present work, within
*Corresponding author Guoqiang Li: Department of Mechanical & Industrial Engineering, Louisiana State
University, Baton Rouge, LA 70803, USA; Department of Mechanical Engineering, Southern University, Baton
Rouge, LA 70813, USA; Tel/Fax: 001-225-578-5302/001-225-578-5924; E-mail: lguoqi1@lsu.edu
Yucheng Liu (Ed.)

154 Frontiers in Aerospace Science, Vol. 1 Ji et al.
the framework of nonlinear fracture mechanics, comprehensive experimental

studies are conducted to investigate the effect of adhesive layer thickness on the
local nonlinear interfacial behaviors. The fracture tests of adhesive joints with
various adhesive layer thicknesses were conducted under different fracture modes:
pure Mode-I (peel fracture), pure Mode-II (shear fracture), and mixed Mode I/II.
The experimentally determined interfacial traction-separation laws provide
valuable baseline data for parameter calibrations in numerical models. The current
experimental results may also facilitate the understanding of adhesive thickness
dependent interface fracture of bonded joints.
Keywords: Adhesive thickness, Bonded joints, Cohesive law, Cohesive

strength, Cohesive zone model, Interfacial strength, Plastic zone,
Toughness.
INTRODUCTION
Background
Adhesive Bonding Technology

The most primitive form of adhesive bonding technology can be traced
back to 200,000 BC when hominins in central Italy glued spear stone flakes
to a wood with birch-bark-tar [1]. Today adhesive bonding technology has
been used in a wide variety of modern industries, such as automotive,
aerospace, marine, construction, and military. Many components and
structures, from microchips to large aircrafts, are made of materials
arranged in layers through adhesive bonding [2].
The strength of attachment, or adhesion, between an adhesive and its

substrate depends on the mechanisms of adhesion and the surface area
over which the two materials contact. The mechanisms of adhesion can be
categorized as mechanical, chemical, dispersive, electrostatic, or diffusive.
Mechanical adhesion occurs when adhesive materials fill the voids or
pores of the surfaces and the surfaces are held together by interlocking;
Effect of Bondline Thickness Frontiers in Aerospace Science, Vol. 1 155
chemical adhesion is developed when the two materials form a compound

at the joint; dispersive adhesion is realized through the attraction between
two molecules with regions of slight positive and negative charge (i.e.,
van der Waals forces); electrostatic adhesion happens when some
conducting materials pass electrons to form a difference in electrical
charge at the joint; finally, diffusive adhesion is realized when the
molecules of both materials are mobile and soluble in each other and
merge at the joint by diffusion [3].
Three Basic Modes of Fracture
Failure of the adhesive joint imposes a major threat to the reliability of

adhesively bonded structures. There are three basic modes of fracture,
which are opening (mode I), shear (mode II), and tearing (mode III), as
shown in Fig. (1.1). A fracture is considered as Mode I when the tensile
stress is normal to the plane of the crack. When the shear stress acts
parallel to the plane of the crack and perpendicular to the crack front, the
fracture is considered Mode II. Mode III is characterized by a shear stress
acting parallel to the plane of the crack and parallel to the crack font. In
reality, however, all three modes occur simultaneously. In other words, it
is a mixed mode.
Fig. (1.1). Three basic fracture modes [4].

Classical Fracture Mechanics
During the World War I, many steel structures failed at low temperature.
English aeronautical engineer, A. A. Griffith, provided a theory to explain
the failure of brittle materials and built up the fundamental research of the
Linear Elastic Fracture Mechanics [5]. A major limitation of Griffith’s
theory is that it assumes infinite stress at the initial crack tip point as the
radius “r” approaches zero. This is impossible in reality. Therefore, in
1954, Irwin modified the Griffith theory and proposed a plastic zone
developed from the initial crack tip. As the applied load increases, the
plastic zone increases in size until the crack grows and the material
behind the crack tip unloads. The plastic loading and unloading cycle near
the crack tip leads to the dissipation of energy as heat. Hence, a
dissipative term has to be added to the energy balance relation devised by
Griffith for brittle materials. In physical terms, additional energy is
needed for crack growth in ductile materials when compared to brittle
materials [6]. However, one basic assumption in Irwin's linear elastic
fracture mechanics is that the size of the plastic zone is small compared to
the crack length. This assumption is quite restrictive for certain types of
failure in structural steels though such steels can be prone to brittle
fracture, which has led to a number of catastrophic failures [6].
In the following three decades, numerous studies had been contributed to
failure problems of bonded joints [7]. During this stage, most efforts were
focused on the classical linear elastic fracture mechanics (LEFM). The
advantage of LEFM lies in the obvious simplicity and decent accuracy,
especially for relatively brittle materials and interfaces. Within the
framework of LEFM, the remote loadings can be correlated to the critical
conditions (crack growth) by a local parameter: stress intensity factor (SIF)
or a global parameter: strain energy release rate (ERR). Despite the huge
success of LEFM, the assumption of small-scale yielding beyond the crack
tip has been identified as one of the major limitations. Specifically, with the
increased use of modern toughened adhesives, the cohesive fracture
associated with plastic zones along the adhesive interlayer, in many cases,
could be comparable to or even larger than the thickness of the adherends.
Nonlinear Fracture Mechanics
Cohesive Zone Method
Motivated by this limitation of LEFM, a number of efforts have been

made along the direction of nonlinear fracture mechanics (NLFM). In the
late 1950s and early 1960s, Barrenblatt [8] and Dugdale [9] independently
proposed the concept of cohesive zone. Specifically, the stresses across a
potential crack path were bounded, and a localized traction-separation law
may be able to describe the fracture behaviors within the cohesive
processing zone ahead of the apparent crack tip, as shown in Fig. (1.2).
Fig. (1.2). Cohesive zone at mode I test.
The stresses vary with the relative surface separations according to the
assumed interface law during the fracture process.
J-Integral
Also during this period of time, Rice [10] proposed a path independent
integral method to calculate the energy release rate (ERR), which is
referred to as the well-known J-integral:
u
J   (Wdy  T ds) (1.1)

x
where W(x, y) is the strain energy density; x and y are the coordinate
directions, T= n   is the traction vector; n is the normal to the curve or
path Γ, σ is the Cauchy stress, and u is the displacement vector [10].
This method can describe the case when there is sufficient crack-tip
deformation and the part no longer obeys the linear-elastic approximation.
Rice's analysis, which assumes non-linear elastic (or monotonic
deformation-theory plastic) deformation ahead of the crack tip, is
designated as the J integral [10]. This analysis is limited to some
situations. First, plastic deformation at the crack tip cannot extend to the
furthest edge of the loaded part. Second, the assumed non-linear elastic
behavior of the material is a reasonable approximation in shape and
magnitude to the real material's load response. Also note that the J
integral approach can be reduced to the Griffith theory for linear-elastic
behavior [10]. This method can measure the ductile material with plastic
deformation. Therefore, it can be used in this study to measure the
fracture energy of adhesive layer.
Application of CZM in Adhesive Joint
Encouraged by these two achievements, intensive studies have been

conducted to investigate the nonlinear fracture behaviors of adhesively
bonded joints in the past three decades. The1990s might be the most
rapidly developing period of CZMs. The realistic demands for nonlinear
simulation become urgent due to the wide applications of modern
toughened adhesive. With such a situation, the classical LEFM, which is
basically a single-parameter method, may have difficulties of meeting the
increased demands in the accurate fracture simulations and predictions.
Instead of the single-parameter model in LEFM, it has been reported that
two or even three parameters were needed in CZMs for the sake of
accurate modeling. For instance, these constitutive parameters could be

chosen as the fracture toughness JIC/JIIC, the cohesive strength σmax/τmax
and/or the characteristic length δ at which the cohesive tractions vanish [2,
11, 12].
Various cohesive zone models (cohesive laws) were proposed to model

the fracture process [13-18]. The main difference between these models
lies in the shape of the traction-displacement response, and the parameters
used to describe that shape. All of them start from the assumption that one
or more interfaces can be defined, where crack propagation is allowed by
the introduction of a possible discontinuity in the displacement field. One
feature of CZMs is that they can be conveniently incorporated in the
traditional finite element analysis (FEA) to model the fracture behaviors
in various materials and structures, including adhesive joints [13, 15-31].
It is worth noting that CZMs can be also incorporated in analytical
models to derive the theoretical solutions for some types of specimens or
structures with relatively simple geometries [2, 12, 32-39].
Although CZM based methods showed considerable difference, all

parameters required calibrations by test data. Importantly, due to the
effects of adhesive thickness on plastic dissipations, calibrations are
highly preferred to be conducted with different thicknesses of adhesive
layer. There might be two means for the parameter characterizations:
global method and local method. For the global method, one may
compare measured global loadings, deflections and toughnesses to those
by numerical models embedded with the parameters to be calibrated, at
different adhesive thicknesses. The global behaviors of bonded joints have
been widely tested with different adhesive thicknesses [3, 26, 40-46], to
name a few. Their study clearly showed that the fracture energy varies
with the thickness of the adhesive layer. On the contrary, fewer attentions
have been paid to the local test on the interfacial traction-separation laws
across the interlayer. However, this is also a quite important issue. First, it
may provide direct thickness-dependent equivalent interfacial laws for
simple models. Meanwhile, it can cross-calibrate the parameters required

by the “CZM+elastoplastic continuum” model. One may argue that global
tests are sufficiently good to calibrate the parameters. It is believed that
the local separation between the two adherends may be even a more direct,
rigorous and reliable means to calibrate the parameters in a numerical
model.
Several local experimental tests have been conducted on the local

interfacial traction-separation laws for bonded joints. For instance,
Søensen [47] and Andersson and Stigh [48] experimentally determined
the interfacial traction-separation laws of bonded joint under Mode-I
loadings. Andersson and Biel [49] and Leffler et al. [50] experimentally
obtained the interfacial traction-separation law of bonded joint under
Mode-II loadings. Recently, interfacial traction-separation laws were also
tested under mixed mode loadings [27]. Most recently, Zhu et al. [51]
conducted a local fracture test to extract the loading rate-dependent
traction–separation laws.
The Mode II, or in-plane interface shear fracture, loading mode is also of
particular importance for adhesive joints [33]. Mode II loading may be
induced when a cracked adhesive joint or a layered composite is subjected
to bending. The various experimental fracture mechanics approaches to
Mode II usually utilize some form of test specimen which is subjected to
applied bending loads in order to determine the values of the Mode II
toughness GIIC [52].
To measure the Mode II interlaminar fracture toughness, Russell and

Street [53] introduced the customary end notched flexure (ENF) test, a
three-point bending test. The compliance in bending is measured at
several positions of the crack front with respect to the loading pin, and the
load for the crack growth is determined for one definite crack position.
The critical interlaminar fracture toughness is computed using the linear
elastic fracture mechanics (LEFM) combined with data reduction
techniques.
One of the main problems that have been encountered with Mode II
loading has been the poor reproducibility of the values of the measured
Mode II toughness GIIC [54]. The experimental loading-unloading cycling
test conducted by Russel and Street [53] indicated a maximum error of
around 2% in GIIC in composites if friction was ignored. More recently,
Davidson and Sun [55] and Davidson et al. [56] considered the effects of
friction in the ENF specimen and 4-ENF tests were conducted on
composites. They concluded that friction accounted for only about 2%
and 5%, respectively, of the measured values of GIIC from their tests. The
effects of friction in the specimen could be a possible reason for this poor
reproducibility.
Another possible major cause of scatter and inconsistency in Mode II

toughness may be a difficulty in determining the location of the crack tip
according to some recent studies [57-59]. The difficulty in determining
the true crack length has also been observed during Mode I peel tests in
composites when extensive fiber-bridging and microcracking occurs. This
has been shown to cause variations and errors in the calculation of
toughness when corrected beam theory was employed [58].
Additionally, numerous studies in the past several decades revealed an

important experimental phenomenon: the global fracture behaviors are
dependent on the loading modes [3, 19-21, 27, 41]. For instance, the
toughness of Mode-I is usually much less than that of Mode-II, while
those of mixed modes are between them depending on mode mixities.
Although there are still arguments regarding the mechanism of fracture
mode dependence, one thing for sure is that the independent fracture tests
under pure Mode-I and Mode-II loading conditions will facilitate the
understanding of the mode dependent phenomenon by eliminating the
coupling of stress and plastic dissipation.
Some local tests have been conducted to measure the critical CZM
parameters. For instance, with J-integral based form, the local interfacial
traction-separation laws were experimentally tested under pure Mode-I
and pure Mode-II loading conditions [47, 48, 50]. It is noted that these
studies focused on the interface fracture of bonded joints with identical
adherends through standard fracture test methods (for instance, double
cantilever beam (DCB) and end notched fracture (ENF). Recently, an
experimental study was also conducted to investigate the local mixed
mode interfacial constitutive behaviors [27]. Most recently, Zhu et al. [51]
conducted a local fracture test to extract the loading-rate dependent
traction–separation laws. These recent efforts on the local test of interfacial
traction-separation and traction-slipping laws across the adhesive interlayer
provided valuable information.
Effects of Adhesive Thickness
Interfacial laws became dependent on the geometries, such as thickness of

the adherend and adhesive. Because they can affect the size of plastic
zone and the magnitude of plastic strain, they change the equivalent
interfacial traction-separation laws.
In load-bearing structures, the adhesive layers are usually weaker than the
adherends. Therefore, the adherends can be treated as rigid constraints on
the adhesive layer. This means that the region where a crack can
propagate is almost in a state of prescribed deformation [49]. During
crack propagation, a zone where the adhesive is damaged precedes the
crack tip. This zone is denoted as the damage zone. With a tough
engineering adhesive, it is substantially longer than it appeared in a bulk
test and typically several times larger than the thickness of the layer [49].
Thus, fracture behavior depends on the thickness of the adhesive layer.
The fracture energy for a specific adhesive can reach a maximum value at
a certain layer thickness [3, 40]. Therefore, the effect of adhesive
thickness is a very important parameter to design the safety factor of
structure.
However, this parameter still lack of sufficient research and attention.

These recent efforts on the local test of interfacial traction-separation laws
across the adhesive interlayer provided valuable information. Most of
them were conducted with a fixed thickness of adhesive layer. As
discussed before, in order to more efficiently calibrate the parameters, the
local traction-separation laws tests may be desired with different adhesive
thicknesses. However, very few tests have been conducted on the
thickness-dependent local interfacial traction-separation laws. Among the
very few local tests regarding effects of adhesive thickness, Kafkalidis et
al. [43] investigated the interfacial fracture by considering three different
thicknesses of adhesive layer and Søensen [47] tested two thicknesses.
Research Objectives
The purpose of the present work is thus to conduct a systematic study on

the effects of adhesive thickness and obtain an entire picture about its
effects by using Double Cantilever Beam (DCB) test, End Notched
Flexure (ENF) test, and Single Leg Bending (SLB) test with different
thicknesses of adhesive layer. Not only its effects on the fracture energy,
but also its effects on the local interfacial traction-separation laws are
investigated.
In order to reduce the workload, the thickness of the adherend remains

identical in this work. The thickness of steel adherends is designed to be
relatively thick to prevent from any plastic deformation. Thus, all the
nonlinear behaviors are limited in the adhesive interlayer. For the cases
that plastic deformations are involved in the adherends, one may refer to
the studies [22, 23, 60]. Both local and global test are conducted
simultaneously. With these test results, the dependency of the interface
strength and the shape of the constitutive laws on adhesive thickness can
be revealed. The current test results may provide valuable test data for the
research communities and facilitate the characterization of the parameters

required by the CZM based numerical methods discussed before.
THEORETICAL BACKGROUND
Current design approaches for predicting the fracture of adhesively-bonded

materials are still somewhat empirical, and improving these approaches is a
critical issue for furthering the engineering applications of this technology.
Before the physical macro-crack is formed, these two surfaces are held
together by traction within a cohesive zone. The interfacial stresses vary
according to the relative displacement of the surfaces, and an interface
cohesive law describes the activities in the cohesive zone in terms of the
separation and the traction of the interface to be formed during the fracture
process. The fracture process of most layered or bonded structures is
commonly under the control of mixed mode cracking combined with mode
I and mode II component. Under mixed mode cracking, the failure process
zone is subjected to both normal (mode I) and shear stresses (mode II). In
order to investigate the mixed mode interface fracture of bonded joints,
pure Mode I and pure Mode II fracture will be studied, respectively.
This chapter focuses on developing a theoretical model which is able to

characterize the nonlinear interface laws for pure Mode I, pure Mode II and
mixed mode interface fracture test. With this developed model, the effects
of adhesive thickness on global and local interface fracture behaviors can
be revealed under different loading conditions.
In this theoretical model, it is assumed that the crack opens in a monotonic

fashion; neither unloading nor crack closure is permitted and the cohesive
(or bridging) laws are assumed to be the same for each point along the bond
length. Since the cohesive stresses represent the failure process zone, it may
be also reasonable to assume that a certain crack opening exists, at which
the cohesive stresses vanish when the failure criterion is satisfied. The
classical beam theory is adapted for the adherends in adhesive joints.
Meanwhile, it is assumed that the thickness of adhesive layer is very thin

compared to that of the adherends. It is also assumed that the bond length is
sufficiently long for all adhesively bonded joints.
Theoretical Derivations
Let’s consider an infinitesimal section of two adherends bonded with a thin

adhesive layer as shown in Fig. (2.1). It is noted that this infinitesimal
section is located within the bonded length. With the classical beam theory,
the displacements field of the two beams U(x, z) and W(x, z) have the form
(each adherend is modeled by a beam),
dw1 ( x) dw2 ( x)
U 1 ( x1 , z1 )  u10 ( x)  z1 ; U 2 ( x, z 2 )  u 20 ( x)  z 2 (2.1)
dx dx
W1 ( x, z1 )  w10 ( x) ; W2 ( x, z 2 )  w20 ( x) (2.2)
where subscripts i=1, 2 correspond to the upper and lower adherends,
respectively; coordinate x and z correspond to longitudinal and vertical
direction of adherends; ui0 represents the longitudinal displacements of the
neutral axis in the adherends; wi 0 is the vertical displacements of the
neutral axis in the adherends.
Fig. (2.1). Free-body diagram of an infinitesimal section of the bonded joint.

The longitudinal displacement of the bottom fiber of the upper beam u1

and that of the top fiber of the bottom beam u2 can be described as follows,
respectively,
h1 dw1 ( x) h2 dw2 ( x)
u1  u10  ; u 2  u 20  (2.3)
2 dx 2 dx
where w1 represents the vertical displacement of the bottom fiber of the
upper beam; and w2 is the vertical displacement of the top fiber of the
lower beam.
According to the classical beam theory, w1 = w10, and w2 = w20.

Associated with Eq. (2.2), obviously, the normal interface separation w
(normal relative displacement) between the bottom fiber of the upper
adherend and top fiber of the lower adherend can be defined by
w  w1  w2 (2.4)
And,
dw dw1 dw2
 
dx dx dx (2.5)
According to Eq. (2.3), the relative tangential sliding δ (interface cohesive
slip) between the upper adherend and lower adherend can be expressed by
h1 dw1 h2 dw2
  u1  u 2  (u10  u 20 )  (  ) (2.6)
2 dx 2 dx
The constitutive equations are written by
dui 0 d 2 wi
N i  Ai ; M i  Di (2.7)
dx dx 2
in which
E xi hi E xi hi3
Ai  ; Di 
1  v 2zxi 12(1  v 2zxi )
Ni and Mi are the axial force and bending moment per unit width of the
beam i (i=1, 2), respectively; Ai and Di are the axial and bending stiffness
of the beam i (i=1, 2) per unit width under the plane strain condition.
The equilibrium equations of each beam within the bonded region are
written in the conventional way as
dN1 dN 2
  ( x) ;   ( x) (2.8)
dx dx
dQ1 dQ2
  ( x) ;   ( x) (2.9)
dx dx
dM 1 h dM 2 h
 Q1  1  ( x) ;  Q2  2  ( x) (2.10)
dx 2 dx 2
where σ(x) and τ(x) are the interface normal and shear stress, respectively;
N1 and N2 are the internal axial forces in beam 1 and 2, respectively; M1
and M2 are the internal bending moments in beam 1 and 2, respectively;
Q1 and Q2 are the internal shear forces in beam 1 and 2, respectively (see
Fig. 2.3). The resultant forces and moment in the two beams are denoted
as follows:
N1  N1  N T ; Q1  Q2  QT ; M1  M 2  M T (2.11)
By substituting Eq. (2.7) into the derivative of Eq. (2.6), it can be derived
that
h N h N (2.12)
    1 M1  1  ( 2 M 2  2 )
2 D1 A1 2 D2 A2
With Eq. (2.7) and note that w=w1-w2, we have follows:
M1 M 2
w  w1  w2  (  ) (2.13)
D1 D2
1 dM 1 1 dM 2
w  w1  w2  (  ) (2.14)
D1 dx D 2 dx
With Eqs. (2.4) through (2.14), the following two equations can be
derived that,
h1 D2  h2 D1 h h 1 1 (h  h ) 2
   w  1 2 QT  [(  )  1 2 ] (2.15)
2( D1  D2 ) 2( D1  D2 ) A1 A2 4( D1  D2 )
h1 h2 h12 h22 1 1
   ( Q1  Q2 )  (    )  (2.16)
2D1 2D2 4D1 4D2 A1 A2
By further taking the derivative of both sides of Eqs. (2.15 and 2.16), the
follows can be derived, respectively,
h1 D2  h2 D1 ( 4) h  h dQT 1 1 (h  h ) 2
w  1 2  [(  )  1 2 ]     (2.17)
2( D1  D2 ) 2( D1  D2 ) dx A1 A2 4( D1  D2 )
h1 h h2 h2 1 1
(  2 )  ( 1  2   )       (2.18)
2D1 2D2 4D1 4D2 A1 A2
The resultant shear force QT=Q1+Q2 in the two adherends can be
determined by global analysis with the given boundary conditions. For the
sake of simplicity, in the current theoretical model, we only consider the
conditions that QT is constant or piecewisely constant.
Therefore, the derivative of QT (or dQT/dx) in Eq. (2.17) becomes zero.

The division of Eq. (2.17) by Eq. (2.18) gives
 
1 
d 4w
DT  4      ( w) (2.19)
dx  
2 

where
1 1 (h1  h2 ) 2 h2 h2 1 1
DT 
D2 D1 ; 1    ; 2  1  2   (2.20)
( D1  D2 ) A1 A2 4( D1  D2 ) 4D1 4D2 A1 A2
Governing Equations
In this dissertation, two adherends materials: steel and glass fiber

composites, are considered. However, the adherends in the adhesive joints

have identical thickness, width and material, although the thickness of
adhesive layer may vary.
Therefore, we have the follows for all conditions:

A1  A2 ; h1  h2 ; D1  D2 (2.21)
By substituting Eq. (2.21) into Eq. (2.20), it is not difficult to verify that
1   2 (2.22)
For the sake of clarify, we denote that
h1  h2  h ; A1  A2  A ; D1  D2  D ; h1 D1  h2 D2  h D (2.23)
By inserting Eqs. (2.22) and (2.23) into Eq. (2.15), we have
h  2 h2 
   QT       (2.24)
2D  A 2D 
By substituting Eqs. (2.22 and 2.23) into Eq. (2.19), we obtain
D d 4w
   ( w) (2.25)
2 dx 4
It is worth noting that Eq. (2.24) represents the governing equation of
interfacial shear behaviors, while Eq. (2.25) represents the governing
equation of interfacial peel behaviors. It is also noted that if only: (i) the
resultant shear force QT in the two adherends is a constant or piecewisely
constant within the bonded length; and (ii) the two adherends are
identical.
J- Integral for Interface Shear Behaviors
Strictly, the adhesive material must be nonlinearly elastic during the test.
However, for a monotonic loading process (no unloading occurs), the
cohesive separation as well as plastic dissipation in the adhesive layer
might still be considered by the well-known path independent integral or

J-integral as follow [10]:
u
J   (Wdy  T x ds)

(2.26)
where W(x, y) is the strain energy density; x and y are the coordinate
directions, T=n  σ is the traction vector; n is the normal to the curve or path
Γ, σ is the Cauchy stress, and u is the displacement vector.
Let us consider an arbitrary tangential cohesive law τ=τ(δ). Note
Q1+Q2=QT is a constant along the x coordinate. Let us integrate both sides
of the governing Eq. (2.24) as follow:
h  2 h2 
  d  2D
  QT d        ( )d
 A 2D 
(2.27)
Eq. (2.27), which is the equivalent integral form, can be rewritten by
  h  2 h2 
( 
 )d 
2D
 QT  d        ( )d
 A 2D 
(2.28)
By integrating   with respect to δ from 0 to δ, it can be derived that,
1 h  2 h2  δ
[( ) 2  ( m ) 2 ]   QT  (   m )        ( )d (2.29)
2 2D  A 2D  δm
where δʹm represents the derivative of δ=0, and δm represents the value of
δ when dδ/dx=0.
It should be noticed that the integral limits on both sides of Eq. (2.29)
must be corresponded to each other. However, when the joint’s overlap
length and the length between the support and concentrated middle load
are sufficiently long relative to the cohesive zone, we may approximate
the solution with δm→0 and δʹm →0.
Therefore, for the specimens with sufficiently long bond length, by

denoting the tangential cohesive slip at the crack tip by δ0, Eq. (2.29) can
be reduced to,
1 h  2 h 2  δ0
( 0 ) 2   QT   0        ( )d (2.30)
2 2D  A 2D  0
Consequently, the mode II energy release rate JII can be written by

1 h
( 0 ) 2   QT   0
J II ( 0 )  
δ0
 ( )d  2 2 D (2.31)
0 2 h2

A 2D
The value of δʹ0 at the crack tip can be readily determined as a function of
the external forces, bending moments through global analysis as described
by Eq. (2.12). In other words, once the crack tip slip δ0 is experimentally
measured, the mode II energy release rate JII is then determined.
J- Integral for Interface Peel Behaviors
Similar to the discussion in interface tangential behavior, let us consider

an arbitrary normal cohesive law σ=σ(w). It can be observed that the
governing Eq. (2.25) is equivalent to the equation as follow,
d 3 w dw 1 d 2 w 2 2 w
  ( )     ( w)dw  0 (2.32)
dx 3 dx 2 dx 2 D 0
One may verify the equivalence between Eqs. (2.25 and 2.32) by taking
derivative of Eq. (2.32) with respect to x and then comparing it to Eq.
(2.25).
Let us denote the beam’s rotations by θi (i=1, 2) for adherend 1 and

adherend 2, respectively. The relative crack tip rotation between the two
beams is denoted by θ0=θ10-θ20. Meanwhile, the crack tip normal opening
is correspondingly denoted by w0. Obviously, the crack tip relative
rotation θ0=−dw0/dx. Simultaneously, by expressing d2w0/dx2 and dw03/dx3

in terms of the bending moments and the shear forces at the crack tip, one
may see that
w0 1 D1 D2 M 1 M 2 2 D1 D2 P P
JI  
0
 ( w)dw   [
2 D1  D2 D1

D2
]  ( 1  2 ) 0
D1  D2 D1 D2
(2.33)
Since D1=D2=D, we may further obtain

w0 1 1
J I    ( w)dw  [ M 1  M 2 ]2  ( P1  P2 ) 0 (2.34)
0 4D 2
Determination of Interface Cohesive Laws
It is important to note that the interfacial separation w and δ represent the

normal opening and tangential slip between the bottom fiber of the upper
adherend and the top fiber of the lower adherend. A typical nonlinear
equivalent cohesive law is shown in Fig. (2.2). For most CZMs in the
literature, the traction-separation laws are such that with increasing
interfacial separation, the traction across the interface reaches a maximum,
then decreases and eventually vanishes. This typical nonlinear
separation-traction law has three segments: (a) elastic stage when the
normal interfacial separation δ≤δ0. The normal interfacial stress σ
increases with separation until the maximum interfacial stress σmax
(interfacial strength) is reached; (b) softening stage when δ0≤δ≤δf. The
normal traction σ decreases with separation δ; and (c) complete debonding
stage. There is no interfacial stress when δ≥δf. By conducting the fracture
tests with different adhesive thicknesses, the effects of adhesive layer on
the constitutive laws, such as the interfacial strength, shape of the
interfacial traction-separation law, and fracture energy can be revealed.
 1 1 
  [ M 1  M 2 ] 2  ( P1  P2 ) 0 
J I ( w)  4D 2 
 ( w)   (2.35)
w w
1 h 
 (  ) 2   QT   0 
2 0 2D 
 2 
 2 h 
  
J II ( )  A 2 D 
 ( )   (2.36)
 
We shall discuss more details regarding how to measure the cohesive laws
in the next section. Meanwhile, it is noted that this interface constitutive
relationship is the equivalent interface cohesive law, not necessarily the
intrinsic cohesive law. This is because in addition to the intrinsic cohesive
separation, possible plastic deformations in the adhesive layer contribute
to the entire separation between the two adherends during the fracture test.
Fig. (2.2). A typical nonlinear interfacial traction-separation law.
Fracture Tests by DCB, ENF and SLB Specimens
By this point, the general theoretical models have been derived for the
interface peel behavior or normal separation (Mode-I fracture) and the
interface shear behavior or tangential slip (Mode-II fracture). It is worth
noting that the derived model can be also applied to the in-plane mixed
mode fracture (Mode I/II fracture) only if the two adherends are identical.
However, as mentioned before, the specific formula might be different for
different fracture test configurations due to the different boundary
conditions. Therefore, in the work, three common standard fracture test
methods: double cantilever beam (DCB), end notched flexure (ENF) and
single leg bending (SLB) are adopted for Mode-I, Mode-II and in-plane
mixed Mode I/II, respectively.
DCB Specimens for Mode-I Fracture Test
DCB specimens have been widely adopted to investigate the pure Mode-I
interface fracture of bonded joints in previous studies [2, 3, 16-18, 21, 26,
41-48, 61].
Consider a typical double cantilever beam (DCB) specimen as shown in Fig.
(2.3). It is assumed that the adherends are linearly elastic during the entire
fracture test process. A thin adhesive layer is used to bond the two
adherends.
P
h
ha
h
P
a L
Fig. (2.3). Schematic of DCB test specimen.
Obviously, with the DCB configuration, the bending moments M1 and M2 at

the cross section of crack tip are M1=Pa and M2= -Pa, respectively (note
P1=-P2=P). Note that the relationship between the relative rotation and each
beam’s rotation at the crack tip is θ0=2θ10=-2θ20. Therefore, Eq. (2.34) can
be rewritten for DCB specimen as follow:
w0 P2a2
J I    ( w)dw   2 P10 (2.37)
0 D
Further rewriting Eq. (2.37) as follow:
w0 Pa 2
J I    ( w)dw  2 P  [  10 ] (2.38)
0 2D
The first term (Pa2/2D) in the bracket reflect the variation of rotation in
each beam, the sum in the bracket can be replaced as follow:
Pa 2
  10   1P (2.39)
2D
where θ1P is the upper beam’s rotation at the cross section of loadline. In
addition, we may denote the lower beam’s rotation at the cross section of
loadline by θ2P. One may readily see that θ1P =-θ2P. We further introduce a
concept: relative rotation θP, which reflects the relative rotation between
the two beams at the cross section of loadline.
Obviously, for the DCB specimen, one can see

 P  21P  2 2 P (2.40)
Finally, with Eq. (2.38) through (2.40), we have
w0
J I ( w0 )    ( w)dw  P   P (2.41)
0
Eq. (2.41) indicates that the energy release JI for pure Mode-I fracture of
DCB specimen equals the product of peel force P and relative rotation θP
between the two beams at the loadline.
According to Eq. (2.35), the interfacial normal stress σ of DCB specimen
(at the crack tip) can be determined as follow:

J I ( w0 ) P   P  (2.42)
 ( w)  
w0 w0
Eq. (2.42) indicates that once the crack tip opening w0 and JI are
simultaneously measured (or JI-w0 curve is measured), the interface
normal stress σ is experimentally determined.
ENF Specimens for Mode-II Fracture Test
End notched flexure (ENF) specimens have been widely applied to

investigate the pure Mode-II interface fracture of bonded joints [35, 48].
Consider a typical end notched flexure (ENF) specimen as shown in Fig.
(2.4). It is assumed that the adherends are linearly elastic during the entire
fracture test process. A thin adhesive layer is used to bond the two
adherends.
Fig. (2.4). Schematic of ENF specimen test.
Obviously, with the ENF configuration, the resultant shear force QT in the
section from left support to the mid-span loading P is a constant. Let’s
assume that the shear forces in the upper and lower beam at the cross
section of crack tip are Q1 and Q2, respectively. Evidently, Q1 and Q2 will
not vary within the region from the left support to the cross section of the
crack tip.
Since the axial forces N1=N2=0 at the cross section of crack tip, with Eq.
(2.12), h1/D1=h2/D2=h/D, the parameter δʹ0 in Eq. (2.31) can be expressed

for ENF specimens as follow:
h h
   （M 1  M 2）  MT (2.43)
2D 2D
Note that the resultant MT at the cross section of crack tip is MT = M1 + M2
= QT .a (QT is also equal to the support reaction force, and QT=P/2).
Associated with Eq. (2.31), the energy release rate JII of ENF specimen can
be expressed by,
1 ha P 2 h P
δ0
(  )   0
J II ( 0 )    ( )d  2 2 D 2 2 D 2 (2.44)
0 2 h2

A 2D
According to Eq. (2.36), the interfacial shear stress τ of ENF specimen (at
the crack tip) can be determined as follow:
 1 ha P 2 h P 
 (  )   0 

 2 2D 2 2D 2
2 
 2 h 
  
J II ( 0 ) A 2 D
 ( )   (2.45)
 0  0
Eq. (2.45) indicates that once the crack tip slip δ0 and JII are
simultaneously measured (or JII-δ0 curve is measured), the interface shear
stress τ (δ0) is experimentally determined.
SLB Specimens for Mixed Mode I/II Fracture Test
Single leg bending (SLB) specimens have been widely applied to

investigate the in-plane mixed Mode I/II interface fracture of bonded joints
in previous studies. Let’s consider a typical SLB specimen as shown in
Fig. (2.5). It is noted that for the SLB specimen, both Mode-I fracture
component and Mode-II fracture component exist. Therefore, we will give
the formula for JI and JII, as well as the corresponding formula for interface
normal stress σ and shear stress τ, separately.
Fig. (2.5). Schematic of SLB specimen test.
Note that for the SLB specimen, the resultant MT at the cross section of
crack tip is still MT = M1 + M2 = QT .a = Pa/2 (QT is still equal to the
support reaction force, and QT=P/2), although M1 = Pa/2 and M2=0,
respectively. The energy release rate component JII of SLB specimen can be
expressed by,
1 ha P 2 h P
δ0
(  )   0
J II ( 0 )    ( )d  2 2 D 2 2D 2 (2.46)
0 2 h2

A 2D
Similarly, the interfacial shear stress τ(δ0) of SLB specimen (at the crack
tip) can be determined as follow:
 1 ha P 2 h P 
 (  )    0 

 2 2D 2 2D 2
2 
 2 h 
  
J II ( 0 ) A 2D
 ( )   (2.47)
 0  0
In order to determine the energy release rate component JI of SLB

specimen, let’s recall Eq. (2.34). Meanwhile, for SLB specimen, M1 = QT .a
Pa/2, M2=0; P1=QT=P/2, P2=0. Therefore, Eq. (2.34) can be rewritten by
w0 1 1
J I    ( w)dw  [QT a]2  (QT ) 0 (2.48)
0 4D 2
With the similar approach to the above section, the lower beam’s rotation
at the crack tip is equal to that at the loadline (θ20=θ2P) because there are
no bending moment in the lower beam within the region from left support
to the crack tip. Thus the Mode-I energy release rate component JI can be
expressed as follow:
QT QT a 2 Q Q a2
JI  [   0 ]  T  [ T  10   20 ]
2 2D 2 2D
2
Q Q a Q (2.49)
 T  [ T  10   2 P ]  T  [1P   2 P ]
2 2D 2
QT
  P
2
where θP is the relative rotation between the two beams at the loadline. It
is worth noting that in a real experimental SLB, it is not convenient to
measure the rotation of lower beam’s rotation at the loadline due to the
presence of the support, neither at the crack tip. However, one may simply
choose any location between the crack tip and support where is more
convenient, since the lower beam’s rotations within this region are
identical.
Since P=2QT, with Eq. (2.49), Mode-I energy release rate component JI
can be expressed as follow:
w0 P
J I    ( w)dw   P (2.50)
0 4
According to Eq. (2.35), the interfacial normal stress σ of SLB specimen

(at the crack tip) can be determined as follow:
1 
 P  P 
J ( w )
 (w)  I 0   
4
(2.51)
w0 w0
Eqs. (2.46, 2.47, 2.50 and 2.51) indicate that once the crack tip slip δ0,
loading force P, and the rotation of the two adherends are simultaneously
recorded, JI and JII can be calculated (or JI-w0 and JII-δ0 curves can be
obtained). Then, the interface shear stress τ (δ0) and interface normal
stress σ (w0) can be experimentally determined, simultaneously.
EXPERIMENTAL METHOD
Raw Material
Adhesive Material
The adhesive, LOCTITE Hysol 9460, is a modified structural epoxy

adhesive. This material has high peel strength, good impact resistance,
and good fatigue resistance. The mix ratio of resin and hardener is 1:1 by
weight. According to the manufacturer, its elastic modulus, tensile
strength and elongation at break are 2.76 GPa, 30.3 MPa and 3.5%,
respectively. The glass transition temperature is 68oC.
Adherends Material
General purpose 1018 low carbon steel bars with yield strength of 372.3
MPa and elastic modulus of 209 GPa were used to fabricate metal based
adherends of Double Cantilever Beam (DCB), End Notched Flexure (ENF),
and Single Leg Bending (SLB) specimens. Before the adhesive was applied,
the surface of the adherends were grounded by sand paper, and then cleaned
by using acetone.
Fabrication of Specimen
There are two different methods to manufacture the DCBs. One is that the
adherends are cut before the adhesive curing processes. The other one is
that the two adherends with adhesive experience the curing processes and
then are cut to small DCB specimens. The major problem of this method is
that during the milling and cutting processes, the adhesive layer could be
damaged by metal chips. Therefore, this study cut the adherends before
bonding them with adhesive.
The thickness of the adherends was measured before the adhesive was
applied and compared with the thickness of the specimen after curing. In
order to obtain different and desirable thickness of the adhesive layer, six
different thickness feeler gage inserts were inserted in between the two
adherends. The feeler gages can also control the edge of the adhesive and
help in keeping the adherends parallel. It is also noted that the surface of the
feeler gauges were sprayed with a super-thin layer of mold release agent so
that the feeler gauges could be easily removed after the curing.
A very thin mylar tape with thickness of 0.035 mm was carefully inserted
from the edge of adhesive layer by 2 mm at the middle height of the
adhesive layer immediately after the application of the adhesive layer to
create a sharp initial crack, as shown in Fig. (3.1).
Fig. (3.1). microscope image shows the inserted 0.035 mm thick mylar sheet located at the middle
plane of the adhesive layer.
The prepared specimens were then pressed by the same weights and cured
for 24 hours at room temperature. The specimens were further put into an
oven at 60.5oC (141oF) for 1 hour for post-curing. After that, they were
cooled down to room temperature before test.
Steel Based DCB Specimen for Mode I Test
The actual average adhesive thicknesses of the six groups of specimens

were 0.09 mm, 0.2 mm, 0.4 mm, 0.6 mm, 0.8 mm, and 1.0 mm. After
cutting, a hole (11.11mm in diameter), which was 50.8 mm from one end of
the adherend, was drilled coaxially in the upper and lower adherends and
coaxially machined threads inside the hole to apply the peel load using our
specially designed loading fixture as will be discussed later. Once the
surface of the adherends was cleaned by using acetone, the adherends were
ready for preparing the DCB specimens, as shown in Fig. (3.2).
Fig. (3.2). DCB specimen bonded with polished/cleaned adherends and uniform thickness of
adhesive layer through standard feeler gauge.
Table 3.1. Geometry of steel based adherends of specimens in mode I test.
Group Average Adhesive Length Height Width Initial Crack

Number Thickness (mm) (mm) (mm) (mm) Length (mm)
1 0.09 254.10 6.44 25.41 52.12
2 0.2 254.11 6.55 25.44 52.22
3 0.4 254.09 6.75 25.42 52.18
4 0.6 254.08 6.95 25.39 52.44
5 0.8 254.10 7.15 25.46 52.37
6 1.0 254.08 7.35 25.37 52.11
Since the distance between the loadline and the edge of the adhesive layer is
50 mm, the total initial crack length “a” is 52mm. The geometry
information of each group of specimens was given in Table 3.1. After the
adhesive was applied, a steel pin with the same diameter as the threaded
hole was inserted into the hole to hold the adherends in place.
Steel Based ENF Specimen for Mode II Test
In the current study, low carbon steel bar were used to fabricate the 9.5mm
thick, 25.4mm wide and 609.6 mm long adherends of enf specimens, as
shown in Fig. (3.3). A total of five groups of specimens were numbered
from group 1 to group 5 for the average adhesive thicknesses of 0.1 mm,
0.2mm, 0.4mm, 0.6mm, and 0.8mm, respectively. Each group had 3
effective specimens, and a total of 15 effective specimens were fabricated.
The specimen geometries were given by Table 3.2.
Table 3.2. Geometry of steel based adherends of specimens in mode II test.
Group Average Adhesive Length Height Width Initial Crack

Number Thickness (mm) (mm) (mm) (mm) Length (mm)
1 0.1 609.61 19.16 25.41 103.62
2 0.2 609.66 19.26 25.44 103.11
3 0.4 609.62 19.46 25.42 103.35
4 0.6 609.66 19.66 25.39 103.56
5 0.8 609.65 19.86 25.46 103.19
After curing, the wire shape feeler gages with relevant thickness were
inserted to the end of the specimen. Therefore, during the three-point
bending test, the upper and lower adherends can rotate with the same angle.
Steel Based Single Leg Bending Specimen for Mixed Mode I/II Test
In the current study, steel based adherends has the same geometry as
laminated composite based adherends. A total of five groups of specimens
were numbered from ha=0.1mm to ha=0.8mm with the average adhesive
thicknesses of 0.1 mm, 0.2mm, 0.4mm, 0.6mm, and 0.8mm, respectively.
In order to consider the effect of the initial crack length, three initial crack
length were selected from a=20mm, 50mm, and 80mm. Specimen design
was shown in Fig. (3.4).
For each different initial crack length, 5 groups of specimens were

prepared. Each group had 3 effective specimens, and a total of 45
effective steel based specimens and a total of 45 effective laminated
composite based specimens were fabricated, geometry of slb specimens
was provided in Table 3.3.
Fig. (3.3). Schematic of steel based adherends of ENF specimen.
Table 3.3. Geometry of SLB specimens in Mixed Mode I/II test.
Average Initial Initial Initial

Adhesive Length Height Width Crack Crack Crack
Thickness (mm) (mm) (mm) Length Length Length
(mm) (mm) (mm) (mm)
0.1 254.0 12.80 25.4 20.0 50.0 80.0
0.2 254.0 12.90 25.4 20.0 50.0 80.0
0.4 254.0 13.10 25.4 20.0 50.0 80.0
0.6 254.0 13.30 25.4 20.0 50.0 80.0
0.8 254.0 13.50 25.4 20.0 50.0 80.0
Fabrication of Self-Aligned Ball Pin
In order to maintain coaxial peel force during the testing, two self-aligned,
free-rotating ball pins were designed and fabricated using tool-grade steel,
as schematically shown in Fig. (3.5).
Fig. (3.4). Schematic of single leg bending test.
Fig. (3.5). Schematic of self-aligned free-rotating ball joint.
One end of the ball pin with threads was mated with the prefabricated
threads within the holes in the test specimen, and the other end was
connected with the MTS machine. The ball pin had an ability to rotate 360o
in the XZ plane and 30o to -30o in the XY and YZ planes.
Instrumentation and Test Method
The MTS 810 machine was used to conduct the test and collect the loading
force data “P” and the displacement “Δ” of the DCB specimens at the
loading point. The fracture test was conducted under the displacement
controlled mode. The loading rate was set as 1 µm/s and the data collecting
frequency was 1 Hz.
In order to measure the rotating angle “θ” of the adherends� during the�peel
test, a digital inclinometer and sensor were attached at the free end of the
adherends to collect the data during the test. The accuracy of the
inclinometer is 0.01o. The test range is from -70o to 70o. The data
acquisition frequency is 1 Hz.
Sony XCD-CR90 High resolution CCD camera with a resolution of

3.7×3.7 µm/pixel was used in this experiment. The position of the camera
was adjusted to be perpendicular to the side of the test specimen and the
deformation images of test specimen during the test was shot, with focus on
the adhesive layer. The camera shooting rate was 1 Hz. The collected
images were input to an image processing toolkit, ImageJ, to post-analyze
the recorded images and thus obtained the local separation of the crack tip
“δ”. The Mode I, Mode II, and Mixed Mode (Mode I/ Mode II) test setup
was shown in Figs. (3.6 - 3.8), respectively.
Fig. (3.6). DCB specimen attached with inclinometer during the peel test.
Fig. (3.7). ENF specimen during bending test.
Fig. (3.8). Mixed Mode I/II fracture test of two adhesively bonded steel joint with mini inclinometer
and high resolution CCD camera.
RESULTS AND DISCUSSION OF MODE I TEST
This section includes two parts: one is the results and discussion based on
the Mode I test data of steel adherends DCB specimens; the other is based
on the data of laminated composite DCB Specimens.
Results and Discussions of Mode I Test
Global Test Results
The loading rate (at loading point) was equal to 1 µm/s. The preliminary
studies showed that this rate could lead to a very stable descending branch
when the crack was propagated, and it could be safely treated as a
quasistatic loading condition. In the current study, a total of six groups of
specimens which were numbered from group 1 to group 6 for the average
adhesive thicknesses of 0.09 mm, 0.2mm, 0.4mm, 0.6mm, 0.8mm, and
1mm, respectively, were prepared and tested. Each group had 5 effective
specimens, and a total of 30 effective specimens experienced the double
cantilever beam (DCB) test by using the MTS 810 machine. The geometric
parameters of each specimen have been given in Table 3.1. Based on the
analytical results of Ouyang and Li [2], with the current geometric
configuration and the experimental data, it was found that all adherends
(tool-grade steel) remained linear elastic during the entire test process
without any plastic deformation. This indicates that all energy dissipations
were contributed by cohesive fracture and the plastic dissipations in the
adhesive layer. A typical force vs. displacement curve (a specimen in group
2 with an adhesive thickness of 0.2 mm) at the loading point is shown in Fig.
(4.1).
The force linearly increased until the local damage occurs. The load
nonlinearly increases until the crack initiation process is completed. After
that, the crack tip starts propagating. A continuous drop in the peel force
was seen as the crack continuously propagated until the crack reached the
DCB specimen’s end.
Fig. (4.1). A typical relationship between loadline displacements Δ and loadline peel force P for
Group 2 (ha=0.2mm).
Although there was no TEM, AFM or spectroscopy (XPS) examination of

the morphology of fractured surfaces, with the help of high resolution
microscope, it was found that the failure mode is cohesive fracture for all
cases. Even for the thinnest adhesive layer of 90µm, an extremely thin
adhesive layer was remained on the surfaces of the two separated
adherends. It was thus believed that even for the thinnest layers, failure
mode is still in the form of cohesive fracture instead of adhesive failure.
This failure mode may be because careful surface cleaning and treatment
were conducted during the specimen preparation associated with the good
bonding between the adhesive and adherends.
The data of the rotation angel at the loading point were collected by the
inclinometer. A typical experimental curve between the rotation angle θP at
the loadline and the displacement Δ at the loadline (a specimen in group 2
with thickness of 0.2 mm) is shown in Fig. (4.2). With Eq. (2.41), the
experimental energy release rate J is determined by combining the
measured θP and P (or combining Figs. 4.1 and 4.2). A typical relationship
(a specimen in group 2 with thickness of 0.2 mm) between global energy
release rate J and loadline displacement Δ is given in Fig. (4.3). In a real test,
we found that it was fairly difficult to exactly define when the initial crack
tip was propagated. Therefore, a characteristic strain energy release rate J0
was defined which represented the J value when the maximum peel force P
was reached. Note that with the growth of the crack, the global strain energy
release rate keeps increasing as shown in Fig. (4.3) (J-Δ curve). This
implies that the plastic dissipations beyond the crack tip must keep
increasing during this stable growth process. However, the increase rate of J
becomes slower and slower as the crack grows, which seems nearly stable
even by the end of the test. Therefore, an approximate asymptote is added in
Fig. (4.3) to estimate the interfacial toughness of the bonded joints. And this
estimated asymptotic value is denoted by JC or fracture energy.
Fig. (4.2). A typical relationship between loadline displacements Δ and loadline rotation of adherend
θP for Group 2 (ha=0.2mm).
Fig. (4.3). A typical relationship between loadline displacements Δ and energy release rate J for
Group 2 (ha=0.2mm).
Local Test Results
In this study, crack tip local deformations along the entire adhesive layer
were recorded using high resolution CCD camera. It was also observed
that a whitening zone appeared near the initial crack tip and was
becoming more visible as the load was increased, as shown in Fig. (4.4).
Fig. (4.4). Local plastic deformation as shown by the whitening region.

It was believed that the whitening zone consisted of numerous micro-

cracks. When a lot of micro cracks were merged together and formed a
macro crack, the initial crack started to propagate with the continuously
increasing load. Digital images of the displacement field at the initial crack
tip region were taken by the high resolution CCD camera. The value of the
crack tip separation δ was measured as the relative normal displacement
between the two adherends at the location of the initial crack tip through the
recorded digital images. Fig. (4.5) gives a typical relationship between the
loadline displacement Δ (global displacement) and the local crack tip
opening δ. By combining (Figs. 4.3 and 4.5), a typical experimental J-δ
curve was obtained as shown in Fig. (4.6) (a specimen in group 2 with
thickness of 0.2 mm). Based on Eq. (2.42), the experimental J-δ curves
were used to determine the equivalent interfacial traction-separation laws
or σ=σ(δ) at different adhesive thicknesses.
Fig. (4.5). A typical relationship between loadline displacements.

Fig. (4.6). A typical relationship between energy release rate J and local crack tip separation δ for
Group 2 (ha=0.2mm).
Effects of the Thickness of the Adhesive Layer
With the global test results of different groups, the average maximum peel
loads Pcr of the five specimens in each group (with error bar) are plotted as
a function of the adhesive thicknesses (0.09 mm, 0.2 mm, 0.4 mm, 0.6
mm, 0.8 mm, and 1.0 mm) in Fig. (4.7). It can be observed that the
average peak load was increased from 578.8 N (Group 1) to 791.12 N
(Group 6) when the thickness of the adhesive layer increased (from 0.09
mm to 1.0 mm). These critical loads correspond to the characteristic
energy release rate J0. The average estimated fracture energy JC (by the
approximate asymptotes) and the average characteristic J0 (at the
maximum peel load) of the five specimens in each group (with error bar)
were plotted as a function of the adhesive thickness in Fig. (4.8). One may
see that the fracture energy JC at the thickness of 0.09 mm is increased by
approximately 160% when the adhesive thickness becomes 1.0 mm. The
significant increase of J0 and JC are responsible for the increased load
capacity when the adhesive thickness becomes thicker. Meanwhile, with

Fig. (4.8), one may see that when the adhesive thickness is thin, J0 and JC
are relatively close to each other. On the other hand, when it is relatively
thick, J0 and JC are relatively departed from each other. Thus, we denote
the difference between J0 and JC by ΔJ=JC─J0. Let’s consider a parameter
ξ, which represents the ratio of the average ΔJ over the pertinent average
J0 of the five specimens in each group. This ratio ξ is then plotted as a
function of the adhesive thickness in Fig. (4.9). Obviously, the ratio ξ
represents the relative increase in J to its initial value of J0 as J value
becomes a nearly stable value (JC). Thus, Fig. (4.9) actually reflects the
effects of adhesive thickness on the contributions by geometry a/h to the
increase of plastic dissipations as crack grows.
Fig. (4.7). The average maximum peel load Pcr at different adhesive thicknesses ha.
Fig. (4.8). The estimated fracture energy JC and the characteristic energy release rate J0
(corresponding to Pcr) with different adhesive thicknesses ha.
Fig. (4.9). The ratio of ΔJ (ΔJ=JC─J0) over J0 as a function of the adhesive thickness ha.
Fig. (4.10). Typical shape of the equivalent interfacial traction-separation laws with different
thicknesses of adhesive layer.
After data collecting and curve-fitting process of the experimental J-δ

curves, by applying Eq. (2.42), one can see that the interfacial
traction-separation laws were determined by numerically differentiating the
experimental J-δ curves. The cohesive law was calculated for each
specimen. The typical equivalent cohesive law of each group is given in
Fig. (4.10) at different adhesive thicknesses. From Fig. (4.10), one can see
that three major effects of the adhesive thickness. First, the local
characteristic separation (δ0) corresponding to the interfacial strength σmax,
decreases with the decrease in the adhesive thickness. Second, the
interfacial strength increases with the decrease of the adhesive thickness.
Finally, the total area under the δ-σ curve, which represents the strain
energy release rate at crack propagation, increases with the increase in the
adhesive thickness. The complete thirty interfacial traction-separation
laws (six groups with various adhesive thicknesses) are given in Fig.
(4.11). One may see the five specimens in each group (each adhesive
thickness) present fairly consistent results.
Fig. (4.11). The shapes of all equivalent interfacial traction-separation laws with different
thicknesses of adhesive layer: (a) adhesive thickness ha=0.09 mm; (b) adhesive thickness ha=0.2 mm;
(c) adhesive thickness ha=0.4 mm; (d) adhesive thickness ha=0.6 mm; (e) adhesive thickness ha=0.8
mm; (f) adhesive thickness ha=1.0 mm.
The average interfacial strength σmax of the six groups (with error bar) are
plotted as a function of the adhesive thickness in Fig. (4.12). According to
the manufacturer, the tensile strength of the bulk adhesive material is
approximately 30.3 MPa (its yield strength is slightly lower than its tensile
strength). One can see that the interfacial strength was 88 MPa at the
thickness of 0.09 mm, which is approximately 3 times as large as the yield
strength of the bulk adhesive material. The authors expects that with further
decrease in the adhesive thickness, the measured equivalent interfacial
strengths should keep increasing. This tendency is correct because the
lower bound is when the adhesive thickness is reduced to the atomistic level,
where the traction-separation response of a discrete atom pairs shows a very
high strength but very low separation. On the contrary, with the increase in
adhesive thickness, the measured interfacial strengths asymptotically
approach a constant value. This asymptote seems being the yield strength of
bulk adhesive material. Another phenomenon is that the equivalent
interfacial strength seems increasing dramatically when the adhesive
thickness was thinner than a certain value (such as 0.2 mm).
Fig. (4.12). The effect of the adhesive thickness on the interfacial strength σmax.
RESULTS AND DISCUSSIONS OF MODE II TEST
Global and Local Experimental Results
Based on the analytical results of Ouyang and Li [35], with the current
geometric configuration and the experimental data Table 3.2, it was found
that all adherends (low carbon steel) remained linear elastic during the
entire test process without any plastic deformation. This indicates that all
the energy dissipation during the crack initiation and propagation were
contributed by the intrinsic cohesive fracture and the possible plastic
dissipation in the adhesive layer. A typical force vs. displacement curve at
the loading point is shown in Fig. (5.1) (a specimen in group 1 with
thickness of 0.1 mm). The force linearly increases until the local damage
occurs. However, the load increases in a nonlinear pattern until the crack
initiation process is completed. After that, the crack tip starts propagating.
The crack continuously propagates with a continuous drop in the force until
the crack propagates to the end of the ENF specimen. With Eq. (2.44), the
experimental value of the strain energy release rate JII can be obtained by
combining experimental values of δ0 and QT.
Fig. (5.1). A typical relationship between loadline displacements Δ and loadline peel force P for
Group 1 (ha=0.1 mm).
The value of the shear slip δ between the two adherends at the location of
the initial crack tip is calculated based on the digital images, and they are
plotted as a function of the loadline displacement Δ (global displacement)
in Fig. (5.2) (a specimen in group 1 with thickness of 0.1 mm). By
combining Figs. (5.1 and 5.2), a typical experimental J-δ curve is obtained
as shown in Fig. (5.3) (a specimen in group 1 with thickness of 0.1 mm).
With Eq. (2.45), the experimental J-δ curves are used to determine the
equivalent interfacial traction-separation laws or τ=τ(δ) at different
adhesive thicknesses. A characteristic strain energy release rate J0 is
defined as the J value when the crack starts propagating immediately after
the maximum fracture load Pcr is reached. It can be seen that J0 is generally
smaller than the critical strain energy release rate or the fracture energy JIIC,
which reflects that the strain energy release rate JII keeps increasing after
the crack propagation. This is because the plastic dissipation (or plastic
zone) can be further increased after the crack initiation.
Fig. (5.2). A typical relationship between loadline displacement Δ and local crack tip slip δ for Group
1 (ha=0.1 mm).
Fig. (5.3). A typical relationship between energy release rate J and local crack tip slip δ for Group 1
(ha=0.1 mm).
Fig. (5.4). Local damage, fracture and deformation in the bondline.

In this study, the crack initiation and propagation processes were recorded
by using the high resolution CCD camera. It was observed that the initial
crack tip started opening as the shear load increased. It was also observed
that the whiting zone appeared near the initial crack tip and was becoming
more visible as the load was increased, as shown in Fig. (5.4) (red circle
area).
It was believed that the whiting zone consisted of numerous microcracks.

When a lot of micro cracks were merged together and formed a macro crack,
the initial crack started to propagate with the continuously increasing load,
as shown in Fig. (5.5). Based on Eq. (2.45), these experimental J-δ curves
are used to determine the equivalent interface cohesive laws or τ=τ(δ) at
different adhesive thicknesses. Fig. (5.6) gives a typical shear stress vs.
local slip displacement curve (a specimen in group 1 with thickness of 0.1
mm).
Fig. (5.5). Crack propagates during the test.

10
9
8
7
6
τ (MPa)
5
4
3
2
1
0
0 100 200 300 400
δ (µm)
Fig. (5.6). Typical interface constitutive relationship (equivalent cohesive law) for Group 1 (ha=0.1
mm).
With the global experiment results for different groups, the maximum
peel loads are plotted as a function of the adhesive thicknesses (0.1 mm,
0.2 mm, 0.4 mm, 0.6 mm, and 0.8 mm) as shown in Fig. (5.7). It can be
observed that the average peak load is increased from 4912.24 N (Group
1) to 5836.33 N (Group 5) when the thickness of adhesive layer increases
(from 0.1 mm to 0.8 mm). However, the rate of increase in the peak load
reduces as the adhesive layer thickness increases.
Fig. (5.7). The average maximum shear load Pcr at different adhesive thicknesses (ha=0.1 mm).
The increase in the adhesive layer thickness might increase the bending
stiffness of the ENF specimen. However, this contribution is very small
since the adherends were made of low carbon steel with a much higher
thickness of 9.535 mm and much higher bending stiffness. The estimated
average fracture energy JC and the average characteristic J0 (at the
maximum peel load) were plotted as a function of the adhesive thickness
in Fig. (5.8) (0.1 mm, 0.2 mm, 0.4 mm, 0.6 mm and 0.8 mm). The
significant increase in J0 and JC are responsible for the increased load
capacity when the adhesive thickness becomes thicker. Meanwhile, one
may see that when the adhesive thickness is thin, J0 and JC are relatively
close to each other. When it is relatively thick, J0 and JC are relatively
departed. After data collecting and curve-fitting process of the
experimental J-δ curves, by applying Eq. (2.45), one can see that the
interfacial traction-slip laws are determined by numerically differentiating
the experimental J-δ curves.
Fig. (5.8). The estimated fracture energy JC and the characteristic energy releaserate J0
(corresponding to Pcr) with different adhesive thicknesses.
A typical equivalent interfacial law of each group is given in Fig. (5.9) at

different bondline thicknesses. One may see that the effects of the
bondline thickness on the interfacial traction-slip laws. First, the
interfacial strength increases for steel ENF joint with the increase in the
bondline thickness. Second, the initial stiffness of the interfacial laws (the
initial slope of the τ-δ curve) is identical for steel ENF joints with various
bondline thicknesses, which suggests that the shear deformation is
controlled by the adhesive layer. Finally, the fracture energy (the total area
under the δ-σ curve) increases with the bondline thickness. As shown in
Fig. (5.9), the fracture energy JC of the bonded joints is plotted as a
function of the adhesive thickness. The fracture energy JC at the thickness
of 0.1 mm is increased by approximately 238% when the adhesive
thickness is 0.8 mm. This significant increase in fracture energy can
explain the increase in load capacity when the adhesive thickness is
increased.
Fig. (5.9). Typical shapes of the interfacial traction–separation laws at different adhesive layer
thicknesses ha.
Fig. (5.10). The effect of adhesive thickness on the maximum interface stress τf.
Fig. (5.11). The shapes of all interfacial traction-slip laws with different thicknesses of bondline: (a)
bondline thickness ha=0.1 mm; (b) bondline thickness ha=0.2 mm; (c) bondline thickness ha=0.4 mm;
(d) bondline thickness ha=0.6 mm; (e) bondline thickness ha=0.8 mm.
The average interfacial strength τmax of the five groups (with error bar) are
plotted as a function of the adhesive thickness in Fig. (5.10). According to
the manufacturer, the shear strength of the adhesive material is
approximately 13.8 MPa (when the thickness is 0.75 mm with steel
adherends). One can see that the interfacial strength was 14.9 MPa at the
thickness of 0.8 mm, which is approximately the yield shear strength of the
bulk adhesive material. Finally, the complete fifteen interfacial
traction-separation laws of the bonded steel ENF joints (three specimens in
each of the five groups at various bondline thicknesses) are given in Fig.
(5.11). One may see the fairly consistent results of the three specimens in
each group at various bondline thicknesses.
RESULTS AND DISCUSSION OF MIXED MODE I AND MODE II

TEST
Global and Local Test Results
The geometric parameters of each specimen have been given in Table 3.3. A
typical force vs. displacement curve (a specimen with ha=0.1mm and an
initial crack length a=80mm) at the loading point is shown in Fig. (6.1).
The force linearly increased until the local damage occurs. The load
nonlinearly increases until the crack initiation process is completed. After
that, the crack tip starts propagating. A continuous drop in the bending force
was seen as the crack continuously propagated until the crack reached the
specimen’s end.
Fig. (6.1). A typical relationship between loadline displacements Δ and loadline bending force P for
ha=0.1mm with initial crack length a=80mm.
The data of the rotation angel at the loading point were collected by the
inclinometer. With Eq. (2.50), the experimental energy release rate J is
determined by combining the measured θP and P. A typical relationship (a
specimen with ha=0.1mm and initial crack length of 20mm) between global
energy release rate J and loadline displacement Δ is given in Fig. (6.2). In a
real test, we found that it was fairly difficult to exactly define when the
initial crack tip was propagated. Therefore, a characteristic strain energy
release rate J0 was defined which represented the J value when the
maximum peel force P was reached. Note that with the growth of the crack,
the global strain energy release rate keeps increasing as shown in Fig. (6.2)
(J-Δ curve). This implies that the plastic dissipations beyond the crack tip
must keep increasing during this stable growth process. However, the
increase in J becomes slower and slower as the crack grows, which seems
nearly stable even by the end of the test. Therefore, an approximate
asymptote is added in Fig. (6.2) to estimate the interfacial toughness of the
bonded joints. And this estimated asymptotic value is denoted by JC or
fracture energy.
Fig. (6.2). A typical relationship between loadline displacements Δ and energy release rate J for
specimen with ha=0.1mm (a=80mm).
In this study, crack tip local deformations along the entire adhesive layer
were recorded using the high resolution CCD camera. The value of the
crack tip separation δ was measured as the relative normal displacement
between the two steel adherends at the location of initial crack tip through
the recorded digital images. By combining JI-δ curve and crack tip
separation δ, a typical experimental JI-δ curve was obtained as shown in Fig.
(6.3) (a specimen with ha=0.1mm and initial crack length of 80mm). With
Eq. (2.51), the experimental JI-δ curves were used to determine the
equivalent interfacial traction-separation laws or σ=σ(δ), as show in Fig.
(6.4).
Fig. (6.3). A typical relationship between energy release rate J and local crack tip separation δ for
specimen with ha=0.1mm (a=80mm).
Fig. (6.4). A typical shape of the equivalent interfacial traction-separation laws with ha=0.1mm
(a=80mm) adhesive layer.
Fig. (6.5). Local damage, fracture and deformation in the bondline.

It was observed that the initial crack tip started opening as the shear load
increased. It was also observed that the whiting zone appeared near the
initial crack tip and was becoming more visible as the load was increased,
as shown in Fig. (6.5) (circle area).
Fig. (6.6). A typical relationship between energy release rate J and local crack tip slip δ for specimen
with ha=0.1mm (a=80 mm).
The value of the shear slip δ between the two adherends at the location of
the initial crack tip is calculated based on the digital images. By combining
P-Δ curve and local shear slip δ, according to Eq. (2.46), a typical
experimental JII-δ curve is obtained as shown in Fig. (6.6) (a specimen wiht
ha=0.1mm and initial crack length 80mm). With Eq. (2.49), the
experimental JII-δ curves are used to determine the equivalent interfacial
traction-separation laws or τ=τ(δ) at different adhesive thicknesses. A
characteristic strain energy release rate J0 is defined as the J value when the
crack starts propagating immediately after the maximum fracture load Pcr is
reached. It can be seen that J0 is generally smaller than the critical strain
energy release rate or the fracture energy JIIC, which reflects that the strain
energy release rate JII keeps increasing after the crack propagation. This is
because the plastic dissipation (or plastic zone) can be further increased
after the crack initiation.
Based on Eq. (2.47), these experimental JII-δ curves are used to determine
the equivalent interface cohesive laws or τ=τ(δ) at different adhesive
thicknesses. Fig. (6.7) gives a typical shear stress vs. local slip displacement
curve (a specimen with ha=0.1mm and initial crack length of 80mm).
Fig. (6.7). Typical interface constitutive relationship (equivalent cohesive law) for specimen with
ha=0.1 mm (a=80mm).
With the global experiment results for different groups, the maximum
bending loads are plotted as a function of the adhesive thicknesses (0.1
mm, 0.2 mm, 0.4 mm, 0.6 mm, and 0.8 mm) as shown in Fig. (6.8). It can
be observed that the average peak load is increased from 3417.67 N to
4791.10 N when the thickness of adhesive layer increases (from 0.1 mm
to 0.8 mm), as shown by the trend line in Fig. (6.8).
Fig. (6.8). The average maximum shear load Pcr at different adhesive thicknesses ha.
The increase in the adhesive layer thickness might increase the bending
stiffness of the specimen. However, this contribution is very small since the
adherends were made of low carbon steel with a much higher thickness of
6.35 mm and much higher bending stiffness. The estimated average fracture
energy JIC, JIIC and the average characteristic energy JI0, JII0 (at the
maximum peel load) were plotted as a function of the adhesive thickness in
Figs. (6.9 and 6.10) (0.1 mm, 0.2 mm, 0.4 mm, 0.6 mm and 0.8 mm). The
significant increase in J0 and JC are responsible for the increased load
capacity when the adhesive thickness becomes thicker. Meanwhile, one
may see that when the adhesive thickness is thin, J0 and JC are relatively
close to each other. When it is relatively thick, J0 and JC are relatively

departed. After data collecting and curve-fitting process of the experimental
J-δ curves, by applying Eq. (2.51), one can see that the interfacial
traction-slip laws are determined by numerically differentiating the
experimental J-δ curves.
Fig. (6.9). The estimated fracture energy JIC and the characteristic energy release rate JI0
(corresponding to Pcr) with different adhesive thicknesses (a=20mm).
A typical equivalent interfacial law of each group is given in Figs. (6.11

and 6.12) at different bondline thicknesses with initial crack length
a=20mm. One may see that the effects of the bondline thickness on the
interfacial traction-slip laws.
Mode I Portion
After data collecting and curve-fitting process of the experimental J-δ

curves, by applying Eq. (2.46), one can see that the interfacial
traction-separation laws were determined by numerically differentiating the

experimental J-δ curves. The cohesive law was calculated for each
specimen. The typical equivalent cohesive law of each group is given in
Fig. (6.11) at different adhesive thicknesses. From Fig. (6.11), one can see
that three major effects of the adhesive thickness. First, the local
characteristic separation (δ 0) corresponding to the interfacial strength σmax,
decreases with the decrease in the adhesive thickness. Second, the
interfacial strength increases with the decrease in the adhesive thickness.
Finally, the total area under the δ-σ curve, which represents the strain
energy release rate at crack propagation, increases with the increase in the
adhesive thickness. Again, the upper bound is the bulk material, which
has the lowest traction but highest separation, and the lower bound is the
atomistic level, which has the highest traction but lowest separation.
Fig. (6.10). The estimated fracture energy JIIC and the characteristic energy release rate JII0
(corresponding to Pcr) with different adhesive thicknesses (a=20mm).
Fig. (6.11). Typical shapes of the interfacial traction-separation laws of bonded steel with different
bondline thicknesses ha (a=20mm).
Mode II Portion
First, the interfacial strength increases for steel joint with the increase in
the bondline thickness. Second, the initial stiffness of the interfacial laws
(the initial slope of the τ-δ curve) is identical for steel joints with various
bondline thicknesses, which suggests that the shear deformation is
controlled by the adhesive layer. Finally, the fracture energy (the total area
under the δ-τ curve) increases with the bondline thickness. As shown in
Fig. (6.10), the fracture energy JIIC of the bonded joints is plotted as a
function of the adhesive thickness. The fracture energy JIIC at the
thickness of 0.1 mm is increased by approximately 143% when the
adhesive thickness becomes 0.8 mm. This significant increase in fracture
energy can explain the increase in load capacity when the adhesive
thickness is increased.
Fig. (6.12). Typical shapes of the interfacial traction-slip laws of bonded steel with different bondline
thicknesses ha (a=20mm).
Mode Mixities
The ratio of the estimated fracture energy JIIC and total fracture energy
(JIC +JIIC) with different adhesive thickness was plotted in Fig. (6.13).
The ratio of the shear strength and total strength with different adhesive
thickness was plotted in Fig. (6.14). It is observed that the ratio of the
estimated fracture energy JIIC and the ratio of the shear strength were
increased as the thickness of the adhesive layer increased. This indicates
that the ratio of mode I within the mixed mode I/II was decreased as the
thickness of the adhesive layer increased. The stiffness of the adhesive
layer is much lower than the stiffness of the adherends. Therefore, as the
thickness of adhesive layer increased, the plastic dissipation increased.
Fig. (6.13). The ratio of the estimated fracture energy JIIC and total fracture energy (JIC + JIIC) with
different adhesive thicknesses (a=20mm).
Fig. (6.14). The ratio of the shear strength and total strength with different adhesive thicknesses
(a=20mm).
Effects of the Initial Crack Tip Length of the Adhesive Layer
In this study, the effects of the initial crack tip length of the adhesive layer
were also considered with different thickness of the adhesive layer. Three
different initial crack lengths a=20mm, 50mm, and 80mm were selected in
this study.
The maximum bending load Pcr at each initial crack length with different
thickness of the adhesive layer was plotted in Fig. (6.15). It is observed that
the maximum load was decreased as the initial crack length increased; the
maximum load was increased as the thickness of the adhesive layer
increased at each initial crack length. As the initial crack length increased at
the same thickness of the adhesive layer, the bending moment was
increased at the initial crack tip. It caused the peeling force increase near the
crack tip area at the same time. In other words, the mode I portion was
increased as the initial crack length increased.
Fig. (6.15). The average maximum bending load Pcr at different initial crack length with different
adhesive thicknesses ha.
When the thickness of the adhesive layer was 0.1mm, as the initial crack
length increased, the mount of the adhesive material between the
adherends decreased and the moment around the initial crack tip area
increased. Additionally, the relative rotation angle between the two
adherends increased and as a result, the smallest peeling force can make
the crack propagate, as shown in Fig. (6.15). According to Eq. (2.50):
JI=P*θ/4, although the loading force increased from 2047N to 3417N as
the initial crack length decreased from 80mm to 20mm, the relative
rotation angle decreased from 3.23rad to 1.97rad. As the thickness of the
adhesive layer increased, the ratio of the increased relative rotation angle
increased. It can be explained that why the increased ratio of JIC with
initial crack length a=80mm is larger than the other two, as the thickness
of adhesive layer increased.
The estimated fracture energy JIC and JIIC at each initial crack length with
different adhesive thickness were plotted in Fig. (6.16 and 6.17). It is
observed that the JIC and JIIC were increased at each initial crack length as
the thickness of the adhesive layer increases.
In Fig. (6.16), the JIC was increased when the initial crack length was
increased from 20mm to 50mm. Further increasing the initial crack length
from 50mm to 80mm leads to reduction in JIC when the adhesive
thickness is 0.1mm. With the increase in the adhesive layer thickness, the
tread line of JIC with the initial crack length of 80mm line crossed the
other trending lines with initial crack length at 20mm and 50mm. The
phenomenon may be due to the combination effect of the thickness of the
adhesive layer and initial crack length. This is a very interesting point that
deserves in-depth investigation.
In Fig. (6.17), the JIIC was decreased as the initial crack length increased
with the same thickness of adhesive layer. It is understandable due to the
decreased loading force with increased initial crack length. Additionally,
as the thickness of the adhesive layer increased, the relative rotation angle
between the two adherends increased and caused the increase in shear
force.
Fig. (6.16). The estimated fracture energy JIC at each initial crack length with different adhesive
thicknesses.
Fig. (6.17). The estimated fracture energy JIIC at each initial crack length with different adhesive
thicknesses.
Typical shapes of the interfacial traction-separation laws and traction-slip

laws of the bonded steel with different bondline thickness were plotted in
Figs. (6.18 and 6.19) (a=50mm), Figs. (6.20 and 6.21) (a=80mm). The
interfacial strength σmax, τmax at each initial crack length with different
thickness of the adhesive layer were plotted in Figs. (6.22 and 6.23). For
each initial crack length, first, the local characteristic separation (δ0)
corresponding to the interfacial strength σmax decreased with the decrease in
the adhesive thickness. Second, the interfacial strength increases with the
decrease in the adhesive thickness. Finally, the total area under the δ-σ
curve, which represents the strain energy release rate at crack propagation,
increases with the increase in the adhesive thickness.
Additionally, the interfacial strength τmax increases for steel joint with the
increase in the bondline thickness. The initial stiffness of the interfacial
laws (the initial slope of the τ-δ curve) is identical for steel joints with
various bondline thicknesses, which suggests that the shear deformation is
controlled by the adhesive layer. The fracture energy (the total area under
the δ-τ curve) increases with the bondline thickness.
Fig. (6.22). The effect of the adhesive thickness on the interfacial strength σmax with different initial
crack length.
Fig. (6.23). The effect of the adhesive thickness on the interfacial strength τmax with different initial
crack length.
According to Figs. (6.22 and 6.23), as the initial crack tip length increased,
the interfacial shear strength was decreased with the increased thickness of
the adhesive layer. Additionally, as the initial crack tip length increased
from a=20mm to a=50mm, the interfacial peel strength decreased with
increased thickness of the adhesive layer. However, when the initial crack
length increased from 50mm to 80mm and the thickness increased from
0.1mm to 0.6mm, the peel strength increased due to the increased moment
and relative rotation angle which caused a compression zone near the crack
tip area and prevented the crack propagation. Therefore, the peel strength
increased. When the thickness increased from 0.6mm to 0.8mm, the
bending stiffness of the adhesive layer increased and caused the decreased
relative rotation angle. Therefore, the peel strength decreased without the
compression zone around the crack tip area.
CONCLUSIONS AND FUTURE WORKS

Three fundamental fracture mechanics test (mode I, mode II, and mixed
mode I/II) were conducted by the Double Cantilever Beam (DCB), End
Notched Flexure (ENF), and Single Leg Bending (SLB) specimens with
different adhesive thicknesses. According to the global test results of load
and displacement and local test results of separation and slipping of initial
crack tip under the quasi-static load, the fracture toughness and cohesive
laws of traction-separation and traction-slipping (σ-δ and τ-δ) were
developed with different thicknesses of adhesive layers.
Mode I Test
DCB specimens with different adhesive thicknesses experienced peel test

(mode I test). Within the boundaries of the used materials parameters and
based on the obtained data results, the following conclusions are yielded:
a. The adhesive thickness has an influence on the critical loading

force. As the thickness of the adhesive layer increases, the critical
force is increased.
b. As the adhesive thickness increases, the fracture toughness JIC and
characteristic energy JI0 (corresponding to critical load) are
increased.
c. The local crack tip separation δ is also increased, as the thickness
of the adhesive layer increases.
d. The thickness of the adhesive layer also has a significant effect on
the interfacial strength (σ). For the adherends made of higher
bending stiffness material (greater than the adhesive’s flexure
stiffness such as steel in this study), as the adhesive thickness
decreases, the fracture strength (σ) is increased due to the
increased constraining force of the adherends which can prevent
the crack propagation around the initial crack tip area.
e. With the J-integral theory, the present study first presents the
experimental evidence that the cohesive strength is 88 MPa which
is about 3 times of the yield strength of the bulk adhesive when
the adhesive thickness is 90 μm (steel based specimen).
f. It is also noted that the significant plastic dissipation will not
necessarily be reflected in the globally measured value of the
fracture energy when the adhesive thickness is relatively thin
and even the intrinsic cohesive strength is much higher than the
bulk adhesive’s yield strength.
Mode II Test
In this study, the steel based adherends of adhesively bonded joints are
subjected to mode II loading using End Notched Flexure (ENF)
specimens. The objective is to measure the fracture toughness of the joint
corresponding to different thicknesses of the adhesive layer and to deduce
the nature of the cohesive (traction-separation) law that governs the crack
growth behavior. Within the boundaries of the used materials parameters
and based on the test results, the following conclusions are obtained:
force. As the thickness of the adhesive layer increases, the critical
force is increased.
b. As the adhesive thickness increases, the fracture toughness JIIC
and characteristic energy JII0 (corresponding to critical load) are
increased.
c. The local crack tip slipping δ is also increased, as the thickness of
the adhesive layer increases.
d. The thickness of the adhesive layer also has a considerable effect
on the fracture strength (τ). As the adhesive thickness increases,
the fracture strength (τ) is increased.
Mixed Mode I/II Test
In this study, the steel based adherends of adhesively bonded joints are
subjected to mixed mode I/II loading using Single Leg Bending (SLB)
specimens. The objective is to measure the fracture toughness (JIC & JIIC)
of the joint corresponding to different thicknesses of the adhesive layer
and different initial crack lengths and to deduce the nature of the cohesive
(traction-separation & traction-slipping) law that governs the crack
growth behavior. Within the boundaries of the used materials parameter
and based on the test results, the following conclusions are derived:

force. As the thickness of adhesive layer increases, the critical
force is increased.
b. As the adhesive thickness increases, the fracture toughness JIC, JIIC
and characteristic energy JI0, JII0 (corresponding to critical load)
are increased.
c. The local crack tip separation and slipping δ are also increased, as
the thickness of the adhesive layer increases.
d. The thickness of the adhesive layer also has a significant effect on
the interfacial strength (σ). As the adhesive thickness decreases,
the fracture strength (σ) is increased due to the increased
constraining force of the adherends which can prevent the crack
propagation around the initial crack tip area.
e. The thickness of adhesive layer also has a considerable influence
on the fracture strength (τ). As the adhesive thickness increases,
the fracture strength (τ) is increased.
f. As initial crack length increases, the critical loading force is
decreased with the same adhesive thickness.
g. The fracture toughness JIIC and fracture strength τ are increased as
the initial crack length decreased with the same thickness of
adhesive layer. It means the mode II portion is decreased as initial

crack length increased.
However, the fracture toughness JIC and fracture strength (σ) are not
always increased or decreased as the initial crack length increases. It is
very complicated due to the combination effect of the initial crack length
and adhesive thickness.
this publication.
ACKNOWLEDGEMENTS
This study is based upon work supported by the NSF under grant number
(NSF/CMMI0900064) and by the NASA/EPSCoR under grant number
NASA/LEQSF (2007-10)-Phase3-01.
REFERENCE
[1] P. Mazza, "A new Palaeolithic discovery: tar-hafted stone tools in a European
Mid-Pleistocene bone- bearing bed", J. Archaeol. Sci., vol. 33, p. 1310, 2006.
[http://dx.doi.org/10.1016/j.jas.2006.01.006]
[2] Z. Ouyang, and G. Li, "Local damage evolution of DCB specimens during
crack initiation process: A natural boundary condition based method. ASME",
J. Appl. Mech., vol. 76, p. 051003, 2009.
[http://dx.doi.org/10.1115/1.3112742]
[3] A.J. Kinloch, and S.J. Shaw, "The fracture resistance of a toughened epoxy
adhesive", J. Adhes., vol.12, pp. 59-77, 1981.
[http://dx.doi.org/10.1080/00218468108071189]
[4] "Twisp", http://en.wikipedia.org/wiki/File:Fracture_modes_v2.svg, 2010.
[5] A.A. Griffith, "The phenomena of rupture and flow in solids", Philos.
Trans. R. Soc. Lond. A, vol. 221, pp. 163-198,
1921.[http://dx.doi.org/10.1098/rsta.1921.0006]
[6] G.R. Irwin, and J.A. Kies, "Critical energy rate analysis of fracture strength",
Weld. J., vol. 33, pp. 193-198, 1954.
[7] J.W. Hutchinson, and A.G. Evans, "Mechanics of materials: top-down

approaches to fracture", Acta Mater., vol. 48, pp. 125-135, 2000.
[http://dx.doi.org/10.1016/S1359-6454(99)00291-8]
[8] G.I. Barenblatt, "The formation of equilibrium cracks during brittle fracture.
General ideas and hypothesis. Axisymmetrical cracks", J. Appl. Math. Mech.,
vol. 23, pp. 622-636, 1959. [PMM].
[http://dx.doi.org/10.1016/0021-8928(59)90157-1]
[9] D.S. Dugdale, "Yielding of steel sheets containing slits", J. Mech. Phys.
Solids, vol. 8, pp. 100-104, 1960.
[http://dx.doi.org/10.1016/0022-5096(60)90013-2]
[10] J.R. Rice, "A path independent integral and the approximate analysis of strain
concentration by notches and cracks", ASME Journal of Applied Mechanics,
vol. 35, pp. 379-386, 1968.
[http://dx.doi.org/10.1115/1.3601206]
[11] Y. Wei, and J.W. Hutchinson, "Interface strength, work of adhesion and
plasticity in the peel test", Int.J. Fract., vol. 93, pp. 315-333, 1998.
[http://dx.doi.org/10.1023/A:1007545200315]
[12] J.G. Williams, and H. Hadavinia, "Analytical solutions for cohesive zone
models", J. Mech. Phys. Solids, vol. 50, pp. 809-825, 2002.
[http://dx.doi.org/10.1016/S0022-5096(01)00095-3]
[13] A. Hillerborg, M. Modéer, and P.E. Petersson, "Analysis of crack formation

and crack growth in concrete by means of fracture mechanics and finite
elements", Cement Concr. Res., vol. 6, pp. 773- 782, 1976.
|[http://dx.doi.org/10.1016/0008-8846(76)90007-7]
[14] J.H. Rose, J.R. Smith, and J. Ferrante, "Universal features of bonding in
metals", Phys. Rev. B, vol. 28, pp. 1835-1845, 1983.
[http://dx.doi.org/10.1103/PhysRevB.28.1835]
[15] A. Needleman, "A continuum model for void nucleation by inclusion

debonding. ASME", J. Appl. Mech., vol. 54, pp. 525-531, 1987.
[http://dx.doi.org/10.1115/1.3173064]
[16] V. Tvergaard, "Effect of fibre debonding in a whisker-reinforced metal",

Mater. Sci. Eng. A, vol. 125, pp. 203-213, 1990.
[http://dx.doi.org/10.1016/0921-5093(90)90170-8]
[17] X.P. Xu, and A. Needleman, "Void nucleation by inclusion debonding in a

crystal matrix", Model. Simul. Mater. Sci. Eng., vol. 1, pp. 111-132, 1993.
[http://dx.doi.org/10.1088/0965-0393/1/2/001]
[18] G.T. Camacho, and M. Ortiz, "Computational modeling of impact damage in

brittle materials", Int. J. Solids Struct., vol. 33, pp. 2899-2938, 1996.
[http://dx.doi.org/10.1016/0020-7683(95)00255-3]
[19] V. Tvergaard, and J.W. Hutchinson, "The relation between crack growth
resistance and fracture process parameters in elastic-plastic solids", J. Mech.
Phys. Solids, vol. 40, pp. 1377-1397, 1992.
[http://dx.doi.org/10.1016/0022-5096(92)90020-3]
[20] A. Corigliano, "Formulation, identification and use of interface models in the

numerical analysis of composite delamination", Int. J. Solids Struct., vol. 30,

pp. 2779-2811, 1993.
[http://dx.doi.org/10.1016/0020-7683(93)90154-Y]
[21] S.R. Chowdhury, and R. Narasimhan, "A finite element analysis of stationary
crack tip fields in a pressure sensitive constrained ductile layer", Int. J. Solids
Struct., vol. 37, pp. 3079-3100, 2000.
[http://dx.doi.org/10.1016/S0020-7683(99)00046-3]
[22] Q.D. Yang, M.D. Thouless, and S.M. Ward, "Elastic-plastic mode-II fracture
of adhesive joints", Int. J. Solids Struct., vol. 38, pp. 3251-3262, 2001.
[http://dx.doi.org/10.1016/S0020-7683(00)00221-3]
[23] Q.D. Yang, M.D. Thouless, and S.M. Ward, "Mixed mode fracture analysis
of plastically-deforming adhesive joints", Int. J. Fract., vol. 110, pp. 175-187,
2001. [http://dx.doi.org/10.1023/A:1010869706996]
[24] G. Alfano, and M.A. Crisfield, "Finite element interface models for the
delamination analysis of laminated composites: mechanical and computational
issues", Int. J. Numer. Methods Eng., vol. 50, pp. 1701-1736, 2001.
[http://dx.doi.org/10.1002/nme.93]
[25] R.H. Andruet, D.A. Dillard, and S.M. Holzer, "Two- and three-dimensional
geometrical nonlinear finite elements for analysis of adhesive joints", Int. J.
Adhes. Adhes., vol. 21, pp. 17-34, 2001.
[http://dx.doi.org/10.1016/S0143-7496(00)00024-5]
[26] T. Pardoena, T. Ferracina, C.M. Landisb, and F. Delannay, “Constraint

effects in adhesive joint fracture", J. Mech. Phys. Solids, vol. 53, pp.
1951-1983, 2005.
[http://dx.doi.org/10.1016/j.jmps.2005.04.009]
[27] J.L. Högberg, "Mixed mode cohesive law", Int. J. Fract., vol. 141, pp.
549-559, 2006. [http://dx.doi.org/10.1007/s10704-006-9014-9]
[28] K. Salomonsson, and T. Andersson, "Modeling and parameter calibration of

an adhesive layer at the meso level", Mech. Mater., vol. 40, pp. 48-65, 2008.
[http://dx.doi.org/10.1016/j.mechmat.2007.06.004]
[29] F. Parrinello, B. Failla, and G. Borino, "Cohesive–frictional interface

constitutive model", Int. J. Solids Struct., vol. 46, pp. 2680-2692, 2009.
[http://dx.doi.org/10.1016/j.ijsolstr.2009.02.016]
[30] Y. Yan, and F. Shang, "Cohesive zone modeling of interfacial delamination in

PZT thin films", Int. J. Solids Struct., vol. 46, pp. 2739-2749, 2009.
[31] M.F. de Moura, R.D. Campilho, and J.P. Gonçalves, “Pure mode II
fracture characterization of composite bonded joints", Int. J. Solids Struct.,
vol. 46, pp. 1589-1595, 2009.
[32] A. Klarbring, "Derivation of a model of adhesively bonded joints by the

asymptotic expansion method", Int. J. Eng. Sci., vol. 29, pp. 493-512, 1991.
[http://dx.doi.org/10.1016/0020-7225(91)90090-P]
[33] B.R. Blackman, H. Hadavinia, A.J. Kinloch, and J.G. Williams, "The use of a
cohesive Zone model to study the fracture of fibre composites and
adhesively-bonded joints", Int. J. Fract., vol. 119, pp. 25-46, 2003.
[http://dx.doi.org/10.1023/A:1023998013255]
[34] J. Pan, and C.K. Leung, "Debonding along the FRP-concrete interface under
combined pulling/peeling effects", Eng. Fract. Mech., vol. 74, pp. 132-150,
2007. [http://dx.doi.org/10.1016/j.engfracmech.2006.01.022]
[35] Z. Ouyang, and G. Li, "Nonlinear interface shear fracture of end notched
flexure specimens", Int. J. Solids Struct., vol. 46, pp. 2659-2668, 2009.
[36] Z. Ouyang, and G. Li, "Cohesive zone model based analytical solutions for
adhesively bonded pipe joints under torsional loading", Int. J. Solids Struct.,
vol. 46, pp. 1205-1217, 2009.
[37] Z. Ouyang, and G. Li, "Interfacial debonding of pipe joints under torsion
loads: A model for arbitrary nonlinear cohesive laws", Int. J. Fract., vol. 155,
pp. 19-31, 2009.
[http://dx.doi.org/10.1007/s10704-009-9318-7]
[38] L.D. Lorenzis, and G. Zavarise, "Cohesive zone modeling of interfacial

stresses in plated beams", Int. J. Solids Struct., vol. 46, pp. 4181-4191, 2009.
[39] C. Nguyen, and A.L. Levy, "An exact theory of interfacial debonding in
layered elastic composites", Int. J. Solids Struct., vol. 46, pp. 2712-2723,
2009.
[40] H. Chai, "Shear fracture", Int. J. Fract., vol. 37, pp. 137-159, 1988.
[41] H. Chai, "Deformation and fracture of particulate epoxy in adhesive bonds",

Acta Metall. Mater., vol. 43, pp. 163-172, 1995.
[http://dx.doi.org/10.1016/0956-7151(95)90271-6]
[42] T. Ikeda, A. Yamashita, D. Lee, and N. Miyazaki, "Failure of a ductile

adhesive layer constrained by hard adherends", J. Eng. Mater. Technol., vol.
122, pp. 80-85, 2000.
[http://dx.doi.org/10.1115/1.482769]
[43] M.S. Kafkalidis, M.D. Thouless, Q.D. Yang, and S.M. Ward, "Deformation
and fracture of adhesive layers constrained by plastically-deforming
adherends", J. Adhes. Sci. Technol., vol. 14, pp. 1593- 1607, 2000.
[http://dx.doi.org/10.1163/156856100742401]
[44] C. Yan, Y. Mai, and L. Ye, "Effect of bond thickness on fracture behaviour in
adhesive joints", J. Adhes., vol. 75, pp. 27-44, 2001.
[http://dx.doi.org/10.1080/00218460108029592]
[45] K.S. Madhusudhana, and R. Narasimhan, "Experimental and numerical

investigations of mixed mode crack growth resistance of a ductile adhesive
joint", Eng. Fract. Mech., vol. 69, pp. 865-883, 2002.
[http://dx.doi.org/10.1016/S0013-7944(01)00110-2]
[46] D. Lee, T. Ikeda, N. Miyazaki, and N. Choi, "Effect of bond thickness on the
fracture toughness of adhesive joints", J. Eng. Mater. Technol., vol. 126, pp.
14-18, 2004.
[http://dx.doi.org/10.1115/1.1631433]
[47] B.F. Søensen, "Cohesive law and notch sensitivity of adhesive joints", Acta
Mater., vol. 50, pp. 1053- 1061, 2002.
[http://dx.doi.org/10.1016/S1359-6454(01)00404-9]
[48] T. Andersson, and U. Stigh, "The stress-elongation relation for an adhesive

layer loaded in peel using equilibrium and energetic forces", Int. J. Solids
Struct., vol. 41, pp. 413-434, 2004.
[49] T. Andersson, and A. Biel, "On the effective constitutive properties of a thin
adhesive layer loaded in peel", Int. J. Fract., vol. 141, pp. 227-246, 2006.
[http://dx.doi.org/10.1007/s10704-006-0075-6]
[50] K. Leffler, K.S. Alfredsson, and U. Stigh, "Shear behaviour of adhesive

layers", Int. J. Solids Struct., vol. 44, pp. 520-545, 2007.

[51] Y. Zhu, K.M. Liechti, and K. Ravi-Chandar, "Direct extraction of

rate-dependent traction–separation laws for polyurea/steel interfaces", Int. J.
Solids Struct., vol. 46, pp. 31-51, 2009.
[52] L.A. Carlsson, J.W. Gillespie, and R.B. Pipes, "On the analysis and design of
the end notched flexure (ENF) specimen for mode II testing", J. Compos.
Mater.,vol. 20, pp. 594-604, 1986.
[http://dx.doi.org/10.1177/002199838602000606]
[53] A.J. Russell, and K.N. Street, Factors affecting the interlaminar fracture
energy of Graphite/Epoxy laminates. Progress in Science and Engineering of
Composites, T. Hayashi, K Kawata & S. Umekawa ed., ICCM-IV, Tokyo, p.
279-286, 1982.
[54] P. Davies, G.D. Sims, B.R. Blackman, A.J. Brunner, K. Kageyama, M. Hojo,
K. Tanaka, G. Murri, C. Rousseau, B. Gieseke, and R.H. Martin, "Comparison
of test configurations for determination of mode II interlaminar fracture
toughness results from international collaborative test programme", Plastics
Rubber Composite, vol. 28, pp. 432-437, 1999.
[http://dx.doi.org/10.1179/146580199101540600]
[55] B.D. Davidson, and X. Sun, "Effects of friction, geometry and fixture
compliance on the perceived toughness from three and four point bended
notched flexure tests", J. Reinf. Plast. Compos., vol. 4, pp. 1611-1628, 2005.
[http://dx.doi.org/10.1177/0731684405050402]
[56] B.D. Davidson, X. Sun, and A. Vinciquerra, "Influences of friction, geometric

nonlinearities, and fixture compliance on experimentally observed
toughnesses from three and four-point bend end- notched flexure tests", J.
Compos. Mater., vol. 41, pp. 1177-1196, 2007.

[http://dx.doi.org/10.1177/0021998306067304]
[57] C. Schuecker, and B.D. Davidson, "Evaluation of the accuracy of the

four-point bend end-notched flexure test for the mode II delamination
toughness determination", Compos. Sci. Technol., vol. 60, pp. 2137-2146,
2000.
[http://dx.doi.org/10.1016/S0266-3538(00)00113-5]
[58] A.J. Brunner, "Experimental aspects of mode I and mode II fracture toughness
testing of fibre- reinforced polymer–matrix composites", Comput. Methods
Appl. Mech. Eng., vol. 185, pp. 161-172, 2000.
[http://dx.doi.org/10.1016/S0045-7825(99)00257-1]
[59] A.J. Brunner, B.R. Blackman, and J.G. Williams, "Calculating a damage
parameter and bridging stress from GIC delamination tests on fibre
composites", Compos. Sci. Technol., vol. 66, pp. 785-795, 2006.
[http://dx.doi.org/10.1016/j.compscitech.2004.12.040]
[60] Q.D. Yang, M.D. Thouless, and S.M. Ward, "Numerical simulations of
adhesively-bonded beams failing with extensive plastic deformation", J.
Mech. Phys. Solids, vol. 47, pp. 1337-1353, 1999.
[http://dx.doi.org/10.1016/S0022-5096(98)00101-X]
[61] S. Li, M.D. Thouless, A.M. Waas, J.A. Schroeder, and P.D. Zavattieri, "Use
of a cohesive-zone model to analyze the fracture of a fiber-reinforced
polymer-matrix composite", Compos. Sci. Technol., vol. 65, pp. 537-549,
2005.
240 Frontiers in Aerospace Science, 2016, Vol. 1, 240-268
CHAPTER 5
Optimization of Geometric Parameters and

Reversing Design Methodology of Investment
Yangqing Dou*, Yangliu Dou, Kun Bu and Yiwei Dong

39762, USA
Abstract: With the continual development of the aircraft industry, aircraft engines
have provoked people’s attention more and more. The turbine blade plays a vital and
critical component of aircraft engines. In order to conform to the dimensional
tolerances of wax pattern die-profile for turbine blade in investment casting process,
this chapter provides an optimization method of geometric parameter for turbine
blades based on inverse adjustment. The geometric parameters for optimizing were
extracted, and the bending and torsional deformation can be compensation.
Therefore the nonlinear deformation compensation during solidification and cooling
procedure can be efficiently realized. This method set the theoretical foundation on
optimization method of die-cavity for turbine blade. The die-profile optimization
system which was developed in this paper proves better effect for the die-cavity
design. This chapter also offers a reverse design methodology for investment die
casting using ProCAST. In industry, the performance of the engine depends not only
on shape, but also on the dimensions of the components. This process is difficult as
super-alloy blade material cannot be easily machined. However investment casting
is an ideal process for such net - shape components, but it still requires an accurate
determination of the casting-die profile. In order to investigate and analyze the
methods that affect the shape and dimensions of the turbine blade most, similar
*Corresponding author Yangqing Dou: Department of Mechanical Engineering, Mississippi State University,
MS, 39762, USA; E-mail: yd120@msstate.edu.
Yucheng Liu (Ed.)

Optimization of Geometric Parameters Frontiers in Aerospace Science, Vol. 1 241
simulations have been conducted by ProCAST. By combining the methods of

simplifying grid files and quick sorting, the efficiency of sorting and matching can
be largely improved. Furthermore, the mold die cavity anti-deformation system can
be easily built by utilizing that reverse design methodology. The optimized die
profile for investment casting can be established with ProCAST.
Keywords: Die Cavity, Geometric Parameter, Inverse Deformation,

Investment Casting Die, Node Matching, Numerical Simulation,
ProCAST, Reversing design methodology, Turbine Blade.
NOMENCLATURE
P The initial shape before deformation

Q The shape after deformation
R The shape after reverse deformation operation
D(x, y, z) Objective function
P(x, y, z) The shape function of turbine blade before investment
casting
Q(x, y, z) The shape function of blade after casting process
W(x, y, z) Displacement field function (deformation function) of the
each node (x,y,z) in the FEM model
D The CAD leaf coordinates of ith node
∆Max The threshold for maximum error
Q The simulated surface coordinate of ith node
n The total number of surface nodes.
M Shape form error
X Average form error of width direction
Y Average form error of thickness direction
Z Average form error of length direction
242 Frontiers in Aerospace Science, Vol. 1 Dou et al.
INTRODUCTION
Investment casting, or traditionally called lost wax casting process, is to

make precise metal product without further machining by pouring the
liquid metal into a pre-shaped mold. This process simplifies production
by casting a single complex-shaped piece instead of manufacturing a
product that requires assembling several pieces together [1]. It is well
known that investment casting is used routinely for fabricating
single-crystal nickel super-alloy turbine blades. The turbine blades are
usually of complex geometries with intricate channels, which allow air to
flow within and along the blades during operation [2-3].
A conventional investment casting procedure includes: A. preparing wax

patterns by injecting wax into previously designed dies; B. making
ceramic shells covering the wax patterns; and C. the alloys are cast into
the de-waxed shell molds. It is obvious that the shape of the casting
significantly depends on the cavity geometry of the metal die. Due to the
shrinkage of the wax and solidification of the alloy material, the size of
the component produced by the investment casting process is smaller than
that of the die cavity. Although the volume changes in the solidification
process are simple in nature, the complex geometry of the turbine blade
makes closed-form solutions for the shrinkage almost intractable.
Therefore, in order to ensure the dimensional tolerance, the geometrical
accuracy and the surface roughness, the design of the die profile for
turbine blade needs to consider the compensation of the shrinkages and
thermal distortion during solidification [4].
Due to the complex, time-consuming and expensive process of

investment casting, traditional methods for designing die profile assumes
constant shrinkage rate [5]. However, the complex shape and structure
causes uneven heat dissipation during cooling, and thus the non-linear and
non-uniform shrinkage distribution. Further considerations have to be
given to the shrinkage of wax pattern during preparation; ceramic shell
shrinkage during drying and expansion during casting; super-alloy

material (if applicable) shrinkage during solidification, as well as cooling
from the solidus to room temperature. The dimensional changes
associated with the wax, the shell mold, or the alloy are thereafter referred
to as wax, shell mold, or alloy shrinkage factors (SFs) (or tooling
allowances), respectively. It is a typical practice to calculate the
dimensions of the die tools by adjusting the nominal casting dimensions
by SFs. Moreover, the casting deforms due to the mold constrain stress
and the thermal stress. Taking the cast steel and cast iron materials for
example, the geometrical shrinkage can be in the range of several
millimeters. Therefore, solid shrinking causes the most accuracy loss
during casting.
The principle of designing the die profile is to compensate geometrical

deformation. At present, the establishment of die profile is based on the
linear scaling methodology including uniform scaling method; chord
length scaling method; mean camber line scaling method; and shrinkage
center scaling method [6]. All these methods assume a constant shrinkage
rate for designing the die tool profile. Consequent discrepancy from the
reality is generated due to the negligence of non-uniform distribution
shrinkage. Ito et al. [7] and Ferreira et al. [8] presented a method by
utilizing different shrinkage rates for various directions. However, such
calculation did not consider the influences of casting structure and various
constraint conditions. It also required experiences and efforts for
establishing an appropriate numerical model.
With the help of commercially developed solidification simulation

software based on finite element analysis, reliable simulation results can
be obtained once the thermal properties data and boundary conditions are
available [9-16]. Therefore, the calculated local deformation, due to the
thermal field and geometrical constraints, can provide the evidence for
designing the die profile. Modukuru et al. [17] proposed a grid
displacement reverse stacking method, in which the calculated

deformations were reversed and stacked into each node, this process was
iterated until the deformed shape had good agreement with the ideal one.
With 3-D graphical output, this method can be regarded as a virtual
process of modifying die profile until the accurate numerical simulation
result is achieved. Recently, Zhang Dan et al. [18] developed a simple
reversing methodology of adjusting parameters for the casting die profile
design of turbine blades, in which the casting shrinkages were considered
as nonlinear at different positions. The die profile obtained by employing
this method can satisfy the design requirement. However, a 2-D
configuration of this method limited its extensive applications; the
accuracy and reliability of die profile require further improvement.
OPTIMIZATION OF GEOMETRIC PARAMETERS FOR TUR-

BINE BLADE
Geometric Parameter Extraction of Cross-section
There are more than 20 kinds of parameters which are related to the
structure, strength, and the manufacturing process of turbine blade [19],
however, this paper mainly studies the method in determining die-profile
of turbine blade. Because the blade-design method can be based on
specifying a thickness distribution around the mean camber line, thus,
those parameters which can be used to represent and re-construct the
mean camber line are discussed.
The deformation of turbine blades can be divided into three aspects:

torsional deformation, bending deformation and shrinkage deformation
[20]. Hence, the geometric parameters can be divided into three categories:
the torsional control parameters, the bending control parameters, and the
shrinkage control parameters. In this paper, assuming that the shrinkage
ratio shows the constant change, which means, the shrinkage ratio is using
the experienced value. Thereafter, only two categories of parameters are

discussed in this paper.
Extraction of Bending Control Parameters
The bending control parameters can be expressed by mean camber line. In

essence, the mean camber line is a two-dimensional free-form curve. If
the length of mean camber line remains constant, the freedom of mean
camber line can be restricted by four parameters: α1, α2, L, and Pmax. The
blade exit angle α2 is the angle between the tangents to the camber line at
the trailing, it represents the slope of blade exit part. If the numbers of
discrete points on mean camber line are sufficient, the angle between the
line connecting the last two points of mean camber line and the front of
blade can be regarded as the blade inlet angle (Fig. 1).
2
Fig. (1). Sketch of geometric parameters.
The way of extraction blade exit angle (α1) is similar to the blade inlet
extraction method. Considering the angle between the line connecting the
first two points and the front of blade as the blade inlet angle, the blade
inlet angle can be established.
If chord is considered as the horizontal axis in coordinate system, the

maximum deflection position (Pmax) can be regarded as the point which
has the maximum distance from the mean camber line to the chord. Thus,
a line parallel to the chord can be drawn which being tangent to the mean
camber line, and the tangency point can be identified as the maximum
deflection position
2  1
Fig. (2). Torsion of the section line.
Fig. (3). Sketch of bending control.

Extraction of Torsional Control Parameters
Using the direction which is perpendicular to the front of blade direction

as X axis direction, the leading edge point is the point which has the
minimum abscissa amount, while the trailing edge point is the point
which has the maximal abscissa amount, and the chord can be define as
the line connecting the leading edge point and the trailing edge point.
Then, the stagger angle can be defined as the angle between chord and the
front of blade.
Optimization of Torsional Control Parameters
The stagger angle β is the parameter which can be used to represent the
torsional deformation of blade, hence, the compensation of the torsional
deformation is to compensate the deformation of β. As indicated in Fig. (2),
the solid line represents the schematic diagram of the blade section and the
dashed line represents the section after torsional deformation. β1 represents
the stagger angle of CAD model, and β2 represents the stagger angle of the
measuring model. The torsional deformation can be expressed as ∆β = β2 -
β1, the equation which can be used for torsion deformation compensation
can be rearranged as:
n 1
 m  (lim   i )D   c   D   c (1)
x  i 1 1  
Where βm the optimized is stagger angle, ∆D is equal to ∆β, βc is stagger

angle of CAD model.
Optimization of Bending Control Parameters
The bending deformation can be denoted by the blade inlet angle α1 , the
blade exit angle α2, the maximum deflection position Pm, and the chord
length L (Fig. 3). The optimization of the bending control parameters is to

compensation each parameter properly. However, these parameters are not
independent and are of interaction and restricted mutually. Therefore, it is
necessary to determine the relationship of geometric parameters firstly. As
is shown in Fig. (3), the coordinate value of A dot (XA, YA) and the last
point of mean camber line B (m, M) need to be determined, the
geometrical meaning of A dot is the intersection point of the tangent lines
of the first and last point of mean camber line. Hence, the establishment of
the coordinate value of dot A can be calculated based on two-point form
of straight-line equation.
Then respectively calculating the coordinate values of A dot in CAD

model and measuring mode of the certain turbine blade, based on the
compensation of discrete points, the optimized coordinate value of A dot
can be determined by optimized coordinate value of A dot can be
determined by
y2  y1
Pxy ( x, y )  (1  )Wxy ( x, y )  P0 ( x0 , y0 ) (2)
x2  x1
Where P0(x0, y0) is the coordinate value of an arbitrary dot of CAD model,
Pxy (x, y) is the optimized coordinate value of this dot, Wxy (x, y) is the
distance between dot $A$ in CAD model and measuring model,
respectively.
The coordinate value of dot B, which is the last point of mean camber line
can be solved in the same way by the distance compensation method. (Eq.
(2).
By solving a quadratic equation with cot α1 as a variant, the dependent

variables are the coordinate value of the last point of mean camber line (m,
M), the x-coordinate value of dot A, and the angle tan  , the relations
between chord length L and cot α1 can be established as:
cot   1/ 2TA[ ( L  M )  MT  ( ( L  M )  2 ( L  M ) MT
(3)
 M 2T 2  4T 2 A2  4TA  4T 2 ( L  M ) A) (1/ 2) )]
Similarly, both the relationship between cot α2 and L, the relationship

between Pm and L can be established. The compensation of the
parameters which can be used for controlling the bending deformation can
be transformed into the compensation of the chord length L. The
compensation method can be written as:
 n
 1
P    lim   i  D  P0  D  P0 (4)
 n 
i 1  1 
 is the shape coefficient, for the form error is related to the structure of
the casting, it is noticed that each form error shows some similarities.
Hence, a shape coefficient noted is introduced.
Fig. (4). Sketch of constant curve length.

Using Eq. (4) to compensate the bending deformation, the shape

coefficient  must be firstly determined. Assuming the length of mean
camber line keeps unchanged after investment casting, based on calculus
relation, the whole mean camber line can be divided into infinite units
equally, shown in Fig (4). Assuming that the length of curve-unit is
sufficiently small, thus, the curve-unit can be interpreted as the straight
line, which can be formed with a right-angled triangle by chord length and
height of segment H.
Fig. (5). Parameters extraction interface.
Fig. (4) shows that the change of chord length is influenced by the height
of segment variation, the shape coefficient can be expressed as ∆Hi+1 /
∆Hi. Where H represents the difference between the height of segments
of measuring model and CAD model. ∆Hi represents the difference of
intervallic deformation between two iterative calculation steps.
Therefore, the optimized chord length can be determined, meanwhile, the

optimized blade inlet angle, optimized blade exit angle and the optimized
maximum deflection position can be obtained, and the bending
deformation can be compensated.
Results and Discussion
The extraction of a type of certain turbine blade is shown in Fig. (5). After
the extraction of parameters, the corresponding data of CAD model and
measuring data are calculated. For the cross- section at height of 50mm,
the results are shown in Fig. (6), and the optimize cross-section is shown
in Fig. (7). The lines shown in Fig. (7) are the mean camber line, the
thicker line represents the optimized mean camber line, while the thinner
line represents the original one.
Fig. (6). Optimized geometric parameters.
Based on the method proposed above, using several (4-6) sections to

reconstruct the die profile. Although the optimized die-profile for the
investment-casting of turbine blade is established, however, the
die-profile's availability obtained by the method proposed still needs to be
further investigated by numerical simulation. Using ProCAST [21], the
displacement of three directions can be obtained.
Fig. (7). Optimized cross-section.
The displacement field of three directions is shown in Fig (8). Where Fig.
(8a) represents X direction, Fig. (8b) represents Y direction, and Fig. (8c)
represents Z direction.
Fig. (8). Displacement field of three directions.

From Fig. (8), the maximum displacement along Z direction is 0.908mm,

the dimension is still can’t satisfy the accuracy requirements. Therefore,
based on the numerical simulation results, using the inverse method to
optimize the die cavity again, calculating the displacement field by
ProCAST based on the same material properties and boundary conditions,
the simulation results are shown in Fig (9). From Fig (9), the maximum
displacement along Z direction is 0.545mm. The deviation is decreased,
which proves the methods proposed in this paper can improve the casting
accuracy efficiently.
Fig. (9). Displacement fields of three directions.
In order to conform to the dimensional tolerances of wax pattern

die-profile for turbine blade in investment casting process, an optimized
method for the determination of geometric parameter of die cavity is
proposed in this chapter. The geometric parameters describing die-cavity
are extracted and optimized, the deformation of the turbine blade occurred
in solidification process can be compensated. This method therefore
exhibits its advantage over the traditional linear scaling method. An
optimization system is developed to realize the optimization process,
which proves better effect on the die cavity design.
OPTIMUM DESIGN METHOD OF DIE PROFILE
The traditional way for designing die profile is to reserve certain

deformations opposite the deformation directions obtained by numerical
simulation or experience. In order to prevent deformations, the shape of
casting would be equal to that of the CAD model after solidification. Fig.
(10) illustrates the basic principle of reverse deformation method, in
which line P represents initial shape, line Q is the shape after deformation,
and line R is the shape after reverse deformation operation.
Fig. (10). Sketch for the basic principles of reverse deformation.
Due to the complicated shape of turbine blade, the accuracy requirement

cannot be satisfied by one step compensation. Consequently, this paper
adopts cyclic iteration to approach the target shape. Assume D(x, y, z),
P(x, y, z), Q(x, y, z) and W(x, y, z) are the objective function, the shape
function of turbine blade before investment casting, the shape function of
blade after casting process, and the displacement field function
(deformation function) of the each node (x,y,z) in the FEM model, the
initial shape of Pi(xi, yi, zi) will evolve into the deformed shape Qi(xi, yi, zi)
after i times inverse iterations. The casting result is then compared with
the target shape and the difference of ∆Di(xi, yi, zi) is calculated as:
Di ( xi , yi , zi )  Qi ( xi , yi , z i )  D( x, y, z ) (5)
If difference ∆Di is less than ∆Max (the threshold for maximum error),
Pi(xi, yi, zi) can be taken as the ideal deformed shape. Otherwise, the
deformed shape must be adjusted by -∆Di (as shown in Eq. 6) and
program iterates until the maximum error threshold is met. Fig. (11)
shows the program flow chart of such iteration. The final reverse
deformation function can be established as Eq. 7.
Pi  Di 1  Pi 1 (6)
P( x, y, z )  Pi ( xi , yi , zi ) (7)
Fig. (11). Flow chart of the solution algorithm of the reverse deformation.
ANALYSIS OF FILES GENERATED DURING NUMERICAL

SIMULATION PROCESS
The general flow chart of ProCAST simulation process is presented in Fig.

(12).
CAD model PreCAST *_shelld.dat
CAE model *.sm DataCAST *_shelld.out
Form shell *_shell.sm ProCAST
Volume
*_shell.mesh ViewCAST *_shellgd.dat
mesh
Fig. (12). Flow chart of the simulation process based on ProCAST.
The primary workflow of ProCAST can be divided into three stages -

“Pre-Processing”; “Problem Solving” or “processing”; and “Date Output”.
In the first stage, the 3-D entity model and gating system are established
by CAD software, such as UG. ParaSolid is employed to deliver the
model information and to convert file format if necessary. After the CAD
model is imported into ProCAST, MeshCAST module - a fully automatic
3-D tetrahedral mesh generator - is called to generate surface mesh, the
shell and the volume mesh. PreCAST module deals with the model
geometry and defines conditions (such as assigning materials, defining
interface, setting boundary conditions, appointing process, and selecting
suitable running parameters). It is noted that each chosen surface can only
be assigned by one boundary condition, therefore the definition of surface
boundary condition is critical. DataCAST module manages data,

including converting the text files generated by PreCAST module to
binary format. FEM calculation takes place in ProCAST module and the
results (such as temperature field, stress field and displacement field
during simulation) can be rigorously visualized by ViewCAST module.
The intermediate files produced in each step during simulation using

ProCAST are also included in Fig. (3). For example, *.sm is the surface
mesh file recording the number of surface mesh nodes and the
corresponding coordinates; * _shell.sm is the casting and shell surface
mesh files; * _shell.mesh is volume mesh file collecting volume mesh
information; *_shelld.dat file records all information in *_shell.mesh files
and interface between shell and casting; *_shelld.out files contain
boundary conditions and thermo-physical properties; and *_shellgd.dat
files record all volume mesh information upon termination.
FEM PROCESSING
The FEM process using ProCAST can be divided into two aspects: A.
collect necessary node information by reading corresponding mesh files
(refer to Fig. 13); and B. generate plenty grid nodes for accuracy during
the simulation process.
During FEM process, sorting nodes has to be optimized for not only
reducing the CPU time but also shortening the node matching process.
Rearrangement of nodes first occurs after ProCAST automatically assigns
a contact node at the interface of casting and shell (those nodes are
generated in MeshCAST module) and put this new node at the end of the
file.
DataCAST module transforms *_shelld.dat from text into binary format

to improve execution efficiency. Node rearrangement also occurs in this
module, in which nodes are sorted with updated coordinate information.
The sorted node information is then exported into *_shelld.out file, which
also contains an index table recording the corresponding relationship
among grid nodes.
Read node
Input mesh file
identification
Read node and Read node

element number coordinate N
Read data characteristic word

code judgement
N Y
characteristic word
code judgement End
Fig. (13). Flow chart of the mesh information acquisition.
Node matching is completed by processing two rearrangements

simultaneously. Shown in Fig. (14), two index tables are created after the
rearrangement and then the corresponding relationship of the nodes can
be obtained by referring to the index table.
Then the mesh files should be simplified and the necessary node
coordinates information is extracted. In current work, the demarcation
searching method is used to search the corresponding nodes. Demarcation
searching method is commonly used to compare the data sizes by
boundary values [18]. First, the data are sorted by initial boundary values
and the boundary value will be adjusted in accordance with demarcation
results. Repeat the comparison process until all data are divided into
orderly small units, as shown in Fig. (15).
Read *sm file Read *shelld.dat file Read *shelld.out file Read *shellgd.dat file
Find corresponding node
substract
Remove
Judgement
shell node
Index tableA--B Index tableB--C
Index tableA--C
Corresponding node
Fig. (14). Flow chart of grid nodes matching information.
Fig. (15). Sketch of demarcation searching process.
Demarcation searching method sorts by employing a divide and conquer

strategy to divide a list into two sub-lists. The steps are: firstly pick an
element, called a pivot, from the list. Then reorder the list so that all
elements which are less than the pivot come before the pivot and so that
all elements greater than the pivot come after it (equal values can go
either way). After this partitioning, the pivot is in its final position. This is
called the partition operation. Finally recursively sort the sub-list of lesser
elements and the sub-list of greater elements. The base cases of the
recursion are lists of size zero or one, which are always sorted.
EXAMPLE VERIFICAITON
DD6 alloy is chosen as the turbine blade material, and the mold material
is silica sand (thermo-physical properties of DD6 alloy and the mould
material can be found in Ref [22-23]). Fig. (16) shows a typical blade
mould design with a gating system. To simplify the geometry of the
turbine blade modeling and to improve the computational efficiency, only
one third blade module group is chosen (shown in Fig. 17). Further, we
adopt gravity filling method and ignore the influence of filling period. We
assume the initial casting temperature is 1550℃, the temperature of the
shell mould (pre-heat) is 1500℃, and the calculation is terminated at a
temperature of 600℃.
Simulation result (including intermediate results) can be extracted from

the data file generated after each iteration step. Fig. (18) illustrates the
deformation of the turbine blade during simulation process at the steps of
80, 150, 300 and 640, respectively. Fig. (18a) represents the deformation
of blade width (X direction); (b) represents the thickness deformation (Y
direction), (c) represents the length deformation (Z direction). It has been
found from the simulation data that severe blade surface deformation is in
the width direction (Fig. 18a). Non-uniformly distributed stress field is
caused by complex heat dissipation in the solidification process. It is
anticipated that the blade back undertakes tensile stress and the
thin-walled part of the blade pressure side undertakes compressive stress.
The shrinkage of the blade turns toward the suction direction and the
maximum deformation.
Fig. (16). Gating system design and mesh generation.
Fig. (17). The simplification of the typical turbine blade.
The displacement field of the turbine blade represents the summation of

nodes displacement. It is determined by the structure of the blade and the
boundary conditions (reflecting the blade and casting process). However,
the casting dimensional change is also associated with wax pattern, wax
property, wax temperature, wax flow and the manufacturing process, etc.
In this paper, we assume the wax pattern is fixed with less distortion or
even no distortion. The influence of wax pattern dimensional changes is
negligible.
S tep=80 S tep=150 S tep=300 S tep=640
(a) Width direction
Step=80 Step=150 Step=300 Step=640
(b) Thickness direction
Step=80 Step=150 Step=300 Step=640
(c) Length direction
Fig. (18). Displacement field based on numerical simulation.

Based on UG, C++ language editor was used for developing "Investment
Casting Blades’ Mould Optimization System". One example of the
system interfaces is shown as Fig. (19).
Fig. (19). Interface of anti-deformation system for investment casting.
This system module reads files generated from each simulation step,
orders and matches the grid nodes while updating the displacement field.
The output of the system is the new surface mesh file. Fig. (20) shows a
new surface mesh of die profile of a turbine blade.
Fig. (20). Optimized die cavity point cloud.

Then the investment casting process should be simulated again so as to

obtain a new turbine blades model. Based on ProCAST, shape error is
used to evaluate the results of displacement field compensation. The
shape form error can be defined as follows:
| Q  D |
i i
M i 1
(8)
n
where D the CAD leaf is coordinates of ith node, Q is the simulated

surface coordinate of ith node, N is the total number of surface nodes.
The shape form error M reflects the coincidence degree of two surface
meshes. The smaller the form error is the more optimized mesh surface
from the original CAD one. Let X,Y,Z denote the average form error of
three directions, respectively, one can see from Table 1 that after 4th
iterations, the synthetic form error decreases from 0.516mm to 0.002mm.
Such convergence effect can meet the error requirement of the die cavity
design for turbine blades.
Table 1. Change of form error with iteration times increasing.

Cavity part 1st iteration 2nd iteration 3rd iteration 4th iteration
X (Width) 0.194366mm 0.001802mm 0.001249mm 0.001326mm
Y (Thickness) 0.054674mm 0.002627mm 0.001771mm 0.001205mm
Z (Length) 0.453754mm 0.002917mm 0.001035mm 0.000964mm
Total form error 0.515815mm 0.004667mm 0.002557mm 0.001978mm
Fig. (21) summaries a turbine blade compensation experiment showing

the form error changes as the iterative times increases. With A
representing the blade width deformation error, B for the thickness
deformation error, C for the length deformation error and D for the
comprehensive deformation errors, the deformation error shows a
descendant trend with iteration times.
Fig. (21). The error of blade with anti-deformation in an iterative way.
CONCLUSION
Based on ProCAST, the mesh files generated during numerical simulation

are used as input for the mesh information processing system. The quick
sorting and matching of nodes are realized, the disorder nodes can be
transformed into ordered ones. By analyzing the characteristics of mesh
files, a reversing design methodology is proposed; the form error can
meet the design requirement after iteration. The investment casting blades’
mould optimization system was developed and verified with an example.
This research has potential applications in optimizing die profile design,
reducing cost and increasing productivity.
this publication.
ACKNOWLEDGEMENTS
I hereby to acknowledge Dr. Kun Bu and her research team for their
contributions to my inspiration and knowledge and other help in creating
this chapter. This work was supported by National Key Technologies
R&D Program and was executed in the Key Laboratory of Contemporary
Design and Integrated Manufacturing Technology, Northwestern
Polytechnical University, China.
REFERENCES
[1] Y. Zhang, W.K. Liu, and H. Wang, “Cast filling simulations of

thin-walled cavities", Comput. Methods Appl. Mech. Eng., vol. 128, no. 3-4,
pp. 199-230, 1995.
[http://dx.doi.org/10.1016/0045-7825(95)00882-9]
[2] P. Carter, D.C. Cox, and C.A. Gandin, "Process modelling of grain selection
during the solidification of single crystal superalloy castings", Mater. Sci. Eng.,
vol. 280, no. 2, pp. 233-246, 2000.
[http://dx.doi.org/10.1016/S0921-5093(99)00701-7]
[3] J. Li, S. Liu, and H. Yuan, "Solidification simulation of investment castings of
single crystal hollow turbine blade", J. Mater. Sci. Technol., vol. 06, pp.
532-534, 2003.
[4] T.C. Chang, and E. Faison, "Shrinkge, behavior and optimization of injection
molded parts studied by the taguchi method", Polym. Eng. Sci., vol. 41, no. 5,
pp. 703-710, 2001.
[http://dx.doi.org/10.1002/pen.10766]
[5] S. Jones, and C. Yuan, "Advances in shell moulding for investment casting", J.
Mater. Process. Technol., vol. 135, no. 2-3, pp. 258-265, 2003.
[http://dx.doi.org/10.1016/S0924-0136(02)00907-X]
[6] C.C. Tai, “The optimization accuracy control of a die-casting product part", J.
Mater. Process. Technol., vol. 103, pp. 173-188, 2000.

[http://dx.doi.org/10.1016/S0924-0136(00)00452-0]
[7] M. Ito, T. Yamagishi, Y. Oshida, and C.A. Munoz, "Effect of selected physical
properties of waxes on investments and casting shrinkage", J. Prosthet. Dent.,
vol. 75, no. 2, pp. 211-216, 1996.
[http://dx.doi.org/10.1016/S0022-3913(96)90101-8] [PMID: 8667282]
[8] J.C. Ferreira, and A. Mateus, "A numerical and experimental study of fracture
in RP stereolithography patterns and ceramic shells for investment casting", J.
Mater. Process. Technol., vol. 134, no. 1, pp. 135-144, 2003.
[http://dx.doi.org/10.1016/S0924-0136(02)01034-8]
[9] M.A. Gafur, M.N. Haque, and K.N. Prabhu, "Effect of chill thickness and
superheat on casting/chill interfacial heat transfer during solidification of
commercially pure aluminium", J. Mater. Process. Technol., vol. 133, no. 3,
pp. 257-265, 2003.
[http://dx.doi.org/10.1016/S0924-0136(02)00459-4]
[10] B.C. Liu, J.W. Kang, and S.M. Xiong, "A study on the numerical simulation of
thermal stress during the solidification of shaped castings", Sci. Technol. Adv.
Mater., vol. 2, no. 1, pp. 82-0708, 2001.
[http://dx.doi.org/10.1016/S1468-6996(01)00041-9]
[11] J.M. Drezet, M. Rappaz, G.U. Grun, and M. Gremaud, "Determination of
thermophysical properties and boundary conditions of direct chill-cast
aluminum alloys using inverse methods", Metall. Mater. Trans., A Phys.
Metall. Mater. Sci., vol. 31A, pp. 1627-1634, 2000.
[http://dx.doi.org/10.1007/s11661-000-0172-5]
[12] T. Hong, J.Y. Ooi, and B. Shaw, "A numerical simulation to relate the shot
peening parameters to the induced residual stresses", Eng. Fail. Anal., vol.
15, no. 8, pp. 1097-1110, 2008.
[http://dx.doi.org/10.1016/j.engfailanal.2007.11.017]
[13] L.M. Galantucci, and L. Tricarico, "A computer-aided approach for the
simulation of the directional- solidification process for gas turbine blades", J.
Mater. Process. Technol., vol. 77, no. 1-3, pp. 160- 165, 1998.
[http://dx.doi.org/10.1016/S0924-0136(97)00413-5]
[14] T. Varady, R.R. Martin, and J. Cox, "Reverse engineering of geometric
models-an introduction", Comput. Aided Des., vol. 29, pp. 255-268, 2008.
[http://dx.doi.org/10.1016/S0010-4485(96)00054-1]
[15] G.H. Liu, Y.S. Wong, and Y.F. Zhang, "Han Tong Loh,“Adaptive fairing of
digitized point data with discrete curvature", Comput. Aided Des., vol. 34,
no. 4, pp. 309-320, 2002.
[http://dx.doi.org/10.1016/S0010-4485(01)00091-4]
[16] S.C. Modukuru, N. Ramakrishnan, and A.M. Sriramamurthy, "Determination
of the die profile for the investment casting of aerofoil-shaped turbine blades
using the finite-element method", J. Mater. Process. Technol., vol. 58, no. 2-3,
pp. 223-226, 1996.
[http://dx.doi.org/10.1016/0924-0136(95)02127-2]
[17] M. Fackeldey, A. Ludwig, and P.R. Sahm, "Coupled modelling of the
solidification process predicting temperatures, stresses and microstructures",
Comput. Mater. Sci., vol. 7, no. 1-2, pp. 194-199, 1996.
[http://dx.doi.org/10.1016/S0927-0256(96)00080-8]
[18] Z. Dan, W. Zhang, and W. Min, "Reversing design methodology of the die
profile in investment casting based on the simulation of displacement field and
identification of featured parameters", Hangkong Xuebao, vol. 03, pp. 509-
514, 2006. [In Chinese].
[19] D.A. Reed, J. Wind Eng. Ind. Aerodyn., vol. 33, no. 3, pp. 487-494, 1990.
[http://dx.doi.org/10.1016/0167-6105(90)90003-U]
[20] Z. Bing, and L. Zhiqiang, "Three dimensional FEM simulation of Titanium
hollow blade forming process", Rare Metal Mat. Eng, vol. 39, no. 6, pp.
963-968, 2010.
[http://dx.doi.org/10.1016/S1875-5372(10)60106-3]
[21] A. Krimpenis, and P. Benardos, "Simulation-based selection of optimum
pressure die-casting process parameters using neural nets and genetic
algorithms", Int. J. Adv. Manuf. Technol., vol. 27, pp. 509- 517, 2006.
[http://dx.doi.org/10.1007/s00170-004-2218-0]
[22] X.H. Li, W. Mao, and W.L. Guo, "Transient liquid phase diffusion bonding of
a single crystal superalloy DD6", China Welding, vol. 458, pp. 281-289,
2005.
[23] W. Bonilla, S.H. Masood, and P. Iovenitti, "An investigation of wax patterns
for accuracy improvement in investment cast parts", Int. J. Adv. Manuf.
Technol., vol. 18, no. 5, pp. 348-356, 2001.
[http://dx.doi.org/10.1007/s001700170058]
CHAPTER 6
Vibration Induced Fatigue Analysis of

Aerospace Structures
Murat Aykan1,* and Mehmet Çelik2
1
Department of Mechanical Engineering, SST Division-ASELSAN Inc.,
06172, TURKEY
2
Department of Mechanical Engineering, MGEO Division-ASELSAN Inc.,
06011, TURKEY
Abstract: This chapter discusses the structural analysis aspects of aerospace

structures focusing on vibration induced fatigue. Metal Fatigue of dynamically
loaded structures is a very common phenomenon in engineering practice. Several
aerospace structures are used in environments where they experience dynamic
loading. Furthermore, when the loading is dynamic, the response of the structure is
affected by the structural resonances. Thus, the structural response to the loading will
amplify at the regions of resonance. As a case study, an antenna (12-59 V/UHF)
integration process on helicopter is investigated where the importance of the dynamic
response is highlighted. Flight tests and finite element analyses (FEA) are carried out
to ensure the safety of the integration process. Furthermore, another important aspect
of the structural analysis of aerospace structures is the testing strategy. In this study,
the fatigue tests performed by uni-axial tests which is a procedure defined in military
standards are compared to multi axial testing. The results are shown for a helicopter
Chaff/Flare Dispenser Bracket. The military standards assume that cumulatively uni-
axial testing has equal multi axial testing fatigue damage. In this study, the uni-axial
*Corresponding author Murat Aykan: Department of Mechanical Engineering, SST Division-ASELSAN

Inc., 06172, TURKEY; Tel/Fax: +903125921000; E-mail: m.aykan@hotmail.com.
Yucheng Liu (Ed.)

270 Frontiers in Aerospace Science, Vol. 1 Aykan and Çelik
fatigue tests were compared to multi axial fatigue tests which were performed by
FEA simulations. Furthermore, the effects of various loading conditions and
geometries were investigated. The study showed that the assumption of multi axial
testing can be represented by uni-axial testing does not hold for various cases.
Keywords: Computer modeling and simulation, Multi-axial testing,

Uni-axial fatigue testing error, Vibration induced fatigue.
INTRODUCTION
Mechanical structures usually work in a dynamic loading environment

and most of the time the loading is repeating itself in time. This repetitive
loading doesn’t have to be above the yield stress of the structure for
failure. Repetitively loading the structure will ultimately cause failure.
This phenomenon is called as Fatigue Failure. Typically, fatigue failure

starts at the highest stress zones by forming cracks and then propagates
under cyclic loading, where the stress state can be still under the yield
point of the material. When a limit is reached for the number of cycles of
loading the component fails at a fatigue failure surface. Unfortunately,
most of the time fatigue failures cannot be detected until catastrophic
accidents occur. In the past, many accidents due to the fatigue failures in
metals have occurred. Fig. (1) shows some of them.
Furthermore, according to the survey [3] for helicopter component fatigue

failures, these failures are approximately 55% of all premature failures in
helicopter components as shown in Table 1.
This chapter will consist of two case studies which will focus on different
analysis and testing aspects of aerospace structures. The first section will
investigate an antenna integration process including flight tests and stress
analysis where the structure will not require fatigue analysis [4]. The
antenna is part of the MXF-484 V/UHF transmitter, which is used to
safely communicate without any enemy interference. Fig. (2) shows the
integration area on the helicopter.
Vibration Induced Fatigue Analysis Frontiers in Aerospace Science, Vol. 1 271
(a) (b)
(c) (d)
Fig. (1). (a) Comet fuselage failure (b) Stress concentrations at window corners [1] (c) Aloha
Airlines Boeing 737 fuselage failure due to multiple cracks at rivet holes (d) Crack growth on
fuselage [2].
(a) (b)
Fig. (2). (a) Antenna location on helicopter, (b) Measurement locations when antenna is not
installed.
Table 1. Main causes of failure of helicopter components [3].
Main Category Description Percentage
Fatigue High and Low Cycle, Pitting & Spalling 54.8
Corrosion Intergranular, Galvanic & Stress Corrosion 11.7
Wear Adhesive - Fretting, Scuffing & False Brinelling 10.3
Overload Component Abuse 10
Manufacturing Laps, Cold Shuts, Grinding Abuse, Heat Treatment

6.4
Defects Cracking & Arc Burns
Debond Metal to Metal or Metal to Composite 3.7
Miscellaneous Loss of Lubrication 1.9
Embrittlement Hydrogen & Metallic 1.2
The second section will be focusing on vibration induced fatigue analysis

and testing of a helicopter’s Chaff/Flare Dispenser Bracket which is an
extended version of the study given by Aykan et al. [5]. The Flare (Fig. 3)
is used to misguide missiles where small, very hot particles are disposed
from the helicopter whereas Chaff hides the radar signature of the
helicopter where electromagnetic particles are disposed.
Fig. (3). Application of flare.
FLIGHT TESTS AND NUMERICAL ANALYSES OF AN

ANTENNA FOR SAFETY OF FLIGHT
Operational Flight Tests before Integration
In order to assess the safety of the integration process, vibration analysis

and testing of the antenna integration region must be performed. First of
all, the loading to the antenna has to be measured. The measured values of
pre-integration state will also be used to compare the test results that are
measured after integration. Flight tests are performed according to a flight
profile which was defined by test pilots. It includes steady flight sections
as well as rough maneuvers that the helicopter will encounter through its
life. The vibration levels of the antenna integration location are measured
by using 7 ICP type tri-axial accelerometers (Fig. 2b) where the sampling
rate was chosen to be 5000 S/s.
Numerical Structural Analyses
The structural analyses consist of modal analysis and vibration analysis.

In order to have accurate results, a wider portion compared to the antenna
integration region of the helicopter is modeled. The cowling model
including the antenna region is given in Fig. (4). 50859 Shell elements
with 90537 nodes are used for modeling the cowling in ANSYS [6].
The edge numbered by "1" as shown in Fig. (4a) is a locked edge and the
edge numbered with "4" is far away from the location analyzed. Therefore,
the degrees of freedoms of these edges are fixed. Quarter turn fasteners
are used close to the edges "2" and "3" and these edges are not fixed but
simply supported. The materials used in the analysis consist of aluminum
2024-T3 and steel 4140.
Modal analysis of the cowling model is performed for with/without
antenna configurations. The mode shape for the first natural frequency of
the cowling changed from bending to rotation and the first natural
frequency of the antenna integration region has decreased from 154.3 Hz
to 67.9 Hz as shown in Fig. (5).
(a) (b)
Fig. (4). Cowling boundary conditions, (a) 1 and 4 fixed, 2 and 3 simply supported, (b) antenna
model (point mass element).
(a) (b)
Fig. (5). Mode shape of the antenna integration region for the first natural frequency, (a) without
antenna, (b) with antenna.
Similarly, the stress distributions of the cowling with and without the
antenna are also needed to assess the safety of integration. A smaller
region of the cowling Fig. (6) is included in the vibration analysis as the
modal analysis results showed that the first 3 modes of the antenna region
are local modes. The analysis model has 19959 elements and 60935
nodes.
The random time history signals are usually defined in frequency domain
as Power Spectral Density (PSD) functions [7]. PSD presents this
information as a statistical spectrum where the area under the curve
represents the mean square amplitude of the wave other than the
amplitude. Since PSD values are needed for the vibration analysis, the
tri-axial vibration measurements are transformed to PSD and applied as
loading boundaries at the measurement locations in the FEA.
Fig. (6). Simplified model created for PSD analyses.
Maximum equivalent von Mises stresses are obtained at the cowling

mounting holes when the antenna is not present as shown in Fig. (7a) and
at antenna mounting holes when the antenna is present (Fig. 7b). Since
PSD stresses are statistically 1σ results, maximum stress values should be
tripled to achieve 3σ results. That is, three times larger stresses may occur
on the given location statistically (Fig. 8).
(a)
(b)
Fig. (7). Stress distributions of vibration (PSD) analysis, Max stress location at the antenna
integration area, (a) without antenna, (b) with antenna.
Fig. (8). Gaussian distribution of stress values.

Table 2 summarizes the results of the numerical modal and vibration

analyses. It is clear that a significant increase in maximum stress and
decrease in first natural frequency results from integration of the antenna.
However, as even the 3σ stress values are small compared to the fatigue
strength of the aluminum alloy used, the antenna integration is not fatigue
critical.
Table 2. Summary of antenna location numerical analysis.
1st Natural Frequency of the Max. Stress Max. Stress

Antenna Region (1σ) (3σ)
With Antenna 67.9 Hz 26.4 MPa 79.2 MPa
Without Antenna 154.3 Hz 1.2 MPa 3.6 MPa
Operational Flight Tests after Integration
Numerical results showed that there is little risk for the installation of the
antenna. In order to validate the integration, flight tests with the antenna
are performed according to the same flight profile. A total of seven
tri-axial accelerometers are placed on the cowling portion on which the
antenna is mounted. Accelerometer locations for this configuration are
shown in Fig. (9).
The acceleration data collected during flight tests are compared with the
previous test where the antenna was not installed. According to the PSD
comparison, antenna integration has increased the overall vibration levels
at some of the measurement locations. A sample comparison is given for
location 5 in Fig. (10).
Fig. (9). Measurement locations when antenna is installed.
Fig. (10). PSD comparison of location-5 on cowling, Y (lateral) and Z (vertical)-axes, with antenna
and without antenna.
PSD graphs are also compared with the levels given in the military
standard MIL-STD-810F [8]. Fig. (11) shows the comparison of
measured lateral vibrations at location-5 with a sine-on-random graph
which is specified in MIL-STD-810F. Fig. (11) shows that the peaks
occur at the same frequencies as in the standards’ profile. In addition,
floor levels do not exceed the random portion of the standards’ profile.
This is valid for all the locations where vibration data is collected. This
comparison gives confidence that the integration process does not
significantly change the overall vibration characteristics of the helicopter
even at the locations where the vibration levels have increased.
Fig. (11). Comparison of experimental vibration level of location-5 at lateral axis with military
standard specification (MIL-STD-810F).
VIBRATION INDUCED FATIGUE ANALYSIS AND TESTING OF

A CHAFF/FLARE DISPENSER BRACKET
The design of a bracket for a chaff/flare dispenser has many requirements

to fulfill. Static, modal, transient and fatigue analyses must be performed
to complete the design with required safety factors. In this study, fatigue
analysis and testing of an aerospace structure exposed to broadband
loading is highlighted. Traditionally, fatigue analyses have been
performed in the time domain where the response is calculated by static
analysis. However, studies [9-12] show that when the loading excites the
natural frequencies of the structure, time domain approach lacked the
dynamics of the structure. Therefore, the analyses in this study are
performed in the frequency domain using random vibration theory. Fig.
(12) shows the basic flow for both time and frequency domain methods.
Moreover, it is common to perform fatigue tests under uniaxial loading

conditions. Even when multi-axial analyses are performed, the S-N
diagrams used are obtained from uni-axial tests. There are numerous
studies in the literature which point out that multi-axial fatigue analyses
are necessary when higher accuracy is required [13-16]. However, the
importance of multi-axial testing in fatigue analysis has yet received
much attention. Recent studies showed that multi-axial testing affects the
failure duration, mode and the distributions even on very simple
specimens [17].
TIME DOMAIN
TIME RAINFLOW
Steady HISTORY COUNT PDF
state
or FATIGUE
LIFE
Transient
Analysis
FREQUENCY DOMAIN
PSD FATIGUE PDF
MODELLER
Transfer M0
M1
FATIGUE
Function M2
BLACK
BOX
LIFE
M
4
Fig. (12). Time and frequency domain methods [7].

Vibration Fatigue Analysis Method
Vibration fatigue method, which is a frequency domain method, is

applicable when the encountered loading is random in nature and the
statistical information of the process has to be evaluated. The random
processes can be categorized as broad band and narrow band processes.
Furthermore, it may consist of a broad band spectra with periodic waves
on top of it (i.e. sine- on-random), which is the case in helicopters. The
random time histories are usually expressed as PSD functions. Sample
time histories and corresponding PSD functions are given in Fig. (13).
Fig. (13). Time histories and corresponding PSD graphs [7].
Given a measurement for a duration T of the random time series y(t), its
PSD, Gyy(f,T) can be defined [18] as Eq. 1 where all phase information is
lost in the process.
G yy ( f , T ) 
2
T

 E Y ( f ,T )
2
 (1)
and Gyy(f) can be defined as;
G yy ( f )  lim
T 
2
T

 E Y ( f ,T )
2
 (2)
where f is the frequency, E the averaging operator and Y(f,T) is the Fourier
Transform of y(t).
Furthermore, in order to obtain the response of the structure to random

loading, transfer functions are used. (Eq. 3).
Y ( f )  H ( f ). X ( f ) (3)
If the PSD of response is obtained by using Eq. 2 and Eq. 3 where the
input is written as a PSD function,
Gyy ( f )  H ( f )  H * ( f )  Gxx ( f ) (4)
where, ‘*’ shows complex conjugate of the functions.
Furthermore, when the system is excited by multiple partially correlated

inputs where m is the number of partially correlated inputs, the PSD of the
response becomes,
m m
G yy ( f )   H i ( f ).H j * ( f ).Gij ( f ) (5)
i j
Thus, using Eq. 5, the stress response PSD values can be calculated for a
specific location of the structure where the loading induced is partially
correlated.
Furthermore, in Vibration Fatigue analysis, the Probability Density

Function (PDF) of Rainflow ranges [19] has to be extracted for damage

calculations. By using such a method, the PDFs will be used in the stress
ranges counting, instead of the classical Rain flow counting method
which is used in time domain. The PDF is derived from the response
stress’ PSD. The elements of the loads' counting process are given in Fig.
(14).
Fig. (14). Schematic description of the elements of the loads’ counting process.
The number of stress ranges of size ∆S (N∆S) during a certain loading

duration T can be evaluated from the relation;
N S  E[ P]  T  p (S  d S  S  S  d S )
(6)
where;
E[P]: the number of stress ranges of all levels per time unit, p(∆S-d∆S<∆S
<∆S +d∆S ): the probability that the stress range ∆S will appear in interval
(∆S-d(∆S) : ∆S+ d(∆S)).
The key element of vibration fatigue analysis is the PDF and PDF can be
obtained from the PSD of stresses [20]. The characteristics of PSD that
are used to obtain this information are the spectral moments of the PSD
function. The nth spectral moment of the PSD is defined as shown in Fig.
(15) and expressed as in Eq. 7.
Fig. (15). Spectral moment calculation of a PSD function.
 M
mn   f n  G yy ( f )  df   f k  G yy k ( f k )  f
n
(7)
0 k 1
The spectral moments can be used to calculate the expected value of

peaks per second, E[P], which is used in the calculation of fatigue life. In
order to derive the fatigue damage equation for a random process, let
p(∆S) to be defined as the Probability Density Function (PDF) of the
random process. For a resonant system subjected to a broadband
excitation, in time T there will be on average E[P]·T stress cycles of
which p(∆S)d∆S will have peak values in the stress range ∆S to ∆S +d∆S.
Thus, the damage at this stress level is calculated as
1
E  P   T  p (S )d S  .
N (S )
Furthermore, the damage from all stress cycles is


1
E  P  T   p(S )d S . By further manipulating and substituting the
0
N (S )
S-N curve definition into this damage equation, the following relation can
be obtained [20],

T
E  D  E  P
C 0
S b p(S )d S (8)
From this equation one can find the life of a component by setting the
E[D] equal to unity and obtain the life, T in seconds. For the integrations
given above a cut-off value for the upper limit is necessary. This value is
typically given in terms of Root Mean Square (RMS) values of stress. It is
common to set the cut-off value to 3 RMS in amplitude or 6 RMS in
range, but practice has shown that it should be at least set to 4.5 RMS in
amplitude in order not to miss fatigue damage.
There are many empirical solutions for the probability density function of
Rainflow stress ranges, p(∆S), like Tunna, Wirsching, Hancock,
Chaudhury and Dover [7] but the best correlation was obtained by Dirlik
[21] after computer simulations to model random signals using the Monte
Carlo method.
 Zi Z 2  Zi 2
D1 D2 Z i 2 Ri2
e Q
 2 e  D3 Z i e 2
Q R
p (Si )  (9)
2 m0
where,
Si E 0
,   , E0  , EP 
m2 m4 m1 m2
Zi  xm 
E P 
,
2 m0 m0 m2 m0 m4
2( x m   2 ) 1    D1  D1
2
D1  , D2  , D3  1  D1  D2
1  2
1 R
1.25 (  D3  D2 R )   xm  D1 2
Q ,R 
1    D1  D1
2
D1
The damage equation (Eq. (8)) with the probability density function (Eq.
(9)) is used primarily in Aerospace, Military and in some industrial areas.
The ability to include broad band frequency loading environments to
fatigue damage calculations makes this method more powerful than time
domain methods when the dynamic characteristics of the structure cannot
be ignored.
Fatigue Analysis and Tests
In this study, due to the combined broad band and narrow band frequency
loading encountered which is typical for a helicopter; vibration fatigue
method is chosen as the analysis method. This section will include the
analysis of three different designs in order to investigate different
parameters effecting fatigue life and damage location.
The analysis method requires transfer functions of the structure and the
vibration loading. It is possible to measure or calculate the transfer
functions from prototypes or from numerical analysis of virtual
prototypes. In this study, they are obtained from a FEA software
(ANSYS). Verification of the numerical model was made by modal
testing and analysis. The modal tests and analysis showed that the first 3
modes of the structure had a maximum error of 2% between the
numerical and test model. The details of the modal testing and analysis
performed will not be discussed in this study. Furthermore, for the

vibration fatigue analysis calculations, MSC Fatigue [22] software is
used.
The installation location of the bracket was chosen to have the maximum
dispersion of Chaff/Flare particles (Fig. 16). In order to find the load time
histories for this location, military standards (MIL-STD-810E/F,
GAM-EG-13 etc.) were investigated for the AH-1W helicopter but
unfortunately very coarse information was obtained. Therefore, the
loading time histories were obtained from flight tests performed on the
helicopter using two tri-axial accelerometers where the system will be
installed.
Fig. (16). Location of the Chaff/Flare dispenser bracket on the AH-1W Helicopter [23].
The vibration fatigue analysis requires the PSDs’ of the loading.

Therefore, a loading PSD matrix with cross correlation components and
corresponding transfer function numbers are calculated and given in Table
3 and Figs. (17-19).
Table 3. Transfer functions and PSD input loading table.
Frequency Response Case PSD matrix (i,1) PSD matrix (i,2) PSD matrix (i,3)
1 X axis PSD X-Y axis cross PSD X-Z axis cross PSD
2 X-Y axis cross PSD Y axis PSD Y-Z axis cross PSD
3 X-Z axis cross PSD Y-Z axis cross PSD Z axis PSD
(a) (b)
Fig. (17). PSD plots from flight tests, (a) X-axis, longitudinal axis of the helicopter, (b) Y-axis,
transverse axis of the helicopter.
The material properties are assigned to the elements as aluminum alloy.

Vibration fatigue analysis is performed to the problem by the given
approach due to the excitation of resonant modes by the loading using
Dirlik’s approach for the probability density function.
(a) (b)
Fig. (18). PSD plots from flight tests, (a) Z-axis, vertical axis of the helicopter, (b) X-Y cross axis.
(a) (b)
Fig. (19). PSD plots from flight tests, (a) X-Z cross axis, (b) Y-Z cross axis.
During the fatigue analysis, the mean stress is not corrected due to the fact
that mean stresses cannot be directly defined in vibration fatigue analysis
method. In this method, the response always oscillates around zero stress
level. However, it is always possible to calculate the mean stress level for
the structure and then shift the S-N curve by applying any mean stress
correction theories (Goodman, Soderberg). Furthermore, a better and
faster solution is to calculate the life under oscillating stresses only and
then the life of the critical points can be recalculated with the added mean
stress, as used in this study.
Analysis of Design-1
The fatigue life results of the first designed Chaff/Flare dispenser bracket
are shown in Fig. (20). This design is produced from aluminum blocks
and assembled using bolts. In order to validate the calculations, fatigue
tests had to be performed by testing the bracket three axes simultaneously.
However, the testing hardware was not suitable for the multi axial testing.
Thus, tests were performed on each axis separately as shown in Fig. (21).
The laboratory vibration tests were performed in 4 hours where the
operational loading was accelerated from 12000 flight hours for the
fatigue tests. The fatigue tests ended with no visual damage.
Fig. (20). Fatigue analysis results of the Chaff/Flare dispenser bracket showing the most critical
location under multi axis loading (Fatigue Life 8.59e-3 sec).
Fig. (21). Fatigue testing of the Chaff/Flare dispenser bracket (z-axis).
However, this is not the same as applying all of the vibration profiles at
the same time as it is occurring in real life. Fatigue analyses were
recalculated for each axis separately which is encountered in accelerated
tests for test and analysis comparison. From the analysis results, it was
observed that x-axis (longitudinal axis of the helicopter) loading has
nearly no effect on the fatigue damage accumulation. Y-axis (transverse
axis of the helicopter) has the biggest effect and z-axis (vertical axis of the
helicopter) has a lower effect compared to the y-axis.
As for each axis a different critical location can exist, 445 locations (with
gradual damage distribution) are obtained from the first fatigue analysis,
in which all the three axes loading is applied at the same time, and then
the damage analyses are performed for these locations. In this analysis,
damage values from each axis are summed for these locations which gave
the cumulative damage for axis by axis loading. The analysis results were
compared with the simulation results of multi axial loading. Furthermore,
the percent deviation values of uni-axial loading from multi axis loading
are calculated and a moving average algorithm is applied on the data. The
mean percent deviation of damage was obtained as ranging from 35% to
72%.
From this analysis it is evident that test time has to be extended in order to
fully obtain the multi axis loading effect when performing an axis by axis
test procedure. However, the increase in test time cannot be the same for
all locations due to having different percent deviations of damage. In
order to be safe, the testing duration for sequential testing can be
increased by 72%.
In order to investigate the effect of different geometries on the deviation

in damage, a model without four stiffener corner components of the
Chaff/Flare dispenser is analyzed.
The new model where the stiffeners are removed is re-analyzed for 6061
HV T6 material at operational and MIL-STD-810F loading environment.
However, it must be noted that this profile is accelerated for 2600 flight
hours whereas the operational loading is used in the previous analyses.
The results of this analysis showed that the life distribution changed and
most damaged locations are at the bottom of the dispenser as shown in
Fig. (22). The mean percent deviations for these new locations are given
in Fig. (23).
In this case study, it is found out that when operational loading is applied
most of the damage is due to the z axis loading. However, for MIL STD
810F loading case, all of the axes have more or less equal amount of
contribution to damage. This phenomenon is dominant for locations
between "10" and "35" which causes a decrease in the mean percent
deviation of MIL STD 810F loading. The faster accumulation of damage
in those locations for MIL STD 810F loading decreases the mean percent
deviation. However, for MIL STD 810F loading case different critical
locations are also formed. This clearly shows that the conservativeness of
military standards can change the critical location for some cases.
(a) (b)
Fig. (22). The new Chaff/Flare dispenser fatigue life distribution without the four stiffener
components (Fatigue Life 3.62e5 sec), (a) Isometric top view, (b) Isometric bottom view.
Fig. (23). Mean percent deviation values of sequential loading from multi axis loading for two case
studies.
In the case study given above, it was shown that slight geometric change
affected the damage distributions considerably. In this case study, a
completely different design is analyzed for fatigue. In order to reduce the
weight of the dispenser bracket, the bracket was produced from sheet
metal where the connecting elements are mostly rivets as shown in Fig.
(24).
Fig. (24). Sheet metal chaff/flare dispenser bracket [23].
Fig. (25). The sheet metal Chaff/Flare dispenser fatigue life distribution (Fatigue Life 1.48e6 sec),
Isometric view [23].
The sheet metal model is analyzed for operational loading environment

and the result of this fatigue analysis is given in Fig. (25). The mean
percent deviation of damage for sequential loading from multi axis
loading for 8 damage critical locations and the result is given in Fig. (26).
Fig. (26). Mean percent deviation values of sequential loading from multi axis loading for sheet
metal model [23].
The last analysis emphasizes the geometry dependence of the mean percent
deviation of damage for sequential loading from multi axis loading.
In order to compare the fatigue analysis results with tests for the sheet
metal model, fatigue tests were performed on each axis separately as
shown in Fig. (27). The operational loading is again accelerated to 4 hours
of laboratory test time which corresponds to 12000 flight hours for the
fatigue tests.
The failure criterion for these tests is defined as the formation of a visible
crack. The x and y axes concluded with no visual damage. However, at
the third hour in z axis, the dispenser loosened. As the x and y axis were
already completed, the corresponding flight hour where the dispenser
loosened cannot be accurately determined. A coarse approximation can be
that these deformations occurred at the 9000th flight hour for three axes
loading.
-Y +Y
-X
+X
(a) (b)
+Z
-Z
(c)
Fig. (27). Sheet metal bracket fatigue test, (a) X-axis, (b) Y-axis, (c) Z-axis. [23].
At the end of the tests, the prototype is disassembled and each part is
examined. The following plastic deformations were observed Fig.
(28-29).
(a) (b)
Fig. (28). (a) Plastic deformation on the bolt hole of the magazine housing, (b) Zoomed view. [23].
(a) (b)
Fig. (29). (a) Plastic deformation on the rivet/rivet hole of the side carrier legs, (b) Zoomed view
[23].
The fatigue analyses resulted in similar locations having high damage

values but the observed deformations at these locations cannot be referred
as fatigue failures. The damage type that can be analyzed with vibration
fatigue method is only fatigue damage. Other types of damage such as
plastic deformations cannot be identified with this approach. However,
due to the plastic deformations observed, there is high probability that
fatigue failures will occur at these locations which is consistent with
sequential uni-axial fatigue analyses.
CONCLUSION
Structures used on helicopters are exposed to a broad frequency band of

excitation which consist of random and periodic loading caused by
aerodynamics and the rotors and transmissions of the helicopter,
respectively. Structures containing dynamic behavior as resonance are
indeed difficult to analyze for fatigue. The response of the structure is
amplified when the loading has frequency components which drive the
structure to resonance. Due to this phenomenon, the time domain fatigue
methods where it is assumed that the structure response can be solely
obtained from static analysis cannot be used. One more point to highlight
is that, although the loading is multi-axial, vibration fatigue method was
used as the dynamics of the structure dominate the response, instead of
multi axial fatigue methods. If stress data for the multi-axial loading can
be obtained, which for large and complex structures in FEM is impractical
(requires transient dynamic analysis) then multi-axial fatigue theories can
also be used. But this method, without using FEM, requires experimental
methods.
It should be noted that the accurate stress results obtained from the finite
elements analysis play a critical role for fatigue life of the structure.
Approximately 7% increase in stress value causes 50% reduction of life.
Therefore, it may not be enough that the finite element model simulates
the dynamic characteristics correctly. Stress concentrations occurring in
the finite element is the main reason for inaccurate stress results, which
can only be validated by tests. In the analyses performed, the damage
distribution of the structure was obtained which is a more valuable
information than the life. As a result, the most damaged areas can be
modified to be stronger. Furthermore, it should be noted that the critical
location found from a static analysis is different from the critical location
obtained from the dynamic analysis. This is due to the fact that when the
natural frequencies of the structure are excited, the response of the
structure is controlled by the mode shapes.
The fatigue tests performed also contain some assumptions. First of all,
the tests are sequential and cross correlations are not included in the
fatigue tests. Secondly, the moments are not applied. The parametric case
studies showed that the fatigue damage accumulating for a structures
varies with changes in loading, material and geometry and a formulated
factor that can take these parameter changes into account is not available.
This sensitive structure when combined with nonlinear response in the
frequency domain yields multi axis testing to be an industrial and military
standard. Also from (Fig. 23), it is evident that accelerating the vibration
loading can cause a further deviation from multi axis testing. The most
realistic case would be to use a six-DOF vibration testing machine where
the loading is not accelerated.
In the case where sequential testing is necessary, a fatigue analysis has to

be performed for sequential and multi axis loading in order to calculate
the safety factor to compensate the damage deviation.
A better way of obtaining fatigue life is obtained by using strain gages at

critical locations, which can be obtained by the fatigue analysis performed
in this work, and then performing operational flight tests to get the stress
history. Then this stress history can be used to get the fatigue life of the
strain gage locations. Due to the fact that the stress values obtained during
flights will be more accurate than the finite element stress results a more
accurate life value can be achieved.
Structures designed for the aerospace applications must ensure flight

safety requirements. This is indeed a very difficult objective to achieve
with the presence of wide band frequency loading along with structural
nonlinearities. In such cases, experimental analysis is still the best method
to secure the durability of these aerospace structures.
this publication.
ACKNOWLEDGEMENTS
Declared none.
REFERENCES
[1] "Federal aviation administration: Lessons learned", Available from:

http://lessonslearned.faa.gov/ll_main.cfm?TabID=1&LLID=28&LLTypeID=2,
[Accessed: 8th Aug. 2015].
[2] M. Ewing, “Aircraft structures design and analysis/component design”,
Course Notes, University of Kansas: Seattle, 2004.

[3] D.P. Davies, S.L. Jenkins, and F.R. Belben, "Survey of fatigue failures in
helicopter components", Eng. Fail. Anal., vol. 32, pp. 134-151, 2013.
[http://dx.doi.org/10.1016/j.engfailanal.2013.03.005]
[4] M. Çelik, M. Aykan, and S. Özsoy, “Operational Flight Tests and Numerical
Analyses of an add-on Aerospace Component for Safety of Flight”, In: 4th
Ankara International Aerospace Conference-AIAC07, Ankara: TURKEY,
10-12 September 2007.
[5] M. Aykan, and M. Çelik, "Vibration Fatigue Analysis and Multi-axial Effect in
Testing of Aerospace Structures", Mech. Syst. Signal Process., vol. 23, no. 3,
pp. 897-907, 2009.
[http://dx.doi.org/10.1016/j.ymssp.2008.08.006]
[6] ANSYS Release 8.1 User’s Manual. ANSYS Inc.: USA, 2003.
[7] N.W. Bishop, and F. Sherratt, Finite Element Based Fatigue Calculations.
NAFEMS: Germany, 2000.
[8] "Environmental Engineering Considerations and Laboratory Tests",
MIL-STD-810F, Department of Defense Test Method Standard: USA, 2000.
[9] N.W. Bishop, “Vibration Fatigue Analysis in the Finite Element Environment”,
In: XVI Encuentro Del Grupo Español De Fractura, Spain, 1999.
[10] X. Pitoiset, A. Preumont, and A. Kernilis, "Tools for Multiaxial Fatigue
Analysis of Structures Submitted to Random Vibrations", In: Proceedings
European Conference on Spacecraft Structures Materials and Mechanical
Testing, 1998 Germany.
[11] W.F. Wu, H.Y. Liou, and H.C. Tse, "Estimation of fatigue damage and fatigue
life of components under random loading", J. Pres. Ves. & Piping, vol. 72, pp.
243-249, 1997.
[http://dx.doi.org/10.1016/S0308-0161(97)00053-7]
[12] Y.S. Shih, and G.Y. Wu, "Effect of vibration on fatigue crack growth of an
edge crack for a rectangular plate", Int. J. Fatigue, vol. 24, pp. 557-566, 2002.
[http://dx.doi.org/10.1016/S0142-1123(01)00110-4]
[13] H. Zenner, A. Simbürger, and J. Liu, “On the fatigue limit of ductile
metals under complex multiaxial loading", Int. J. Fatigue, vol. 22, pp.
137-145, 2000.
[http://dx.doi.org/10.1016/S0142-1123(99)00107-3]
[14] Z. Xia, and F. Ellyin, "Multiaxial fatigue of an Alumina particle reinforced
aluminum alloy", Int. J. Fatigue, vol. 20, pp. 51-56, 1998.
[http://dx.doi.org/10.1016/S0142-1123(97)00114-X]
[15] A. Ahmadi, and H. Zenner, "Lifetime simulation under multiaxial random
loading with regard to the microcrack growth", Int. J. Fatigue, vol. 28, pp.
954-962, 2006. [http://dx.doi.org/10.1016/j.ijfatigue.2005.09.015]
[16] C.C. Chu, "Multiaxial fatigue life prediction method in the ground vehicle
industry", Int. J. Fatigue, vol. 19, pp. 325-330, 1997.
[http://dx.doi.org/10.1016/S0142-1123(97)00018-2]
[17] R.M. French, R. Handy, and H.L. Cooper, "A comparison of simultaneous and
sequential single- axis durability testing", Experimental Techniques,
September/October, pp. 32-37, 2006.
[http://dx.doi.org/10.1111/j.1747-1567.2006.00083.x]
[18] J.S. Bendat, and A.G. Piersol, Random Data: Analysis and Measurements
Procedures. John Wiley & Sons, 1971.
[19] Y.L. Lee, J. Pan, R. Hathaway, and M. Barkey, Fatigue Testing and Analysis:
Theory and Practice. Elsevier Butterworth-Heinemann, 2005.
[20] C. Lalanne, Mechanical Vibration & Shock, Fatigue Damage, Vol IV. Taylor
& Francis Books: London, 2002.
[21] T. Dirlik, “Application of computers in Fatigue Analysis”, PhD Thesis,
University of Warwick, 1985.
[22] MSC Fatigue Version 2003 User’s Manual. MSC Software Inc.: USA, 2003.
[23] M. Aykan, “Vibration Fatigue Analysis of Equipments Used in Aerospace”,
MSc. Thesis, Middle East Technical University, 2005.
CHAPTER 7
Fatigue Life Optimization of Laser Peened

Aircraft Components
Anoop Vasu* and Ramana V. Grandhi
Department of Mechanical and Materials Engineering, Wright State

University, Dayton, OH 45435, USA
Abstract: Surface treatments, such as laser peening, can increase the life of the
component by generating compressive residual stresses on the surface. Laser peening
of an already peened component, termed as re-peening, can further increase the
fatigue life of the component. Re-peening has several applications in the aerospace
industry. The huge population of ageing aircraft components is one such application,
which can benefit significantly from the re-peening process. However, this process
is not optimized for maximum fatigue life due to the presence of many design
variables and the complex nature of the problem which requires a large number of
experimental testing to reach conclusions. Therefore, a computationally efficient
optimization strategy needs to be developed to conduct large-scale laser peening
simulations for problems related to fatigue life, such as aircraft lug failure, a problem
that requires consideration of component curvature and residual stress relaxation
effects. Deciding the time to peen an already peened component (re-peening time) is
another variable which makes the problem further complicated. The ultimate goal of
this research is to construct the framework to predict the optimum parameters for
maximum fatigue life on structural components. A two-step optimization strategy
is adopted for the fatigue life optimization of an aircraft lug component. The strategy
*Corresponding author Anoop Vasu: Department of Mechanical and Materials Engineering, Wright State
University, Dayton, OH 45435, USA; Tel/Fax: 937-775-5040; E-mail: anoop1984@gmail.com.
Yucheng Liu (Ed.)

Fatigue Life Optimization Frontiers in Aerospace Science, Vol. 1 303
employs laser peening process parameters, residual stress relaxation, and

re-peening schedule as design variables.
Keywords: Fatigue life optimization, Laser peening, Re-peening,

Residual stress relaxation.
NOMENCLATURE
R Relaxation coefficient
N Number of load cycles
σa Alternating stress
σm initial mean stress
σ1 Mean stress after one cycle
σy Material yield strength
σu Ultimate strength
σr Residual stress
σNf Fatigue strength
Nf Number of cycles to failure
D Cumulative damage
P Peak pressure of laser pulse
O Percentage overlaps of sequential laser spots
t Mid-span duration of laser pulse
σCRS Compressive residual stress
σTRS Tensile residual stress
INTRODUCTION
The potential reason for fatigue failure on structural components is the

generation of tensile stresses on component surface. These surface regions
become the hot spot to initiate cracks. Generation of compressive residual
stresses can inhibit or delay the crack initiation and growth. Surface
enhancement techniques can generate compressive residual stresses and
hence play an important role in improving the life of the peened
304 Frontiers in Aerospace Science, Vol. 1 Vasu and Grandhi
components [1]. Shot peening is the most commonly used mechanical

surface treatment method. Other widely applied methods in the industry
are low plasticity burnishing, water jet peening, laser peening, roller
burnishing, ultrasonic peening, peen forming, cavity peening, etc. Laser
peening generates deeper compressive stresses compared to other surface
treatments like shot peening, making it appealing to the industry. Fig. (1)
shows the fatigue effect of the material Al 7075-T7351 treated by laser
peening with comparison to unpeened and shot peened materials (bending
tests, stress ratio=0.1) [2]. This shows that laser peened material has an
improved fatigue life.
Fig. (1). Fatigue life comparison for Al 7075-T7351.
A pictorial representation of a typical laser peening method is shown in

Fig. (2). The target material is typically coated with an ablative overlay
and confined by a transparent overlay. When the laser pulse impacts the
material, the absorbent material vaporizes and creates plasma. This
plasma confined by the transparent overlay creates a pressure pulse which

is propagated into the material as a shock wave. When the stress created
by shockwave exceeds dynamic yield strength of the material, plastic
deformation of the surface microstructure occurs and compressive stresses
are generated inside the material, significantly more near the surface. The
ablative overlay protects the material surface from temperature effects and
makes laser peening a purely mechanical process. The transparent overlay
confines the plasma from rapid expansion away from the material surface
which helps to obtain a high intensity pressure pulse into the material.
Black paint or tape is usually considered as the opaque overlay while
flowing water along the peening surface is generally taken as the
transparent overlay.
Fig. (2). Schematic of laser peening.

Although the conventional laser peening process takes advantage of the

sacrificial overlay, laser peening without coating (LPwC) has been
employed for the underwater maintenance of nuclear power plants in
Japan [3]. This process results in surface roughness and may have thermal
effects due to direct ablation, but can increase in the fatigue life of the
component. The method’s advantages in the application in remote
environments such as nuclear facilities make it a preferable choice for
certain operations.
Fig. (3). Cracking at frame lug.
Laser peening has extensive applications in the aerospace industry. One

such example is the F-22 Raptor, which is a twin engine fifth-generation
fighter aircraft that uses stealth technology. In the fatigue life testing
carried out on this aircraft, cracks in the wing attachment lugs were
identified, as shown in Fig. (3) [4]. Location 1 indicates the region at the
lower lug profile that experienced the initial cracking approximately 1.2
lifetimes of testing. This region was repaired during the test and after 2.5
lifetimes of testing another crack emerged around the lug bore region
(Location 2). Replacing the lugs with new ones requires removing the
part from the entire assembly. This process is very cost prohibitive. Laser
peening serves the purpose of imparting favorable residual stresses in the
material surface thereby improving the fatigue life of such components.
Other aerospace applications include usage on engine parts, turbine blades,
fasteners, valves, transmission components, etc.
Crack initiation often occurs along the curved regions or fillets of

structural components because of the presence of high stress
concentrations. Laser peening should be conducted in these critical
regions to improve the life of the components. There are several
parameters which will affect the peened regions. However, the physics of
laser peening involves shock wave propagation and is very complicated.
Simulation models are very useful in predicting the structural behavior
and hence, save the costly experiments. This research concentrates on
predicting the material behavior after laser peening using computational
models. The residual stress induced in a curved geometry cannot be
predicted based on the trends obtained from a flat model, since neglecting
the curvature effect can lead to an incorrect assumption of residual
stresses and thereby the fatigue life [5]. Therefore, it is pivotal to
understand the behavior of laser peening application along component
curvature. The complicated nature of residual stress behavior on curved
geometry makes this task very difficult for researchers [6]. Laser peening
of an already peened component can increase the fatigue life during the
service life of the component. Peening the component at the correct time
can also be key to increasing the fatigue life. However, the residual
stresses generated by surface enhancement techniques can relax
(reduction in compressive nature) during the fatigue life cycle. Therefore,
the effect of stress relaxation needs to be considered for better fatigue life
prediction and re-peening decision. In addition, the actual loading
scenarios can also affect the fatigue life significantly. Surface treatments
are generally not optimized for maximum fatigue life due to the presence
of many design variables and the complex nature of the problem which
requires a large number of experimental testing to reach conclusions.
Therefore, a computationally efficient optimization strategy needs to be
developed to conduct large-scale laser peening simulations for problems
related to fatigue life, such as aircraft lug failures, which is a problem that
requires consideration of curvature and relaxation effects.
In this research, a two-step optimization strategy is implemented to

maximize the fatigue life of a laser peened component. In the first step, a
DOE-based response surface model is created with the novel surrogate
modeling technique, the Sorted k-fold Approach (SKA). This step will
maximize the compressive residual stress (CRS) while satisfying the
maximum allowable tensile residual stress (constraint). In the second step,
the maximum CRS will be input into the re-peening formulation for
obtaining the maximum fatigue life when the re-peening time is used as
the design variable.
INFLUENCING FACTORS FOR FATIGUE LIFE PREDICTION
The prediction of fatigue life can be improved significantly by

considering all the important parameters related to laser peening of
structural components. Five factors have been deemed critical for accurate
prediction of laser peened components; they are the peening-induced
residual stresses, the applied load, the relaxation of residual stresses, the
re-peening mechanism, and the fatigue failure. The framework for
predicting the fatigue life is shown in Fig. (4). An indirect approach is
adopted where the load and the residual stresses are analyzed separately.
In a realistic scenario, it is a sequential process where residual stresses are
generated by laser peening and then the load is applied. However, the
computational cost of doing it is very high that the residual stress can be
assumed to be a mean stress added to the applied load. The fatigue
loading can cause the residual stress to relax; hence, the relaxation step
should be considered in order to give a good prediction of the fatigue life.
By adding re-peening to the existing framework, the increase in the
fatigue life of the structure can be predicted. The significance of each
factor has been explained previously by the authors [7]. A summary of

predicting these factors are discussed in this section.
Fig. (4). Computational framework for conducting fatigue life prediction.
Applied Load
The fatigue load on the components is usually approximated by simplistic

constant amplitude loading to save the experimental cost. However,
realistic load scenarios can have a significant impact on the fatigue life of
the component. Using simulation models, fatigue load on the component
can be applied similar to what the component experience during its life
cycle. For aerospace components, a loading standard was introduced
pertaining to a fighter aircraft wing bending mechanism primarily
governed by maneuver loadings. FALSTAFF (Fighter Aircraft Loading
Standard for Fatigue) represents uniquely defined sequence of normalized
numbers ranging from 1 to 32 representing the peaks and troughs in load

history generated from 200 flight data as shown in Fig. (5a) [8]. The
FALSTAFF has much more pronounced peaks due to aggressive
maneuvers of fighter aircrafts. Although the FALSTAFF spectrum
overestimates the fatigue life, many aircrafts (not just fighters) use it as a
standard for a worst case scenario, thus promoting safety. Based on the
stress induced on the critical component, a stress spectrum can be created
from the load spectrum as shown in Fig. (5b). This stress spectrum can be
utilized in the fatigue analysis.
Fig. (5). Loading for fatigue analysis (a) load spectrum (b) stress spectrum.
Laser Peening Induced Residual Stresses
The residual stresses generated by laser peening process can be predicted

by finite element analysis (FEA) [9]. The geometry of the critical region
is an important factor to consider while modeling laser peening process
with FEA. Assuming the geometry to be a simplified flat model can
significantly affect the residual stress prediction [10]. The comparison of
compressive residual stress prediction of spherical models to cylindrical
models for concave and convex geometry is shown in Fig (6). The
difference in compressive residual stress is due to the difference in
plasticity generated by the shock waves.
Fig. (6). Effect of cylindrical and spherical geometries on component curvature.
Residual Stress Relaxation
Many empirical and analytical models, which can be used to evaluate the
effect of relaxation during cyclic loading, have been formulated by
researchers over the years. Some of the most prominent empirical models
are given in this section. Morrow and Sinclair modeled residual stress
relaxation based on mean stress relaxation in axial fatigue tests [11].
Kodama proposed a linear logarithmic relationship based on surface
stresses on a shot peened material [12]. Their empirical models are given
by Eqs. 1, and 2 respectively.
5.7
  
R 1  a  log N (1)
m  
 y 
R  C1  C2 log N (2)
C1, and C2 are constants which are determined by curve fitting the
experimental data performed under various loading conditions. More
sophisticated physics-based models have also been developed [13]. Any
of these models can be utilized in predicting the residual stress relaxation
of laser peened components based on the requirements of the application.
Fig. (7). Flowchart for fatigue life prediction.
Fatigue Life
In this research, a stress-life approach is used to predict fatigue life of

laser peened components. Other fatigue prediction methods, such as
strain-life approach, can also be used as an alternative based on the
availability of material parameters and the nature of the problem [14]. Fig.
(7) shows the step-by-step procedure for life prediction of components
undergoing the laser peening process. The load on the aircraft component,
represented by the FALSTAFF spectrum, is used to create stress spectrum
on the component by conducting a unit load FEA. The alternating stress
(σa) and the mean stress (σm) due to the fatigue loading are obtained from
the stress spectrum. The residual stress (σr) from the laser peening
simulation is added to the mean stress. Using Modified Goodman
Equation, fatigue strength (σNf) is calculated. In the next step, number of
cycles to failure (Nf) is calculated for individual stress segments using
Basquin’s equation. The Cumulative Damage caused by the complete
stress spectrum is obtained by palmgen-miner linear damage rule. In the
last step, the fatigue life is calculated by inverting the cumulative damage.
Fig. (8). Flowchart for predicting re-peened stress state.

Re-Peening
A flow of predicting re-peening stress state is shown in Fig (8). The

predicted residual stress state is also shown. In the first step, laser peening
is conducted on the base material. The generated residual stresses will
undergo relaxation when the component is subjected to fatigue loading.
Another round of laser peening on the relaxed component can bolster the
fatigue life by increasing the magnitude of residual stresses around the
surface regions. As seen from the figure, re-peening can also increase the
compressive residual stress depth, which can further delay the
propagation of crack along the peened regions.
Fig (9). Critical location on the lug (a) experimental results [4] (b) FEA prediction.
OPTIMIZATION PROCEDURE
Application of laser peening on an aircraft lug component is used as an
example to illustrate the fatigue life optimization process. The critical
region predicted by unit load FEA (concave geometry with angle of

curvature 400) matched the crack location in the lug as shown in Fig (9).
Therefore, laser peening FEA was conducted for concave geometry to
predict the residual stresses. Johnson-Cook material model is often used
to predict the effects of laser peening process on components due to the
high strain rate behavior [15]. The material used in this research is 4340
Steel. The details of the FEA model, material properties, and the analysis
results have been published by the authors recently [7].
Design Variables
The design variables considered are re-peening time (R), peak pressure of
the laser pulse (P), mid-span duration of the laser pulse (t), and percentage
overlapping (O) of the laser spots. A brief description of the design
variables and the choice of design bounds are explained in the following
paragraphs:
Re-Peening Schedule
Re-peening is believed to increase the fatigue life by delaying the crack

nucleation as well as the propagation of structural components. Since
peening multiple times during a component’s life can be a costly process,
we assume that re-peening is conducted once during the lifetime of the
component. Therefore, the bounds for re-peening time can be any time
(once) during the component’s life time and is represented in problem
formulation. Cyclic loading leads to relaxation and redistribution of the
residual stress field. These factors influence the fatigue life during
re-peening. Two stress relaxation models are created to predict the
residual stresses after re-peening. The first one is a conservative model
which takes the maximum stress amplitude from the stress spectrum to
predict the re-peened residual stress state. The second relaxation model
uses all the information from the stress spectrum to predict the residual
stresses using a weighted average method [7]. The effects of the two
models for an initial residual stress of -493 MPa is shown in Fig (10).
Fig. (10). Re-peening effect caused by maximum and weighted relaxation models.
Mid-Span Duration
In the experiments, the laser beam mid-span duration normally varies

between 10-30 ns. The pressure pulse mid-span duration is approximately
2 to 2.5 times that of the laser beam mid-span duration. Therefore 20 ns
and 60 ns are taken as upper and lower limits for the mid-span duration.
The temporal pressure profile input to the simulation box can be
approximated by seven representative points as shown in Fig. (11). Point
5 is used to define the mid-span duration while keeping point 4 constant.
The bounds for point 5 are calculated with a two-point formula and are
found to be 30 ns and 90 ns.
Peak
Fig. (11). Temporal pressure profile taken for optimization.
Peak Pressure
Plastic deformation occurs only if the peak pressure pulse is at least 1

HEL (Hugoniot Elastic Limit), which is a function of dynamic yield
strength. A very high peak pressure can lead to spallation of the material
resulting in reduction of fatigue life. Hence, the bounds for the peak
pressure (3 GPa - 6 GPa) have been chosen accordingly.
Percentage Overlap
Percentage overlapping is another important parameter to be considered

while modeling for maximum fatigue life [16]. Laser peening without
proper overlapping instructions can lead to the formation of tensile
stresses and can even cause premature failure of the laser peened
component. Increasing the percentage of overlapping has been found to
have a direct correlation with higher CRS and hence a bound of 40% to
80% is considered. A 75% overlapping pattern is shown in Fig. (12).
Fig. (12). Sequential laser shot with 75% overlapping.
Optimization Strategy
The choice of cost function is based on the major output of the simulation,
which is to obtain the maximum fatigue life. A two-step optimization
strategy is followed in this chapter as shown in Fig (13).
Fig. (13). Multi-step optimization strategy.

RESULTS AND DISCUSSION
Compressive Residual Stress Optimization Results
The problem statement is given by:

Maximize f(x), f(x) = Compressive Residual Stress
subject to g(x) ≤ 100 MPa, g(x) = Allowable Tensile Residual Stress
where the bounds of the variables are given by:
3 GPa ≤ P ≤ 6 GPa, P – Peak pressure
30 ns ≤ t ≤ 90 ns, t – Mid-span duration
40% ≤ O ≤ 80%, O – Percentage Overlap
A DOE-based solution method (Sorted k-fold Approach) is developed to
create the surrogate models for f(x) and g(x). Peak pressure (P),
percentage overlapping (O), and mid-span duration (t) are used to
construct the surrogates for compressive residual stress optimization [17].
All these variables are quantitative in nature and have been proven to
make a significant difference in the compressive residual stress (σCRS)
generated on the surface of the component as well as the tensile stress
generated inside the material (σTRS). The data set to be utilized in the
metamodels is obtained through latin hypercube sampling (lhs) to obtain
30 data points. The tabulated laser peening simulation outputs with design
input parameters based on the design bounds are shown in Table 1.
A quadratic polynomial response surface model is considered for this
analysis, which requires determining ten regression coefficients. The
surrogates for the objective function and constraints are constructed with
the ten-fold SKA, and are given by Eqs. 3 and 4 respectively.
Table 1. Design of experiments.
Test P O t σCRS σTRS

No. (GPa) (%) (ns) (MPa) (MPa)
1 5.79 68.94 44.86 -793 206
2 3.87 44.18 37.90 -434 84
3 3.48 74.22 65.65 -434 108
4 3.58 56.82 45.35 -469 97
5 3.96 47.07 73.57 -166 119
6 5.31 66.06 63.25 -725 195
7 5.27 64.74 55.03 -490 188
8 4.81 69.67 56.45 -620 179
9 5.68 58.04 49.06 -554 218
10 5.75 60.18 51.68 -623 209
11 5.07 51.38 35.20 -419 190
12 3.64 40.67 70.40 -305 87
13 3.32 55.70 40.20 -395 57
14 4.66 78.35 83.51 -719 139
15 3.20 78.85 32.03 -299 55
16 5.87 53.42 81.70 -250 208
17 4.77 63.47 86.11 -369 142
18 5.58 42.70 39.55 -139 184
19 5.97 62.02 41.14 -744 210
20 3.12 71.98 69.97 -362 71
21 4.04 59.17 72.10 -465 142
22 4.13 75.30 66.47 -472 130

23 3.21 41.86 60.82 -275 54
24 5.61 72.76 48.24 -820 178
25 4.65 45.39 33.36 -423 130
26 3.05 50.62 54.79 -234 49
27 4.43 66.82 76.21 -516 165
28 5.42 53.32 80.72 -253 203
29 3.75 77.62 88.73 -459 95
30 4.68 72.38 76.18 -541 136
 C R S  447  190 P  24 O  282 t  80 P 2  192 O 2  40 t 2

 1175 PO  224 Pt  300 Ot (3)
 T R S  33  332 P  157 O  155 t  102 P 2  88 O 2  95 t 2
 98 PO  55 Pt  50 Ot (4)
The optimized results are given by: P = 3.97 GPa, O = 80%, t = 30 ns,
σCRS = -542 MPa, and σTRS = 100 MPa. An extra simulation is run at the
optimized design point which yields σCRS = -511 MPa, σTRS = 103 MPa,
which is very similar to the predicted result. Therefore, this result is used
for the second phase of the optimization problem.
Optimization of the Re-Peening Schedule

The maximum CRS (σCRS = -511 MPa) obtained for the aircraft lug
problem can be directly used as the mean stress for fatigue analysis. Since
analytical equations are used in fatigue life calculation with the stress-life
approach, exhaustive optimization scheme is employed. Fig. (14) shows
the fatigue life versus re-peening schedules for the maximum and
weighted relaxation models. The results from the life predictor models
with the relaxation models show that the best time to re-peen the
component is ~ 55% of the component’s expected life.
Fig. (14). Life optimization with re-peening (a) maximum relaxation model (b) weighted relaxation
model.
CONCLUSION
The optimum parameters to laser peen a structural component in order to
maximize the component’s fatigue life have to be determined. The fatigue
loading can cause the residual stress to relax; hence, the relaxation step
should be considered to give a good prediction of the fatigue life. By
conducting re-peening, we will be able to increase the fatigue life of the
component. A computational framework is created incorporating the key
constituents impacting the life such as load spectrum, residual stress
relaxation, and re-peening for the laser peening process to predict an
optimal re-peening schedule to improve the fatigue life. A two-phase
optimization strategy is employed to obtain prolonged fatigue life of an
aircraft lug. In the first phase, a three-variable RSM is constructed with
the SKA for both objective function and the constraint to conduct CRS
optimization. The predicted CRS matches the simulation results closely.
In the second phase, the optimum CRS is input into re-peening
optimization routine, which predicts the best time to re-peen the
component for maximum fatigue life. It is observed that re-peening the
component around 55% of its initial fatigue life yields the maximum
fatigue life for the structural component.
this publication.
ACKNOWLEDGEMENTS
Declared none.
REFERENCES
[1] A.T. Bozdana,"On the mechanical surface enhancement techniques in

aerospace industry–a review of technology", Aircr. Eng. Aerosp. Technol., vol.
77, no. 4, pp. 279-292, 2005.
[http://dx.doi.org/10.1108/00022660510606349]
[2] P. Peyre, R. Fabbro, L. Berthe, and C. Dubouchet, "Laser shock processing of
materials, physical processes involved and examples of applications", J. Laser
Appl., vol. 8, no. Issue. 3, pp. 135-141, 1996.

[http://dx.doi.org/10.2351/1.4745414]
[3] Y. Sano, K. Akita, K. Masaki, Y. Ochi, I. Altenberger, and B. Scholtes, "Laser
peening without coating as a surface enhancement technology", J. Laser
Micro Nanoen., vol. 1, no. 3, pp. 161-166, 2006., vol. 1, no. 3, pp. 161-166,
2006.
[http://dx.doi.org/10.2961/jlmn.2006.03.0002]
[4] M. Cayton, J. Bunch, P. Walker, J. Brown, T. Brussat, J. Ransom, T. Poast, W.
Garcia, and R. Bair, "Test demonstrated damage tolerance of F-22 wing attach
lugs with forceMate bushings", USAF Aircraft Structural Integrity Program,
Palm Springs, CA, December 2007.
[5] A. Vasu, Y. Hu, and R.V. Grandhi, "Differences in plasticity due to curvature
in laser peened components", J. Sur. Coatings Tech., vol. 235, no. 25,
November, pp. 648-656, 2013.
[http://dx.doi.org/10.1016/j.surfcoat.2013.08.043]
[6] A. Vasu, and R.V. Grandhi, "Effects of curved geometry on residual stress in
laser peening", J. Sur. Coatings Tech, vol. 218, no. March, pp. 71-79, 2013.
[7] A. Vasu, K. Gobal, and R.V. Grandhi, "A computational methodology for
determining the optimum re-peening schedule for an increased fatigue life of
laser peened aircraft components", Int. J. Fatigue, vol. 70, no. January, pp.
395-405, 2015.
[http://dx.doi.org/10.1016/j.ijfatigue.2014.07.008]
[8] G.M. Van Dijk, and J.B. De Jonge, "Introduction to a fighter aircraft loading
standard for fatigue evaluation", Proc. 8th ICAF Symposium, Lausanne,
Switzerland, June 2–5; 1975.
[9] Y. Hu, Z. Yao, and J. Hu, "3-D FEM simulation of laser shock processing", J.
Sur. Coatings Tech., vol. 201, pp. 1426-1435, 2006.
[10] A. Vasu, and R.V. Grandhi, "Impact of plasticity generated by rayleigh waves
on the residual stress behavior of structural components subjected to laser

peening", International Journal of Structural Integrity, vol. 6, no. Issue. 1, pp.
107-123, 2015.
[http://dx.doi.org/10.1108/IJSI-06-2014-0028]
[11] J. Morrow, and G.M. Sinclair, "Cycle-dependent stress relaxation",
Symposium on Basic Mechanisms of Fatigue, , 1958pp. 83-109
[12] Kodama, S., “The behavior of residual stress during fatigue stress cycles,” In:
Proceedings of the International Conference on Mechanical Behavior of
Metals II, Society of Material Science, Kyoto, 1972, pp. 111–118.
[13] W.Z. Zhuang, and G.R. Halford, "Investigation of residual stress relaxation
under cyclic stress", Int. J. Fatigue, vol. 23, suppl. Suppl. 1, pp. 31-37, 2001.
[http://dx.doi.org/10.1016/S0142-1123(01)00132-3]
[14] T.J. Spradlin, R.V. Grandhi, and K. Langer, "Experimental validation of
simulated fatigue life estimates in laser-peened aluminum", Int. J. Struct.
Integrity, vol. 2, no. 1, pp. 74-86, 2011.
[http://dx.doi.org/10.1108/17579861111108635]
[15] H.K. Amarchinta, R.V. Grandhi, K. Langer, and D. Stargel, "Material model
validation for laser shock peening process simulation", Model. Simul. Mater.
Sci. Eng., vol. 17, no. 1, pp. 1-15, 2009.
[http://dx.doi.org/10.1088/0965-0393/17/1/015010]
[16] G. Singh, and R.V. Grandhi, "Mixed-variable optimization strategy employing
multifidelity simulation and surrogate models", AIAA J., vol. 48, no. 1, pp.
215-223, 2009.
[http://dx.doi.org/10.2514/1.43469]
[17] A. Vasu, and R.V. Grandhi, "Response surface model using the sorted k-fold
approach", AIAA J., vol. 52, no. 10, pp. 2336-2341, 2014.
[http://dx.doi.org/10.2514/1.J052913]
CHAPTER 8
Prediction of Residual Stress Relaxation in

Ti-6Al-4V subjected to Laser Peening
Anoop Vasu* and Ramana V. Grandhi
Department of Mechanical and Materials Engineering, Wright State

University, Dayton, OH 45435, USA
Abstract: Laser peening is an advanced surface enhanced method which induces

compressive residual stress on the critical regions of components prone to fatigue
failure. However, the residual stresses relax under the fatigue loading conditions.
Constitutive models have to be robust enough to predict the residual stress
relaxation mechanism. Although tensile cold working increases the tensile yield
strength, the compressive yield strength is reduced. As a result of this, a lower
compressive load can relax the initial compressive residual stress. This
phenomenon, termed as Bauschinger Effect, can be represented by an analytical
stress-strain model to predict the relaxation effects based on the cold working of
the material. Three dimensional finite element (FE) models are created to represent
residual stress relaxation in a low cycle fatigue regime for Ti-6Al-4V material. The
creation of the numerical model for simulating stress relaxation model involves
two phases. The first phase is modeling the stress gradient effect which relates to
the effect of cold working. This process utilizes a laser peening simulation model.
The second phase is modeling the stress-strain response of the material by creating
a mixed hardening model.
Keywords: Finite element analysis, Laser peening, Low cycle fatigue,
*Corresponding author Anoop Vasu: Department of Mechanical and Materials Engineering, Wright State
University, Dayton, OH 45435, USA; Tel/Fax: 937-775-5040; E-mail: anoop1984@gmail.com.
Yucheng Liu (Ed.)

Prediction of Residual Stress Relaxation Frontiers in Aerospace Science, Vol. 1 327
Mixed hardening, Residual stress relaxation, Surface enhancement

technique.
NOMENCLATURE
HEL Hugoniot elastic limit

dyn
y
Dynamic yield stress
v Poissons ratio
σvm Equivalent von mises stress
A Yield stress
B Strain hardening coefficient
n Strain hardening exponent
C Strain rate sensitivity constant
εp Equivalent plastic strain

*
Dimensionless strain rate
 Strain rate from high strain rate experiments
 0 Reference strain rate
dp Heat penetration depth
D Heat diffusion coefficient
 Laser pulse width
C1 Kinematic hardening modulus
σij stress tensor
αij Backstress tensor
σ0 Equivalent stress defining size of yield surface

p Equivalent plastic strain rate
γ Material constant from cyclic test data
 ij
p
Rate of plastic flow
R Stress ratio
INTRODUCTION
How to prevent failure in materials has been a great research interest all
over the world for centuries. It has been found out that cyclic loading is a
major cause of failure in many situations. There can be other contributing
factors, such as manufacturing defects and the presence of unfavorable
residual stresses. But residual stresses can also play a positive role and
can enhance the life of the component, if applied properly. There exist
methods named Surface enhancement techniques (SET), which can
induce favorable residual stresses (mechanical SET) on surface regions of
peened components to improve the fatigue life of the component [1]. Fig.
(1) depicts a pictorial representation of the impact of SET on fatigue life.
Fig. (1). Extended fatigue life in peened components.
Fig. (2) shows the mechanism behind the mechanical surface

enhancement methods. When the material is peened under loading, the
surface region in contact would be plastically deformed. The surrounding
bulk material, which is elastically deformed, tries to come back to its

initial state generating compressive stresses in the surface region while
tensile stresses are created in the subsurface to attain static equilibrium in
the material. These methods are effective on solving problems such as
tensile stress buildup, the main reason of failure in many components.
Surface enhancement techniques play an important role in improving the
life of the peened components by creating compressive stress field on the
surface, and thus, delaying or eliminating the crack initiation and
propagation inside the material [2].
Fig. (2). Mechanism of surface enhancement techniques.
Laser peening has emerged to be a popular surface treatment method

because it can significantly improve the fatigue performance of structural
components by inducing compressive residual stresses [3]. However, the
process parameters should be carefully chosen to obtain optimal
performance; application without proper strategy can result in unwanted

tensile stresses and even reduction in the fatigue life [4]. The residual
stresses induced by different peening methods may also relax due to
cyclic loading [5]. The material still shows an overall beneficial effect on
the fatigue life of the specimen in spite of the relaxation, but neglecting
the relaxation effect would lead to over-prediction of the fatigue life [6].
James proposed that the relaxation behavior due to cyclic loading can be
separated into three regimes based on the yield strength and the endurance
limit [7]. When the stress exceeds the yield strength of the material, a
complete redistribution of the residual stress takes place and can cause
rapid relaxation of initial compressive residual stress. When the stress is
below the endurance limit, a relatively weak relaxation can take place in
the material. In the case of intermediate stress amplitudes, relaxation has
been found to be a function of cyclic stress amplitudes.
Morrow and Sinclair created an empirical model to predict the relaxation

effect based on axial fatigue test experiments [8]. Many other researchers
followed up the initial work to generate more experimental models to
quantify cyclic stress relaxation. Jhansale and Topper suggested an
exponential relationship between mean stress relaxation and the number
of cycles [9]. Kodama proposed a linear logarithmic relationship based on
the surface stresses on a shot peened material, and claimed that the
relaxation rate was proportional to stress amplitude [10]. Zhuang and
Halford proposed an advanced physics-based relaxation model which
could take into account parameters like degree of cold working, stress
ratio and stress gradient, and the results were comparable to those
predicted by finite element analysis [11]. Smith et al. conducted finite
element analysis to compare the relaxation effect of the two material
hardening models [12]. They found out that complex multi-linear model
had more relaxation than a simple linear hardening model. But the
predictions were poor compared to measurements in the cyclic loading
case. Farrahi et al. showed that residual stresses decreased and their
reduction rate was a function of the cyclic stress level and an inverse
function of the depth of the plastically deformed surface layer on shot
peened spring steel specimens [13]. Iida et al. found out that repeated
cyclic loading caused more relaxation when compared to reverse cyclic
loading [14]. Boyce et al. investigated relaxation effect due to foreign
object damage [15]. They found out that damage sites with high stress
concentrations which were initially under a state of strong compression
relaxed, and cracks tend to form in these regions. Vasu et al. developed a
computational methodology to incorporate the relaxation effect into
fatigue life prediction models [16]. However, very little information is
available to comprehend the relaxation behavior in laser peened
components subjected to low cycle fatigue loading conditions.
In this research, we predict the relaxation behavior when material yielding

occurs due to the combined effect of residual stresses and applied loading.
Finite element method is used to predict the relaxation behavior. The first
step is to choose an efficient hardening law. Constitutive models have to
be robust enough to represent the relaxation mechanism. The mixed
hardening model has been very useful in accounting the Bauschinger
effect, a phenomenon that refers to the reduction of flow stress after a
reversal in strain direction that makes even lower applied loads able to
generate additional plastic deformation in the material. Experimental test
for the chosen material is used to obtain the material constants, and can be
directly input into the simulation to replicate the stress-strain behavior.
Once experimental data is obtained, finite element laser peening
simulation is conducted on the fatigue specimen to obtain the residual
stress profile for analysis. In this research, an axial fatigue problem is
considered to illustrate the implementation of residual stress relaxation
occurring on laser peened components.
LASER PEENING MECHANISM

A pictorial representation of laser peening process is shown in Fig. (3).
The target material is coated with ablative overlay and confined by
transparent overlay. When the laser pulse impacts the material, the
absorbent material vaporizes and creates plasma. This plasma which is
confined by the transparent overlay creates a pressure pulse which is
propagated into the material as a shock wave. When the stress created by
shockwave exceeds dynamic yield strength of the material, this results in
plastic deformation of the surface microstructure generating compressive
stresses inside the material. Black paint or tape is considered most of the
time as the opaque overlay while flowing water along the peening surface
is usually taken as the transparent overlay. Some special features observed
in the process of laser peening are discussed in this section. The features
are unique of laser peening among all surface enhancement methods.
Fig. (3). Laser peening process.
Pressure Pulse Generation
The vaporization of heated ablative material by laser beam generates

plasma. Confined plasma creates shock wave that is propagated into the
material. The dynamic nature of the shock wave generated by the laser
beam is represented as a pressure pulse having a temporal as well as a
spatial profile. The laser pulse generated by the laser source and the
pressure pulse generated by the shock wave are monitored and the
temporal profile representing them is shown in Fig. (4) [17]. This pressure
pulse has unique properties such as high peak pressure and very short
pulse duration.
Fig. (4). Temporal profile of the laser pulse and the generated pressure pulse.
High Strain Rate Phenomenon
The shock waves generate high strain rates inside the material during the
laser peening process on the order of 106/sec. These high strain rates can
play an important role in determining the material behavior. It has been
experimentally proven that many materials are sensitive to high strain rate
effects. It requires a higher stress to yield the material in case of high
strain rates when compared with a quasi-static process. The stress-strain
curve showing this high strain rate effect can be represented as shown in
Fig. (5) [18].
Fig. (5). Stress-strain curve showing the strain rate effect.
Hugoniot elastic limit (HEL) is a measure to represent the dynamic yield

stresses due to high strain rate effect. The HEL is the axial stress required to
plastically deform a material in a uniaxial strain state and is represented as a
function of dynamic yield strength and Poisson ratio of the material as
shown in Eq. 1 [19].
 ydyn 1  
HEL  (1)
1  2 
When the shock wave propagates into the material, the pressure pulse
amplitude drops (shock wave attenuation) as it goes deeper into the material.
When the stress wave magnitude goes below the HEL, no further plasticity
occurs and the surrounding material representing the elastically deformed
field compresses the plastically affected region to achieve static
equilibrium.
Plasticity in Laser Peening
Plasticity experienced in a high strain rate process like laser peening can be
explained by means of the dependence of plastic strain on impact pressure
with the use of the HEL as the measure, as illustrated in Fig. (6) [19]. This
model assumes a Von Mises yield criterion along with a uniform pressure
along the impact region. The first step involves the elastic loading until the
HEL is reached. The second step involves plastic loading where plastic
deformation occurs in the material. The third step is the elastic unloading
phase and the final step is the plastic unloading phase. When the impact
pressure is between 1 and 2 HEL, plastic strain can occur with a purely
elastic reverse strain. When the impact pressure is more than 2 HEL, elastic
reverse strain gets saturated and plastic reverse strain occurs. If the impact
pressures are of the order of 2.5 HEL or more, this can result in the
spallation on the material. Hence impact pressure should be optimized to
obtain best results. Theoretically, a load of 2 HEL should give the
maximum compressive residual stress according to Fig. (1.15). If the load is
beyond 1 HEL, permanent plastic deformation occurs. When the load is
between 1 and 2 HEL, as the plastic strain increases, the compressive
residual stress also increases.
Fig. (6). Plastic strain under pressure loading.
FINITE ELEMENT MODELING OF LASER PEENING
2D Axisymmetric finite element models can predict the residual stresses for
one shot laser peening process. However, 3D models are required to predict
the residual stresses for a realistic application scenario. The details of the
2D and 3D FEA models have been described by the authors before [20, 21].
This section discusses the important parameters considered while

conducting the modeling and analysis of laser peening.
Pressure Pulse Input
The temporal profile of the shock wave is constructed similar to the sharp
rise time pulse profile obtained from experimental testing as shown in Fig.
(7). Although non-homogeneity is observed along the spatial profile of the
laser shot impact, the effect on the peak pressure is fairly small [22]. This
research considers the initial condition of laser power density, which is
uniformly distributed along the laser spot with the same area dimensions for
all three geometries. A uniform spatial profile over the spot is used for the
current investigations.
Fig. (7). Temporal pressure profile.
Simulation Procedure
Laser peening simulation process can be divided into two stages. The first
stage involves the generation of pressure pulses by high-energy plasma,
which can be explained by laser physics. The second stage incorporates
the plastic deformation on the surface of the material induced by the

shock waves generated by the pulse pressure [23]. Our interest is to model
the second stage of the process using a finite element model to obtain the
residual stress field induced in the material. The conventional method
involves an integrated approach using both explicit and implicit solving
techniques to predict the shock wave phenomenon imparted to the
material. However, this “explicit-implicit” procedure becomes difficult to
implement for 3D finite element simulations with a large number of laser
shots. Each shot takes more than 1 day (~ 40 CPU hours) to simulate a
finite element model encompassing 0.5 million elements at the Ohio
Supercomputing Center (OSC) Facility. At OSC multiple nodes of quad
core 2.5 GHz Opteron 24 GB RAM are available to perform the analysis.
A modified explicit procedure is a good approximation to reduce the
computational time, as shown in Fig. (8). In this “modified explicit”
procedure, the explicit solver is run for a short duration until a steady
dynamic stress state is reached for each laser peening shot (when kinetic
energy approaches zero). For the final shot, an extended explicit
simulation time replaces the time consuming force equilibrium analysis
(implicit part). The modified explicit procedure is suggested based on the
observation that the redistribution of residual stress field becomes weak
when a steady dynamic stress state is reached. The details of the
procedure and the comparison of its performance have been discussed
before [21].
Fig. (8). Laser peening simulation procedure.

Constitutive Model
The strain rate for laser peening is in the order of 106 /second. In order to
give an accurate prediction of the material response, temperature effects
as well as the effect of strain rate on the flow stress should be included in
the constitutive model. The Johnson-Cook model has been an efficient
material model to deal with high strain rate processes and is used for all
the finite element simulations in the present work [24]. According to the
Johnson-Cook strength model, the equivalent Von Mises stress is given
by:
 vm  A  B pn 1  C ln  *  (2)
Constant A is the yield stress at 0.2% offset strain; Constant B and
exponent n represent the strain hardening effect. The expression in the
second bracket represents the strain rate effect through constant C. Model
constants A, B, n and C are determined from experimental data in the
form of room temperature stress-strain curves at specific strain rates. It
can be observe that, the thermal term has been neglected from the
Johnson cook material model. The absorbent layer can protect the
material surface from thermal effects of laser irradiation. It not only
prevents the part surface from the direct laser ablation, but also prevents
the thermal effect propagating into the specimen; hence the thermal term
is dropped. More discussion will be given in the next section.
Effect of Temperature in Laser Peening Process
The heat effect is not considered in the simulation because the absorbent
layer can protect the material surface from thermal effect during laser
peening. Sources of heat in laser peening to consider include laser
irradiation of the material and temperature rise due to the plastic strain
caused by shock waves. The heat penetration depth dp shown in Fig. (9)
due to the nanosecond laser irradiation is given by:
d p  D 
1/ 2
(3)
Taking black paint as an example of the absorbent layer, the heat diffusion
coefficient is D = 2.1*10-7 m2/s. If it is irradiated by a 20 ns laser beam,
the heat penetration depth would be on the order of 0.065 μm. The layer
thickness used in the laser peening process is usually tens of micrometers.
Therefore, the ablative layer not only prevents the part surface from the
direct laser ablation, but also prevents the thermal effect propagating into
the specimen.
Fig. (9). A pictorial representation of the heat penetration depth.
Shock waves generated by the laser pulse create compression in the

material resulting in temperature rise. A thermo-mechanical finite element
analysis is conducted on Ti-6Al-4V material (flat geometry, spot radius =
2.5 mm, initial temperature = 25°C, melting temperature = 1660°C). A 2D
axisymmetric model is used to perform this analysis [25]. As the pressure
pulse goes beyond 8 GPa, we can see the spike in temperature in the
material as shown in Fig. (10). We can also observe that, in the laser
peening operating range (3-6 GPa related to HEL), maximum temperature

observed is around 121 °C. The Johnson-Cook material constants for
Ti-6Al-4V are shown in Table 1.
Table 1. Johnson-Cook material model constants for simulation.
Material A B 𝜀̇0 E ρ
n C ν
(MPa) (MPa) (/s) (GPa) (Kg/m3)
Ti-6Al-4V 950.2 603.4 0.1992 0.0198 0.0009 113.8 0.342 4500
Fig. (10). Effect of temperature rise due to shockwaves.
To see how the temperature rise affects the residual stress in the material,
a peak pressure of 5.5GPa (~2HEL) is applied to the material. The
residual stress imparted to the material on the surface as well as the depth
direction are shown in Fig. (11). The results indicate that the effect of
temperature due to laser peening in its operating conditions is small and
hence temperature effect needs not be considered to model the laser
peening process.
Fig. (11). A comparison of mechanical and thermo-mechanical FEA in laser peening (a) Surface
residual stress (b) Residual stress along depth.
Curvature Effect
The finite element modeling simulation conducted to demonstrate the

significance of curvature effects is illustrated in Fig. (12) for Ti-6Al-4V
(Rc = 7.64 mm for convex and concave geometries). A difference in peak
compressive residual stress of 300 MPa (greater than 30%) is observed.
This big difference in magnitude should not be neglected since the
assumption of flat geometry can lead to incorrect residual stress
prediction, which can be detrimental in fatigue life estimation. Note that

the differences, though significant, only concern the surface layer of the
depth mostly within 0.05 - 0.1 mm; beyond this depth, the differences
start to become less significant or even negligible. We can also observe
the reduction of compressive residual stress on surface around the spot
center, which can be attributed to the stress focusing due to Raleigh waves
[23].
Fig. (12). Curvature effect on Ti-6Al-4V (a) Stress on the surface of the material (b) Stress in the
depth direction at 1.5 mm away from the center of the spot.
Fig. (13). Cyclic stress relaxation.
RESIDUAL STRESS RELAXATION
The residual state just after peening should not be directly taken into
fatigue analysis. Residual stresses generated by laser peening or any other
surface enhancement method on structural components will have to deal
with complex loads during their life cycle. The cyclic fatigue loading, as
well as elevated temperature, can take off the beneficial effect of the
surface enhancement methods significantly. This phenomenon of stress
instability is called stress relaxation. There are three different ways to
impart relaxation to peened components. The first one is termed as
thermal stress relaxation, which occurs when the surface treated
components are exposed to elevated temperature [26]. The second
mechanism is called the static relaxation, which arises when the sum of
the residual stress and the applied load exceeds the yield criterion of the
material [27]. The third way to yield stress relaxation is through repeated
fatigue cycles and is the most prominent method among all three. This
process is called cyclic stress relaxation and can happen even when there
is no macroscopic plastic deformation. It should also be noted that
relaxation can happen as a combination of these three relaxation

mechanisms, but the scope of this research is confined to stress relaxation
resulting from a cyclic load. Depending on the magnitude of loading, the
residual stress field can undergo relaxation and redistribution if the
loading exceeds yielding criterion. Relaxation can also occur due to micro
plasticity effects. A pictorial representation of the cyclic stress relaxation
effect on a surface treated component is shown in Fig. (13).
Fig. (14). Two phases in determining the relaxation effect.
The main parameters which can influence stress relaxation are the number
of cycles, load amplitude, stress ratio, and degree of cold working [11].
Compared to shot peening, the effect of cold working is low in the case of
laser peening [28]. A framework has been constructed to determine the
relaxation behavior as shown in Fig. (14). It can be divided into two
phases. The first phase is the peening phase where residual stresses are
generated in the model by the laser peening. An advanced
modified-explicit procedure is employed for simulating the multi-shot
laser peening process. The second phase involves the relaxation effect
where the residual stress is subjected to cyclic loading. Based on the
number of cycles applied onto the material, we can solve the problem
using FE models or available empirical/analytical models. The division is
based on two categories of fatigue failure, namely Low Cycle Fatigue
(LCF) and High Cycle Fatigue (HCF). LCF relaxation can be predicted
by finite element simulation models while HCF relaxation can be
predicted by already existing analytical models. The prediction of stress
relaxation with FE models has been detailed in this section.
Obtain Empirical Data for Calibration

In order to conduct the numerical simulation, we require material
parameters defining the hardening model. An axial fatigue problem is
considered. Boyce conducted strain-controlled test to measure the cyclic
stress-strain behavior of Ti-6Al-4V [15]. The experimental specimen and
the stress-strain behavior are shown in Fig. (15).
Fig. (15). Strain controlled tests (a) Specimen geometry (b) Cyclic stress-strain behavior of
Ti-6Al-4V.
Boyce measured the stress-strain behavior for 5 cycles, and a functional

fit to the data yields the parameters which can be directly input in to
ABAQUS to define the hardening behavior of the material. The hardening
law is given by Eq. 4.
 ij   ij  p
 ij  C1   ij  p (4)
0
2 p p
where equivalent plastic strain rate is given by  p  ij ij . C1 is the
3
kinematic hardening modulus, γ is the material constant which determines
the rate at which the kinematic hardening modulus decrease with
increasing plastic deformation. The material hardening parameters are
calibrated from the cyclic testing data and are given by C1 = 1.38x1011 and
γ=635.
FE Modeling of Stress Relaxation
The numerical modeling can be broken down into two steps. The first step
is to laser peen the specimen to obtain the residual stress field. The
importance of curved geometry is taken into consideration while
modeling the residual stress imparted to the specimen. To make a good
representation of the real problem, multi-shot laser peening with
overlapping is conducted on a 3D convex model and the individual stress
components for each element are imported into the 2D axisymmetric FE
model as prestress. This task of modeling the initial residual stress field is
performed with the use of SIGNI.f, a user subroutine in abaqus. Step 1 is
shown in Fig. (16).
The second step is to apply fatigue load on the laser peened specimen.
Variable amplitude loading such as that obtained from FALSTAFF can be
used to describe the loading spectrum. But for the simplicity of the
procedure of calculating the cyclic response due to loading magnitudes,
constant amplitude loading is considered for this research. Step 2 is
shown in Fig. (17).
Fig. (16). Initial residual stress state.
Fig. (17). Numerical prediction of stress relaxation due to applied cyclic loading.
RESULTS AND DISCUSSION
Residual stress response for a maximum stress of 200 MPa and

completely reversible loading are shown in the Fig. (18) along the
longitudinal and transverse directions. The red line in both graphs
represents the initial residual stresses. We can observe significant
relaxation in both directions. The relaxation phenomenon is dominant in
the initial cycles and gradually there is no change in the residual stresses
after a certain number of cycles. We can not only see relaxation of
compressive residual stresses, but also the relaxation of tensile stresses as
we move along the depth. These trends correlate well with the
experimental results conducted on shot peening fatigue specimen.
Fig. (18). Cyclic response of residual stresses undergoing fatigue loading along (a) the longitudinal
direction (b) the transverse direction.
Fig. (19) shows the residual stress in the longitudinal and transverse
directions for a completely reversible loading condition with different
maximum stresses. We can observe that, as the maximum stress gets
higher, relaxation becomes significant in both directions. For a repeated
loading case scenario, the relaxation effect is minimal. This indicates that
compressive loading is the driving force for the relaxation of residual
stresses produced by methods like laser peening.
Fig. (19). Stress relaxation under different loading scenarios.
CONCLUSION
Numerical models are constructed in ABAQUS to simulate the relaxation

behavior of laser peened residual stresses in Ti-6Al-4V. Three
dimensional multi-shot laser peening is conducted on a concave specimen
representative of the gauge section of the fatigue specimen. The residual
stresses obtained from laser peening are used in the two-dimensional
axisymmetric model as an initial residual stress field. The experimental
results obtained from a strain-controlled test for an axial fatigue problem
are used to construct a mixed hardening model representing the cyclic
hardening behavior of the material. The results from the simulation

models indicate that relaxation has a significant effect, especially in the
initial cycles for the axial fatigue problem, at higher load levels, while
minimal relaxation is observed at lower load levels. The effects of
relaxation in different loading conditions are investigated. It is found that
the compressive fatigue cycles play an important role in the relaxation of
compressive residual stresses.
this publication.
ACKNOWLEDGEMENTS
Declared none.
REFERENCES
[1] A.T. Bozdana, "On the mechanical surface enhancement techniques in

aerospace industry–a review of technology", Aircr. Eng. Aerosp. Technol., vol.
77, no. 4, pp. 279-292, 2005.
[http://dx.doi.org/10.1108/00022660510606349]
[2] Y. Ochi, K. Masaki, T. Matsumura, and T. Sekino, "Effect of shot-peening
treatment on high cycle fatigue property of ductile cast iron", Int. J. Fatigue,
vol. 23, no. 5, pp. 441-448, 2001.
[http://dx.doi.org/10.1016/S0142-1123(00)00110-9]
[3] C.S. Montross, T. Wei, L. Ye, G. Clark, and Y.W. Mai, "Laser shock
processing and its effects on microstructure and properties of metal alloys: a
review", Int. J. Fatigue, vol. 24, no. 10, pp. 1021- 1036, 2002.
[http://dx.doi.org/10.1016/S0142-1123(02)00022-1]
[4] T.J. Spradlin, R.V. Grandhi, and K. Langer, "Experimental validation of
simulated fatigue life estimates in laser-peened aluminum", Int. J. Struct.
Integrity, vol. 2, no. 1, pp. 74-86, 2011.

[http://dx.doi.org/10.1108/17579861111108635]
[5] I. Altenberger, R.K. Nalla, Y. Sano, L. Wagner, and R.O. Ritchie, "On the
effect of deep-rolling and laser-peening on the stress-controlled low and
high-cycle fatigue behavior of Ti–6Al–4V at elevated temperatures up to 550
C", Int. J. Fatigue, vol. 44, no. November, pp. 292-302, 2012.
[6] R.L. Mattson, and W.S. Coleman, "Effect of shot peening variables and
residual stresses on fatigue life of leaf spring specimens", Transactions,
Society of Automotive Engineers, vol. 62, pp. 546-556, 1954.
[7] M.R. James, "Relaxation of residual stresses-An overview, Pergamon Press,
Advances in Surface Treatments", Technology-Applications-Effects, vol. 4, pp.
349-365, 1987.
[8] J. Morrow, and G.M. Sinclair, “Cycle-dependent stress relaxation,”
Symposium on Basic Mechanisms of Fatigue, Sixty-first Annual Meeting,
American Society for Testing Materials, Boston, Massachusetts, June, 1958,
pp. 83-109 .
[9] H.R. Jhansale, and T.H. Topper, "Engineering analysis of the inelastic stress
response of a structural metal under variable cyclic strains", Symposium on
Cyclic Stress–Strain Behavior-Analysis, Experimentation, and Failure
Prediction, 1971pp. 246-270 Bal Harbour, Florida
[http://dx.doi.org/10.1520/STP38034S]
[10] S. Kodama, "The behavior of residual stress during fatigue stress cycles",
Proceedings of the International Conference on Mechanical Behavior of
Metals II, Society of Material Science, Kyoto, 1972, pp. 111–118.
[11] W.Z. Zhuang, and G.R. Halford, "Investigation of residual stress relaxation
under cyclic stress", Int. J. Fatigue, vol. 23, suppl. Suppl. 1, pp. 31-37, 2001.
[http://dx.doi.org/10.1016/S0142-1123(01)00132-3]
[12] D.J. Smith, G.H. Farrahi, W.X. Zhu, and C.A. McMahon, "Experimental
measurement and finite element simulation of the interaction between residual
stresses and mechanical loading", Int. J. Fatigue, vol. 23, no. Issue.4, pp.
293-302, 2001.
[http://dx.doi.org/10.1016/S0142-1123(00)00104-3]
[13] G.H. Farrahi, J.L. Lebrun, and D. Couratin, "Effect of shot peening on residual
stress and on fatigue life of a spring steel", Fatigue Fract Eng. Mat. Struct.,
vol. 18, no. 2, pp. 211-220, 1995.
[http://dx.doi.org/10.1111/j.1460-2695.1995.tb00156.x]
[14] K. Iida, and M. Takanashi, "Relaxation of welding residual stresses by
reversed and repeated loadings", Welding in the World/Le Soudage dans le
Monde, vol. 14, no. 4, pp. 314-327.
[15] B.L. Boyce, X. Chen, J.O. Peters, J.W. Hutchinson, and R.O. Ritchie,
"Mechanical relaxation of localized residual stresses associated with foreign
object damage", Mater. Sci. Eng., vol. 349, no. Issues.1-2, pp. 48-58, 2003.
[http://dx.doi.org/10.1016/S0921-5093(02)00543-9]
[16] A. Vasu, K. Gobal, and R.V. Grandhi, "A computational methodology for
determining the optimum re-peening schedule for an increased fatigue life of
laser peened aircraft components", Int. J. Fatigue, vol. 70, no. January, pp.
395-405, 2015.
[17] P. Peyre, R. Fabbro, P. Merrien, and H. Lieurade, “Laser shock
processing of aluminum alloys. Application to high cycle fatigue behavior",
Mater. Sci. Eng. A, vol. 210, no. 1-2, pp. 102-113, 1996.
[http://dx.doi.org/10.1016/0921-5093(95)10084-9]
[18] M.A. Meyers, Dynamic Behavior of Materials, J.. Wiley: New York, 1994.
[http://dx.doi.org/10.1002/9780470172278]
[19] P. Ballard, J. Fournier, R. Fabbro, and J. Frelat, “Residual stresses induced by
laser-shocks,” Le Journal de Physique IV, Vol. 1, No. C3, pp. 487-494.
[20] A. Vasu, Y. Hu, and R.V. Grandhi, "Differences in plasticity due to curvature
in laser peened components", J. Surf. Coat. Technol., vol. 235, no. November,
pp. 648-656, 2013.
[21] A. Vasu, and R.V. Grandhi, "Effects of curved geometry on residual stress in
laser peening", J. Surf. Coat. Technol., vol. 218, no. March, pp. 71-79, 2013.
[22] Y. Hu, C. Gong, Z. Yao, and J. Hu, "Investigation on the non-homogeneity of
residual stress field induced by laser shock peening", J. Surf. Coat. Technol.,
vol. 203, no. 23, pp. 3503-3508, 2009.
[23] A. Vasu, and R.V. Grandhi, "Impact of plasticity generated by rayleigh waves
on the residual stress behavior of structural components subjected to laser
peening", Int. J. Struct. Integrity, vol. 6, no. Issue. 1, pp. 107-123, 2015.
[http://dx.doi.org/10.1108/IJSI-06-2014-0028]
[24] H.K. Amarchinta, R.V. Grandhi, A.H. Clauer, K. Langer, and D. Stargel,
"Simulation of residual stress induced by laser peening process through
inverse optimization of material models", J. Mater. Process. Technol., vol. 210,
no. 14, pp. 1997-2006, 2010.
[http://dx.doi.org/10.1016/j.jmatprotec.2010.07.015]
[25] Vasu, A., Influence of Curved Geometries on the Fatigue Life of Laser Peened
Components, Ph.D Dissertation, Wright State University, 2014.
[26] Z. Zhou, A.S. Gill, D. Qian, S.R. Mannava, K. Langer, Y. Wen, and V.K.
Vasudevan, "A finite element study of thermal relaxation of residual stress in
laser shock peened IN718 superalloy", Int. J. Impact Eng., vol. 38, no. Issue. 7,
pp. 590-596, 2011.
[27] O.S. Zaroog, A. Ali, B.B. Sahari, and R. Zahari, "Relaxation of compressive
residual stress. Part 1: Relaxation of stage 1", J. Sci. Ind. Res. (India), vol. 68,
no. Issue. 12, pp. 1035-1037, 2009.
[28] P. Prevey, J. Telesman, T. Gabb, and P. Kantzos, "FOD resistance and fatigue
crack arrest in low plasticity burnished IN718".
CHAPTER 9
Non-Destructive Evaluation (NDE) of Welded

Structures for Aerospace Applications
Mohammad W. Dewan, Daniel J. Huggett, and Muhammad A. Wahab*

University, Baton Rouge, LA 70803, USA
Abstract: The aerospace industry has utilized high strength aluminum alloys to
propel the production and manufacturing of advanced aerospace technology;
however, welding of Aluminum and its constituent alloys introduce challenges which
affect the structural integrity of the welded area if conducted indecorously. For this
reason, Non-destructive Evaluation (NDE) of welded areas is conducted to ascertain
defective regions to ensure structural integrity of the aerospace structure. NDE
techniques are noninvasive and can determine whether the object contains
irregularities, discontinuities, or flaws. Inspecting weld areas allows for cost
reduction by detecting discontinuities in the early stages of manufacturing;
consequently, reducing the time and money to rework the error and allows for the
validation of sound welds. A variety of NDE techniques are available depending on
the applications, each with its own advantages and disadvantages. Among the
number of NDE techniques, radiography and ultrasonics are the most widely utilized
for inspection of weld defects. In this study, a detailed analysis was thereby
conducted to ascertain the critical phased array ultrasonic testing (PAUT) parameters
for the detection of weld defects, more specifically with Friction-Stir-Welding (FSW).
Consequently a comparison to X-ray radiography is also included. It was observed
*Corresponding author Muhammad A. Wahab: Department of Mechanical and Industrial Engineering,

Louisiana State University, Baton Rouge, LA 70803, USA; Tel/Fax: (1) 225 578 5823/(1) 225 578-5924;
E-mail: wahab@me.lsu.edu.
Yucheng Liu (Ed.)

Non-Destructive Evaluation (NDE) Frontiers in Aerospace Science, Vol. 1 355
that both techniques produced similar detection results for defects in the range of
1.0mm; however, it was found that PAUT was the only technique able to discover
defects in the range of 0.15 mm.
Keywords: Non-destructive evaluations, Phased array ultrasonic testing,

Radiographic testing, Welding, Welding defects.
INTRODUCTION
Welding is an essential manufacturing process performed in almost every

major industry; however, in many welding techniques flaws or defects are
quite common in the welded joints. These defects can be found in the
form of surface or sub-surface cracks, undercut, porosity, or inclusions
[1-3]; and consequently, failure can occur from these flaws [4]. An
important decision must be made regarding severity of these weld defects
and their effect on the strength; therefore, weld quality and integrity are
critical to safety in an extremely wide range of products and structures,
especially in the aerospace industry. To ensure sound welds have been
forged, Non-destructive Evaluation (NDE) techniques can be employed to
determine if defects have formed [5]. Different NDE methods can identify
cracking, porosity, incomplete penetration, misalignment, inclusions, and
lack of fusion which all can compromise weld strength and eventually
structural integrity under dynamic loading conditions.
NDE techniques are utilized in a multitude of scenarios including:

Determination whether an object is acceptable after each fabrication step
(in-process inspection), determining whether an object is acceptable for
final use (Final inspection), and lastly, determining whether an existing
object already in use is acceptable for continued uses (in-service
inspection). To summarize, NDE is applied to find welding defects as well
as quality assurance/quality control (QA/QC) of welded structures [6].
The most common NDE techniques to conduct various inspections are:
356 Frontiers in Aerospace Science, Vol. 1 Dewan et al.
Ultrasonic Testing (UT), Radiographic Testing (RT), Liquid Penetrant

Testing (LPT), Magnetic Particle Testing (MT), Eddy Current Testing
(ET), and Acoustic Emission (AE) testing. Each NDE technique discussed
previously has distinct advantages and disadvantages; consequently,
depending on the application one technique may be better suited than
another. Table 1 briefly illustrates NDE techniques and their principle of
operation, applications, limitations, advantages, and welding defects that
can be determined. It is noted that NDE techniques rely heavily on human
skills and knowledge to correctly assess and interpret results. Proper and
adequate training, developing confidence, and appropriate certifications
are required to perform non-destructive testing (NDT) [7, 8]. Therefore,
flaws or defects are often dictated by a code or requirement which
indicates acceptable tolerances, i.e. The American Society of Mechanical
Engineers Pressure Vessel Code [9] and American Welding Society
Structural Welding Code [10-12].
Among the number of NDE techniques, ultrasonics and radiography are

the most widely used for checking for weld defects. In recent years, the
conventional ultrasonic testing technique has been replaced with more
reliable and technologically advanced techniques of Phased Array
Ultrasonic Testing (PAUT) and Time of Flight Diffraction (TOFD).
Alternatively, radiography is utilized providing a range of techniques
from traditional X-ray generators and film to newer technologies such as
Computed Radiography (CR), Direct Radiography (DR), and 3D
Computed Tomography (CT). These new technologies allow for remote
visual inspection and enhancement of data visualization.
Table 1. Common NDT techniques and their applications [9].
Method Principle of Operation Application Limitations Advantages Welding Defects

Penetrant Liquid dye penetrant into Will not find Easy and can be Burn Through, Surface
Surface defects
Testing cracks and make visible subsurface or used in complex Porosity, Surface crack,
(PT) once developer is applied volumetric geometry undercut

defects
Not applicable
Can detect
Magnetic Magnetic particles are Surface and to
flaws up to ¼ Cracks, Incomplete
Particle attracted to breaks in near subsurface non-magnetic
inch below penetration, overlap
(MT) magnetic lines of force defects metals or
surface
alloys
High frequency Sensitivity is

Defects Can detect
ultrasound vibrations are reduced by Cracks, Incomplete
Ultrasonic anywhere defects with
introduced into sample. rough-surface fusion, Incomplete
Testing within the dimensions half
Waves are reflected or parts; a skilled penetration, porosity,
(UT) examined of the excitation
scattered by operator is undercut
volume wavelength
discontinuities required
Economic
Penetrating rays (X-ray or limit to depth
Burn Through, Excessive
gamma rays) cast Defects penetration; Permits visual
Radio-gra /inadequate
shadows on the other side anywhere hazardous analysis of
phic reinforcement,
of “solid” objects; film within the operation; buried defects
Testing Incomplete penetration,
radiography records examined complex or components
(RT) Misalignment, Porosity,
shadow on photographic volume shapes are in assembly
Root concavity, undercut
film difficult to
analyze
Constantly measures Cannot give

Eddy impedance of the probe Surface or absolute
Can be adapted
Current coil; coil impedance slightly measurement Cracks, Inclusions,
to high-speed
Testing changes with material subsurface only Incomplete penetration
production lines
(ET) properties and constituent flaws qualitative
variations comparison
ULTRASONIC TESTING (UT)
Ultrasonic testing has become a widely used NDE technique with many
advancements and variations. Ultrasonic testing can be used to detect
cracks, voids, and changes in geometric and material parameters such as:
thickness, stress concentration, and modulus [13]. In ultrasonic testing,
high frequency ultrasound vibrations are generated from piezoelectric
elements and thereby transmitted into test specimens. The transmitted high
frequency waves are reflected or scattered by discontinuities inside the
material as the waves propagate. The piezoelectric elements also act as
receivers which detect the ultrasonic reflections from the discontinuities.
FUNDAMENTALS OF ULTRASONIC TESTING
In solids, sound waves can propagate in four principle methods that are
established by fundamental particle oscillation principles. Sound can pro-
pagate as longitudinal waves, shear waves, surface waves, and plate
waves. Longitudinal and shear waves are the two modes of sound wave
propagation most commonly used in ultrasonic testing. In longitudinal
waves, the oscillations occur in the longitudinal direction or the direction
of wave propagation. In the transverse or shear wave, the particles
oscillate at a right angle or transverse to the direction of wave propagation.
Three important properties of sound waves propagating in isotropic solid
materials are: wavelength (𝜆), frequency (𝑓), and velocity (𝑉). The
wavelength is directly proportional to the velocity of the wave and
inversely proportional to the frequency of the wave. This relationship is
shown below in Eq. (1). The wavelength is related to defect detection
capabilities, which vary with ultrasonic transducer capabilities. Generally
the measurable smallest defect size should be larger than one half of the
wavelength. In general, a defect size must be larger than one-half the
wavelength in order to be detected. Therefore, if a material’s velocity
remains constant, by increasing the frequency the size of λ will decrease
which will result in smaller defects that can be found; hence, low
frequency probes can overlook defects and can cause an operator to
believe that a defect-free weld has been conducted.
𝑉𝑒𝑙𝑜𝑐𝑖𝑡𝑦 𝑉
𝑤𝑎𝑣𝑒𝑙𝑒𝑛𝑔𝑡𝑕, 𝜆 = (1)
𝐹𝑟𝑒𝑞𝑢𝑒𝑛𝑐𝑦 𝑓
The speed of sound (𝑉) within a material is a function of the properties

of the material and is independent of the amplitude of the sound wave.
The general relationship between the speed of sound in a solid and its
density and elastic constants is shown in Eq. (2). Ultrasonic velocity is
constant for each material.
𝐶𝑖𝑗
𝑉=√ (2)
𝜌
Where, elastic constant Cij is Young's modulus (E), Poisson’s ratio(𝜇), or

shear modulus ( 𝐺) , depending on the type of sound wave. When
calculating the velocity of a longitudinal wave, Young's Modulus and
Poisson's Ratio are commonly used. When calculating the velocity of a
shear wave, the shear modulus is used.
Sensitivity and resolutions are two important ultrasonic properties

generally used to describe the ability to locate flaws during testing.
Sensitivity is the ability to find small defects and resolution is the ability
to detect flaws that are close together within the materials. Generally
sensitivity and resolution of an ultrasonic probe increases with the
increase of frequency. On the other hand, as frequency increases, sound
tends to scatter from large or coarse grain structure and from small
imperfections within a material. Before selecting an inspection frequency,
the material's grain structure and thickness, and the discontinuity's type,
size, and probable location should be considered.
During ultrasonic testing, the sound waves generally originate from a

number of the piezoelectric transducer surface rather than a single point.
Due to multiple waves, the ultrasound intensity along the beam is affected
by constructive and destructive wave interference. The wave interference
results extensive variations in the sound intensity near the source and it is
known as near field. Because of intensity fluctuations within near field, it

is difficult to accurately estimate defect size when a defect is located
within this area. The area beyond the near-field where the ultrasonic beam
is more uniform is called the far-field [14-16]. In the far-field, the beam
spreads out in a pattern originating from the center of the transducer and
sound waves have maximum strength. Therefore, optimal detection
results will be obtained when flaws occur in the far-field area. Near-field
distance (𝑁) can be expressed as Eq. (3). Ultrasonic wedges helps to focus
the defect in far-field area for thin metal inspection. Schematic of near-field
and far-field is shown in Fig. (1).
𝐷2 𝐷2 𝑓
𝑁= = (3)
4𝜆 4𝑉
Where, 𝐷 is the probe diameter, 𝑓 is the probe frequency, and 𝑉 is the

material sound velocity.
Fig. (1). Radiated fields from an ultrasonic transducer: near-field and far-field [17].
The energy in the ultrasonic beam does not remain in a cylinder, but
instead, spreads out as it propagates through the material. The
phenomenon is usually referred to as beam-spread but is sometimes also
referred to as beam-divergence or ultrasonic diffraction. Beam-spread is

twice the beam-divergence (see Fig. 2 below). It occurs because the
vibrating particles of the material do not always transfer all of their
energy in the direction of wave propagation. Beam-spread is largely
determined by the frequency and diameter of the transducer. Beam-spread
is greater when using a low frequency transducer than when using a high
frequency transducer. As the diameter of the transducer increases, the
beam-spread will be reduced. Eq. (4) is used to calculate the beam-
divergence angle (θ). Larger beam-divergence results overestimation of a
defect size. Fig. (2) illustrates beam-divergence and beam-spread during
ultrasonic inspection.
𝑉
𝑆𝑖𝑛𝜃 = 1.2 (4)
𝐷𝑓
Fig. (2). Beam-spread and beam-divergence during ultrasonic testing [17].
Ultrasonic technology in general, can be classified into three categories:

conventional ultrasonic testing (UT), time-of-flight-diffraction (TOFD)
ultrasonic testing, and phased array ultrasonic testing (PAUT). Fig. (3)
illustrates PAUT, TOFD, and UT configurations designed with
ESBeamTool software. The basis of all UT inspection is A-scan data.
A-scan waveforms represent the reflections from one sound beam
position in the test piece (various illustrations for A-scan views are
discussed later). A-scan data are used to generate other scan views
including B, C, and S-scan views which allow better representation of

inspection data and defects. The B-scan is an image showing a
cross-sectional profile through one vertical slice of the test piece, showing
the depth of reflectors with respect to their linear position. The C-scan is a
two dimensional presentation of data displayed as a top or planar view of
a test piece. The S-scan or sectorial scan image represents a
two-dimensional cross-sectional view derived from a series of A-scans
that have been plotted with respect to time delay and refracted angle. Fig.
(4) illustrates B, C, and S-scan patterns of ultrasonic testing.
Fig. (3). Schematic of conventional UT, TOFD, and PAUT, three different ultrasonic set-up created
with ultrasonic simulation software (ESBeam Tool).
Fig. (4). Phased array ultrasonic scan pattern and different scan views [18].
CONVENTIONAL ULTRASONIC TESTING (UT)
A typical conventional ultrasonic inspection system consists of several

functional units, such as a pulser/receiver, transducer, and display device.
A pulser/receiver is an electronic device that produces high voltage
electrical pulses. Driven by the pulser, the transducer utilizing the
piezoelectric elements converts the electrical pulses into mechanical
energy in the form of ultrasonic vibrations. When there is a discontinuity
in the wave path, part of the energy will be reflected back from the flaw
surface. The reflected wave signal is transformed into an electrical signal
by the transducer and is displayed on a screen. The reflected signal
strength is displayed versus time. Signal travel time can be directly related
to the distance that the signal traveled. From the signal, information about
the reflector location, size, and orientation are acquired. The conventional
UT method scanning process is time consuming. The presentation of
defects is obtained from A-scan data and documentation of the defect

position has to be calculated manually. Inspections with more than one
angle require sequential operations, which also increases the inspection
time. One important aspect of conventional ultrasonic testing is to calculate
the size and location of the inspected defect. Fig. (5) illustrates a schematic
of weld defect detection using angle beam ultrasonic testing. Fig. (6) shows
the corresponding A-scan view.
Fig. (5). Schematic ultrasonic method to calculate defect location [17].
Fig. (6). Typical A-scan signal indicating received signal voltage vs. time.
For A-Scan data, an ultrasonic echo will be found in between the front
and back wall echoes versus travel time (∆𝑡). For the particular material,
sound velocity ( 𝑉) is known. Accordingly, the sound path can be
calculated using Eq. (5).
𝑠𝑜𝑢𝑛𝑑 𝑝𝑎𝑡𝑕 = (∆𝑡 × 𝑉)/2 (5)
Now, if the reflection angle (𝜃𝑅 ) and material thickness (𝑇) is known, the
location of the defect can be calculated using the following equations:
𝑆𝑢𝑟𝑓𝑎𝑐𝑒 𝑑𝑖𝑠𝑡𝑎𝑛𝑐𝑒 = 𝑆𝑖𝑛𝜃𝑅 × 𝑠𝑜𝑢𝑛𝑑 𝑝𝑎𝑡𝑕 (6)
𝐷𝑒𝑝𝑡𝑕 = 𝐶𝑜𝑠𝜃𝑅 × 𝑠𝑜𝑢𝑛𝑑 𝑝𝑎𝑡𝑕 (7)
TIME-OF-FLIGHT-DIFFRACTION (TOFD) ULTRASONIC

TESTING
Time-Of-Flight-Diffraction (TOFD) is another important version of

ultrasonic testing. In TOFD, a single element angled beam with one pulser
and receiver is used to obtain A and B scan views to represent defects. Fig.
(7) illustrates a schematic of the TOFD process and the A and B scan
views that are acquired. TOFD images can be corrupted by incorrect setup
parameters such as gain value, gate setup, transducer separation,
transducer positions from weld center or other problems such as electrical
noise [9]. At lower gain levels diffracted signals can become undetectable.
At high gain values noise levels increase which obscures the diffracted
signals leading to reduced probability of detection and poor sizing.
Correct gate settings are critical, because in TOFD A-scans are not that
easy to interpret since there are multiple visible signals. As a minimum,
the gates should cover the lateral wave and longitudinal wave’s first
back-wall signal. If the transducer separation distance is too large the
back-wall and lateral wave signal weaken, causing the diffracted signal
amplitudes to drop. If the transducer separation distance is too close, the

lateral wave signals become large and the back-wall signals become weak.
If the transducers are not centered on the weld, the diffracted signal
amplitudes will decline to the point where flaw detection is seriously
impaired. As discussed previously, noise levels can seriously impair
TOFD interpretation. The generation of noise can come from a number of
sources such as electrical interference, ultrasonic focusing depth, gain
values, and coupling. One main challenge of TOFD analysis is the data
interpretation and classification of defects. Table 2 represents some
typical welding defects and corresponding TOFD scan views.
Fig. (7). Schematic of TOFD scanning process and typical TOFD A and B scan views showing
different echoes [9].
Table 2. Interpretation of Time-Of-Flight-Diffraction (TOFD) ultrasonic images [9].
Type of
UT Signals Schematic and TOFD Scan View
Defect
Two signals (2, 3)
Incomplete
from top and
root
bottom of the
penetration
defect
More back wall
signal and
Lack of root
inverted as
penetration
compared to
lateral wave (LW)
Side wall Two signals from

lack of top and bottom of
fusion the defect
One signals
Interphases
in-between LW
lack of
and back-wall
fusion
signal
Many small
signals
Cluster in-between LW
Porosity and back-wall
echoes, looks like
noise
Concave Distortion of
root back-wall echo
PHASED ARRAY ULTRASONIC TESTING (PAUT)
PAUT is relatively a new ultrasonic testing technology and greatly

decreases the level of difficulty in determining welding defects.
Ultrasonic phased arrays eventually replaced conventional ultrasonic
methods in many nondestructive evaluation applications [19-21]. Phased
arrays offer increased inspection sensitivity and coverage, as well as
decreasing inspection times compared with the conventional ultrasonic
techniques using single and multi-element transducers. Another benefit of
phased array systems is their ability to produce immediate images
allowing straightforward visualization of the internal structure of a
component thereby simplifying data interpretation. By using PAUT
technology an operator is better prepared to determine the size, shape, and
orientation of a defect versus traditional pulse echo UT techniques [22,
23].
PAUT has three distinct components. Firstly, the probe (transducer) which
functions as the transmitter and receiver of the high frequency sound
waves. Phased array ultrasonic probes utilize an array of piezoelectric
elements that in sequence generate high frequency sound waves.
Transducer frequencies are most commonly found to be in the range of 2
MHz to 10 MHz [24]. The probe is mounted on top of a wedge
component which couples the sound waves generated by the probe to the
work piece. Wedges are also used to steer the sound waves at a certain
angle as they enter the work-piece according to Snell’s Law. The couplant,
typically a moderately viscous, nontoxic liquid or gel, is used between the
wedge and the work-piece to facilitate the transmission of sound energy
between the wedge and the test piece. Lastly, a data acquisition unit is
used to generate images that can utilize the sound waves to determine
defects of a material. These data acquisition units are very sophisticated
and have many input variables that will affect the images produced. In Fig.
(8), an illustration is given of a probe/wedge unit that is situated on top of
a Friction-Stir-Weld with a defect located on the upper advancing side.

During one firing sequence, each element is engaged beginning in the
direction of the arrow starting from A and ending at B (shown on the
probe). With the associated wedge angle and firing sequence, a wave front
is created following Huygen’s interference patterns [24, 25] allowing
defects to be found in a large area. For simplicity only the “first leg” of
the wave front is illustrated. In practical applications, one can use
multiple legs and can view larger swept areas.
Fig. (8). A typical probe and wedge configuration with illustration of wave propagation inside a
friction stir (FS) welded specimen.
For PAUT inspection, multiple input parameters must be entered into the
data acquisition unit to accurately determine defect sizes and locations.
These variables include the characteristics of the transducer, wedge, and
welded area. Setup variables also include parameters that assist in
generating an accurate focal law for the high frequency sound waves.
These parameters include scan type, wave type, number of elements to be

viewed, beam angle, and gate settings. Lastly, one of the most important
procedures that need to be conducted is “calibration”. Different types of
calibration blocks (ASTM, PACS, IIW, NAVSHIP, etc.) are utilized
depending on the application and defect code. Calibration blocks should
be made of the same material that is to be inspected [26]. In the
calibration procedure, many steps are required that have to be followed in
a particular order to allow for the system to calculate correct wave
characteristics. In general, three procedures have to be conducted. These
include the material velocity, wedge delay, and sensitivity calibrations. In
the calibrations, three different types of reflectors can be used to verify
velocity and wedge delay variables which include radius, depth, and
thickness. If one is using an angle-beam wedge, the radius and depth
reflectors must be utilized; however, if a 0 degree wedge is employed then
the thickness reflector (back-wall reflection) can be used.
The sound waves are predictable due to wave physics principles. These
wave principles are classical physics phenomena and have been well
documented as they pertain to ultrasonics, as seen in [27-34]. Work has
also been conducted to determine the effect of element size in linear
phased arrays [35]. Optimization of PAUT probe design has likewise been
discussed in [36, 37]. Simulations of PAUT systems have been conducted
to predict phased array ultrasonic wave interactions and behavior inside
materials in [38-40]. With this new technology, simulation software has
been developed to aid in the scanning process to create the required
PAUT software to accurately detect defects. A requirement is to describe
the inspection task and enter fundamental information such as site
location, number of welds, pipe diameter, thickness, material properties,
weld preparation, procedure, and ultrasonic method to be used. The
software will then calculate and generate all the UT set-up parameters
required to perform Phased Array inspection of the welds.
As mentioned earlier the most common scan views (images) that a PAUT
system utilizes are A, B, C, and S scans. If there is a defect present in the
scanned samples, it can be visualized with 6-dB color change [25, 41, 42].
Fig. (9) shows typical A, S, and C-scan views obtained from phased array
ultrasonic testing. The S-scan is perhaps the most useful asset to current
PAUT post-weld evaluation due to the ability to steer the sound waves in
a range of angles which allows for easy visualization of a specimen [20].
The horizontal axis corresponds to location of the defect from probe
position, and the vertical axis to depth of the test piece. During inspection,
the inspection data for each weld is stored into the memory. The location
can be identified by incorporating an encoder or scanner. After inspection,
a sophisticated software package is employed allowing a qualified
ultrasonic inspector to review and analyze the inspection data utilizing
advanced analysis tools such as: real time volume corrected imaging, as
well as, conventional image analysis, image enhancement, and
measurement determination.
Fig. (9). PAUT A, S, and C-scan view acquired using Olympus Omniscan MX2.0 data acquisition
unit [18].
Defect Sizing Using PAUT
Defect sizing capabilities of the PAUT unit was examined using the
Olympus OmniScan MX2.0 unit and TomoView software version 2.10. For
defect sizing, two important parameters need to be considered: (i) gain
and (ii) index offset. In this study, defects were measured with (-2 dB), (-3
dB), and (-6 dB) color drop techniques and was compared with actual
defect sizes. Higher gain value results in higher A-scan amplitude; and
higher A-scan amplitude leads to oversize estimation. It is also observed
that the higher color drop (-6 dB) technique implies larger defect sizing
compared to smaller color drop (-2 dB) techniques. Hence, during
post-weld inspection the defect size is dependent on the peak A-scan
amplitude and color drop technique (Fig. 10). The index offset is the
distance of the tip of the wedge from the weld-centerline. For constant
A-scan amplitude, a higher index offset results in higher defect size
estimation (Fig. 11).
Fig. (10). Effect of A-scan amplitude and color drop on defect size estimation.
Another important calibration parameter is Time-Corrected-Gain (TCG).

If the TCG calibration is not performed, the A-scan amplitude will
decrease with the increase of index offset, which results in defect size
estimation errors. After TCG calibration, A-scan peak amplitude remains
constant with the variation of index offset (Fig. 12).
Fig. (11). Effect of index offset and color drop on defect size estimation (fixed gain value).
Fig. (12). Variations of A-scan amplitude with index offset to illustrate the effect of TCG calibration.
To find the detectability and defect sizing capability of PAUT, varying

sizes (0.79 mm, 1.58 mm, 1.98 mm, 2.38 mm, 2.78 mm, 4.76 mm, and
6.35 mm diameters) of seven holes were prepared on an AA6061-T651
plate and scanned with a calibrated PAUT unit. Fig. (13) shows the
C-scan view of the plate with varying-hole sizes. The PAUT unit can
precisely detect the sizes and locations of the holes. To identify the defect
sizing precision and accuracy of PAUT, different size side-drilled-holes
(SDH) were bored into an Aluminum Alloy (AA) 2xxx block at varying
depths and were measured. The variation in measurement is shown in
Table 3. Absolute error in height, depth, and location sizing was
0.24±0.20, 0.28±0.18, and 0.46±0.17 respectively. It is generally accepted
that the tolerance for PAUT is 0.5-1.0 mm depending on how stringent the
application requirement [13]. It is observed that there is overestimation
for thicker plates and underestimation for thinner plates. This exemplifies
one of the characteristics of waves propagating through a material. For a
defect at larger thicknesses, the sound waves travel a larger distance
which causes the wave to expand.
Fig. (13): Aluminum alloy plate with seven varying-hole sizes with associated C-scan and
eco-dynamic A-Scan images.
This results in the elongation of the defect image; alternatively, for defects
at smaller depths, the wave has not reached its focal length and thereby
underestimates the defect. In other words the defect image depends on the
near and far fields focusing distances of the wave. These results indicate
that PAUT has the ability to accurately detect defect sizes and locations.
Table 3. Comparison of defect sizing of PAUT.
Thick Actual Meas. Δ Actual Meas. Δ Actual Meas. Δ
-ness Dia. Dia. Dia. Depth Depth Depth Dist. Dist. Dist.
(mm) (mm) (mm) (mm) (mm) (mm) (mm) (mm) (mm) (mm)
25.5 2.8 3.1 0.3 8.1 8.25 0.15 16.1 16.6 0.5
25.5 2.8 3 0.2 14.1 14.2 0.1 10.8 11.4 0.6
25.5 1.6 2.2 0.6 11.8 11.9 0.1 17.9 18.3 0.4
25.5 3.5 3.6 0.1 12.5 11.9 -0.6 18.5 19.0 0.5
8.35 1.6 1.5 0.1 4.2 3.9 -0.3 4.1 3.6 -0.5
8.35 2.4 2.0 -0.4 4.2 3.8 -0.4 3.6 3.0 -0.6
8.35 2.8 2.8 0 4.5 4.2 -0.3 12.5 12.4 -0.1
Case Study: Welding Defect Detection with PAUT
In a previous study, PAUT was applied to find and classify welding

defects in Tungsten-Inert-Gas (TIG, also known as GTAW) welded
aluminum alloy AA6061-T651 plates [43]. Non-destructively measured
defects were classified into three different categories and compared with
microscopic defect sizing. PAUT defect sizing correlates well with
destructive measurements (Fig. 14). PAUT was also applied to detect
flaws in Friction-Stir-Welds (FSW). These friction stir (FS) welds were
executed utilizing varying weld parameters (plunge or forge force, travel
speed, and spindle speed). Fig. (15) illustrates a FS weld that has incurred
a surface defect located on the advancing side (AS) of the weld. Another
example in Fig. (16) illustrates a FS weld that has incurred an internal

defect located on the advancing side (AS) of the weld. The defect found
from Fig. (16) was measured to be in the range of 0.15 mm. After
scanning with the PAUT unit, the defect sizes and locations were verified
with optical microscopy (OM) which confirms that PAUT can be
successfully applied to find ‘on the surface’ and embedded internal
defects.
Fig. (14). Welding defect sizes measured with optical microscope and PAUT [43].
Fig. (15). PAUT’s S and A-scan views, defect location, and cross section optical micrograph of a
FS welded panel illustrating a surface-defect.
Fig. (16). PAUT’s S-Scan view, A-Scan View, defect location, and cross sectional optical
micrograph of a FS welded panel illustrating an internal defect.
RADIOGRAPHIC TESTING (RT)
One of the widely used NDT methods for volumetric examination is

radiography. In radiography, X-rays or gamma-rays are used to produce a
radiographic image obtaining differences in thickness, defects (internal
and surface), and changes in structure. The procedure for producing a
radiograph is to have a source of penetrating (ionizing) radiation (X-rays
or gamma-rays) on one side of a specimen to be examined and a detector
of the radiation (radiographic film) on the other side, as shown in Fig.
(17).
Fig. (17). Schematic of basic set-up for film radiography [9, 44].
FUNDAMENTALS OF RADIOGRAPHIC TESTING
Different radioactive materials and X-ray generators produce radiation at

different energy levels and at different rates. It is important to understand
the terms used to describe the energy and intensity of the radiation. The
energy of a single wavelength X-ray can be calculated by the Eq. (8).
𝑕𝑐
𝐸 = 𝑕. 𝜗 = (8)
𝜆
Where, 𝐸 is the energy, 𝑕 is Planck’s constant, 𝜗 is frequency, and c is

velocity of electromagnetic radiation, and 𝜆 is the wavelength.
The energy of the radiation is responsible for its ability to penetrate matter.
Higher energy radiation can penetrate more and higher density matter
than low energy radiation. Radiation intensity is the amount of energy
passing through a given area that is perpendicular to the direction of
radiation travel in a given unit of time. The penetrating intensity can be
expressed as Eq. (9).
𝐼 = 𝐼0 . 𝑒 −𝜇𝑡 (9)
In which, 𝐼0 is intensity at material entry, 𝐼 is intensity at materials exit, 𝜇

total absorption coefficient, and 𝑡 is materials thickness.
The relation between intensity and distance according to the Newton's

Inverse Square Law can be expresses as Eq. (10).
𝐼1 𝑑1 2 = 𝐼2 𝑑2 2 (10)
Where, 𝐼1 and 𝐼2 are intensities at distances 𝑑1 and 𝑑2 from the source,

respectively. Source to film distance (𝐹) can be calculated from the Eq.
(11).
𝑡(𝑈𝑡 +4𝑠)
𝐹= (11)
𝑈𝑡
Where, s is effective source size, t is the object thickness, and 𝑈𝑡 is total

un-sharpness which can be expressed (see Eq. 12) in-terms of film
un-sharpness (Uf) and geometric un-sharpness (Ug).
𝑈𝑡 = √𝑈𝑓2 + 𝑈𝑔2 (12)
Presently, a wide range of industrial radiographic equipment, image

forming techniques, and examination methods are available. Similar to
ultrasonic NDE methods, knowledge and experience are needed to select
the appropriate method for a particular application. Generally for weld
inspection, the methodology will follow ASTM-E1032 standards [45]
with associated safety requirements [10]. The radiographic sensitivity
shall be judged based on a penetrometer or wire image quality indicators
(IQI). The results of the X-ray image produce a two dimensional
representation of the specimen. In a radiographic film, darkness of the
film vary according to the amount of radiation that has reached the film
through the test object. In radiographic film, darker areas indicate more
exposure and lighter areas indicate less exposure. The processed film (or
image) is usually viewed by placing it in front of a screen providing white
light illumination of uniform intensity so that the light is transmitted
through the film and the image can be clearly seen. The term
“radiographic density” is a measure of the degree of film darkening.
Radiographic density is the logarithm of two measurements: the intensity
of light incident on the film (𝐼0 ) and the intensity of light transmitted
through the film (𝐼𝑡 ).
𝐼
𝑅𝑎𝑑𝑖𝑜𝑔𝑟𝑎𝑝𝑕𝑖𝑐 𝑑𝑒𝑛𝑠𝑖𝑡𝑦 = log( 0 ) (13)
𝐼𝑡
One important challenge of radiographic analysis is the data interpretation

and defect classification. For this reason, X-ray images that are acquired
are compared with standard radiographic images [9, 44]. Table 4
illustrates typical welding defects and corresponding radiographic images.
Table 4. Different welding defects and corresponding RT images [9, 44].

Welding Weld configuration and Radiographic Image
Description
Defect
An abrupt change in film

Offset or
density across the width of the
mismatch
weld image
An abrupt density change

Offset or
across the width of the weld
mismatch
image with a straight
with lack of
longitudinal darker density
penetration
line at the center
External The weld density is darker
concavity or than the density of the pieces
insufficient welded and extending across
fill the full width of the weld
A lighter density in the center
Excessive of the width of the weld
penetration image, either extended along
the weld or in isolated circular
drops
An irregular darker density
along the edge of the weld
External
image. The density will
undercut
always be darker than the
density of the pieces being
welded
Internal An irregular darker density
(root) near the center of the width of
undercut the weld image and along the
edge of the root pass image
Internal An elongated irregular darker

concavity density with fuzzy edges, in
(suck back) the center of the width of the
weld image
Localized darker density with

Burn
fuzzy edges in the center of
through
the width of the weld image.
Incomplete
A darker density band, with
or Lack of
very straight parallel edges, in
Penetration
the center of the width of the
(LOP)
weld image
Interpass Irregularly shaped darker

slag density spot, usually slightly
inclusions elongated and randomly
spaced.
Elongated
Elongated parallel or single
slag lines
darker density lines, irregular
(wagon
in width and slightly winding
tracks)
lengthwise
Elongated parallel, or single,

Lack of side darker density lines sometimes
wall fusion with darker density spots
(LOF) dispersed along the LOF-lines
which are very straight in the
lengthwise direction
Small spots of darker

Interpass densities, some with slightly
cold lap elongated tails in the welding
direction
Rounded spots of darker

Scattered
densities random in size and
porosity
location
Rounded or slightly elongated

Cluster darker density spots in clusters
porosity with the clusters randomly
spaced
Rounded and elongated darker

Root pass density spots that may be
aligned connected in a straight line in
porosity the center of the width of the
weld image
Feathery, twisting lines of

Transverse
darker density running across
crack
the width of the weld image
Feathery, twisting line of

darker density running
Longitudina
lengthwise along the weld at
l crack
any location in the width of
the weld image
Feathery, twisting lines of

darker density along the edge
Longitudina of the image of the root pass.
l root crack The “twisting” feature helps to
distinguish the root crack from
incomplete root penetration
Irregularly shaped lower

Tungsten
density spots randomly
inclusions
located in the weld image
Case Study: Welding Defect Detection with RT
In the current study FS welded plates were analyzed employing digital

X-ray radiography and compared with PAUT results. Fig. (18) illustrates
a representation of the X-ray images of internal and surface defects in the
FS welded plates. It is shown that X-ray radiography can be successfully
applied to locate the surface-defects and embedded internal defects.
Fig. (18). Digital Radiographic images of FS welded AA 2219-T87 plates showing (a) internal
defect and (b) surface-defect.
Comparison of PAUT and RT to Obtain Welding Defect
A comparison of the two most prominent techniques, PAUT and X-ray

radiography, defect detection capabilities was conducted utilizing various
FS welded specimens. It was observed that both methods easily found
large surface-defects as well as internal welding defects. Fig. (19)
illustrates a representation of internal defects of the X-ray image
compared with the B-scan and C-scan views of the PAUT system. This
shows that both the NDE techniques correlate very well for large defects,
which is expected; however, practically in industry smaller defects are
more common. Therefore, a very small defect was analyzed, see Fig. (20).
The defect found was measured to be in the range of 0.15 mm; however,
fascinatingly the X-ray image does not detect the small defect. The
associated PAUT’s B-scan and C-scan images produce the defect which is
located in two portions of the weld.
Fig. (19). Radiographic and PAUT’s B-Scan and C-Scan images illustrating a surface-defect in a
FS welded AA 2219-T87 panel.
Fig. (20). Radiographic and PAUT’s B-Scan and C-Scan images of a FS welded AA 2219-T87
panel.
CONCLUSION
The objective of this study was to increase understanding of the post-weld

NDE techniques that are currently used in industry, especially in the
aerospace industry which heavily relies on welding. For further
verification, NDE results were compared with microscopic analysis.
X-ray radiography requires detailed knowledge to interpret correctly;
although, in a sense, is more straightforward due to the permanent image
that is created. PAUT post-weld inspection is complex and requires
significant time and understanding in order to be applied effectively. It
has been shown that the PAUT technique is an excellent choice for this
undertaking and also the more reliable of the two techniques. It was found
that the essential methods to conduct PAUT include setup and calibration.
The calibration process requires that sensitivity, velocity, and wedge delay
values must be accurately determined to precisely display defect sizes and

locations. The importance of A-scan amplitude, index offset, gain value,
TCG calibration, PAUT imaging, and defect sizing were discussed. It was
observed that in order not to overlook small defects a higher frequency
phased array probe is required. Depending on the application, the PAUT
system may need to be calibrated to find small defects with a satisfactory
color drop.
this publication.
ACKNOWLEDGEMENTS
Declared none.
REFERENCES
[1] T.M. Morton, "Defects in 2021 aluminum alloy welds", Weld. J., vol. 7, pp.
304-312, 1971.
[2] J. F. Rudy, and E. J. Rupert, "Effects of porosity on mechanical properties of

aluminum welds", Weld. Res. Suppl., vol. 49, pp. 322s-336s.
[3] F.V. Lawrence Jr, W.H. Munse, and J.D. Burk, "Effects of Porosity on the
Fatigue Properties of 5083 Aluminum Alloy Weldments", Welding Research
Council Bulletin, vol. 206, pp. 95s-98s, 1975.
[4] H.D. Greenberg, Applications of Fracture Toughness Parameters to Structural

Metals. Gordon and Breach: New York, NY, 1966.
[5] R.A. Youshaw, "Tensile properties of HY 80 steel welds containing defects

correlated with ultrasonic and radiographic evaluation", Weld. J., vol. 4, pp.
198-206, 1972.
[6] R. Halmshaw, Introduction to the Non-Destructive Testing of Welded Joints.

Abington Publishing: Cambridge, England, 1998.
[7] ASNT Standard for Qualification and Certification of Nondestructive Testing

Personnel “, ANSI/ASNT CP-189. American Society for Non-destructive
Testing: Columbus, OH, 2011.
[8] I.V. Rivero, "Fundamentals of nondestructive testing at Texas tech university",

Mater. Evil. vol. 64, pp. 765-768, 2006.
[9] "ASME Boiler and Pressure Vessel Code (BPVC)," In: Section V-
Nondestructive Examination, ASME International, New York, NY, 2013.
[10] "Structural Welding Code – Steel", In: AWSD1.1/D1.1M. American Welding

Society: Miami, FL, 2010.
[11] "Structural Welding Code - Aluminum", In: AWSD1.2/D1.2M. American

Welding Society: Miami, FL, 2008.
[12] "Specification for Friction Stir Welding of Aluminum Alloys for Aerospace
Applications", In: AWS D17.3/D17.3M:200X. American Welding Society:
Miami, FL.
[13] D.E. Bray, and D. McBride, Nondestructive Testing Techniques.

Wiley-Interscience: New York, 1992.
[14] G.S. Kino, Acoustic Waves: Devices, Imaging, and Analog Signal Processing.
Prentice Hall PTR: New Jersey, 1987.
[15] K.B. Ocheltree, and L.A. Frizzel, “Sound field calculation for rectangular
sources", IEEE Trans. Ultrason. Ferroelectr. Freq. Control, vol. 36, no. 2, pp.
242-248, 1989.
[http://dx.doi.org/10.1109/58.19157] [PMID: 18284974]
[16] A. Weyns, "Radiation Field Calculations of Pulsed Ultrasonic Transducers:

Part 2: Spherical Disc- and Ring-Shaped Transducers", Ultrasonics, vol. 18,
pp. 219-223, 1980.
[http://dx.doi.org/10.1016/0041-624X(80)90124-9]
[17] “Ultrasonic Transducers Technical Notes,” Tech. Report. Olympus NDT:

Waltham, MA, 2006.
[18] “Phased array testing: Basic theory for industrial applications,” Tech. Report.
Olympus NDT: Waltham, MA, 2010.
[19] D. Choqueuse, and A. Lamarre, "Use of Phased Array Ultrasonic Equipment

for Fatigue Crack Characterization for Underwater Inspection of Offshore
Structures", 8th International Offshore and Polar Engineering Conference,
1998 Montreal: Canada
[20] P. Crowther, "Practical Experience of Phased Array Technology for Power

Station Applications", Insight: Non-Destructive Testing and Condition
Monitoring, vol. 46, pp. 525-528, 2004.
[http://dx.doi.org/10.1784/insi.46.9.525.40843]
[21] M. Moles, N. Dube, and F. Jacques, "Ultrasonic Phased Arrays for Thick
Section Weld Inspections," In: Joining of Advanced and Specialty Materials
VI: Proceedings from Materials Solutions 2003 on Joining of Advanced and
Specialty Materials, 2004, pp. 112-118.
[22] A.S. Birks, R.E. Green, and P. McIntire, Nondestructive Testing Handbook. 2nd
ed American Society for Nondestructive Testing: Columbus, OH, 1991.
[23] B. Li, Y. Shen, and W. Hu, "The study on defects in aluminum 2219-T6 thick
butt friction stir welds with the application of multiple non-destructive testing
methods", Mater. Des. vol. 32, pp. 2073-2084, 2011.
[http://dx.doi.org/10.1016/j.matdes.2010.11.054]
[24] “Introduction to Phased Array Ultrasonic Technology Applications: R/D

Tech Guideline,” Tech. Report. Olympus NDT: Waltham, MA, 2004.
[25] “Advances in Phased Array Ultrasonic Technology Applications” Tech.

Report. Olympus NDT: Waltham, MA, 2007.
[26] "Standard Guide for Evaluating Performance Characteristics of Phased-Array

Ultrasonic Testing Instruments and Systems", In: ASTM E2491-13. ASTM
International: PA, 2013.
[27] W. Shi-Chang and S. Yijun, "Optimum Beam Steering of Linear Phased

Arrays," Wave Motion, vol. 29, April 1999.
[28] W. D. Bruce and D. W. Paul, "Ultrasonic Arrays for Non-destructive

Evaluation: A review," NDT & E International, vol. 39, October 2006.
[29] J. L. Rose, "A baseline and vision of ultrasonic guided wave inspection
potential," J. Press. Vess. Tech., vol. 124, August 2002.
[http://dx.doi.org/10.1115/1.1491272]
[30] J.L. Rose, Ultrasonic Waves in Solid Media. Cambridge University Press:
New York, 2004.
[31] J.L. Rose, "Guided wave nuances for ultrasonic nondestructive evaluation",
IEEE Trans. Ultrason. Ferroelectr. Freq. Control, vol. 47, no. 3, pp. 575-583,
2000.
[http://dx.doi.org/10.1109/58.842044] [PMID: 18238584]
[32] L. Azar, Y. Shi, and S. C. Wooh, "Beam Focusing Behavior of Linear Phased
Arrays," NDT & E International, vol. 33, April 2000.

[http://dx.doi.org/10.1016/S0963-8695(99)00043-2]
[33] J.D. Achenbach, Wave Propagation in Elastic Solids. North-Holland

Publishing Co.: New York, 1984.
[34] B.A. Auld, Acoustic Fields and Waves in Solids, vol. Vol. 2. 2nd ed Kreiger
Publishing Co.: FL, 1990.
[35] S. Wooh and Y. Shi, "Influence of phased array element size on beam steering
behavior," Ultrasonics, vol. 36, April 1998.
[http://dx.doi.org/10.1016/S0041-624X(97)00164-9]
[36] V. Lupien, "Phased array ultrasonic probe design: From art to science through
optimization", AIP Conf. Proc., vol. 700, pp. 761-768, February 2004.
[http://dx.doi.org/10.1063/1.1711697]
[37] R.A. Roberts, "Efficient design of ultrasonic phased array transducer surface
geometry", AIP Conf. Proc., vol. 820, pp. 797-804, March 2006.
[http://dx.doi.org/10.1063/1.2184608]
[38] L. Satyanarayan, C. Sridhar, C.V. Krishnamurthy, and B. Krishnan,

"Simulation of ultrasonic phased array technique for imaging and sizing of
defects using longitudinal waves", Int. J. Press. Vessels Piping, vol. 84, 2007.
[http://dx.doi.org/10.1016/j.ijpvp.2007.08.002]
[39] Y. Jing, K. Hak-Joon, S. Sung-Jin, K. Sung-Sik, K. Kyungcho, and S.

Myung-Ho, "Model-based simulation of focused beam fields produced by a
phased array ultrasonic transducer in dissimilar metal welds", NDT Int., vol.
44, 2011.
[40] X. Na, and Z. Zhenggan, "Numerical simulation and experiment for inspection
of corner-shaped components using ultrasonic phased array", NDT Int., vol. 63,
2014.
[41] M.W. Dewan, J. Liang, M.A. Wahab, and A.M. Okeil, "Effect of post-weld
heat treatment and electrolytic plasma processing on tungsten inert gas welded
AISI 4140 alloy steel", Mater. Des. vol. 54, pp. 6-13, 2014.
[http://dx.doi.org/10.1016/j.matdes.2013.08.035]
[42] “Standard Guide for Evaluating Performance characteristics of phased-array

ultrasonic testing instruments and systems,” ASTM-E2491. ASTM
International: PA, 2008.
[43] M.W. Dewan, M.A. Wahab, and A.M. Okeil, "Effect of weld defects on tensile
properties of light weight materials and correlations with phased array
ultrasonic non-destructive evaluation", In: ASME 2014 International
Manufacturing Science and Engineering Conference, Detroit: Michigan, 2014.
[44] “Introduction to the overview of “Industrial Radiography”- Image forming

techniques,” Tech. Report, General Electric (GE). Inspection Technologies:
Fairfield, CT, 2007.
[45] "Standard Test Method for Radiographic Examination of Weldments", In:

ASTM E1032 – 12. ASTM International: PA, 2012.
CHAPTER 10
Development of a Remotely Piloted Helicopter

for Civil Applications
N. Goudarzi1,*, R. M. Ziazi2, F. Kermanshahi3, S. Sadati4, M. S. Sajedi5,
and M. Mohagheghi6
1
Department of Engineering Technology and Construction Management,
University of North Carolina at Charlotte, NC, USA
2
School of Mechanical, Industrial, and Manufacturing Engineering,
Oregon State University, OR, USA
3
Ecole Polytechnique de Montreal, Quebec, Canada
4
Sustainable Environment and Energy Systems Department, Middle East
Technical University, Northern Cyprus Campus, Turkey
5
Aerospace Engineering Department, Amirkabir University of Technology,
Tehran, Iran
6
Aerospace Engineering Department, University of Tehran, Tehran, Iran
Abstract: An applicable procedure for design, optimization, and manufacturing of a

remotely piloted helicopter (RPH) is studied analytically and experimentally. The
procedure is presented in four main phases of conceptual design, detailed design,
manufacturing and assembly, and test and reliability analysis. Using this procedure,
an RPH case study, called Parvan, for an arbitrary traffic monitoring mission is
*Corresponding author Navid Goudarzi: Department of Engineering Technology and Construction

University of North Carolina at Charlotte, NC, USA; Tel/Fax: 704-687-5056; E-mail: navid.goudarzi@uncc.edu.
Yucheng Liu (Ed.)

394 Frontiers in Aerospace Science, Vol. 1 Goudarzi et al.
designed and manufactured. Different subsystems of an RPH such as rotor assembly,

engine and fuel systems, power transmission system, and control system for the
proposed design are studied at each phase accordingly. Finally, 3-D
multiplatform software of CATIA is used to simulate the RPH and making the
full-scaled prototype. The successful performance of Parvan at hover, climb, and
forward flight modes showed the effectiveness of the proposed procedure in
developing an RPH.
Keywords: Aerodynamic derivatives, Design experiments method,

Dynamic modeling, 3-D simulation, Manufacturing, Optimum design,
Reliability, Remotely piloted helicopter, Stability and control, Unmanned
aerial vehicle.
NOMENCLATURE
Symbol Description
Acronyms
FMEA Failure Modes and Effect Analysis
FMECA Failure Modes, Effects, and Criticality
Analysis
MTBF Mean Time between Failures - s
MTTF Mean Time to Failure - s
PDF Probability Density Function
RPH Remotely Piloted Helicopter
UAV Unmanned Aerial Vehicle
Greek Symbols
βc,CR Control rotor longitudinal flapping - rad
βs,CR Control rotor lateral flapping - rad
λs(t) Failure rate - s
Ω Rotational speed - rad/s
ρ Air density - kg/m3
σ Solidity factor
ΘOM Main rotor blade pitching angle - rad
Development of a Remotely Piloted Helicopter Frontiers in Aerospace Science, Vol. 1 395
Θ0T Tail rotor blade pitching angle - rad

Roman Symbols
V Acceleration - m/s2
A Rotor swept area – m2
A1 Lateral cyclic pitch - rad
B1 Longitudinal cyclic pitch - rad
C Criticality number
CD Drag coefficient
C1 Lift coefficient
Cp Power coefficient
CT Thrust coefficient
E Hinge offset - m
F Force exerted on the body - N
Ib f Moment of inertia for body frame –
kg.m2
M Pitching moment - N.m
m Mass - kg
P Power - W
p Pitching rate - rad/s
Pi Induced power - W
Po Profile power - W
Q Torque - N.m
q Rolling rate - rad/s
R Rotor radius - m
Ry Rotor moment - N.m
ROC Rate of climb - m/min
wbf The matrix of rotational rates of body
frame
X Force acting in x direction – N
Y Force acting in y direction – N
Z Force acting in z direction - N
INTRODUCTION
In recent years, Unmanned Aerial Vehicles (UAVs) with enhanced

control systems and electronics have been developed to improve their
performance and reliability for civil applications. Among different
UAVs that include an extensive range of aircrafts, from small
hand-launched to strategic high-tech ones, Remotely Piloted
Helicopters (RPHs) with the capability of performing different flight
modes and maneuvers such as hovering and backward flight in a small
radius of turn have drawing more interest on the market. To name a
few, traffic monitoring [2], communication relays [3], forest fire
surveillance [4], sports reporting [5], and delivering goods [6] are
among civil applications of RPHs. In recent years, there is a strong
support of implementing renewable energies across the world [7, 8].
Employing UAVs in civil applications would help the global motion for
CO2 emission reductions.
Besides all advantages of RPHs for civil applications, their dynamics

complexities, complicated aerodynamics, and structural challenges
require precise studying to develop reliable and robust models. The
statistical information achieved from an earlier study on distribution of
operational RPHs all around the world, showed the worldwide interest
in developing these UAVs [1]; USA is the pioneer among more than 20
active countries in developing this technology. Monitoring roadway
networks with large distances require either a large fixed infrastructure
or large amount of labor. Also, intelligent transportation system
technologies require quick and real-time aerial monitoring for
highway/city traffic information and management, and roads surface
condition monitoring (in mountain regions) [9, 10]. In this chapter, a
novel procedure for developing an RPH is proposed in four major
phases:
1. Conceptual design
2. Detailed design
3. Manufacturing and assembly
4. Tests and reliability analysis.
Each phase is studied in details through applying this procedure on an

RPH case study, called Parvan, with a specific application in traffic
monitoring. It should be noted that the emphasis here is on providing a
designing procedure of the development of an RPH. Hence, further
studies on different aspects of an RPH with the traffic monitoring
mission such as flight stability, camera selection, data transmission, and
high quality surveillance are required.
CONCEPTUAL DESIGN
The main goal in the conceptual design phase is to specify the desired
application of the RPH and determine the proper range of weights,
dimensions, and stability and performance characteristics at different
flight modes including hover, climb, and forward flights. There are
different methods to obtain these characteristics. Two different methods
can be implemented: 1) deriving theoretical equations based on
aerodynamic and dynamic principles of helicopter performance, such as
blade element theory and momentum theory [11-14], and 2) deriving
empirical equations based on similar helicopters characteristics. Here, we
employ a combination of both methods to obtain the desired RPH
characteristics for the defined mission. A strong database that includes
dimensions and performance characteristics of similar operational
unmanned helicopters is needed. Table 1 summarizes the studied parameters
of more than 100 RPHs all around the world. These RPHs are classified in
four major groups of power unit, weight, performance, and dimension [1]. In
order to have a complete database the performance characteristics of the
studied unmanned helicopters at sea level (SL) are also used.
Table 1. Studied parameters for the database.

Power Unit Weight Performance Dimension
Engine type Empty weight Max level speed @ SL Overall length
Max available
Max combat load Max cruise speed @ SL Overall height
power
C.G. location Max ROC @ SL Overall width
Max takeoff weight Service ceiling Blade chord
Takeoff weight Hovering ceiling Main rotor diameter
Payload weight Combat radius Tail rotor diameter
Fuel capacity Range Tail rotor to C.G.
Max disk loading Endurance Fuselage length
A five-step procedure using the mentioned two methods would be

employed to obtain initial estimations of the desired RPH characteristics
(Parvan in this study): weight sizing, configuration selection, rotor
sizing, aerodynamic sizing, and power sizing.
2.1. Weight Sizing
Using the obtained trends [1], one can estimate the empty weight (WEmpty),
payload weight (WPayload ), fuel weight (WFuel ), and overall weight (WOverall)
of an RPH for a desired mission. Mission requirements affect the RPH
performance (altitude, range, endurance, and climate), design,
monitoring requirements (real time or post flight, bandwidth for real
flight), and subsystems characteristics. The configuration of the RPH is

affected by the type of the payload (imaged based surveillance,
environmental sampling, payload delivery, communi-cations,
serviceability, and so on). Table 2 summarizes the average performance
characteristics (takeoff weight (WTO), operation altitude (HOperation),
operation speed (VOperation), rate of climb (ROC), and service ceiling
altitude (HService) of studied RPHs for four missions and the desired values
for Parvan.
Table 2. RPHs performance characteristics for different missions.

WTO WPayload HOperation VOperation ROC Range HService
Mission
(kg) (kg) (m) (m/s) (m/min)) (km) (m)
(m)
Traffic
4 0.8 50 14 120 30 2100
monitoring
Remote
5 1.2 80 18 140 43 2100
sensing
Forest fire
7 1.4 300 5 150 15 1500
surveillanc
e
Rescue
14.3 2 300 3.9 120 100 1650
and relief
Parvan
8.5 2 50 10 150 80 1500
(traffic
monitoring
)
2.2. Configuration Selection
In order to get the insight on the configuration design a rough sketch of

the RPH configuration should be drawn for the specific selected
mission. The designer has the option to follow a regular design such as
conventional main and tail rotor, coaxial rotor [15, 16], NOTAR (no
tail rotor) [17], and tandem rotor [18], or comes up with a new
innovative design. Design and manufacturing complexities, cost,
mission requirements, and aesthetic aspects of the design should be

considered for the final configuration [1]. Fig. (1) illustrates the final
configuration of Parvan for the defined mission of traffic monitoring. A
conventional main and tail rotor design, a camera system as its
payload, a fuel tank, and other parts as will be described in the following
are modeled in CATIA to obtain this configuration at this phase.
Fig. (1). The final Parvan configuration after conceptual design.
2.3. Rotor Sizing
The next step is to characterize the main and tail rotor geometric
dimensions. Interpolating in the trends of different RPH design geometric
characteristics in the created database can be performed through defining
a set of dimensionless parameters [1]. One of the well-used
non-dimensional parameters is the solidity factor σ, the ratio of blades
area to rotor area which has a significant role in rotor sizing. Using σ
helps the designer to interpolate between three main design parameters:
number of blades b, blade chord c, and rotor radius R. The interpolation
of c and R with two blades for the Parvan gave the initial σ value of
0.081, the blade chord of 0.093 m and the blade radius of 0.73 m.
Besides defining dimensionless parameters that reduce the number of
design variables and simplify the interpolation of geometric
characteristics, there are certain relationships between some design
parameters. Fig. (2) illustrates the relationship between main and tail
rotor diameters and Fig. (3) illustrates the relationship between RPH
body height and main rotor diameter using the database [1]. These
relationships were used to design Parvan. Finally, initial guess on other

geometric characteristics such as the overall length and tail rotor length
can be made based on the empirical trends obtained from the database.
Table 3 summarizes the Parvan components dimensions.
1.4
1.2
Tail rotor diameter (m)

1
0.8
0.6
0.4
0.2
0
0 2 4 6 8 10
Main rotor diameter (m)
Fig. (2). The trend line of tail rotor and main rotor diameter relationship: the data collected for several
RPHs ( ), and fitted curve (-).
3.5
2.5
RPH Body Height (m)
1.5
0.5
0
0 1 2 3 4 5 6 7 8 9
Main rotor diameter (m)
Fig. (3). The trend line of RPH body height with respect to the main rotor diameter: the data collected
for several RPHs ( ), and fitted curve (-).
Table 3. Parvan components dimensions.

Main rotor diameter 1.54 m
Tail rotor diameter 0.31 m
Main rotor chord 0.063 m
Maximum body height 0.50 m
Overall length 1.85 m
Body width 0.25 m
2.4. Aerodynamic Sizing
The blade element theory together with the created database are the
basis for aerodynamic sizing, thrust sizing, and rotational speed (in
RPM) selection. Due to the importance of this step, three different
methods are performed and compared to each other to obtain the final
values:
1. Using the database to estimate the required power, select the RPM,
and obtain the lift/drag coefficients based on the required thrust;
2. Using the database to estimate the lift/drag coefficients and RPM
through the power and thrust relations;
3. Using conventional airfoil databases to select an airfoil, determine
the thrust, required power, and RPM.
2.5. Power Sizing
All the generated and lost powers should be calculated in order to estimate
the required power and choosing the right engine type. Through the Parvan
project, it was found out that the most important parts of the power
contributions are the induced power (Pi) and the profile power (Po) which
are defined by [13]:
Pi  C pi  A  R 
3
(1)
CT3 2
C pi   (2)
2
 CD  A  R 
3
Po  (3)
8
where Cpi is the induced power coefficient, ρ is the air density, A is the rotor
area, R is the rotor radius, ω is the rotor blade angular velocity (in RPM), κ
is an empirical factor equal to 1.15, CT is the thrust coefficient, and CD is the
drag coefficient. The total power at different altitudes, rate of climbs,
speeds, and different flight modes including hover, climb, and forward
flights should be calculated [1]. Different software packages or self-coding
can be employed for performance analysis and calculating performance
parameters such as maximum sustainable speed, maximum ROC, hover
ceiling, service ceiling, and power matching. These parameters were
calculated through a novel coding developed by the Parvan group. The
changes in the main rotor power, total power, and the maximum ROC at
different forward speeds have been shown in an earlier study [1]. It was
seen that at a specific altitude, the main rotor power decreases with
increasing the forward speed. Also, at a specific altitude, the total power
(the summation of induced power, profile power, parasite power, and main
rotor power) has its minimum for maximum forward speed in the range of
32-72 km/hr and it increases as the maximum forward speed increase.
Finally, the maximum ROC decreases with altitude.
DETAILED DESIGN
Upon completion the estimations made on initial design parameters

through conceptual design phase, further refinements on design
parameters should be performed to obtain the final RPH design. Hence,
final component design characteristics for Parvan were determined,
detailed component sketches got prepared, and the material selection
and manufacturing steps were performed accordingly. Before final

manufacturing and assembling phase, different subsystem components
were assembled and tested separately. A brief review on the studied
subsystems such as engine and fuel system, blade selection, power
transmitting system, control command system, and modeling of the
developed RPH is provided here.
3.1. Engine and Fuel System

The power plant unit, the heart of small-scale aircraft, requires certain
considerations in different design phases. Selecting the best engine in
the design process asks for the compatibility with the required
performance for hovering and forward flights that needs to consider the
advantages and disadvantages of each type of the available options. Four
common types of engines used in UAVs and especially RPHs are: 1)
electrical engines, 2) small-scale turbofan engines, 3) gas engines, and 4)
turbine engines [19]. The engine of such small aircrafts is selected to be
sufficiently efficient in terms of the weight and the available power.
However, although a piston engine is less efficient in the start-up
process and more costly than electric engines, two important
characteristics made it an appropriate choice for Parvan for the
predefined mission: higher power to weight ratio and higher endurance
[20]. Note that jet engines are used in larger classes and they are much
heavier than the conventional piston ones [20]. There is not a solid
decision on the best engine choice; the defined mission and the RPH
class would determine the most appropriate choice.
The very first choice for an RPH would be an electric engine. These
engines are the most convenient ones to setup and obtain a moderate
power to weight ratio compared to internal combustion engines [19].
The process of tuning carburetors, fuel systems and exhaust does not
need to be tuned. They are more efficient than piston engines and they
are less likely to crash during the flight as a result of the fuel/air mixture
or ignition problems. However, these engines are not economically as
efficient as nitro-gas powered engines due to the price of large-capacity
lithium batteries. The batteries are not durable in the severe maneuvers
due to the power required for these regimes. An advantage of these
types of engines would be the quick start up compared to other engine
types such as the gas powered engines and turbine powered engines.
One of the most profitable factors of using electric engines would be the
constant center of gravity during the flight compared to the other types
which use the chemical fuels. Moreover, the noise created from these
engines would be quite negligible and their operation at high frequencies
results in low vibration levels. On the other side, to name a few of
their main disadvantages: 1) the flight time can be short, 2) the cooling
time for the batteries between flights would be essential, and 3) large
lithium batteries can be dangerous and easily ruined if not cared
properly.
Model engines utilize glow-ignition combustion, rather than spark

ignition or fuel injection. Fuel and air are premixed and an incandescent
platinum wire is used to catalyze the combustion. Combustion analysis
was carried out to investigate the combustion process [21].
Nevertheless, the nitro powered RC-engines are more advantageous in
terms of the time required for charging the batteries or cooling down the
system of speed controllers is rather high for electric engines.
RC-engines can easily operate by just refueling the aircraft after each
flight mission [19]. These engines are cheaper than electric powered
engines. However, these engines are very noisy with many vibrations.
As it was mentioned before, the vibration is a common problem in
many piston engines; hence, there are many researches dealing with the
noise reduction in such engines since helicopters are almost huge
vibrating systems due to their main rotors [22]. Also, as it is proposed by
Whitelaw [23] the internal combustion engines like what is used in
Parvan has less efficient combustion process compared to other types of

engines such as electric ones. These types of engines are not
environmentally friendly due to the emission of nitro-based fuel.
Moreover, the change of center of gravity of the power system due the
fuel consumption requires a real-time or predictive estimation about the
fuel consumption in these engines. These engines are difficult and tricky
in appropriate tuning and setup since the air to fuel ratio is not
programmed and is controlled through the throttle position directly.
Gas engines are very powerful and run at lower RPMs. Their operational
procedure is easier than the nitro engines, whilst they provide higher
flight endurance, and are less expensive [24]. The main disadvantages of
such engines are: 1) lower power to weight ratio, 2) higher price
than electric engines, and 3) high voltage igniters which requires a
powerful battery [19].
Turbine engines are one of the most reliable ones to be used in the
unmanned helicopters. There is no requirement for tuning the system
manually since they are controlled using their built-in computers. These
engines are very noisy and are the heaviest type of helicopter engines.
Moreover, they are more expensive than the electric ones and possess a
higher operating cost and maintenance than any other type. However, the
startup and stop processes of turbine engines are much easier than the
other types. Although these engines are powerful, the power to weight
ratio is not as efficient as other types such as electric ones.
Considering different types of engines, a piston engine was selected for

Parvan [1]. The selected OS-91 SZH engine with a nominal power of 3.3
hp [25] which is 2460.8 Watt was compatible with the flight mission
and performance requirements. However, as it is illustrated in Fig. (4)
the shaft power required for hovering condition is almost less than half
of the engine nominal power and the power measured in the detailed
design process is sufficient for such a flight.
1.2
1200
1
1000
0.8
Power (Watt)
Torque (N.m)
800
0.6
600
400 0.4
200 0.2
0 0
7000 8000 9000 10000 7000 7500 8000 8500 9000 9500 10000
 (rpm)  (rpm)
(a) (b)
Fig. (4). The Parvan performance charts derived for the OS-91 SZH Engine; (a) estimated power using
the experimental data obtained from P = 4.47Ω3 − 1.3Ω2 + 1.28Ω − 4.15, and (b) estimated torque
using the same data form and the torque equation of Q = 6.81Ω3 − 1.96Ω2 + 1.89Ω − 5.98.
Since the defined mission for Parvan, requires a high endurance in hovering
regime, as opposed to other modes of flight, the design team decided to
choose the main rotor rpm according to the thrust required for hovering
flight. Nevertheless, this number might be changed for different flight
missions. Hence, the gear ratio should be selected to maintain the hovering
flight (based on the defined mission), while other flight regimes are
considered as well. The best way to choose the optimal gear ratio is based
on measuring different torque and angular velocity of the engine while it is
running the power transmission system, main rotor, and tail rotor. A thrust
test stand can be designed and manufactured to measure the main rotor
thrust and power to find the gear ratio. Due to the complexities of using the
OS-91 SZH engine on the test stand, an electric engine was used in place of
the piston engine with a higher gear ratio than the one used on Parvan
(Parvan’s power transmitting system is explained in section 3.3). The gear
train was designed to rotate the main rotor with an angular velocity of 1500
rpm. Fig. (5) illustrates the required power and thrust of main rotor which
are obtained experimentally.
3000
rpm = 1000 rpm = 1000
rpm = 1100 rpm = 1100
2500 rpm = 1200 200 rpm = 1200
rpm = 1300 rpm = 1300
rpm = 1400 rpm = 1400
rpm = 1500 rpm = 1500
2000
Power (watts)
rpm = 1600 rpm = 1600
150
Thrust (N)
1500
100
1000
50
500
0 0
6 7 8 9 10 11 12 13 6 7 8 9 10 11 12 13
 (degrees)  (degrees)
(a) (b)
Fig. (5). Main rotor power and thrust for evaluating the optimum gear ratio (a) the main rotor power,
and (b) the main rotor thrust measured using the experimental test stand for different angular velocity
of the main rotor.
The required thrust for stabilizing the RPH in hovering flight can be
calculated using the existing models [1], hence, a total collective pitch is
picked for neutralizing the weight; and the power associated with the
pitching angle gives us the required rpm for the main rotor. This angular
velocity is found to be 1500 rpm which gives an optimum power for the
Parvan engine based on the power equation shown in Fig. (4). Hence, the
total gear ratio is determined through the aforementioned procedure.
3.2. Blade Selection

In order to meet the performance goals such as the maximum ROC and
the required flight ceiling at different flight modes, the blade chord and
length were determined based on the available power and the required
main rotor power. As the next step, parameters including the required
thrust, rotor RPM, and solidity factor should also be considered. Hence,
a database of blades’ thrust at different chord and length values, but
similar pitch angle for a required power was obtained. A blade with
optimum characteristics is determined by performing this analysis [1]. It

should be noted that the aim of this step was blade selection and in case of
fabricating a new composite blade, a bi-level design optimization approach
can be used to determine the stacking sequence while the overall weight is
the objective function.
3.3. Power Transmitting System
The schematic of the power transmission system in a helicopter design

procedure is illustrated in Fig. (6). Upon determining the mission and
weight class of an RPH, choosing the main and tail rotor configuration, and
obtaining the mission torque and power requirements, one may proceed
with designing the gearbox. A gearbox system transfers the engine power
from the main shaft to the main and tail rotors. Using different gear ratios,
different rotational speeds and torques can be obtained. The engine RPM
is higher than the main and tail rotor rotational speeds, and the tail RPM is
higher than the main rotor RPM. Considering all these limitations, the
gearbox system is designed to obtain optimum rotational speeds and
torques. An appropriate transmission system design would help to reduce
the structural loads and prevent failures because of the maximum RPM at
the main rotor. Given the properties and dimensions of the Parvan engine
and blades, keeping the main rotor rotational speed in a range of 6 to 9
times slower than the engine RPM would result in keeping tip Mach
number of blades less than one which is desirable to avoid loud rotor and
heavy structures. On the other hand, higher rotor speeds in the range of 3
to 6 times faster than the main rotor rotational speeds are desirable for the
tail rotor with smaller rotor diameter.
Small engines can drive a broad range of desired rotors by adjusting the
optimal RPM of rotors through a proper transmission design. Indeed, a
high-speed low-torque engine output can be transferred to low-speed
high-torque values via gearbox in order to drive a large and heavy main
rotor. The best engine performance, main, and tail rotors are achieved at
certain rotational speeds. While obtaining this optimal rotational speed is
achieved through a proper gearbox design, a careless design of torque and
RPM setting might result in serious engine damages.
Rotor Thrust/Power Torque

Mission Weight Trend configuration calculation calculation
& & & & &
Weight class Flight envelop Blade profile Engine Tail rotor
& selection configuration
Rotor RPM
Gear ratio
& Vibration
Gearbox analysis
design
Frequency
separation
adequacy
End
Fig. (6). The schematic of the transmission system design in overall helicopter design process.
The optimal performance of the transmitting system is achieved mainly

by choosing the best gear ratio for each component. For main rotor, gear
ratio is the ratio of the main rotor gear speed over engine gear speed
which is usually in the range of 6 to 9. This ratio was obtained by
calculating the ratio of number of teeth on main gear over engine gear,
or equally by ratio of the main gear torque to the engine gear torque. It
means that reducing the speed on the main rotor can be achieved by
choosing different number of teeth on main and engine gears. For
example, for a gear ratio of 10, one can reduce the 15000 RPM at 3.3 hp
from a typical small engine, to 1500 RPM at the main rotor.
Engine gear speed

Gear ratio 
Main gear speed
(4)
Number of main gear teeth Main gear torque
 
Number of engine gear teeth Engine gear torque
The next step in this transmission system design is choosing tail to main
rotor gear ratio which is normally in the range of 3 to 6. This ratio is
defined as the ratio of the tail rotor RPM over main rotor RPM. The
other aspect of transmission system design is choosing a mechanism for
delivering the power at the tail rotor. In conventional small helicopters
this is done via a shaft or strap passing through tail boom. This can also
be done through belts and pulleys which is the case in Parvan. Other
important parameters in transmission system design are gear mesh and
gear module (or gear mod). Gear mesh is the representative of the free
tolerance between two gears. In the case of tight gears, friction
increases and causes overheat and vibration with increased power loss in
the system. On the other hand, using loose gears results in gear stripping.
Hence, an optimal gear mesh should be selected in order to achieve a
good performance in transmission system. Gear module (mod) represents
the width of teeth in millimeters for each gear; the thicker the gear teeth,
the higher the mod number.
Pitch circle diameter

Gear mod  (5)
Number of teeth
Here, pitch circle as shown in Fig. (7) is described as the circle of one gear
that is tangent to the other gear’s pitch circle. In order to choose this
parameter correctly, the effect of gear ratio on gear mesh in a specific range
of gear diameters as well as the amount of power to be transmitted should
be taken into account. Having higher gear mod results in handling more
power and fewer teeth engagement for the same gear ratio. On the other
hand, low gear mod will result in lower power handling capability, but good
teeth engagements.
Fig. (7). Pitch circle [26].
After gear ratio selection for gearbox that is done by analyzing the main and
tail rotors and engine thrust and power values, in order to prove the
adequacy of frequency separation between natural frequency of the power
transmitting system and the rotor RPM for a safe flight, vibration behavior
of the transmission system were studied. Holzer method [27] is used for
calculating the natural frequencies of the power system. According to Fig.
(6), if frequency separation to avoid coupling with the operating modes of
the vehicle is not enough, then gear box should be redesigned in order to
meet the requirements.
As the final note on transmission system design, a centrifugal clutch is

designed in order to prevent the rotation of the blades in starting and warm
up phases. Geometry design and material selection for gears and clutch are
done considering the engine idle RPM and maximum transmitted torque.
STABILITY AND CONTROL OF A REMOTELY PILOTED

HELICOPTER
Stability and control of an aircraft is an inevitable part of a flight; the

safe and appropriate flight modes would not be achieved without an
accurate determination of stability and control of the system. In the
design stages of an aircraft, investigating the performance limits is

required for stability analysis. For manned aircrafts, flight simulation
before the test is a usual process that saves time and costs in the design
and manufacturing phases, and helps avoiding the contingent human
errors during the flight which saves the life of the pilot and prevents
unavoidable accidents.
Advanced control systems with small sizes that are able to

communicate through wireless systems have an emerging progress in
the recent years. UAVs such as remotely piloted helicopters are one
those advanced systems that have an inherent dynamic complexity, and
critical stability that needs to be considered to maintain a controlled
state in the vehicle. Hence, it is required to have a stable vehicle with
a designed control system to avoid any possible deviation from the
prescribed mission of an aircraft during the flight path. A model of the
vehicle describing the aerodynamics and dynamics of the system is
required to design the controllers. According to the seventy years of
experience in the field of helicopters, different studies are made to
optimize these vehicles and find newer applications such as unmanned
helicopters [1].
UAVs have been used extensively during the past decades. New
innovations, improvements, and modifications are made to these types
of aircrafts every year. The flight simulation of an aircraft contributes
the designers to build the mathematical model of the aircraft in order to
test the performance of controlling and navigating devices such as a
controller or an autopilot. The autopilot guides a real aircraft, hence the
mathematical model used for the autopilot must be as accurate as
possible to maintain the sophisticated dynamics of the vehicle
sufficiently. This will arise the importance of the accurately modeling of
the aerodynamics and the dynamics of the aircraft.
4.1. Dynamic Modeling
Helicopter is a complex system with inherent instabilities. Nonlinear

aerodynamic forces and moments, as well as the gravity exerted on the
flexible parts of the system such as blades, require adding the flapping
characteristics of the rotor to the dynamic equations of the model which
is described in the following sections.
4.1.1. Methodology
In this section, an appropriate level of stability of an aircraft, which

contributes a pilot to accomplish the mission, is defined briefly. The
essence of this part of the design process will be beneficial if the
control of an unstable UAV can be improved in the earlier stages of
design before the prototypes being made. Hence, the stability of an
aerial vehicle is one of the most significant and challenging factors that
needs serious attention during its design or optimization. The
determination of the stability characteristics of helicopters is of
importance in two different branches: during the design process, before
manufacturing, and for modifying the stability characteristics of a
manufactured helicopter by designing controllers to match its behavior
with the new defined mission. Indeed, design parameters directly affect
the stability characteristics. Hence, in the first phases of design, the
stability can be improved by some adjustments in those parameters.
These improvements also reduce the complexity of designing controllers
in order to make the aircraft operational. Basically, there are two
techniques to have a desirable stability characteristic: active or passive
techniques. In the active mode, a controller works as a stabilizer and
adds an artificial stability to the aircraft. In the passive one, the aircraft
has been designed to be naturally stable; hence, there would be no need
to be artificially stabilized. For both modes, an appropriate analysis of

stability characteristics of the system is required, so improvement of
these characteristics as much as possible is necessitated. In fact, this job
is usually done by implementing trial and error methods that would not
only be a time consuming process but also the designer cannot find out
which parameter is more effective in a special stability mode and has to
be changed.
A linear mathematical model of an RPH with stabilizer bar is introduced

based on the stability derivatives of these special types of aircrafts.
Although deriving these derivatives for helicopters by theoretical
techniques is not easy, it is an advantageous method in comparison to the
other ones; the influence of all design parameters on the stability
characteristics of the entire system can be directly seen. First, some
simple analytical models of the rotor, fuselage and other components of
the helicopter are built. Then, by combining these models and
linearizing them, the derivatives can be evaluated by analytical methods
in dynamics and aerodynamics of conventional helicopters. Moreover,
the dynamics of a stabilizer bar including its flapping characteristics is
generally described in order to include its behavior in the entire dynamics
of the aircraft. In the next step the trim condition of the aircraft is
calculated by iteration based on a simplified set of equations in hover
and forward flight. Then, after building the state-space equation, it is
solved based on the states estimated in trim condition. After simulation,
the time history of each state under distinct situations such as existence
of a stabilizer-bar or without it and with an external excitement or
individually investigated. As it is mentioned, the mathematical model
describing the stability of a helicopter is dependent on the performance
characteristics.
The helicopter motion, aerodynamics and flight mechanics are highly

influenced by the main rotor. The model of the helicopter asks for
certain requirements for the helicopter performance and then, those

requirements may help to define the behavior of the vehicle. Such a
model can be used for simulations. For instance, the forward flight of the
helicopter is the same as hovering flight, but in larger forward speeds the
parasite drag force has a critical role in the motion of the helicopter.
Hence, a wind tunnel testing will answer the questions about the
accurate identification of the drag force. However, if the wind tunnel is
not available, the analytical model of the aircraft with smaller forward
speed is sufficiently valid for the purpose of deriving the equations of
motion, forces, and accelerations.
(a) (b)
Fig. (8). Rotor-body interaction in (a) climb or hovering flight, and (b) forward flight.
In the present work, the dynamic model of the helicopter is separated

based on the most significant flight regimes in conjunction with the
importance of the regimes according to the mission. These flight
regimes are the forward flight, climb, and hover as it is mentioned in
Fig. (8). In each flight regime, the performance and the trim conditions
of the aircraft should be studied based on the performance in the
forward, climb, and hovering flights. The control of the helicopter is
different in each flight regime.
An active control device which most of the small-scale RPH take

advantage of is control rotor (flybar) which is a small rotor located on
the top of the main frame and below the main rotor. In Fig. (9), the
control rotor is illustrated which is a small rotor with two blades. These
blades are acting as control surfaces providing an analog type of control
that is associated with the flapping motion of the small rotor from which
the cyclic pitch on the main rotor is produced.
Fig. (9). The control rotor stabilizing the main rotor inputs in Parvan.
As shown in Fig. (10), the Bell-Hiller mechanism is used to dampen the

control inputs to the main rotor. This mechanism is a gain to the main
rotor cyclic pitching system through which the dynamic feedback on roll
and pitch rates is damped. In the proposed study, the control rotor has
only a stabilizing character and is not applied to increase handling
qualities.
Clockwise
Rotation
Main Rotor
Blade
Stabilizer Bar
Pitch Input
Roll Input
Fig. (10). Schematic of the Bell-Hiller stabilizer bar used in Parvan [28].
4.1.2. Rigid Body Equations of Motion
Newton’s second law is the key to study the physics of the helicopter
motion. Assuming that the reference frame is the stationary earth, one
can develop the kinetics and kinematics of the rigid body motion to be
expressed in the following format:
F  m(V  wbf  Vbf ) (6)
M  I bf wbf  wbf  wbf (7)
where Ib is the moment of inertia tensor, and wbf is the body frame matrix
of rotational rates which is discussed in detail in [29] and [30]. The

forces and moments denoted as F and M, respectively, are acting on the
system. Expanding forces and moments as well as the acceleration and
inertial terms on the right hand side of Eq. 8 facilitates the linearization
and making the aerodynamic and stability derivatives.
X M  X T  X H  XV  X F  mg sin   m( x  yr  zq )
YaM  YaT  YaV  YaF  mg sin   m( y  xr  zq)
Z M  ZT  Z H  ZV  Z F  mg cos   m( z  xq  zp )
RoM  YaT hM  Z M yM  YaT hT  yV hV  Ya  Fh  F  RoF  I xx p  qr ( I yy  I xx )
RoM  YaT hM  Z M lM  M T  X T hT  ZT lT  XHhH  Z H lH  XV hV  M F  Z F lF
 X F hF  I yy q  pq ( I xx  I yy )
N M  YaM lM  YaT lT  Ya  Vl  V  N F  YaF lF  I zz r  pq ( I xx  I yy )
(8)
where X, Y, and Z are corresponding to the forces in the x, y, and z axes,
respectively, whereas, the pitching, rolling and yawing moments are
denoted by M, R, and N, respectively. The variable h defines the height
from body axis to the point that the force is exerted and so forth for l.
Those variables described by the index F are in association with
fuselage, while H refers to the horizontal stabilizer mounted on the
helicopter tail, and V refers to the vertical stabilizer.
The six degrees of freedom (DOF) of a rigid body is a simple way to

start describing the behavior of a system with higher complexities.
These are the minimum complexity equations summarizing a dynamic
model of the helicopter including the translational velocities known as
u, v, and w; the other three rotational DOFs excluding the flapping
dynamics of the control rotor and the directional degrees of freedom
respect to the earth axis such as Euler angles are p, q, and r. Considering
this model, the equations of motion are evaluated based on the
contribution of each component of the system in the following format:
Studying the cyclic motion of the main rotor disk due to the aerodynamic
forces and moments associated with the gyroscopic behavior of the rotor
which is also affected by the type of control input whether to be
longitudinal or lateral. Hence, the lateral and longitudinal cyclic pitch
of the main rotor is dependent to the flapping dynamics of the control
rotor in the systems that are working with this analogue controller to
dampen the pilot’s control inputs for two cyclic angles in lateral and
longitudinal directions. These two angles create two more degrees of
freedom for the rigid body of the helicopter, which will be added to the
two more degrees of freedom for the body frame rotations respect to the
earth reference frame. The latter 2-DOFs are created based on the
pitching and rolling Euler angles. Hence, the dynamical system has
10-DOF which is investigated in this study.
4.2. Trim and Stability Analysis
4.2.1. Analysis of the Trim
The linear dynamic modeling of an aircraft is complete when the trim

and stability analysis have significant progress. The trim is a condition
when all of the accelerations are removed from the system and the
equations are solved based on some initial conditions. The solution to the
trim condition then is used as an input to the linearized dynamic model
where small perturbations assumption is valid. The linearization is a
process in which: (1) all the components on the left hand side of Eq. 8 is
summarized in one term and the small angle assumption is applied to the
right hand side, (2) then the contribution of each component on the left
hand side (LHS) is considered using the first derivative of the forces or
moments on LHS respect to the component itself, and this derivative
which is called stability derivative is multiplied by the degrees of the
freedom in the system and the four control inputs. The form of the LHS
in the equations of motion will be:
X X X X X X X
X x y z q p r c,CR
x y z q p r c,CR
X X X X X (9)
  s,CR  0M 0T  A1  B1
 s,CR 0 0 A1 B1
M T
where, βc,CR and βs,CR are the longitudinal and lateral flapping angles for
the stabilizer bar, respectively. According to the Eq. 9, there are
nonlinear terms on the right hand side which can be linearized using the
Taylor expansion over the trim point of the system. The requirement to
use the Taylor expansion in linearizing the equations is based upon the
assumption of analytic functions of aerodynamic forces and moments to
be valid over the trim point [11]. Having spikes without smoothness in
the functions with discontinuities breaks this assumption.
4.2.2. Stability
The main problem in analyzing the dynamics of flight and the system
response to the control inputs are the number of degrees of freedom in
the model. The best way to analyze the nonlinear systems linearly is to
divide the longitudinal and lateral modes of flight with the certain
assumption suggesting the small coupling between these modes as it has
been asserted for Parvan in [31]. According to this study, the stability
and aerodynamic derivatives for the case study (Parvan) delineating the
coupling between the longitudinal and lateral modes, as well as roll and
yaw axes [11] are sufficiently negligible compared to the pure
derivatives in each aforementioned mode. Hence, the process of
isolating these modes in the analysis may have only a weak drawback in
maneuvering phases in which the coupled dynamic forces play key role
in the analysis system; a case where the small perturbation assumption is
no longer valid and the system cannot be modeled linearly. The 10-DOF
flight model studied in [31] was tested by the linear method of analyzing
stability proposed by Prouty and Padfield in [11, 29]. In this section, the
stability is studied in hover and forward flight.
4.2.2.1. Stability Analysis in Hovering Flight
Hover is a major part of the flight envelopes in most of the defined

missions for helicopters. The characteristic equation of the 10-DOF
system of Parvan in hovering flight is shown in Eq. 10:
s10  12.04 s 9  35.41s8  299.6 s 7  2291s 6  2606 s 5  532.9 s 4 (10)

640.5s 3  547.7 s 2  85.2 s  3913  0
where the roots are almost stable, while there are two unstable roots that
belongs to the lateral motion of the aircraft as it is illustrated by Eq. 11:
s1  13.83, s2  0.93, s3  0.62, s4  6.7

(11)
s5,6  0.47  0.5i, s7,8  1.95  4.06i, s9,10  0.022  0.27i
The stability in the lateral or longitudinal modes are not determined

unless the modes are studied separately to presume which mode or root
is more coupled with the other mode. The longitudinal characteristics of
the 5-DOF which is explained in Table 4, the system has an almost
neutral state with a period of 23.7 seconds and a large divergence time to
reach the double amplitude at 217 seconds, while the stable oscillations
are short period that are damped very fast in 0.35 seconds. Hence, the
longitudinal mode in hover is showing a stable character which is
uncoupled from the lateral mode with a very good accuracy. The first
root in Table 4 is emanated from the vertical forces and moments in the
hovering flight. As it is shown in [31], the aerodynamic and stability
derivatives in vertical direction are almost negligible in hovering flight.
Hence, by removing the corresponding equation from the analysis, the

roots are just slightly changed which proves the assumption of
uncoupled dynamics in vertical direction in hover with other degrees of
freedom. Hence, the response to the 5-DOF and 4-DOF are almost
similar to an input. Also, it is possible to simplify the system to a
3-DOF (as it is been demonstrated in Fig. (11) or 2-DOF one using the
assumption that the forces acting in the forward direction X are
negligible since there is no forward speed or the speed is controlled to be
very small and stable. Therefore, the only oscillation that has a period of
23 seconds is due to the pitching moment and flybar flapping. The
oscillations are neutrally stable, therefore, the longitudinal mode of
Parvan in hover is stable.
As it was mentioned in this section, the helicopter is slightly unstable in

the lateral modes. It is essential to mention that the rolling inertia is
much less than the pitching inertia in the proposed design of Parvan, so
the magnitude of the stability derivatives are larger than the pitching
derivatives. Hence, the sensitivity of the system response is larger to
instability in this direction since the period of lateral motions has shorter
periods and lower damping ratios compared to longitudinal ones. The
characteristic equation of the 5-DOF system is shown in Eq. 12:
s5  7.189s 4  94.57s3  34.39s 2  7.335s  29.57  0 (12)
where the oscillation is stable but there are two unstable roots that is
illustrated by Eq. 13:
s1  13.84, s2  0.616, s3  6.997, s4,5  0.481  0.514i (13)
As the details of this mode are discussed in [31], the lateral motion is a
little unstable in Parvan unmanned helicopter.
Fig. (11). The block diagram of the 3-DOF simplified longitudinal mode in hover for Parvan [31].
Table 4. Characteristics of longitudinal modes from 5-DOF to simplified models.

Mode Roots Period (s) tdouble (s) thalf (s)
s1  0.951
5-DOF s2,3  0.003  0.265i 23.7 217 -
s4,5  1.954  4.058i 1.5 - 0.35
s1,2  0.004  0.267i 4.5 - 178

4-DOF
s3,4  1.953  4.056i 1.5 - 0.35
2-DOF - 23.5 - -
4.2.2.2. Stability Analysis in Forward Flight
Forward flight is another important part of the flight envelopes in most

of the defined missions for helicopters. The characteristic equation of
the 10 DOF system of Parvan in forward flight with 50Km/hr is shown
in Eq. 14:
s10  24.78s 9  128.6s 8  509.4s 7  7097 s 6  5706s 5  1354s 4 (14)

1912s 3  3477 s 2  4506s  1115  0
where the roots are almost stable, while there are two unstable roots that
belongs to the lateral motion of the aircraft as it is illustrated by Eq. 15:
s1  15.31, s2  10.68, s3  3.195, s4  7.52

(15)
s5,6  0.075  0.176i, s7.8  1.54  5.58i, s9,10  0.057  0.477i
By using the same approach that is indicated in hovering flight, one is

able to distinguish the coupled and uncoupled roots and the couplings
between the modes in forward flight, consequently. The longitudinal
roots of this mode are evaluated by expanding the longitudinal
sub-matrix in the state-space system of equations as it shown by below:
s5  6.225s 4  42.98s3  105.7s 2  4.496s  8.936  0 (16)

s1  3.104, s2,3  0.0008  0.34i, s4,5  1.56  4.68i (17)
There are two oscillations in the longitudinal mode of forward flight;

stable and neutrally unstable. It is possible to stabilize the oscillations by
improving the longitudinal control surfaces such as horizontal stabilizer.
The step response of the 5-DOF system in forward flight to 1 degree of
cyclic pitch B1 is depicted in Fig. (12). The response of the 5-DOF
system to the cyclic pitch consists of two modes of Phugoid and short
period. According to this, one degree in cyclic pitch will cause
approximately 10 degrees of response which is damped after
approximately 5 seconds.
Fig. (12). The step response of the 5-DOF simplified longitudinal mode in forward flight to
cyclic pitch of B1 [31].
The dynamics of the helicopter in lateral flight is based on the lateral
wind, yawing rate, rolling rate, lateral flapping of control rotor, and the
rolling angle. Basically, the lateral dynamics of helicopters is not quite
stable inherently and is controlled with the thrust of the main rotor and
tail rotor to compensate for the discrepancies between their available
torques in the case of the lateral instability such as a lateral wind. One of
the main contributions of the instabilities is emanated from the tail rotor
thrust. The increase in forward speed results in more instabilities since
the torque of main rotor is increased. The dynamics of pure lateral
motion in forward flight is described using Eq. 18.
s5  18.56s 4  30.4s3  1233s 2  91.49s  205.6  0 (18)

s1  15.3, s2  10.71, s3  7.53, s4,5  0.035  0.41i (19)
As it is shown in Eq. 18, Parvan is unstable in lateral flight mode due to

the large positive root. This is something that was expected before the
analysis but it is possible to improve the instabilities using controllers for
this design or by optimizing the design.
Fig. (13). The 3-D schematic of Parvan in CATIA.
CATIA MODELING
There is a transition phase from the conceptual to detailed design, we
called it the “preliminary design”, where the overall configuration of the
helicopter is specified according to the main subsystems. Computer-
aided design (CAD) software would be the best option to create this
preliminary design and also for the following modification, analysis, or
optimization studies. One of the most suitable CAD software packages is
CATIA that has lots of powerful modules (Part Design, Product Design,
DMU Kinematics, Generative Shape Design and Drafting and so on) for
designing mechanical parts and products together with defining the
physical properties of different parts. Fig. (13) illustrates the 3-D
schematic of Parvan in CATIA. The output results from the preliminary

design such as the overall allowed and estimated weight and parts
dimensions were used in the detailed design phase for the final
subsystem designs. Further modifications and studies of different
subsystems were also carried out in CATIA - Final designs of
subsystems were converted to an appropriate format for manufacturing.
MANUFACTURING AND ASSEMBLING
The final technical drawings from the detailed design phase should be
sent for manufacturing. Before considering the required accuracy,
precision, and optimum machining techniques, the best material should
be selected. It should meet the design and mission requirements, and it
meets an optimum value from a trade-off between cost, weight, life
cycle and reliability considerations. The product tree is found the best
tool to show relations of components and subsystems. Concerning a
complex system like a helicopter, for providing a detailed product tree,
Parvan was divided into eight distinct groups from the assembly
viewpoint including engine, fly-bar, hub, power train, tail, and avionics.
Therefore, each group was assembled separately to montage the whole
system. For instance, Fig. (14) illustrates the product tree of the engine
group with its subsystems. It is important to make an accurate
documentation at this step in order to assure the procedure of
assembling and disassembling is followed correctly each time.
Engine Group
Clutch Fan Fuel Engin Y-pipe Engine Governor Linkages &

ball links
Clutch center nut Fan lock nut Ball link nut Mount nut
Fig. (14). Product tree of the engine group.

An extensive research is required to obtain appropriate materials and

manufacturing techniques. In the developed RPH, aluminum alloys are
preferred to carbon fiber sheets for fuselage frame due to its less cost
and manufacturing complexities. Fiber glass was selected for tail boom
and horizontal and vertical stabilizers where certain stiffness and weight
are considered.
(a) (b)
Fig. (15). Manufactured parts, (a) hub, and (b) swashplate.
The main and tail rotor hub and swashplates are made from Al 7075 and
steel is used for the main shaft of Parvan. Al-7075 is a strong aluminum
alloy with high strength, light weight, and less resistance to corrosion
than many other aluminum alloys. Among different manufacturing
techniques, CNC was selected for small and fine aluminum based
components and wire cut technique was used for fuselage and fly-bar
frames. Some parts such as hub Fig. (15a) and swashplate Fig. (15b) are
manufactured for this specific design, and some others are bought from
existing off-the-shelf market such as the OS 91SZ-H ring engine.
Fig. (16) shows major subsystems of Parvan. Fig. (17) shows the final
assembly.
Parvan
Engine Flybar Hub Power train Tail Structure
Fig. (16). Parvan product tree.
Fig. (17). Built model of Parvan with all sub-systems.
TEST AND RELIABILITY ANALYSIS
There are many tests that should be performed on major subsystems

before flight tests in order to assure the properly functioning of them.
Performing the Engine tests the rated power/torque values of the studied
engine at different rotational speeds are obtained. Main/tail rotor tests

provide the aerodynamic characteristics and performance parameters of
the main/tail rotor. The strength of the structure is obtained by
conducting structural and fatigue tests. Wind tunnel tests would provide
estimations on the aerodynamics forces around the helicopter and the
flow interaction between different components such as the main rotor
blades and fuselage. In this study, because of the complexities and
limitations of the various tests, only engine torque/power tests and main
rotor tests are performed for Parvan subsystems which are described in
this section.
As explained in the detailed design section and in Fig. (6), the engine
torque/power rating is the input for designing the gearbox system. Fig.
(18) shows the designed engine torque/power test-stand for the Parvan
project. The engine torque/power values at different rotational speeds
and various blade sizes were calculated using Eq. 20 [13, 29]:
R
R4
P    r 33 dr   3 (20)
0 4
where e is hinge offset representing the specific point on the blade

where a positive lift force starts on the blade, R is the blade radius, α is a
constant value ( 1/2 ρ𝑐̅𝐶̅ l) at certain height (with constant density, ρ) and
a fixed blade (with constant chord 𝑐̅ and lift coefficient 𝐶̅ l), and is the
rotational speed. Fig. (18) illustrates the output power and torque curves
at different rotational speeds from one of the test results.
Having the final gearbox design as shown in Fig. (19), the RPH main
rotor thrust values at different rotational speeds and input powers are
obtained using the designed thrust test stand shown in Fig. (20).
Fig. (18). Engine torque/power test stand.
(a) (b)
Fig. (19). The final gearbox system for main rotor thrust test stand: (a) side view of the main frame,
(b) transmission system on the Parvan.
Fig. (20). Thrust test stand.
Reliability is the probability of desired system operation under a

specified condition for determined time duration [32]. The ground tests
of the subsystems and also the whole system, together with the flight
tests are employed to estimate the mission success and failure rates of
Parvan. Fig. (21) shows the flight of Parvan during one of the flight test.
Fig. (21). Parvan flight test.

It is helpful to review a number of fundamental concepts defined in

reliability studies. Mean Time between Failures (MTBF) states how
long a repairable component will execute without facing any failure
[33]. Definite relationships are defined between the reliability and
MTBF. For non-repairable components, Mean Time to Failure (MTTF)
is used instead of MTBF.
Fig. (22) illustrates a typical failure rate historic diagram of a

component. After initial tests and debugging the system, the failure rate
for a given component tends to be constant until the component
approaches the end of its life time [28], the wear out period. The MTBF
or failure rate is estimated with the reliability function R(t) as follows [34]:
t

R(t )  et  e MTBF (21)
Failure Rate
Normal useful life
Debugging Wear out

1
period  period
MTBF
Lifetime of component
Fig. (22). Typical failure rate history diagram [27].
Failure mode effect analysis (FMEA) describes any studies performed to

obtain the potential failure state in a product or process [35]. Failure modes,
effects and criticality analysis (FMECA) study the probability and criticality
of the failures and identifies potential product weaknesses assuming a

pessimistic viewpoint since the design procedure usually approaches from
an optimistic viewpoint. Table 5 shows a typical developed FMECA table
to find the most critical failure modes and hence, the most critical group of
the components of Parvan. In Table 5, P is the probability of the failure, PN
is the probability number of the failure, n is the severity of the failure and d
is the possibility of its detection. Moreover, C is the criticality of each
component which is a multiplication of PN, n, and d. In the present study,
these numbers are assumed to vary from 1 to 3 with respect to the
component specifications, operation and role. One can see that the possible
minimum and maximum values for C are 1 and 27, respectively. The norm
of the criticality is considered to be 12 which sets the maximum acceptable
criticality for any component. The components with critical numbers larger
than the norm have been revised and redesigned to meet this norm and
ensure safe operation. In order to have a good estimation for the reliability
of Parvan, detailed calculations are performed for most critical subsystems.
The column "preventive and recovery precautions" of the FMECA table is
used to determine some maintenance issues. These precautions include three
actions: a) Scheduled preflight checks(including check of servos, fan casing,
pinion, gear, main shaft, control links, structure screws, tail swash plate, etc.
before each flight and the main shaft lock, bearings, mechanism screws,
pulley and tail belt, etc. every 10 hours of flight), b) Replacement of
distorted parts (including clutch strip and main shaft bearing), c) Some
alarm gauges for undesirable conditions(including low battery and governor
alarm, engine temperature sensor).
Table 6 summarizes the criticality numbers for all the subsystems in the
Parvan project. The control subsystem has the most average critically
number; hence, more detailed reliability calculations are performed in [33].
Finally, an improved reliability of the control system is achieved through
analyzing and modifying the control components and the fabrication
methods.
Table 5. A sample of FMECA table developed for one component of Parvan, Hub.
Failure Preventive
Part Failure Failure Detection and
effects P PN n d C
Name mode cause technique recovery
precautions
M-5 screw
(on the Extra Blade
Preflight
horizontal Undesirable flapping and unbalancing
check (every 1/3000 1 2 2 4
shaft) might vibration Unstable and rotor
get 10hr)
flight vibration
loosened
Hub Bearing of
Delay in
hub Loss of
Hub control Preflight
teetering lubricant;
mechanisms commands; check (every 1/1000 2 2 2 8
damage Undesirable
loose Rotor 10hr)
vibration
vibration
Due to any No power

-
fault in its transmitted
Regulator Damaged 1/1000 2 3 3 18
electronic to the
system avionics
Regulator
Redundant 2 12
(After 2 3
part
revision)
Table 6. Final Parvan subsystems criticality numbers after including design and maintenance
considerations.
Subsystem Criticality number
Engine 5.78
Avionics 7.10
Gear Box 4.00
Control 7.38
Power Train 7.20
Tail 5.33
CONCLUSION
In this chapter, a practical stepwise approach in developing an RPH

proposed and studied analytically and experimentally. The procedure
employed on designing and manufacturing an RPH, called Parvan for a
specific mission of traffic monitoring. The main conclusions of the
proposed procedure can be summarized as: 1) a feasibility study of
more than 100 RPHs provided a strong database with a vast range of
RPH missions and performance and geometric characteristics, 2) the
conceptual design phase provided with performance sizing, weight
sizing, aerodynamics design, structural design, and power system for a
defined mission through either theoretical equations or empirical results
for similar designs, 3) the detailed design phase provided with final
component designs for the specified mission. A 3-D Parvan design was
developed in CATIA for full model formation, 4) the performance of
sub-systems was tested through a series of ground tests. The final
system was assembled and further improvements performed using the
flight tests data, 5) reliability analysis together with tests performed to
enhance the safety and maintainability of Parvan. Hence, component
failures at different flight modes for the specified mission were studied
using the FMEA/FMECA method, and a series of preventive and
recovery precautions were performed on the design, configuration,
assembling, and material selections, and 6) finally, the test flights
demonstrated the success of the proposed procedure in developing a
reliable operational RPH with a sensible cost.
The authors confirm that they have no conflict of interest to declare for this
publication.
ACKNOWLEDGEMENTS
The authors wish to thank the other members of the Parvan group
throughout the course of this project; in particular Dr. Mortazavi for
supervising the project, and Hoofar Pourzand for technical assistance.
REFERENCES
[1] F. Kermanshai, M. Mortazavi, M. Mohagheghi, M.S. Sajedi, R.M. Ziazi, S.

Sadati, H. Pourzand, and N. Goudarzi, "Design, optimization, and building flight
model of an operational unmanned helicopter", IEEE Aerospace Conference,
2010, pp. 1-10
[2] E.A. Carroll, and D.B. Rathbone, Using an Unmanned Airborne Data
Acquisition System (ADAS) for traffic surveillance, monitoring, and
management. ASME conference, pp. 145-157, 2012.
[3] S. Kim, H. Oh, J. Suk, and A. Tsourdos, "Coordinated trajectory planning for
efficient communication relay using multiple UAVs", Control Engineering
Practice Journal, vol. 29, pp. 42-49, 2014.
[http://dx.doi.org/10.1016/j.conengprac.2014.04.003]
[4] F. Nex, and F. Remondino, “UAV for 3D mapping applications: a review",
Applied Geomatics Journal, vol. 6, no. 1, pp. 1-15, 2013.
[http://dx.doi.org/10.1007/s12518-013-0120-x]
[5] R. Feltman, "The future of sports photography: Drones", Available from:
http://www.theatlantic.com/technology/archivfuture-of-sports-photo-
graphy-drones/283896/, 2014.
[6] M. Sadeghzadeh, Abdolhoseini, and Y. Zhang, “Payload drop application using
an unmanned quad-rotor helicopter based on gain-scheduled PID and model
predictive control, Unmanned Systems Journal, vol. 2, no. 1, 2014.
[7] N. Goudarzi, and W.D. Zhu, "A review on the development of the wind turbine
generators across the world", International Journal of Dynamics and
Control. Available from: http://link.springer.com/article/10.1007/s40435-
013-0016-y, vol. 1, 2013, no. 2, pp. 192-202.

[http://dx.doi.org/10.1007/s40435-013-0016-y]
[8] N. Goudarzi, W.D. Zhu, and H. Bahari, "An assessment of the potential of Novel
ducted turbine for harvesting wind power", Journal of Intelligent Material
Systems and Structures. Available from: http://jim.sagepub.com/
content/early/2014/05/18/1045389X14533438, 2014.
[9] K. Ro, J.S. Oh, and L. Dong, "Lessons learned: application of small UAV for
urban highway traffic monitoring", 45th AIAA Aerospace Sciences Meeting and
Exhibit, 2007 Reno, Nevada
[http://dx.doi.org/10.2514/6.2007-596]
[10] J. Irizarry, “Feasibility study to determine the economic and operational benefits
of utilizing unmanned aerial vehicles (UAVs)" PhD dissertation, Georgia Institute
of Technology, 2014
[11] R.W. Prouty, Helicopter performance, stability, and control. Krieger Publishing
Company: Florida, 2002.
[12] J. Seddon, and S. Newman, Basic helicopter aerodynamics. Blackwell Science,
2002.
[13] J.G. Leishman, Principles of helicopter aerodynamics. Cambridge, 2006.
[14] W. Johnson, Helicopter theory. Dover Publications, 1994.
[15] C.P. Coleman, "A survey of theoretical and experimental coaxial rotor
aerodynamic research", NASA Technical Paper 3675.
[16] F. Wang, J. Cui, B.M. Chen, and T.H. Lee, "Flight dynamics modeling of coaxial
rotorcraft UAVs", In: Handbook of Unmanned Aerial Vehicles. Springer, 2014,
pp. 1217-1256.
[17] C. Hwang, "Conceptual design of ducted fan for helicopter anti-torque system",
KSAS International Journal, vol. 6, no. 1, pp. 89-97, 2005.
[18] N. Aldawoodi, An approach to designing an unmanned helicopter autopilot using
genetic algorithms and simulated annealing, PhD Dissertation, University of
South Florida, 2008.
[19] "Choosing the best RC helicopter engine", Available from: http://www.rcheli-

copterfun.com/rc-helicopter-engine.html. (Last Accessed on Dec. 7, 2014).
[20] V.V. Kishore, “Renewable Energy Engineering and Technology: Principles
and practice”, The Energy and Resources Institute. TERI, 2009.
[21] L. Tian, D.B. Kittelson, and W.K. Durfee, "Experimental Tests and Simulations
of A 1.5 cc Miniature Glow-Ignition Two-Stroke Engine", SAE Paper, paper
number 10SETC-0095, 2010
[http://dx.doi.org/10.4271/2010-32-0018]
[22] C. Maucher, B. Grohmann, P. Janker, A. Altmikus, F. Jensen, and H. Baier,
"Actuator Design For The Active Trailing Edge of A Helicopter Rotor Blade",
33rd European Rotorcraft Forum, Kazan, Russia
[23] J.H. Whitelaw, Internal Combustion Engines. Academic Press, 1988.
[24] "IRCHA PIT Crew Training Manual", Available from: http://www.ircha.org/
sites/default/files/Digital%20Library/IRCHA_PIT_Crew_Manual-Gaui_Final_2.
2.pdf, File:pdf, (Last Accessed on Dec. 7, 2014).
[25] OS-91 ZH Instruction Manual, OS Engines M. Co., Ltd. Japan, 2005.
[26] N. Gelpi, "Involute gear plug-in demo", Architectural Design Workshop.
Available from: http://makearchitecture.wordpress.com/people-2/varvara-
toulkeridou/involute-gear-plug-in-demo, 2014.
[27] W.T. Thomson, and M.D. Dahleh, Theory of vibration with applications. 5th ed
Prentice Hall, 1997.
[28] J.K. Cooper, Modeling a Small-Scale Helicopter for Simulation and Control
Development, M.S. Thesis, Department of Mechanical Eng., Virginia Tech.,
Blacksburg, VA, 2006.
[29] G.D. Padfield, Helicopter Flight Dynamics: The Theory and Application of
Flying Qualities and Simulation Modeling.. Blackwell science Ltd, 1996.
[30] L. Raptis, and K. Valavanis, Linear and nonlinear control of small-scale
unmanned helicopters. Vol. 45. Springer, 2011.
[http://dx.doi.org/10.1007/978-94-007-0023-9]
[31] M.S. Sajedi, and R.M. Ziazi, Development of a software for the stability and
performance of unmanned helicopter, B.S. thesis, Aerospace Eng. Dept.,
Amirkabir University of Tech., Tehran, Iran, 2009.
[32] D.C. Lombardo, and K.F. Fraser, Importance of reliability assessment to
helicopter structural component fatigue life prediction. DSTO Aeronautical and
Maritime Research Laboratory, 2002.
[33] F. Kermanshahi, M. Mohagheghi, and S. Sadati, "Reliability analysis, Design
improvements, and maintenance procedure compilation of a novel remotely
piloted helicopter", AIAA Info Tech Conference, 2011 St. Louis, MI, USA.
[http://dx.doi.org/10.2514/6.2011-1545]
[34] Engineering design handbook helicopter engineering. Part one. Preliminary
design. Storming Media, 1997.
[35] H. Pham, Handbook of reliability engineering. Springer Verlag, 2006.
CHAPTER 11
Materials Selection in Design of Structures of

Subsonic and Supersonic Aircrafts
Zainul Huda*
Department of Mechanical Engineering, King Abdulaziz University,

Jeddah 21589, Saudi Arabia
Abstract: This chapter reviews the advances in the materials for applications in
structures of both subsonicas well as supersonic aircrafts. An account of the
operating and ambient environmental conditions during flight is first given and the
resulting material requirements have been discussed. Design relationships have been
established taking into consideration the loading conditions and the strength
requirements. In particular, the aircraft skin temperatures at various mach numbers
have been taken into account for selecting appropriate structural materials for both
subsonic and supersonic aircrafts; and consequently various aerospace aluminum
alloys, titanium alloys, superalloys, and composites have been suggested. Finally, a
new materials-selection chart is presented which would help aerospace designers to
select appropriate materials for structural application in subsonic and supersonic
aircrafts.
Keywords: Aluminum alloys, Composites, Materials selection, Subsonic

aircraft, Supersonic aircrafts, Titanium alloys.
*Corresponding author Zainul Huda: Department of Mechanical Engineering, King Abdulaziz University,
Jeddah 21589, Saudi Arabia; Tel/Fax: +966537701246; E-mail: drzainulhuda@hotmail.com.
Yucheng Liu (Ed.)

Materials Selection in Design of Structures Frontiers in Aerospace Science, Vol. 1 443
NOMENCLATURE
M Mach number
CT Civil transport
SST Supersonic transport
CFRP Carbon fiber reinforced polymer
SCC Stress corrosion cracking
W(a) Weight of a structural member using material (a)
W(b) Weight of a structural member using material (b)
ρ(a) Density of material (a),
ρ(b) Density of material (b),
σy(a) Yield strength of material (a)
σy(b) Yield strength of material (b)
τ(a) Torsional shear stresses acting on materials (a)
τ(b) Torsional shear stresses acting on materials (b)
E Young’s modulus
K Stress intensity factor
KIC Plane strain fracture toughness
σ Stress applied normal to a crack
RMAF Royal Malaysian Air Force
Al-Cu Aluminum-copper alloy
P/M Powder metallurgy
Ti-Al-V Titanium-aluminum-vanadium alloy
UTS Ultimate tensile strength
BMIs Bismaleimides
CEs Cyanate esters
FGM Functionally graded materials
INTRODUCTION
The design of an aircraft and its materials have to be determined based on

classical engineering principles and practices because even a tiny mistake
in the design of any critical aircraft components such as fuselage, skeleton,
444 Frontiers in Aerospace Science, Vol. 1 Zainul Huda
and wing may lead to fatality. A subsonic aircraft is an aircraft with a

maximum speed less than the Mach 1 (speed of sound) i.e. less than 750
mi/h (1207 km/h). On the other hand, supersonic aircrafts have the
capacity of flying at speeds greater than Mach one, 1 < M < 4. Currently,
supersonic aircrafts fly at speeds less than 2000 mi/h (3219 km/h). The
important factors in selecting aircraft materials include specific strength
(strength-to-weight ratio), tensile mechanical properties, fatigue strength,
low-speed impact strength, fracture toughness, notch sensitivity,
manufacturability, resistances to crack propagation, stress corrosion, and
exfoliation corrosion [1-3]. Extra material criterion, exclusively related to
the design of supersonic aircraft, is the resistance to creep. This is because
the long-term operation at Mach 3.5 may lead to heat buildup within the
structure of the aircraft to a temperature around 300 oC [4, 5].
Having established the material requirements for application in structures

of subsonic and supersonic aircrafts in the preceding paragraph, it is
appropriate to discuss the performance the aerospace structural materials;
the latter include aluminum alloys, titanium alloys, and composites. This
paragraph deals with aerospace alloys for subsonic aircrafts. A high-priority
material research focus in today’s civil-transport (CT) aerospace industry is
the development of aluminum alloys with higher specific strength
(strength/weight ratio) for subsonic aircrafts; this research focus is
obviously owing to lower cost of aluminum as compared to titanium and
composites. The 2024 and 7075 aluminum alloys have remained in long use
in CT subsonic aircrafts. For instance, age-hardenable 2024-T3 aluminum
alloy is used in the subsonic structure: an aircraft for military transport
application. The 7075-T6 Al alloy sheets and plate products have
applications in subsonic aircraft structures owing to their notable
combination of high strength with moderate toughness and corrosion
resistance [6, 7]. In addition to Al alloys, titanium aerospace alloys are
known for their high specific strength, moderately high-temperature
stability, and resistance to corrosion. The exceptionally attractive
properties of titanium alloys render them suitable for applications in both

subsonic and supersonic aerospace structures. The aerospace industry is the
single largest market for titanium and its alloys. For instance, the high
corrosion resistance of titanium alloys renders them to be used in hatch and
door edging where moisture is likely to be accumulated. A detail account of
titanium alloy applications in given in section 3.
Now, we focus our discussion on supersonic aircrafts. In one of the first

supersonic transport (SST) aircraft design, Concorde, aluminum alloy was
selected as the primary structural material based on the fact that Mach 2 was
chosen as the design cruise speed. However, such design of Concorde was
rejected by almost every country due to its sonic boom and also because
that the pollutants exhausted from its engine may deplete ozone. Sonic
boom is the “thunder-like” noise produced when a plane is traveling faster
than the speed of sound [8]. This is the reason why almost all today’s
supersonic aircrafts are designed for military use, which are mainly made of
lightweight carbon-fiber reinforced polymer (CFRP) composite materials
that excel in specific strength, fatigue strength, corrosion-resistance, and
has reasonably high creep strength [9, 10]. Another criterion in the
aerospace composites is their ability to be repaired when the skin of the
aircraft panel becomes disbonded [11]. In addition to military aircrafts,
CFRP are also extensively used in SST aircrafts. For example, much of the
fuselage of the new Boeing 787 Dreamliner and Airbus A350 XWB are
made of CFRP.
OPERATING CONDITIONS AND MATERIAL SELECTION
Environmental Service Conditions
In aircraft material’s section, a complete understanding of expected service

conditions is very important. Besides that, other important factors include
the identification of design criteria, and the testing and evaluation of
structural concepts. Typical variables that must be considered in both sub-
and supersonic designs include operating stresses and temperatures, loads,

ambient environmental conditions, moisture and fluid exposures, radiation,
maintenance, and ground handling [12, 13]. In particular, exposure to
extremes in temperature (ranging from about −30°C to 370oC), moisture
(extreme humidity to desert conditions), and radiation (especially
ultra-violet) can either cause degradation of coatings and structural
materials or exacerbate the damaging effects of flight exposure conditions.
Furthermore, discrete damage events, both on the ground and in flight, stay
as a threat to polymer composite and thin-skinned components. The
discrete damage may be caused by hail impact, lightning strike, transport
and handling, and foreign objects [14]. For supersonic aircraft, the
flight-cycle conditions are determined by speed at cruise, altitude, flight
loads, and spike (or failure) conditions. The supersonic civil transport
(SCT) aircraft should be designed for a lifetime of approximately 20,000
flight cycles, which equal about 60,000 flight hours [15]. A major portion of
this flight cycle will be under supersonic cruise conditions at speeds
between Mach 2 and Mach 2.4.
Fig. (1). Fuselage lap joint showing the localized pillowing caused by crevice corrosion occurring
between the two layers.
Stress corrosion cracking (SCC) in a particularly deleterious form of

environmental attack on aircraft structures that will create opportunities
for cracks to nucleate and grow to failure, even under limited fatigue
loading conditions. In fuselage lap joints, the crevice corrosion that occurs
will result in pressure build up between the layers, sometimes to the point
where rivet heads will pop off and the joint will look pillowed, such as
shown in Fig. (1).
LOADING CONDITIONS AND DESIGN RELATIONSHIPS
During its flight, an aircraft may be subjected to four types of loading: (1)
tension, (2) compression, (3) bending, and (4) torsion. Minimization of
aircraft weight while still satisfying strength requirements is an important
objective in material selection for both subsonic and supersonic aircrafts.
For subsonic applications, lightweight structure results in cost-saving on
fuel; whereas high speeds are achieved by lightweight structures for
supersonic aircrafts. The relationship of each of the four loading types and
the overall weight of the aircraft are discussed as follows [16].
Design Equation for Tension
For a tensile loading below the yield limit, the applied stress on a structural
component (e.g. fuselage) should be considered regarding its weight (W). In
comparing specific strengths of two materials: (a) and (b), the design
relationship may be expressed as:
(W(a) / W(b)) = [(ρ(a) σy(b)) / (ρ(b) σy(a))] (1)
where W(a) is weight of a part using material (a), W(b) weight of the part with
material (b); density of materials (a) and (b) are denoted as ρ(a) and ρ(b) ,
respectively; σy(a) and σy(b) are yield strengths of (a) and (b), respectively.
Design Equation for Compression
Under compression, an aircraft part may fail due to buckling; this loading
condition requires the selected material has a high stiffness (Young’s
modulus, E). In comparing two materials (a) and (b), the following
compression relationship is recommended for the aircraft designer:
(W(a) / W(b)) = (ρ(a) /ρ(b))[(E(b)/ E(a))]1/3 (2)
Design Equation for Bending
Under bending, a structural member, such as the aircraft wing, is subjected

to both tensile and compressive stresses, thus the following relationship
should be applied when comparing the two materials:
(W(a) / W(b)) = (ρ(a)/ ρ(b))[σy(b)/σy(a)]1/2 (3)
where the symbols have the same meanings as mentioned in Eq. (1).
Design Equation for Torsion
Thin-walled tube is a popular airframe component, which are subjected to

torsional stresses. Under torsion, the relationship between two materials
may be expressed as:
(W(a) / W(b)) = (ρ(a) τ(b)/ ρ(b) τ(a)) (4)
where τ(a) and τ(a) are torsional shear stresses acting on materials (a) and (b),
respectively.
Eqs. (1-4) provide useful tools for an aerospace designer to select the best
material for a specific aircraft structural member with respect to the type of
loading applied on it. Although Eqs. (1–4) provide useful design
relationships in selection of aircraft materials, they do not completely
satisfy all design requirements. In order to meet all requirements, fracture

mechanics has to be applied during the design process.
MATERIALS SELECTION FOR AIRCRAFT STRUCTURE ----

A DESIGN APPROACH
Design philosophy relationship has to be applied when considering safety
issue in material selection, as explained in [17-19]:
K = KIC = Y σ (πa)1/2 (5)
where K is stress intensity factor, KIC the plane strain fracture toughness, σ
is the stress which acts along a normal direction to the major axis of a
central crack of length 2a (or an edge crack of length a) in the component,
and Y denotes geometric compliance function that describes the geometry
of the structural component. In order to achieve a safe design, material’s
fracture toughness has to be selected to guarantee that KIC> K in order to
prevent crack propagation.
The design philosophy {see Eq. (5)} is of paramount importance for

selecting an appropriate material for safe design of a part for either
subsonic or a supersonic aircraft. For instance, the British de Havilland
Comet 1, G-ALYP Aircraft fractured due to fatigue crack propagation;
which in turn was due to incorrect design approach for material selection.
A lower fracture-toughness 7000 series aluminum alloy was selected for
application in wing skin of the aircraft on the basis of its high strength and
light weight; hence KIC was fixed. The design stress, σ, on the wing was
then set at a high level to achieve a high payload capacity for the aircraft.
Using input values of KIC and σ, Eq. (5) was used to define flaw size, a.
During the crash flight, a fatigue crack grew out from a hidden rivet hole
in one of the wing plates and propagated to the critical size of ac, where
conditions of Eq. (5) are met. The result was fracture. The most
unfortunate aspect of the fracture was the fact that the allowable flaw size
that could be tolerated by the material under the applied stress was
smaller than the diameter of the rivet head; and hence it was impossible
for the inspection people to detect the crack which grew to fracture. In
fact, the designers should have used a different approach. They should
have first fixed the two variables namely high-level stress, σ, and the
critical flaw size, ac; and then should have applied Eq. (5) to define the
required fracture toughness KIC. This approach would have resulted in the
selection of material of appropriate fracture toughness to avoid the failure.
Because of the stringent operating conditions that supersonic flight, it will
be helpful to combine the application of high-performance aluminum alloys,
titanium alloys and polymer matrix composites in the aircraft structure.
Section 3 discusses the performance characteristics of each class of the
three aerospace materials.
In selecting materials for structural parts of modern high-performance

aircrafts Materials Selection Charts, such as Ashby chart, should be used
[20, 21]. An Ashby plot is particularly helpful for designing a stiff and light
part of an aircraft. In an Ashby plot, a candidate material is located as a
point on the plot based on its Young's modulus and density. With that plot,
users can easily find a material with the highest stiffness, the lowest density,
and even the best (E/ρ) ratio. Traditional materials selection charts are 2D
plots and becomes inefficient for advanced aircraft design. In 2011, M.F.
Ashby reported the development of an advanced 3D Materials Selection
Chart, which comprehensively takes Young’s modulus-density-fracture
strength into consideration within a single chart. This chart allows for a
correct and quick material selection for advanced aircraft design [22].
Today’s aerospace designers are expected to efficiently determine materials

for different structural components in a designed aircraft. A remarkable
progress in this field has been made through the development of
computer-based and web-based materials-selection systems. An aerospace
engineer is able to quickly decide materials for both sub- and supersonic
aircrafts with assistance of these IT-based systems (e.g. CES Selector)
[23-25].
MATERIALS SELECTION BASED ON SPEED AND

TEMPERATURE
Materials Selection for Subsonic Structures (for Speeds ≤ 1.0 Mach)
The skin temperature during a subsonic flight is usually below 70 oC. This
non-stringent thermal condition enables aerospace designers to use
non-expensive age-hardened aluminum-copper alloys or aluminum lithium
alloys for subsonic aircraft’s structures. Aluminum (alloy) is the primary
aircraft material, comprising about 80 per cent of an aircraft's un-laden
weight. Because the metal resists corrosion, some airlines do not paint
their aero-planes, saving several hundred of kilograms in weight. In
particular, series 2xxx, 6xxx, and 7xxx Al alloys find wide applications in
subsonic aerospace structures. For instance, the ALCLAD 2024-T3 alloy is
being used in C-130 Hercules (a subsonic aircraft for military transport
applications) by Royal Malaysian Air Force (RMAF) [2]. Additionally, the
age-hardened 2024-T3 aluminum-copper alloy is used in fuselage, wing
tension members, shear webs, and structural areas where stiffness, fatigue
performance, and good strength are required. The selection of heat
treatment approaches for the aerospace Al alloys play an important role in
strengthening the alloys for supersonic applications [26]. The
age-hardening heat treatment for the alloy involves solution treatment
(heating) in the range of 450-550 oC and holding for about 1 hour followed
by quenching. The solution treated alloy is then either naturally or
artificially aged [17]. Fig. (2) illustrates the steps in age hardening of Al-Cu
alloys.
The 7075-T6 alloy (Al-5.6Zn-2.4Mg-1.6Cu-0.3Mn-0.5Fe-0.4Si-0.22Cr-

0.2Ti alloy) is the most commonly used age-hardened alloy for subsonic
structures and highly stressed parts of the aircraft [27]. The 6061-T6
aerospace alloy has considerable application for extrusions requiring thin

sections and excellent corrosion resistance. Additionally, the high-strength
powder-metallurgy (P/M) Al alloys (e.g. CT90, CT91, 9051, etc.) and
aluminum-lithium alloys have been successfully used in subsonic aircrafts
[28].
Fig. (3) shows use of various aerospace aluminum alloys in subsonic

aircraft’s structures.
ALUMINIUM - COPPER EQUILIBRIUM DIAGRAM
800 K
K
SOLUTION TREATED
K K
K
Liquid
K
K
600 K +L
K 5,6%
K
QUENCHED
Temperature (oC)
400
K +0 AGED
200 K + submicroscopic 0
OVERAGED
0
0 2 4 6 8 10
K +0
Wt. % Copper
Fig. (2). Steps in age hardening of Al-Cu alloys.

Wing upper surface Fuslage skin

7068-T7751skin Alclad 2524-T3
7051-T77511 spur chords
Lower wing surface

2524-T39 skin
2026-T3511 stringers Floor
2099-T83
Wing tension members

7075-T6
2024-T3
Fig. (3). Aluminum alloys application in subsonic aircraft’s structure.
Besides aluminum, titanium alloys have wide applications in subsonic

aircraft’s structure. The moderately high-temperature stability of titanium
alloys enables them suitable for application in fuselage skins and flame
preventing fire safety-proof membranes of subsonic aircrafts. Additionally,
the high-strength Ti-6Al-4V is the most commonly used alloy for
application in subsonic-aircraft’s highly-loaded assemblies (e.g. landing
gear, fastening elements of the wing, and the like). Although aluminum
alloys are cost-effective aerospace structural materials, composite materials
offer better weight reduction; and hence are increasingly used in the
structure of modern passenger aircrafts (e.g. Airbus A330) (see Fig. 4) [29].
Composite materials reduce weight (typically 20% less weight than
aluminum), and are known to be more reliable with a longer lifespan than
other traditional metallic materials; thereby reducing aircraft maintenance
costs, and frequency of inspections during service. This is why composite
materials contribute 25% structure (particularly fuselage frames) of the
Airbus A380; however, the composite frames features aluminum strips to
ensure the electrical continuity of the fuselage (for dissipating lightning
strikes).
Fig. (4). Airbus A330 subsonic aircraft; composites contribute to 25% structure [29].
Materials Selection for Structures with Speeds ≤ 2.0 Mach
When supersonic transport aircrafts (SSTA) fly at a speed of about 1.8

Mach, the temperatures are below 100oC. Under that temperature level,
both the aluminum and CFRP composite laminates can be selected for the
fuselage of SST aircrafts. Considering the complexity of manufacturing
process and its cost, aerospace aluminum alloys should be preferred over
composites for SST aircrafts flying at speeds up to 1.8 Mach. The
heat-treated aluminum alloys: 2024-T81, T6, T62, or T81 are more suitable
for extruded members, while 2014-T6, 2024, and 2618-T61 are best for
forged products located in the areas that will be affected by heat in an
aircraft. An aerospace designer can refer to Table 1 for selecting aerospace
aluminum alloys for supersonic applications at speeds up to 2.0 Mach.
Table 1. Mechanical properties of selected aluminum alloys for application in SST aircrafts [30,
31].
Al Alloy Density Yield Strength UTS (MPa) Elongation at
(g/cm3) (MPa) Break
2024-T81 2.78 450 485 6%
2024-T86 2.78 440 515 6%

2090-T83 2.59 483 520 5%
2219-T87 2.84 393 476 10%
8090-T651 2.53 360 465 6.2%
7075-T6 2.81 505 575 11%
New type 2-mm n/a 424 555 12.3%
Al-Zn-Mg-Sc-Zr
alloy
It is evident from the data in Table 1 that although the 2219-T87 aluminum
alloy is easy to manufacture (10% elongation at break), its applicability in
modern supersonic aircrafts is limited by its low yield strength.
Comparatively, the easy-to-fabricate 7075-T6 alloy and the new
(heat-treated) type 2-mm Al-Zn-Mg-Sc-Zr alloy (with the highest
elongation at break of 12%) meet strength requirements for the modern
SCT aircrafts [27, 31]. In fact, 7075-T6 is the primary material for making
honeycomb panel frames, which are machined from plate to eliminate
corner joints. The aluminum honeycomb can also be used in the beaded
areas of skin doublers to stiffen the skin of fuselage. At elevated
temperatures, 2024-T81 foil provides higher strength than the work
hardened alloys, such as 5052-H39 and 5056-H39. Since lithium is the least
dense metallic element, the alloying of Li with Al can substantially reduce
the weight of aerospace alloys. For instance, 2090-T651 can be used in
fuselage bulkhead webs and internal framework parts. Another lightweight
Al-Li alloy is 8090-T651 has a high melting range of 600-655 oC, which
places it the best option for building the extendable nose.
The skeleton of the supersonic aircraft needs to be light but strong enough
so as to withstand high pressure applied on it. Titanium alloy Ti-6Al-4V has
excellent corrosion resistance, machinability, and the strength-to-weight
ratio and is an ideal material for the skeleton [32, 33]. A skeleton made of
Ti-6Al-4V will not be affected by the heat due to the friction because of the
high melting point of titanium (1668 °C). In addition, the very low thermal
expansion (8.6 μm∙m−1·K−1) at 25 °C of titanium provides excellent thermal
stability of the aircraft during supersonic operation.
Materials Selection for Structure for Speeds Between 2.0 and 4.0 Mach
The temperature of skin of the supersonic aircraft is a function of its speed
and also depends on its environment. The skin temperature at first drops
when the aircraft climbs and goes to contact with the decreasing ambient air
temperatures. As the speed exceeds Mach 1, the ambient air temperature
begins to increase, and the skin temperature reaches a maximum of 120 °C
(248 °F) when the aircraft is cruising at Mach 2.2. The skin temperature
stays below 100 °C (212 °F) at a speed of Mach 2.0 and reaches 150 °C
(302 °F) when the speed reaches Mach 2.4. As shown in Table 2, for a
hypersonic aircraft whose speed higher than 4.0 Mach, the skin temperature
may exceed 370 oC.
Table 2. Skin temperatures of aircrafts at different speeds.

Speed (Mach Number) Skin Temperature (oC)
2.0 100
2.5 150
3.0 200
3.5 300
4.0 370
The data in Table 2 clearly suggests that high strength and high
heat-resistant composites should be applied for current supersonic aircrafts
that fly at speeds in the range of 2.5-4.0 Mach. For example, the SR-71
Blackbird jet could fly continuously at Mach 3.1 with the temperature of
some parts reaching or even exceeding 315 °C (600°F). Most recent
high-temperature carbon fiber reinforced polymers (CFRP), such as
graphite fiber/PMR-15 and graphite fiber/PMR-11-55, can properly
function for thousands of hours under temperatures between 290 and
345 °C. Due to this reason, both SCTAs and military aircrafts are made of
CFRP to achieve a higher fuel efficiency.
The data in Table 3 provides a guideline for designers to select best CFRP
for supersonic aircrafts under different operating conditions. For example,
the CFRP with high elastic modulus should be selected for high Mach
speeds.
Table 3. Mechanical properties of three classes of CFRP composites [30].
Tensile Standard Modulus Intermediate Modulus High Modulus

Mechanical CFRP CFRP CFRP
Properties
Tensile strength 3450-4830 MPa 3450-6200 MPa 3450-5520 MPa
Young’s modulus 220-241 GPa 290-297 GPa 345-448 GPa
(E)
Elongation at 1.5-2.2% 1.3-2.0% 0.7-1.0%
break
Specific materials for each part of the SST aircraft are now selected based
on the operating conditions of the parts in supersonic aircrafts and the
temperature resistance of the materials. Fig. (5) indicates that 2650 Al alloy,
a cost effective material which can resist long-term creep at temperatures
ranging from 100 to 130 °C, is the ideal candidate for the fuselage skin of
the SST aircraft. For example, the application of 2650-T8 alloy showed that
it did not deform by more than 0.1% under a 150 MPa stress during a period
of its 60,000 h service life [34].
As shown in Fig. (5), different parts of the aircraft are heated up to different
temperatures during flight, therefore they have to be made of different
materials. Specifically, the temperature at the nose and around wings are
significantly higher than that along the fuselage. It is then decided that
high-temperature CFRP (such as graphite fiber/PMR-15, graphite fiber/
PMR-11-55) should be used for making the fuselage and advanced CFRP
with better heat resistance should be used for the nose and around the wings.
A good candidate is carbon fiber reinforced epoxy, which possesses
excellent specific strength (450 GPa/kg/m3) (see Fig. 6) [7], and is
recommended for speeds in the range of 2.5-3.5 Mach.
Fig. (5). Variation and distribution of skin temperatures for Mach 2.2 and 2.4 SST aircrafts.
Today’s high-speed supersonic flights require advanced aerospace

materials that can properly function under extremely high temperatures. For
example, engine nozzles of military aircraft need to be operated at a
temperature as high as 399 °C (750 oF) for thousands of hours. High-
performance composites have been developed to fulfill that requirement.
Those composites include polyimides, bismaleimides (BMIs), cyanate
esters (CEs), benzoxazines and phthalonitriles [35]. Among them,
polyimides, bismaleimides (BMIs), cyanate esters (CEs), andbenzoxazines
have outstanding temperature resistance and are recommended for speeds
above 3.5 Mach. The phthalonitrile resin system (graphite fiber-
phthalonitrile composite) developed at the United States Naval Research
Laboratory, Washington, DC can survive when exposed to elevated
temperatures up to 371 °C for longer time periods [36]. Comparatively the

epoxy-based systems used in current naval aircrafts start to lose their
rigidity or shear strength (depending on the moisture content of the resin)
when the temperature approaches 150 °C. For those applications asso-
ciated with short exposures times (such as missile applications), it was
shown from high-temperature tension tests that graphite fiber composites
fabricated using the phthalonitrile resin system can maintain their tensile
strength to under 538 °C. Epoxy-based graphite fiber-reinforced composites,
however, will dramatically lose their tensile strength at temperatures
around 260 °C during a similar exposure time (Fig. 7). From Fig. (7) it can
be deduced that that the ultimate tensile strength (UTS) of graphite-epoxy
drops nearly 50% at 316 °C. On the other hand, the graphite-phthalonitrile
does not suffer significant degradation until the temperature exceeds 482 °C.
It is therefore specially recommended for application in nose and around
wings of the supersonic aircraft when its flying speed stays around 3.5
Mach.
Fig. (6). Density normalized stiffness – strength curve for lightweight aerospace materials [7].
Fig. (7). A comparison of normalized ultimate strengths of graphite-epoxy and graphite-

phthalonitrile unidirectional composites [36].
Material-Selection for Military Aircraft Wings

In this section, the material for military aircraft wings is selected following
Weighted Relative Property Rating Procedure, which is illustrated in (Table
4). In that procedure, only metallic materials are considered because the
composites do not have a sharp yield point (σy). As shown in the last
column of Table 4, weighting factors have been allocated as follows: 10 for
mechanical properties, 20 for temperature resistance (if the aircraft speed
reaches Mach 4.0), and unity for cost.
The analysis of data in Table 4 indicates that stainless steel should be the
optimum choice for the specific application. However, with its outstanding
corrosion-resistance capability and low density, titanium alloy should be
preferred if the supersonic aircraft speed is below 3.0 Mach.
Table 4. Weighted relative property rating procedure.
Material σy/ρ Abs Rel. (KIC/σy)2 mm E1/3/ρ Abs. Temp limit Cost (£/ton) Abs Overall
o
=A Abs Rel. = B Rel. = C ( C) Abs Rel Rel =E Rating
=D [10A+10
B+10
C+20
D+(1E)]/51
Al alloy 1 130 0.64 16.5 1.00 1.5 1.0 150 0.38 13,100 0.134 0.683
Al alloy 2 204 1.00 2.1 0.13 1.5 1.0 15 0.38 14,000 0.142 0.583
Ti alloy 196 0.96 4.6 0.27 1.0 0.7 30 0.75 98,000 1.000 0.674
Stainless 115 0.56 12.3 0.75 0.7 0.5 40 1.00 10,900 0.111 0.764
steel
Fig 8. Temperature variation within the military aircraft wings during a supersonic flight [2].
In conclusion, the best scheme of selecting materials for the military aircraft
wings should be a combination of stainless steel (or composite), titanium
alloy, and high-temperature aluminum alloy (2650-T8). This
materials-selection recommendation is made based on the wide
temperature range to which the wings may be heated up during the
supersonic flight (Fig. 8).
A NEW MATERIALS SELECTION CHART FOR SSTAs
Based on the previous discussions of pros and cons of different aerospace

materials, Table 5 is developed which allows aerospace designers to select
appropriate materials in designing structural components for subsonic and
supersonic aircrafts that are cruising at speeds within different Mach
number ranges. In Table 5, a strong dependence between the cruise speed
and the material selection strategy is presented. Table 5, combined with the
material properties, as displayed in Figs. (6 and 7) provide an effective tool
for selection of supersonic aircraft materials.
Table 5. A new materials selection guide for subsonic and supersonic aircraft structural
components.
Current Subsonic with Current Supersonic Current Supersonic Future Hypersonic
Cruise Speed up to 1.0 with Cruise Speed up with Cruise Speed: with Cruise Speed
Mach to 2.0 Mach 2.0-4.0 Mach above 4.0 Mach
Possible Material(s) to Possible Material(s) to Possible Material(s) to Possible
be Selected be Selected be Selected Material(s) to be
Selected
Fuselage: Fuselage: Fuselage: Fuselage:
Al: 2024-T4; 7075-T6; Al: 2090-T651; CFRP: PMR-15, PMR- Polyimides, BMIs,
6061-T6, Al-Li alloys 7075-T6; 8090-T651 11-55, or similar high- CEs, Graphite fiber-
(nose) modulus composites; or phthalonitrile, or
New type 2-mm Al-Zn- FGM (ceramic-
Mg-Sc-Zr alloy metal composite)
Skeleton: Skeleton: Skeleton: Skeleton:

Ti-6Al-4V Ti-6Al-4V or other High perf. Ti alloy Titanium aluminides,
high-perf. Ti alloys or high perf. Ti alloy
Wings: Wings: Wings: Wings:

Fatigue-resistant Al Commercial grade CFRP: PMR-15, Mesh-protected
alloys + Commercial CFRP + Ti alloy PMR-11-55, or similar CFRP, Polyimides,
grade CFRP composites; stainless BMIs, CEs, or
steel; or Ti alloy Stainless steel, or
Ti alloy
CONCLUSION AND RECOMMENDATIONS
In this chapter, methods and guidelines of selecting appropriate materials

for subsonic and supersonic aircraft structural components have been
reviewed. It is concluded that during the material selection, the most
important part is to correctly identify the expected service conditions of the
designed structures, which include operating stresses and temperatures,
compressibility effects, ambient environmental conditions, moisture and
fluid exposures, radiation, lightening-strike, maintenance, ground handling
etc. Strength and thermal properties of a variety of aerospace materials have
been reviewed. The materials-selection tool Weighted Relative Property
Rating procedure is introduced and has been applied to select materials for
a supersonic aircraft wings. A unique materials-selection table is presented
to facilitate the selection of structural materials with respect to different
cruising speeds.
The author confirms that author have no conflict of interest to declare for
this publication.
ACKNOWLEDGEMENTS
Declared none.
REFERENCES
[1] G. Campanile, A. Prisco, A. Squillace, C. Bitondo, G. Dionoro, and P.

Buonadonna, "FSW of AA2139-T8 Butt joints for aeronautical applications",
J. Mater.: Design Appl., vol. 225, no. 2, pp. 103-110, 2011.
[2] Z. Huda, T. Zaharinie, and G.J. Min, "Temperature effects on material
behavior of aerospace aluminum alloys for subsonic and supersonic
aircrafts", J. Aerosp. Eng., vol. 23, no. 2, pp. 124-128, 2010.
[http://dx.doi.org/10.1061/(ASCE)AS.1943-5525.0000015]
[3] P.L. Mangonon, The Principles of Materials Selection for Engineering Design.
Prentice Hall: New York, 1999.
[4] Z. Kun, D. Sheng-long, Y. Shou-jie, and H. Min, "Development of a new
creep resistant Al-Cu-- g-Ag type alloy", J. Aeronaut. Mater., vol. 6, no. 3, pp.
251-257, 2006.
[5] S. Pantelakis, A. Kyrsanidi, E. El-Magd, J. Dunnwald, Y. Bardaux, and G.
Pons, "Creep resistance of aluminum alloys for the next generation supersonic
civil transport aircrafts", Theoretical & Applied Fracture Mechanics., vol. 31,
no. 1, pp. 31-39, 1999.
[6] M. Tajally, Effects of Deformation and Recrystallization on Mechanical
Properties of 7075 Aluminum Alloy, PhD thesis, University of Malaya:
Malaysia, 2010.
[7] J.C. Williams, and E.A. Starke Jr, "Progress in structural materials for
aerospace systems", Acta Mater., vol. 51, no. 19, pp. 5775-5799, 2003.
[http://dx.doi.org/10.1016/j.actamat.2003.08.023]
[8] J.D. Anderson, Aircraft Performance and Design. McGraw-Hill: New York,
1999.
[9] C. Soutis, "Fiber reinforced composites in aircraft construction", Prog.
Aerospace Sci., vol. 41, no. 2, pp. 143-151.

[10] S-L. Lee, and J-H. Kim, "Thermal post-buckling and the stability boundaries
of structurally damped functionally graded panels in supersonic airflows",
Compos. Struct, vol. 92, no. 2, pp. 422-429, 2010.
[http://dx.doi.org/10.1016/j.compstruct.2009.08.022]
[11] H. Subawi, "A practical method for repairing dis-bonded skin in aircraft
panels", In: Proceedings of the Institution of Mechanical Engineers, Part L:
Journal of Materials: Design and Applications, vol. 223, no. 4, pp. 161-168,
2009.
[http://dx.doi.org/10.1243/14644207JMDA250]
[12] "American Society of Automotive Engineers", the ASME International
Manufacturing Science and Engineering Conference, 2009 West Lafayette,
USA
[13] J.E. McCarty, and D.R. Wilson, "Advanced Composite Stabilizer for Boeing
737Aircraft", In: the Proceedings of the 6th Conference on Fibrous Composites
In Structural Design, AMMRC, TELAMS publications, 1983.
[14] C. Blohm, Ed., National Materials Advisory Board: Advanced Composites
Structures: In: Service Experience in the View of a Carrier. Workshop on
Long-Term Aging of Materials and Structures, August 10–12, National
Research Council, Washington, D.C, 1994.
[15] "National Materials Advisory Board", Workshop on Long-Term Aging of
Materials and Structures, August 10–12, 1994, National Research Council,
Washington, D.C
[16] J.C. Lopes, Materials selection for aeronautical structural application. Science
and Technology of Materials (Ciência & Tecnologia dos Materiais); vol. 20,
no.3/4, 78-82, 2008.
[17] Z. Huda, and R. Bulpett, Materials Science and Design for Engineers. Trans
Tech Publications: Zurich-Durnten,2012.
[18] Z. Huda, R. Bulpett, and K.Y. Lee, Design against Fracture and Failure,
Trans Tech Publications: Satafa-Zuerich, 2010.
[19] R.W. Hertzberg, R.P. Vinci, and J.L. Hertzberg, Deformation and Fracture
Mechanics of Engineering Materials. 5th ed John Wiley & Sons Inc: New
York, 2012.
[20] K.G. Budinski, and M.K. Budinski, Engineering Materials: Properties and
Selection. 8th ed Pearson Prentice Hall: New Jersey, 2005.
[21] G. Design, "Materials Decision Support (software)", Cambridge. Available
from: http://www.grantadesign.com, 2012.
[22] M.F. Ashby, Materials Selection in Mechanical Design. 4th ed
Butterworth-Heinemann: Burlington, 2011.
[23] L. Yuan-pei, M. Qing-chun, L. Feng, X. Guang-xing, and G. Zhi-dong,
"Aircraft design material- selection method based on MAUT theory", J.
Aeronaut. Mater., vol. 30, no. 3, pp. 88- 94, 2010.
[24] E. Norman, "Expert systems in the Design Process", Studies in Design
Education Craftand Technology, vol. 20, no. 1, 1987.
[25] Y. LAN, Z. Guan, Q. Jiao, and G. Xu, "A web-based computer-aided
material-selection system for aircraft design", J. Comput. (Taipei), vol. 6, no. 5,
pp. 976-983, 2011.
[26] J. Beddoes, "Design of solution heat treatments for aerospace alloys",
Canadian Metallurgical Quarterly, vol. 20, no. 3, pp. 215-221, 2011.
[27] M. Tajally, Z. Huda, and H.H. Masjuki, "A comparative analysis of tensile and
impact-toughness behavior of cold-worked and annealed 7075 aluminum
alloy", Int. J. Impact Eng., vol. 37, pp. 425- 432, 2010.
[28] R.F. Simenz, and M.K. Guess, "Structural Aluminum Materials for 1980's", J.
Aircr., vol. 17, no. 7, pp. 514-520, 1980.
[http://dx.doi.org/10.2514/3.57933]
[29] AIRBUS, "Innovative Materials: Setting the Innovation Standards", Available
from: http://www.airbus.com/innovation/proven-concepts/in-design/
innovation-materials/, 2014.
[30] W.D. Calister, and D.G. Rethwisch, Materials Science and Engineering. 8
John Wiley & Sons Inc: Singapore, 2011.
[31] Y. Deng, Z. Yin, J. Duan, K. Zhao, B. Tang, and Z. He, "Evolution of
microstructure and properties in a new type 2 mm Al–Zn–Mg–Sc–Zr alloy
sheet", J. Alloys Comp., pp. 118-126, 2012.
[32] M.J. Donachie Jr, TITANIUM: A Technical Guide. ASM International: Metals
Park, OH, 1988.
[33] S.I. Jaffery, and P.T. Mativenga, "Assessment of the machinability of
Ti-6Al-4V alloy using the wear map approach", Int. J. Adv. Manuf. Tech., vol.
40, no. 7-8, pp. 687-696, 2009.
[34] J. Majimel, M.J. Casanove, and G. MolA(c)nat, "A 2xxx aluminum alloy crept
at medium temperature: role of thermal activation on dislocation mechanisms",
Mater. Sci. Eng. A, vol. 380, no. 1-2, pp. 110-116, 2004.
[http://dx.doi.org/10.1016/j.msea.2004.03.050]
[35] V.P. McConnell, "Resins for the Hot Zone, Part II: BMIs, CEs, benzoxazines
and phthalonitriles", Available from: http://www.compositesworld.com/articles
/resins-for-the-hot-zone-part-ii--mis-ces-benzoxazines-and-phthalonitriles,
2009.
[36] H.N. Jones, and T.M. Keller,o High-Temperature Tensile Properties of
Graphite Fiber- Phthalonitrile Composites ??, Tech. Report. United States
Naval Research Laboratory: Washington, DC, 2002.
CHAPTER 12
Potential Aerospace Applications of Carbon

Nanotubes
Omid Gohardani*
Springs of Dreams Corporation, 340 East 1st Street, No. 8, Tustin,

California 92781, United States of America, USA
Abstract: Advanced materials with optimized properties are essential in

addressing the stringent requirements imposed by future aerospace vehicles. The
discovery of carbon nanotubes and their desirable properties, as recognized in
diverse scientific disciplines, have therefore identified these materials as
expedient candidates for usage in aerospace applications. Given the limited
number of overarching review articles encapsulating the usefulness of carbon
nanotubes in aerospace sciences, this chapter explores the prospective
applications of these materials in aerospace applications with their possible
implementation on future aircraft, unmanned aerial vehicles and rotorcraft. The
prospects of carbon nanotube usage intended for fuselage/satellite weight
reduction, aircraft icing mitigation, lightning protection for aircraft, and future
space launch are further explored. Conclusively, present challenges associated
with successful implementations of these materials and existing obstacles
preventing their safe integration in the aerospace industry are revisited, outlined
and discussed.
Keywords: Aerospace nanotechnology, Aircraft icing mitigation, Carbon

nanotubes, Future aerospace materials, Hydrophobicity, Space elevator.
*Corresponding author Omid Gohardani: Springs of Dreams Corporation, 340 East 1st Street, No. 8,
Tustin, California 92781, U.S.A; E-mail: omid.gohardani@springsofdreams.org.
Yucheng Liu (Ed.)

Applications of Carbon Nanotubes Frontiers in Aerospace Science, Vol. 1 469
NOMENCLATURE
ρ Density
ς Electrical resistivity
σmax Tensile strength
dt Nanotube diameter
k Thermal conductivity
m, n Integers
AR Aspect ratio
Ch Chiral vector
E Young’s modulus
INTRODUCTION
Carbon nanotubes (CNTs) have since their relatively recent discovery, been
employed in a multitude of different scientific applications such as sensing
[1], mechanical systems [2], energy storage [3, 4], biological applications
[5, 6], and field emission [7, 8]. The expanding usage of composite
materials on current commercial and military aircraft has in this respect
highlighted these materials as viable candidates that can sustain different
operational requirements in aeronautics and astronautics.
This chapter provides an introduction to CNTs, by means of a brief

historical overview of CNT usage in aerospace applications, viewed
through the lens of the aerospace industry and further investigates their
prospective usage areas in aeronautics and astronautics. In this context,
particular attention is devoted to potential benefits of CNTs intended for
unmanned aerial vehicles (UAVs) and their proposed usage, and other
aspects pertaining to future space launch platforms. The final discussion of
this chapter features a review of obstacles that have prevented a more
frequent utilization of CNTs in the aerospace industry and establishes
470 Frontiers in Aerospace Science, Vol. 1 Omid Gohardani
essential factors for such implementation.
In essence, CNTs are commonly described in the literature in two distinct

forms namely, single-walled carbon nanotubes (SWCNTs) and
multi-walled carbon nanotubes (MWCNTs) [9]. SWCNTs consist of single
tubular sheets of graphite while the latter comprise a family of concentric
graphite tubes [10]. For SWCNTs, the roll direction of the graphite sheet
and the tube structure denoted by the integers n and m, determines the
conductivity. The chiral vector is defined as Ch ≡ (n, m) [11], where the
values of m and n determine the structure of the nanotube as an armchair
(n = m), zigzag (m = 0) and chiral (other n and m values than the previous
two structures). An overview of the properties of SWCNTs and MWCNTs,
in comparison to traditional aerospace materials, is shown in Table 1.
Table 1. Comparison of CNTs to conventional materials. The symbol (*) distinguishes

normalized values, while (†) and (‡) denote theoretical and measured values, respectively.
(Sources: [12, 13]).
Specific Yield Elastic Thermal Electrical Strength-to

Material Gravity Strength Modulus Conductivity Resistivity -Mass
[g·cm-3] [GPa] [GPa] [W·m-1·K-1] [μΩ∙cm] Ratio*
SWCNTs† 1.4 65 1,000 ~ 6,000 30–100 225
SWCNTs‡ 1.4 1.8 80 150 150 7
Conventional
Carbon fiber, 2.2 4 550 70 800 9
M55J
IM7 Carbon
1.6 2.1 152 30 2,000 7
composites
Titanium 4.5 0.9 103 12 127 1
Aluminum 2.7 0.5 69 180 4.3 1
A comparison between density (ρ), nanotube diameter (dt), aspect ratio

(AR), thermal conductivity (k), electrical resistivity (ς), tensile strength

(σmax), and Young’s modulus (E) of SWCNTs and MWCNTs is additionally
presented in Table 2.
Table 2. Comparison of different properties pertaining to SWCNTs and MWCNTs. (Sources:

[2, 12, 14, 15]).
Property Unit SWCNTs MWCNTs
Density g·cm-3 ~ 1.3 ~1.75
Nanotube diameter Nm 0.6-0.8 5-50
Aspect ratio – 100-10,000 100-100,000
Thermal conductivity W(m·K)-1 3,000-6,000 3,000-6,000
Electrical resistivity Ωm 1·10 –1·10

-3 -4
2·10-3–1·10-4
Tensile strength GPa 50-500 10–600
Young’s modulus GPa 1,500 1,000
CARBON NANOTUBES IN AEROSPACE - HISTORICAL

OVERVIEW
Many review articles have been authored on the topic of CNTs, focusing
mainly on their properties [9, 16, 17], present state [18] and future
applications [10, 19]. The myriad of review and research articles describing
CNTs therefore render a complete literature list, to be beyond the scope of
this chapter.
Consequently, a selection of studies based on their actual applications

pertinent to aerospace applications, are discussed herein. The first
nano-sized carbon filaments of tubular nature were studied in 1952 [20].
However, this publication became available in English numerous years later.
Following the aforementioned study, a number of elaborated research
studies were independently carried out on the observations of fullerenes
[21-25]. The scientific impact of the article published by Iijima in 1991 [26],
with its novelty and dissemination to a broader scientific community,
conversely came to be considered as the birth year of MWCNTs. This topic

has however, attracted some scientific discussions, due to existing evidence
of CNTs preceding this date [27]. Given the general acceptance of 1991 as
the inception year of CNTs, this date has been adopted as the
commencement year of CNTs in aerospace sciences [12].
A literature excursion on CNTs in different databases such as Compendex,

Inspec, the National Technical Information Service (NTIS), and Referex,
clearly indicates that the number of publications related to these materials
has increased since the year 2000, as shown in Fig. (1). Within these
publications, the number of articles containing the keyword “aerospace”
has augmented from less than 0.25% to 1.8% in 2013, signifying the
potential prospects of CNT usage in aerospace sciences. In addition, the
low percentage also confirms their limited current implementations in this
discipline.
Number of publications
0 20000 40000 60000 80000 100000 120000
2.00%
Percentage of publications with the keywords "aircraft"
1.80%
and "aerospace" from total CNT publications
1.60%
1.40%
1.20%
CNT
1.00% CNT and aerospace
CNT and aircraft
0.80%
0.60%
0.40%
0.20%
0.00%
2000 2001 2002 2003 2004 2005 2006 2007 2008 2009 2010 2011 2012 2013
Year
Fig. (1). CNT publications with the keywords “aerospace” and “aircraft”, between the years 2000 and
2013. (Source: Adapted from [12]).
National Aeronautics and Space Administration (NASA) published a

significant roadmap for prospective applications of nanomaterials and
particularly CNTs in 2010 [28]. This roadmap identified that potential
benefits of utilizing CNTs in aerospace applications to consist of reduced
vehicle mass, improved functionality and durability, enhanced tolerance to
damage, improved self-healing characteristics, and enhanced thermal
protection. Moreover, this report outlined that CNTs can be utilized for
enhanced energy generation and distribution. Some identified challenges
related to usage of CNTs in aerospace applications comprise large-scale
production methods of CNTs, uniform dispersion of CNTs in composite
materials, alignment and adhesion issues associated with CNTs in
reinforced polymers, and a complete understanding of their toxicity and
production of CNTs, rendering them rice in volume, with a uniform size.
U.S. Army will utilize
Nanocomp Technologies,
CNTs for
declares that the
Nobel Prize in manufacturing of
EMSHIELDTM - a CNT sheet
Chemistry is stronger helicopter
material will be used on
awarded for blades
NASA’s Juno spacecraft
discovery of during its Jupiter mission, 2013
Discovery of MWCNTs fullerenes in order to protect the
1991 1996 vehicle from electrostatic
discharge
2013
NASA publishes its First consolidated
roadmap for future review article of CNTs
Iijima publishes his applications of in Aerospace
discovery in Nature nanomaterials and Applications published
CNTs by Gohardani et al.
2010 2014
CARBON NANOTUBES IN AEROSPACE APPLICATIONS

1990 2000 2010 2020
Iijima and Ichihashi

publish their
discovery in Nature 2011
Lockheed Martin
announces that the
1993 F-35 Lightning II will
Discovery of SWCNTs 2008 2008
Unidym, Inc. be the first mass-
NASA develops a
announces the CNT based chemical produced aircraft to
maiden flight of integrate a
sensing platform for
an aircraft using in-situ chemical thermoset epoxy
CNTs for analysis in space reinforced by CNTs
increased for non-load bearing
and terrestrial
fuselage strength applications airframe
components
Fig. (2). Selected milestones of CNTs and their implementations in aerospace sciences. (Source:
Adapted from [12]).
Despite the numerous original research articles and published reviews

related to CNTs, it would take numerous years since their authenticated
introduction, for a thorough review to emerge on this topic, specifically
intended for the field of aerospace. In 2014, Gohardani et al. [12],
published the very first consolidated review article of the prospects of
utilizing CNTs and nanoparticles in aerospace sciences, based on their
recent implementations and predicted future applications. Fig. (2), depicts
an overview of some milestones related to CNTs and their applications in
aerospace sciences.
CARBON NANOTUBES IN AERONAUTICS
In aeronautics, CNTs have the potential to become utilized in different

implementation areas, such as the airframe, wiring, aircraft icing mitigation,
propulsion systems, lightning protection, electromagnetic interference
shielding, stealth applications, structural damping, and sensing. Lower
weight consideration, has been one of the challenges constantly faced by
the aerospace industry throughout the history of aviation. Stemming from
economic reasons, a light aircraft or rotorcraft is often desired without the
risk of sacrificial structural integrity. For future commercial aircraft,
employing distributed propulsion technology, Gohardani et al. [29]
identified the total weight and propulsive power as the two dominant
factors likely to influence aviation. In addition, Gohardani’s first proposed
official definition for distributed propulsion technology in subsonic
ifxe. -wing aircraft [30, 31], considers materials such as CNTs to be crucial
for future designs in the aerospace industry.
The large surface area for a given volume and reinforcement effects caused
by a high aspect ratio [32] therefore, render nanocomposites suitable for
aeronautical applications. The influence of CNTs on the weight, range and
fuel consumption of commercial aircraft, was in a pioneering study
examined by O’Donnell [33] and colleagues [34, 35]. In these simulations,
a carbon reinforced polymer (CNRP) was used as the primary airframe

material on four conceptual aircraft structures, namely the Boeing 747-400,
Boeing 757-200, Airbus A320, and Embraer E145. By interchanging the
conventional aluminum material with CNRP with 70% by volume
SWCNTs, an average 14.1% weight reduction, with a corresponding 13.2%
range increase and 9.8% decrease in fuel consumption, was obtained.
Despite the unfeasible practical application of this study on the current fleet
of aircraft, the authors have successfully exhibited the weight saving
potential of CNTs, through their research.
Reduced fuel consumption and weight reduction can potentially be

obtained in aeronautics by replacement of the existing copper wiring with
wires made of CNTs [36-38]. For a current commercial airliner, the copper
wiring accounts for approximately 4,000 lbs in weight. Studies with a CNT
cable called 1553B, indicate a 69% weight saving potential, due to
interchanged wiring [39]. Additional research carried out by Northrop
Grumman, as a part of environmentally responsible aviation (ERA), under
contract with NASA on a N+2 advanced vehicle study, indicate that CNT
data cables can contribute to 0.1% improvement of the fuel burn reduction
for the Cargo 2025 preferred system concept (PSC) and a corresponding
value of 0.2%, for the Passenger PSC. This research also predicts the
development risk for the aforementioned cables to be moderate, by the year
2025 [40].
Aircraft icing is another discipline within aeronautics with adverse effects

on aircraft during flight. These effects may comprise a reduction of lift, stall
angle of attack and increased profile drag [41]. This phenomenon
commonly occurs upon existence of super-cooled droplets in the
atmosphere, impinging on the surfaces of an aircraft during flight [42]. A
recent study, carried out by Chu et al. [43] demonstrated a high efficiency
of a carbon nanotube paper (CNP) in deicing applications. The composite
in this study exhibited excellent electric heating properties at a wind speed
of 14 m·s-1 during ambient temperatures, resulting in deicing time intervals

between 220–450 s. Similarly, Zhao et al. [44], utilized functionalized
MWCNT sheets anchored onto carboxyl, by means of a photochemical
reaction. Upon integration by the resin infusion method, a steady-state
value of the sample temperature was obtained within 13 s, at a power
density consumption level of 0.58 W·in-2.
Novel concepts in aeronautics, suggest turboelectric generators to be

implemented at the wing tips and fans across the rear of the fuselage, on the
blended wing body (BWB) aircraft [29-31, 45]. The justification for
utilizing small-diameter motors stems from their high magnetic field
intensity and the produced force in relation to size [46]. CNTs are proposed
to be used on carbon electric motors, with efficiencies in proximity of those
with copper. Given that hydrogen is one of the alternative fuels with
positive environmental impact, Dillon et al. [3] studied a high reversible
hydrogen storage capacity of SWCNTs, hence attracting an increased
interest related to the application of CNT usage in hydrogen storage.
The introduction of composite materials into the airframe of modern

aircraft, imparts certain challenges, besides beneficial aspects related to
increased fuel efficiency, reduction of acoustic emissions and pollutants.
One such challenge, is a lower electrical conductivity in comparison to
aluminum airframe, should the aircraft be subject to a lightning strike threat
[12].
During such event on an unprotected surface, electricity with current levels

up to 200,000 A, traverses the path of least resistance [47], and might ensue
in resistive-heating, acoustic shock, ignition of vapors in the fuel tanks,
arcing, sparking at joints, and melting or burning at the lightning
attachment points [48, 49]. Fig. (3), depicts an aircraft’s lightning zones:
Zone 1 indicates the prospective area to be influenced by the initial
attachment of a lightning strike, Zones 2 and 3 designate a swept/moving
attachment and an area which may experience non-attached conducted

currents of a lightning strike, respectively [48, 49].
Fig. (3). Aircraft lightning zones as designated by SAE aerospace recommended practices 5414.
(Source: Adapted from [12]).
Gou et al. [47] attributed inadequate level of attained electrical conductivity

for lightning strike protection, due to heavy re-agglomeration levels of
CNTs and CNFs at high viscosity of the polymer resin and a limitation in
increased conductivity beyond the percolation threshold, originating in the
absence of effecting connection between the nanofillers. To alleviate these
effects, the authors introduced nickel nanostrands for bridging individual
CNFs and for modifying the electrical pathway within the carbon nanofiber
papers (CNFPs). The subject of lightning strike protection by employment
of CNTs, has also been studied elsewhere [50].
Sensing, is yet another prospective future application of CNTs. Self-sensing

entails that component diagnostic data is passively collected and preventive
signals supplied, in order to avoid control component failures [51, 52].
Moreover, these signals can be employed for optimization of the operating
configurations of the component [53].
For military aircraft, stealth technology is crucial for survival of the system.
Stealth performance is often represented by the radar cross section (RCS),
which is described as the strength of the radar signal backscattered from an
object for an incident electromagnetic wave and equally a measure of the
amount of reflected electromagnetic wave from an incident electromagnetic
wave [54]. The RCS can be minimized by modification of the external
features of the target, utilizing radar absorbing materials [55] and radar
absorbing structures. In one such study, Kim and Lee [54] have utilized
CNTs to achieve absorption of incident electromagnetic waves up to 90% in
the X-band frequency.
Morphing aircraft is an emerging aerospace technology, in which CNTs

usually are mentioned as envisaged materials [12]. Morphing structures
enable the different surfaces of the airframe to adapt to existing conditions.
By way of example the wings may adapt to flight conditions, i.e.
instantaneously unfurl, bend and reshape, based on changes in pressure,
temperature, and exposure to magnetic and electric ifelds [56]. An artist’s.
rendering of the “Morphing Airplane” is shown in Fig. (4).
Fig. (4). NASA’s Morphing Airplane. Source: NASA Dryden Flight Research Center Photo
Collection.
Future rotorcraft and helicopters may benefit from CNTs, for structural
damping applications. Studies carried out by Liu et al. [57] and Bakis and
Wang [58] on CNTs embedded in polymer resins through
micromechanics/molecular dynamics models exhibit that the interfacial
shear strength is ~25 times the value of non-functionalized CNTs, when
approximately 1% of the carbon atoms in CNTs are covalently bonded to
the epoxy resin.
Recent Studies
An extensive research effort has been carried out by Gohardani [41] since
2011, on the prospects of CNTs in aeronautics. A review study conducted
on erosion modeling in particular with materials reinforced with CNTs,
exhibited such modeling to be highly complex in nature [59]. In a different
study, empirical liquid erosion experiments conducted on polymeric matrix
composites reinforced with CNTs, exhibited nominally similar damage
threshold velocities as the non-reinforced composites. The liquid erosion
experiments were conducted upon utilizing Cavendish Laboratory's
Multiple Impact Jet Apparatus (MIJA) [60]. Based on the aforementioned
study, the CNT dispersion level did not appear to result in significantly
higher damage threshold velocities (DTVs), confining the aerospace
applications of the candidate materials to UAVs, with a cruise speed less
than Mach 0.4, such as the generic unmanned air vehicle (GUAV) [61].
Since wettability of materials reinforced with CNTs, has been an interesting

topic for aeronautical applications, Gohardani and Hammond [62] explored
the wetting character of these materials by utilizing the Dynamic and Static
Wettability Scheme for Advanced Materials (DSWSAM). The described
methodology was employed in order to numerically examine, empirically
acquired results based on their wettability. This study also described a
decision matrix that discriminated between hydrophobic and hydrophilic
surfaces based on their static and dynamic wettability properties and
presented idealized wetting character representations of the examined

aerospace materials [62]. The usefulness of CNTs for aircraft icing
applications was further studied by Gohardani and Hammond [63] and
additionally by Gohardani [64].
CARBON NANOTUBES AND UNMANNED AERIAL VEHICLES
The employment of CNTs on electric UAVs, can contribute to favorable

flight characteristics [12]. In a study conducted by Thomas and Qidwai
[65], it was demonstrated that the flight endurance of an electric UAV can
be extended 1.5 times more, for a given vehicle weight reduction versus
an increased battery energy.
Given the numerous future application areas of UAVs, the need for icing
protection systems has recently been more emphasized. Granted that
UAVs exhibit a lower climb rate than other military, rotary- and
ifxed-wing aircraft, ice accretion is more prone to ensue in destabilization
of the aircraft [66]. For this purpose, many different methods for icing
mitigation for UAVs have been undertaken by researchers in recent years
[67, 68].
CNTs are capable of providing additional benefits for tackling the

challenges currently faced by UAVs related to aircraft icing. With
expansion of the mission envelope and the operational usability of UAVs,
Battelle [69] has developed a Resistive Heating Coating (RHC)
technology, which alleviates the aforementioned issues by means of a
lightweight coating system applicable as a spray-on system to areas where
ice formation mitigation is desired. The hindrance of ice accretion and its
shedding is attained by a stack of coatings, in which the dispersion level
of CNTs in combination with the barrier coating and paint layers enable a
high conductivity, upon employment of electrical current, which results in
a heated surface. This technology stands out due to its lower power
densities in line with the limited available power for UAVs and traditional
coating application methods, enabling the retrofitting of this system onto
existing airframes.
Recent Studies
Pertaining to aircraft icing challenges on UAVs, a recent study [70] on

small UAVs addresses the issues associated with structural change, caused
by ice accretion by integration of a power control system and an
electrically conductive coating based on carbon nano material, providing
temperature control of the UAV’s airfoil surface. Results obtained in this
study indicate that a coating covering the entire length of the wing
provides both de-icing and anti-icing capabilities when needed, at
moderate energy consumption levels.
CARBON NANOTUBES IN ASTRONAUTICS
The Space Elevator concept has been one of the envisioned ideas for space
transportation that has attracted a considerable amount of attention, since
the discovery of CNTs [12, 71-81]. A Space Elevator is essentially a
physical connection from Earth’s surface to Geostationary Earth Orbit
(GEO) located approximately at an altitude of 36,000 km, by means of a
proposed cable made of CNTs, on which magnetic levitating vehicles travel
[12].
There are many benefits associated with the realization of the Space
Elevator. Upon its establishment, the Space Elevator may reduce the
current high costs associated with placement of small payloads into Low
Earth Orbit (LEO), enabling equipment to be brought back to Earth and a
rapid cargo launch [82]. Equally, numerous challenges have been identified
with this concept consisting of susceptibility to lightning strikes, potential
meteorite damages, LEO cable impacts, atomic oxygen damage, and micro
scale defects [12, 74, 77].
The benefits of CNTs in large satellite applications consist of weight

reduction of a satellite to one-third of its current weight [28]. A weight
saving potential is also evident for other spacecraft upon CNT usage, as
illustrated by the work of Harris et al. [83]. For terrestrial and satellite
communications systems, patch antennas fabricated from CNTs exhibit
advantageous features such as a planar structure, low-profile, gain ~ 7 dBi,
and bandwidth of 5–20% [84].
Immediate applications of CNTs in space applications can be beneficial in

order for onboard computer screens to operate in the vicinity of other
electronic equipment at the International Space Station and in future space
exploration missions [85]. These missions will require that computer
screens onboard the space vehicles have coatings with at least 90%
transmittance and a shielding effectiveness ~40 dB [12].
Recent Studies
In a recent study [86], polyimide (PI) infiltration into entangled cup-stacked
CNT (CSCNT) sheets preserved the beneficial properties of CNTs, while
preventing agglomeration and enabling CNT patterning. The obtained
CNI-PI films exhibited low sheet resistivity values in both lateral and
vertical direction, in line with as-grown CNT sheets, while preserving a
high conductivity during thermal and mechanical manipulations. The
developed CNT-PI films are hence rendered suitable as an alternative for
electrostatic discharge (ESD) protection layer in spacecraft thermal blanket,
given their durable exhibited character.
CHALLENGES RELATED TO CARBON NANOTUBE USAGE IN

AEROSPACE SCIENCES
A number of factors are essential in addressing the challenges currently

faced by the aerospace industry prior to full implementation of CNTs.
These factors consist of health and safety concerns, large scale production,
adequate quality at a low cost, achievement of optimized properties upon

CNT implementation, as discussed in further details elsewhere [12]. In
particular, it is essential for CNTs to be available at full large-scale
production levels, in order to meet the demands of the industry [87].
Current advancement of 3D printing may further open new avenues for
usage of CNTs in aerospace sciences, with aims to realize optimized
material properties at a low cost.
The most important aspect that needs to be explored prior to full-scale

implementation of CNTs in aerospace applications, is according to
Gohardani et al. [12], an adequate assessment of the health and safety
concerns related to these materials. Given that CNT exposure has exhibited
adverse health effects on animals [12], a complete health risk assessment is
mandatory prior to introduction of these materials in the aerospace industry.
Fig. (5). An overview of the benefits of using CNTs in commercial aircraft, military aircraft,
rotorcraft, and space applications. (Source: Adapted from [12]).
Table 3, provides an overview of some unexplored aspects and challenges

raised by Gohardani et al. [12], related to CNT implementation in
aeronautical and aerospace applications. Fig. (5), provides an overview of
the potential benefits of CNT usage in commercial aircraft, military aircraft,
rotorcraft, and space applications.
Table 3. Unexplored aspects related to CNT implementation in aeronautical and aerospace

engineering. (Source: Adapted from [12]).
Discipline CNT implementation caveats and unexplored aspects
 Does CNT implementation on the airframe affect
the respiratory systems of passenger and crew?
 Is CNT implementation safe, in terms of fire
retardance?
 What are the possible effects of CNT dispersion in
the atmosphere due to an accident or incident?
 Can composites reinforced with CNTs withstand
erosion for prolong time periods?
 Can aircraft manufacturers utilize CNTs in the
manufacturing process without health hazards to
workers arising from CNT exposure?
Aeronautical Engineering  Will accelerated laboratory testing of CNTs
represent actual flight conditions?
 How accurate can CNT implementation on
composite materials be simulated
computationally?
 Can the superior material properties of CNTs stand
the trial of time, or is there a grace period for these
materials in comparison to an estimated operating
period of a commercial airliner ~ 30 years?
 Are there any potential hazards with utilizing
CNTs in airborne applications in terms of
human/wild life exposure to nanomaterials?
 Will CNTs exhibit terrestrial-like properties in
Aerospace Engineering
space?
 How susceptible are CNT-manufactured parts to

solar radiation in space?
 Are there any health hazards or safety concerns
upon utilizing CNTs in space?
 Can the empirical space environments be
simulated in terrestrial laboratories?
 What methods of CNT exposure assessment will
be available upon full implementation of these
materials in aerospace sciences?
 Will a long-term exposure to an environment
where CNTs have been utilized as interior
materials, in conjunction with a pressurized cabin
and material degradation expose less desirable
aspects of CNTs?
 What climate change impacts are possible due to
the traverse of aircraft and spacecraft with CNT
reinforced fuselages?
 How well do CNTs serve as fire retardant
materials when implemented in aerospace
materials, in case of a fire onboard an aerospace
vehicle?
DISCUSSION AND FUTURE OUTLOOKS
Many aspects associated with the implementation of CNTs in aerospace

applications, still remain unexplored. Some of these have briefly been
highlighted in Table 3. The ambiguous health and safety effects and hazards,
in conjunction with long time exposure effects on humans, necessitates
further research, prior to a well-deifned implementation of CNTs in.
aerospace applications. Based on the conducted study, it is determined that
certiifcation of.new products and technologies for the aerospace industry
featuring CNTs, will be prolonged due to challenges, such as those
associated with health and safety originating from the novelty of these
materials. Once recognized standards and a certiifcation fra. ework for
aeronautics and astronautics have been established, future accreditation of

systems involving CNTs for aerospace applications will be facilitated.
A glance on the covered topics in this chapter portrays a multifaceted image

of CNTs in aerospace applications and their potential revolutionary benefits.
Equally, the prospective implementation of CNTs in aerospace sciences
unwraps multi-layered terrestrial and celestial health and safety concerns,
and performance evaluations during extended time periods. The limited
current existing applications of CNTs in aerospace sciences may in addition
also change with new breakthroughs in nanotechnology and in-depth
understanding of these novel materials.
this publication.
ACKNOWLEDGEMENTS
Any opinions, findings and conclusions expressed in this article are those of
the author only and do not necessarily reflect the corporate views of any
organization(s) affiliated with the author. Furthermore, the author is not
aware of any affiliations, memberships, funding, or financial holdings that
might be perceived as affecting the objectivity of this study.
REFERENCES
[1] J. Gooding, "Nanostructuring electrodes with carbon nanotubes: A review on
electrochemistry and applications for sensing", Electrochim. Acta, vol. 50, no.
15, pp. 3049-3060, 2005.
[http://dx.doi.org/10.1016/j.electacta.2004.08.052]
[2] "U. Khan, Y. Gun’ko, “Mechanical reinforcement of polymers using carbon

nanotubes", Adv Mater., vol. 18, no. 6, pp. 689-706, 2006.
[http://dx.doi.org/10.1002/adma.200501851]
[3] A. Dillon, K. Jones, T. Bekkedahl, C. Kiang, and D. Bethune, "M. Heben M,

“Storage of hydrogen in single-walled carbon nanotubes", Nature, vol. 386,
pp. 377-379, 1997.
[http://dx.doi.org/10.1038/386377a0]
[4] G. Che, B. Lakshmi, E. Fisher, and C. Martin, "Carbon nanotubule

membranes for electrochemical energy storage and production", Nature, vol.
393, no. 6683, pp. 346-349, 1998.
[http://dx.doi.org/10.1038/30694]
[5] A. Bianco, K. Kostarelos, and M. Prato, "Applications of carbon nanotubes in

drug delivery", Curr. Opin. Chem. Biol., vol. 9, no. 6, pp. 674-679, 2005.
[http://dx.doi.org/10.1016/j.cbpa.2005.10.005] [PMID: 16233988]
[6] N.W. Kam, M. O’Connell, J.A. Wisdom, H. Dai, and H. Dai, "Carbon
nanotubes as multifunctional biological transporters and near-infrared agents
for selective cancer cell destruction", Proc. Natl. Acad. Sci. USA, vol. 102, no.
33, pp. 11600-11605, 2005.
[http://dx.doi.org/10.1073/pnas.0502680102] [PMID: 16087878]
[7] S. Fan, M.G. Chapline, N.R. Franklin, T.W. Tombler, A.M. Cassell, and H.
Dai, "Self-oriented regular arrays of carbon nanotubes and their field emission
properties", Science, vol. 283, no. 5401, pp. 512- 514, 1999.
[http://dx.doi.org/10.1126/science.283.5401.512] [PMID: 9915692]
[8] W. De Heer, A. Chatelain, and D. Ugarte, “A carbon nanotube

field-emission electron source", Science, vol. 270, no. 5236, pp. 1179-1180,
1995.
[http://dx.doi.org/10.1126/science.270.5239.1179]
[9] V. Popov, "Carbon nanotubes: properties and application", Mater. Sci. Eng.
Rep., vol. 43, no. 3, pp. 61-102, 2004.
[http://dx.doi.org/10.1016/j.mser.2003.10.001]
[10] R.H. Baughman, A.A. Zakhidov, and W.A. de Heer, "Carbon nanotubes--the
route toward applications", Science, vol. 297, no. 5582, pp. 787-792,
2002.
[http://dx.doi.org/10.1126/science.1060928] [PMID: 12161643]
[11] M. Dresselhaus, and M. Endo, "Relation of carbon nanotubes to other carbon

materials". In: Carbon nanotubes, vol. 80, pp. 11-28, 2001.
[http://dx.doi.org/10.1007/3-540-39947-X_2]
[12] O. Gohardani, M. Chapartegui Elola, and C. Elizetxea, "Potential and

prospective implementation of carbon nanotubes on next generation aircraft
and space vehicles: A review of current and expected applications in aerospace
sciences", Prog. Aerosp. Sci., vol. 70, pp. 42-68, 2014.
[http://dx.doi.org/10.1016/j.paerosci.2014.05.002]
[13] J. Baur, and E. Silverman, "Challenges and opportunities in multifunctional

nano-composite structures for aerospace applications", MRS Bull., vol. 32, no.
4, pp. 328-334, 2007. [http://dx.doi.org/10.1557/mrs2007.231]
[14] M. Al-Saleh, and U. Sundararaj, "A review of vapor grown carbon nanofiber/
polymer conductive composites", Carbon, vol. 47, pp. 2-22, 2009.
[http://dx.doi.org/10.1016/j.carbon.2008.09.039]
[15] M. Al-Saleh, and U. Sundararaj, "Review of the mechanical properties of

carbon nanofiber/polymer composites", Compos., Part A Appl. Sci. Manuf., vol.
42, pp. 2126-2142, 2011.
[http://dx.doi.org/10.1016/j.compositesa. 2011.08.005]
[16] P.M. Ajayan, "Nanotubes from Carbon", Chem. Rev., vol. 99, no. 7, pp.
1787-1800, 1999.
|[http://dx.doi.org/10.1021/cr970102g] [PMID: 11849010]
[17] E. Thostenson, Z. Ren, and T. Chou, "Advances in the science and technology
of carbon nanotubes and their composites: A review", Compos. Sci. Technol.,
vol. 61, pp. 1899-1912, 2001.
[http://dx.doi.org/10.1016/S0266-3538(01)00094-X]
[18] P. Ehrburger, "Carbon nanotubes: the present state", Carbon, vol. 40, no. 10,
pp. 873-1006, 2002.
[19] F. Komarov, and A. Mironov, "Carbon nanotubes: present and future", Phys
Chem Solid State, vol. 5, no. 3, pp. 411-429, 2004.
[20] L. Radushkevich, and V. Lukyanovich, “O strukture ugleroda,

obrazujucegosja pritermiceskom razlozenii okisi ugleroda na zeleznom
kontakte", Zurn Fisic Chim, vol. 26, pp. 88-95, 1952.
[21] A. Oberlin, M. Endo, and T. Koyama, "Filamentous growth of

carbon through benzene decomposition", J. Cryst. Growth, vol. 32, no.
3, pp. 335-349, 1976.
[http://dx.doi.org/10.1016/0022-0248(76)90115-9]
[22] R. Bacon, and J. Bowman, "Production and properties of graphite whiskers",

Bull. Am. Phys. Soc., vol. 2, p. 131, 1957.
[23] "N. Lange, “The structure of graphite filaments", Z. Kristallogr., vol. 111, pp.
24-34, 1958.
[24] H. Boehm, "Carbon from carbon monoxide disproportionation on

nickel and iron catalysts: morphological studies and possible growth
mechanisms", Carbon, vol. 11, pp. 583-590, 1973.
[http://dx.doi.org/10.1016/0008-6223(73)90323-0]
[25] R. Baker, P. Harris, R. Thomas, and R. Waite, "Formation of filamentous

carbon from iron, cobalt and chromium catalyzed decomposition of acetylene",
J. Catal., vol. 30, pp. 86-95, 1973.
[http://dx.doi.org/10.1016/0021-9517(73)90055-9]
[26] S. Iijima, "Helical microtubules of graphitic carbon", Nature, vol. 354, pp.
56-58, 1991. [http://dx.doi.org/10.1038/354056a0]
[27] M. Monthioux, and V. Kuznetsov, "Who should be given the credit for the
discovery of carbon nanotubes?", Carbon, vol. 44, pp. 1621-1623, 2006.
[28] M. Meador, B. Files, J. Li, H. Manohara, D. Powell, and E. Siochi,

“Nanotechnology Roadmap Technology Area 10”, Technical report. National

Aeronautics and Space Administration.. NASA, 2012.
[29] A.S. Gohardani, G. Doulgeris, and R. Singh, "Challenges of future aircraft

propulsion: A review of distributed propulsion technology and its potential
application for the all electric commercial aircraft", Prog. Aerosp. Sci., vol. 47,
no. 5, pp. 369-391, 2011.
[30] A.S. Gohardani, “Distributed propulsion technology", In: Mechanical

Engineering theory and applications. Nova Science Publishers: New York,
USA, 2014.
[31] A.S. Gohardani, "A synergistic glance at the prospects of distributed

propulsion technology and the electric aircraft concept for future unmanned air
vehicles and commercial/military aviation", Prog. Aerosp. Sci., vol. 57, pp.
25-70, 2013.
[32] J. Luo, and I. Daniel, “Characterization and modeling of mechanical

behavior of polymer/clay nanocomposites", Compos. Sci. Technol., vol. 63,
no. 11, pp. 1607-1616, 2003.
[http://dx.doi.org/10.1016/S0266-3538(03)00060-5]
[33] S. O’Donnell, "Impact of nanomaterials in airframes on commercial aviation",

In: AIAA third annual aviation technology, integration, and operations (ATIO)
technology conference. American Institute of Aeronautics and Astronautics,
Denver, CO, AIAA-2003-6774, 2003.
[34] S. O’Donnell, and D. Smith, Potential impact of carbon nanotube reinforced

polymer composite on commercial heavy aircraft. Technical report. MITRE
Corporation, Project no. 02MSR055-F3, 2004.
[35] S. O’Donnell, K. Sprong, and B. Haltli, Potential impact of carbon nanotube

reinforced polymer composite on commercial aircraft performance and
economics, In: AIAA fourth aviation technology, integration and operations
(ATIO) forum, Chicago, IL, Project no. 02MSR055-F3, 2004.
[36] M. Roco, C. Mirkin, and M. Hersam, "Nanotechnology research directions for

societal needs in 2020: summary of international study", J. Nanopart. Res.,
vol. 13, pp. 897-919, 2011.
[http://dx.doi.org/10.1007/s11051-011-0275-5]
[37] R. Van Noorden, "Chemistry: The trials of new carbon", Nature, vol. 469, no.
7328, pp. 14-16, 2011.
[http://dx.doi.org/10.1038/469014a] [PMID: 21209637]
[38] T. Kukowski, "Lightweight CNT cables for aerospace", In: NESC NDE
technology assessment NASA NDI workshop. Johnson Space Center Houston:
TX, 2012.
[39] E. McKenna, "Product focus: wire and cable", Avionics Magazine.

Available from: http://www.aviationtoday.com/av/military/Product-Focus-
Wire-and-Cable-79536.html, 2013.
[40] A. Drake, "NASA environmentally responsible aviation (ERA) N + 2

advanced vehicle study", In: 50th Aerospace Sciences meeting including the
new horizons forum and aerospace exposition, 2012 Nashville, TN, USA
[41] O. Gohardani, The influence of erosion and wear on the accretion and
adhesion of ice for nano reinforced polymeric composites used in aeronautics,
Ph.D. thesis, Cranifeld University, United Kingdom, 2011�
[42] A. Amendola, and G. Mingione, "On the problem of icing for modern civil
aircraft", Air Space Eur.,vol. 3, no. 3–4, pp. 214-217, 2001.
[http://dx.doi.org/10.1016/S1290-0958(01)90098-X]
[43] H. Chu, Z. Zhang, Y. Liu, and J. Leng, "Self-heating fiber reinforced polymer
composite using meso/macropore carbon nanotube paper and its application in
deicing", Carbon, vol. 66, pp. 154-163, 2014.
[44] W. Zhao, M. Li, Z. Zhang, and H. Peng, "Carbon nanotube based composites
film heater for de-icing application", In: 14th European conference on
composite materials, Budapest, Hungary, Paper ID: 545-ECCM14, 2010.
[45] J. Felder, H. Kim, and G. Brown, "Turboelectric distributed propulsion engine

cycle analysis for hybrid-wing-body aircraft". In: 47th AIAA Aerospace
Sciences Meeting, Orlando, FL, January, pp. 5-8. 2009.
[http://dx.doi.org/10.2514/6.2009-1132]
[46] B. Ruff, W. Li, R. Venkatasubramanian, D. Mast, A. Sowani, and M. Schulz,

"Development of lightweight sustainable electric motors", In: Nanotube
superi�b er material�, pp. 595–626, 2014.
[http://dx.doi.org/10.1016/B978-1-4557-7863-8.00021-9]
[47] J. Gou, Y. Tang, F. Liang, Z. Zhao, D. Firsich, and J. Fielding, "Carbon

nanofiber paper for lightning strike protection of composite materials",
Compos., Part B Eng., vol. 41, no. 2, pp. 192-198, 2010.
[http://dx.doi.org/10.1016/j.compositesb.2009.06.009]
[48] F. Fisher, J. Plumer, and R. Perala, "Aircraft lightning protection handbook",

Technical report. Lightning Technologies Inc, DOT/FAA/CT-89/22, 1989.
[49] E. Rupke, "Lightning direct effects handbook", Technical report. Lightning

Technologies Inc, AGATE-WP3.1-031027-043-Design guideline, 2002.
[50] C. Wu, H. Lu, Y. Liu, and J. Leng, "Study of carbon nanotubes/short carbon
fiber nanocomposites for lightning strike protection", Proceedings of the SPIE
7644, behavior and mechanics of multifunctional materials and composites,
2010
[51] S. Wang, D. Chung, and J. Chung, "Self-sensing of damage in carbon fiber

polymer–matrix composite by measurement of the electrical resistance or
potential away from the damaged region", J. Mater. Sci., vol. 40, no. 24, pp.
6463-6472, 2005.
[http://dx.doi.org/10.1007/s10853-005-1708-2]
[52] D. Chung, "Damage detection using self-sensing concepts", J. Aerosp. Eng.,

vol. 221, no. G4, pp. 509- 520, 2007. [Proc Inst Mech Eng Part G].
[53] A. Sehra, and J. Shin, "Revolutionary propulsion systems for 21st century
aviation", Technical report. National Aeronautics and Space Administration
(NASA), NASA/TM-2003-212615, 2003.
[54] P. Kim, and D. Lee, "Composite sandwich constructions for absorbing the
electromagnetic waves", Compos. Struct., vol. 87, no. 2, pp. 161-167, 2008.
[55] K. Vinoy, and R. Jha, Radar absorbing materials: from theory to design and
characterization. Vol. 1. Kluwer Academic Publishers London: United
Kingdom, 1996.
[http://dx.doi.org/10.1007/978-1-4613-0473-9]
[56] B. Rosenberg, "Product focus: wire and cable", Avionics Magazine. Available
from: http://www.aviationtoday.com/av/military/Product-Focus-Wire-and-
Cable-79536.html, 2013.
[57] A. Liu, C. Bakis, and K. Wang, "Multiscale damping model for polymeric
composites containing carbon nanotube ropes", J. Compos. Mater., vol. 44, pp.
2301-2323, 2010.
[http://dx.doi.org/10.1177/0021998310365176]
[58] C. Bakis, and K. Wang, "Structural damping and health monitoring

enhancement via multifunctional carbon nanotube-based composites tailoring",
In: Technical report. Pennsylvania State University, 2011.
[59] O. Gohardani, "Impact of erosion testing aspects on current

and future flight conditions", Prog. Aerosp. Sci., vol. 47, no. 4, pp.
280-303, 2011. [http://dx.doi.org/10.1016/j.paerosci.2011.04.001]
[60] O. Gohardani, D.M. Williamson, and D.W. Hammond, "Multiple liquid

impacts on polymeric matrix composites reinforced with carbon nanotubes",
Wear, vol. 294, pp. 336-346, 2012.
[http://dx.doi.org/10.1016/j.wear.2012.07.007]
[61] A.S. Gohardani, and O. Gohardani, "Ceramic engine considerations for future
aerospace propulsion", Aircr. Eng. Aerosp. Technol., vol. 84, no. 2, pp. 75-86,
2012.
[http://dx.doi.org/10.1108/00022661211207884]
[62] O. Gohardani, and D.W. Hammond, "Dynamic and static wettability of

advanced materials used in aeronautical applications", Int. J. Comput.
Methods Exp. Meas, vol. 1, no. 3, pp. 298-320, 2013.

[http://dx.doi.org/10.2495/CMEM-V1-N3-298-320]
[63] O. Gohardani, and D.W. Hammond, "Ice adhesion to pristine and eroded
polymer matrix composites reinforced with carbon nanotubes for potential
usage on future aircraft", Cold Reg. Sci. Technol., vol. 96, pp. 8-16, 2013.
[http://dx.doi.org/10.1016/j.coldregions.2013.09.005]
[64] O. Gohardani, "The exploration of icephobic materials and their future

prospects in aircraft icing applications", J Aeronaut Aerospace Eng, vol. 1, no.
e116, 2012.
[65] J. Thomas, and M. Qidwai, "The design and application of multifunctional

structure- battery material systems", JOM, vol. 57, no. 3, pp. 18-24, 2005.
[http://dx.doi.org/10.1007/s11837-005-0228-5]
[66] L. Nascimento, "UAV lessons learned from OPERATION ALLIEDFORCE",

Technical report. Center for Army Lessons Learned, January- February, 2000.
[67] "Unmanned aerial vehicle reliability study", Technical report. Ofifce of the.
Secretary of Defense, 2003.
[68] L. Haslim, and R. Lee, Electro-expulsive separation system, Technical US

8,336,830 B2, 1985.
[69] "Battelle’s anti-icing system for remotely piloted aircraft (RPAs)", Available
from: http://www.battelle.org, [accessed 15.09.13].
[70] K.L. Sørensen, A.S. Helland, and T.A. Johansen, "Carbon nanomaterial- based
wing temperature control system for in-flight anti-icing and de-icing of
unmanned aerial vehicles", IEEE Aerospace Conference, Big Sky, 2015.
[http://dx.doi.org/10.1109/AERO.2015.7119206]
[71] K. Tsiolkovsky, Grezy o Zemle i Nebe (i) Na Veste (Speculations about Earth
and Sky and on Vesta). Izd-vo AN SSSR: Moscow, 1895.
[72] Y. Artsutanov, V kosmos na Elektrovoze, Komosomolskava Pravda. The

contents described in English by Lvov in Science, vol. 158, pp. 946-947, 1967.
[73] J. Pearson, "The orbital tower: a spacecraft launcher using the

earth's rotational energy", Acta Astronaut., vol. 2, pp. 785-799, 1975.
[http://dx.doi.org/10.1016/0094-5765(75)90021-1]
[74] B. Edwards, "Design and deployment of a space elevator", Acta Astronaut., vol.
47, no. 10, pp. 735- 744, 2000.
[http://dx.doi.org/10.1016/S0094-5765(00)00111-9]
[75] D. Smitherman Jr, "Tether G. Space elevators", NASA. Technical report.

NASA, NASA/CP-2000-210429, 2000.
[76] A. Bolonkin, "Non-rocket earth-moon transport system", Adv. Space Res., vol.
31, no. 11, pp. 2485- 2490, 2003.
[http://dx.doi.org/10.1016/S0273-1177(03)00565-9]
[77] N. Pugno, "On the strength of the carbon nanotube-based space elevator cable:
from nanomechanics to megamechanics", J. Phys. Condens. Matter, vol. 18,
no. 33, pp. S1971-S1990, 2006.
[http://dx.doi.org/10.1088/0953-8984/18/33/S14]
[78] J.M. Wörle-Knirsch, K. Pulskamp, and H.F. Krug, "Oops they did it again!
Carbon nanotubes hoax scientists in viability assays", Nano Lett., vol. 6, no. 6,
pp. 1261-1268, 2006.
[http://dx.doi.org/10.1021/nl060177c] [PMID: 16771591]
[79] B. Iannotta, "Nanotubes lift hope for space elevator", Aerosp. Am., vol. 44, no.
3, pp. 30-35, 2006.
[80] M. Avnet, "The space elevator in the context of current space exploration
policy", Space Policy, vol. 22, no. 2, pp. 133-139, 2006.
[http://dx.doi.org/10.1016/j.spacepol.2006.02.005]
[81] J. Dempsey, System and method for space elevator, US patent no. US
6,981,674 B1, 2006.
[82] A. Ketsdever, M. Young, and J. Mossman, "Overview of advanced concepts

for space access", J. Spacecr. Rockets, vol. 47, no. 2, pp. 238-250, 2010.
[http://dx.doi.org/10.2514/1.46148]
[83] C. Harris, M. Shuart, and H. Gray, A survey of emerging materials for

revolutionary aerospace vehicle structures and propulsion systems, Technical
report National Aeronautics and Space Administration, ASA/TM-
2002-211664; 2002.
[84] S. Keller, and A. Zaghloul, "Carbon nanotube yarn and sheet antennas", In:
Nanotube superfiber materials – changing engineering design.. Elsevier,
William Andrew: Boston, 2014, pp. 749-787.
[http://dx.doi.org/10.1016/B978-1-4557-7863-8.00026-8]
[85] M. O’Rourke, G. Nelson, M. Waid, P. Moloney, P. Nikolaev, and E. Sosa,

“Electromagnetic interference shielding via carbon nanotubes”, Biennial
Research and Technology Development Report. NASA Johnson Space
Center, 2007, pp. 65-66.
[86] N. Atar, E. Grossman, I. Gouzman, A. Bolker, and Y. Hanein, "Reinforced

carbon nanotubes as electrically conducting and flexible films for space
applications", ACS Appl. Mater. Interfaces, vol. 6, no. 22, pp. 20400-20407,
2014.
[http://dx.doi.org/10.1021/am505811g] [PMID: 25366559]
[87] H. Tennent, J. Barber, and R. Hoch, Carbon fibrils, method for producing same
and adhesive compositions containing same, US patent no. 5,578,543, 1996.
CHAPTER 13
Combination of Carbon Fiber Sheet Molding

Compound and Pre-Impregnated, Tailored
Carbon Fiber Reinforcements
Marc Fette1,3,*, Nicole Stöß2, Jens Wulfsberg1, Axel Herrmann3,
Gerhard Ziegmann4 and Georg Lonsdorfer3
1
Institute of Production Engineering, Helmut Schmidt University, 22043
Hamburg, Germany
2
Polynt Composites Germany GmbH, Kieselstraße 2, 56357 Miehlen,
Germany
3
Composite Technology Center GmbH (An Airbus Company),
Airbusstraße 1, 21648 Stade, Germany
4
Institute of Polymer Materials and Plastics Engineering, TU Clausthal,
Agricolastrasse 6, 38678 Clausthal, Germany
Abstract: The current use of fuel efficient and environmentally friendly aircraft is
only possible by the development of innovative lightweight constructions and the use
of lightweight materials, such as carbon fiber reinforced plastics. With the rising
demand on fiber reinforced components in the aerospace industry new production
processes have been built up. However, current production technologies for
composites cause higher costs and obtain longer process cycle times in comparison
to the manufacturing processes of metals. Moreover raw materials, such as carbon
fibers and resin, and semi-finished products are very expensive. In contrast to this
and compared with other manufacturing technologies for fiber reinforced plastics
*Corresponding author Marc Fette: Institute of Production Engineering, Helmut Schmidt University,
Hamburg, Germany; E-mail: marc.fette@hsu-hh.de.
Yucheng Liu (Ed.)

498 Frontiers in Aerospace Science, Vol. 1 Fette et al.
Sheet Molding Compound compression processes are characterized by cost

efficiency, high productivity, the option of full automation and the possibility for
the realization of complex shapes and integrated functions. However there are also
some disadvantages like a low level of stiffness and strength in comparison to
continuous fiber reinforced plastics. The reasons for these facts are the short fibre
length, a lower fibre-volume fraction and an isotropic fibre distribution.
Consequently, the combination of sheet moulding compound and pre-impregnated,
tailored carbon fibre reinforcements in an one-shot compression moulding and
curing process merges the advantages of both groups of composite materials.
Therefore the creation of load-bearing, complex, functional and autoclave-quality
parts without an autoclave can be realised. In this chapter, this innovative
technology and its potentials are presented. This paper will also deal with the
resulting material characteristics.
Keywords: Aircraft, CFRP, Hybrid, Light weight, SMC.
NOMENCLATURE
ATH Aluminum hydroxide or alumina trihydrate

CFRP Carbon fiber reinforced plastic
FRP Fiber reinforced plastic
SEM Scanning electron microscope
SMC Sheet Molding Compound
UP Unsaturated polyester resin
INTRODUCTION
In the recent decades, the demand for efficient and fuel-saving aircraft for
the commercial air traffic has led to the development of innovative
lightweight solutions and to an increasing use of modern lightweight
Tailored Carbon Fiber Reinforcements Frontiers in Aerospace Science, Vol. 1 499
materials, such as carbon fiber reinforced plastics (CFRP). In this context

the aerospace industry developed new manufacturing processes and
enhanced existing technologies to produce such high-performance
materials. However, the today's production of aerospace components
made of FRP materials is time-consuming and costly in comparison to the
production of metallic constructions. Reasons for this are relatively high
costs of raw materials and semi-finished products, long processing times
and a partially low degree of automation. As a result there exist lots of
aspirations for the development of more efficient technologies, the
optimization of existing processes or the smart combination of different
technologies.
One promising technology is the combination of carbon fiber Sheet

Molding Compound (SMC) and pre-impregnated, continuous carbon fiber
reinforcements in a one-step compression and curing process. In this
context the SMC technology provides the implementation of light,
geometrically complex and highly functional composite components
considering short cycle times, an optimum usage of material, the
possibility of complete automation and high cost efficiency. The locally
integrated continuous carbon fiber reinforcements obtain tailored
mechanical properties depending on the load characteristics of the
respective component and with regard to the required high lightweight
potential. In addition, this hybrid composite technology promises the
possibility of using of recycled carbon fibers as long fiber reinforcements
in the SMC mass and the direct integration of metallic components. This
material and process combination can realize light weight, complex and
functional composite components for aerospace applications in a highly
efficient way.
The main objectives of this article are researches and analysis of the
potentials of this technology in reference to aerospace applications with
the requirements for secondary structures of commercial aircraft.
STATE OF THE ART
Today, some successful developments and research efforts in the field of

thermoplastic composites and hybrid material already exist. In this regard
the combination of injection molding and organo-sheets obtain the best
opportunities for hybrid thermoplastic composite materials. Over-molded
organo-sheets combine short and continuous carbon fiber reinforcements
to realize functional thermoplastic composite parts with complex shapes,
excellent mechanical performance, lightweight properties and complex
shapes. Appropriate components have recently appeared in electronics
and automotive industry. In addition, these over-molding technologies can
be extended by integrating metal components [1].
In contrast to these developments there are no comparable technologies

for thermoset composites and appropriate material combinations available
on the market. In general Sheet Molding Compounds are characterized by
lower mechanical properties than composites reinforced by continuous
fibers. The reasons are the random fiber distribution and lower fiber
volume contents. Therefore the mechanical properties of Sheet Molding
Compounds can be increased by using fiber reinforcements like
embedded unidirectional carbon fiber layers.
This kind of material is used for a few industrial applications, especially

in the automotive sector. Typical Sheet Molding Compound applications
for vehicles or trucks are fenders, hoods, deck lids, spoilers, scuttle panels,
floor panels, tailgates, wheel arches and many more [2, 3]. In contrast to
the mentioned applications Sheet Molding Compounds are not well
established in aerospace industries. Reasons are absent dimensioning
methods, purer mechanical properties, and lower weight-saving potentials
in comparison to typical aircraft materials. Nevertheless, there are several
developments on cabin, cargo and other secondary structure parts because

of the excellent fire safety properties, high flame retardancy and
high-quality surface finishing of some special Sheet Molding Compound
formulations based on unsaturated polyester resin. Some aerospace
applications for primary structure parts in the field of compression
molding, for example by using HexMC®, were already commercialized.
However, the combination of tailored, continuous carbon fiber
reinforcements and carbon fiber Sheet Molding Compounds with the
possibility of the direct integration of metal components represent a new
technology approach for the aerospace industry. Therefore, highly
integrated parts could be designed and produced by means of this hybrid
composite technology in a cost-efficient way [4-10].
AIM AND PURPOSE OF THE PRESENT WORK
The present work follows the intention to develop a new hybrid

composite technology for the production of future aircraft. This
technology shall include innovative material combinations, new
multi-material designs and an efficient production process. The
combination of carbon fiber Sheet Molding Compounds and
pre-impregnated, tailored carbon fiber fabrics in a one-step compression
and curing process permits the production of lightweight, load-bearing
and highly integrated aircraft parts, shown in Fig. (1). This technology
combines the flowability of long fiber reinforced Sheet Molding
Compounds and the tailored mechanical properties of continuous fiber
fabrics, which are impregnated by the thermoset matrix system. Such a
combination allows the production of components with high functionality,
complex shape, high freedom of design and excellent mechanical
properties. Therefore, the level of fiber content and mechanical
characteristics like flexural strength and tensile strength could be adapted
exactly to the load cases and the requirements of the appropriate part [11].
Fig. (1). Process cycle of the combination of pre-impregnated carbon fiber fabrics and carbon fiber
SMC.
In addition these thermoset composite combinations can be directly

combined with metallic components, e.g. inserts, nuts and plates,
according to the part design and the required material complexity in the
compression process. On the one hand these composite parts can be used
for load-bearing secondary structure components which should have a
complex shape or should include a lot of different functions, e.g.
functional cabin monuments, fittings to structural components,
attachments for system installations or complex housings. On the other
hand parts of the primary structure with a complex shape, an exact
adaption to the load cases and with integrated fittings, fasteners and
openings can be realized in one step. Moreover innovative grid structures
on inner shells of aircraft could be put into practice with this technology.
Besides the possibility of integral construction, tailored carbon fiber
reinforcements for load path optimized mechanical properties and a
reduction of adhesives, rivets, shim or additional fittings can minimize
weight, production times and costs of the parts. In addition, lower
material costs and a better buy-to-fly ratio, as a result of an optimized
material usage, are causing an additional decrease of part costs.
Furthermore the proposed hybrid manufacturing technology creates
higher productivity and lower production costs due to low cycle times,
less time for part assembly or finishing and the possibility of complete
automation. In this context such a compression process lead to less energy
consumption in aerospace production.
Possible materials for reinforcements are carbon fibers, glass fibers and
natural fibers which can be combined with different thermoset matrix
materials, for example unsaturated polyester resin, vinyl ester resin or
epoxy resin. Besides it is possible to use carbon fiber recyclates,
especially as long fiber reinforcements in SMC formulations. Using
recycled carbon fibers made out of end-of-life cycle aircraft parts or
cutting scrap of the current aerospace production can close the ecological
value chain referring to future aircraft. These fibers can be final products
from recycling processes like pyrolysis. These carbon fibers keep up to 95
percent of the original mechanical properties and obtain a lower price
level than new carbon fibers. In this context, this technology realizes the
reuse of expensive and energy-intensive resources as reinforcement
materials for aircraft parts again. Possible areas of application can be
cabin and cargo structures. As a result the ecological sustainability and the
resource efficiency of aerospace production for composite aircraft
components will be improved.
Therefore an interdisciplinary project for several investigations on

material combinations, process parameters and lightweight constructions
using the described technology has been started. The aim is the realization
of different cabin and cargo components of commercial aircraft. In this
context these secondary structure parts have to fulfil appropriate
aerospace requirements. To obtain some information about technical,
economic and ecological performance these components will be
compared according to relevant requirements with the currently
assembled ones. In addition a complete feasibility analysis shall relate this
production process to current composite technologies to highlight its
expected advantages for aerospace industries. Finally special simulation
methods and FEM-models will be developed for designing, dimensioning
and calculating hybrid composite aircraft parts which will be produced
with this combination technology.
In a first feasibility study the substitution of a cargo foot step for a current
commercial short and middle range aircraft, which is currently made of
titanium, has been analyzed. Lead time reduction, cost savings and weight
reduction as results of increased mechanical properties, especially tensile
strength, flexural strength, stiffness and impact strength, are the key
aspects for the planned optimization. Subsequently, the industrial
introduction of this manufacturing process and the substitution of other
aircraft components will follow. Therefore completely new possibilities
for aircraft production and a new generation of hybrid composite
structures for future aircraft will be created.
Experimental Procedures
In a cooperation project Polynt Composites Germany GmbH and Airbus

Operations GmbH have developed a special Sheet Molding Compound
formulation called HUP 27 GF25 (Table 1). It is based on an un-
saturated polyester resin (UP) and contains many additives, especially the
flame-retardant additive aluminum hydroxide or alumina trihydrate
(ATH). Because of a high percentage of flame-retardant components HUP
27 meet the requirements on fire safety and flame retardancy of the
aircraft cargo section. In this regard, aerospace materials have to pass the
following Airbus specific tests, which are required in the cargo area of
commercial aircraft:
x AITM 2.0002: Flammability vertical, 60 s test A;
x AITM 2.0007: Measurement on smoke density;
x AITM 3.0005: Determination of the toxic components on

combustion products and measurements on toxicity;
x AITM 2.0010: Flame penetration resistance of cargo compartment

liners.
Table 1. Material properties of HUP 27 determined by different investigations.
Material characteristics Method / Notation Value Unit
Surface weight 3000 g/m²

Physical properties
Specific density 1.93 g/cm³
Flexural modulus 9000 N/mm²
Flexural strength 150 N/mm²

Mechanical properties
Tensile strength 60 N/mm²
Impact strength 60 kJ/m²
Fiber content 25 %
Glass fiber reinforcement
Fiber length 25 mm
UP thermoset resin system 27 %
Flame-retardant components 66 %
Thermoset matrix components Curing system and process
6 %
additives
Thickening agents 1 %
Curing time 90 sec.
Molding time 25 sec./mm
Processing conditions Molding pressure 100 bar
Molding temperature 145 °C
Shrinkage 0.08 %
Further processing additives are necessary, because the flame-retardant

formulations, compared to the standard formulations with 150 phr
non-volatiles, contain a high amount of non-organic materials. The

proportions of non-organic components are a big challenge in comparison
to the organic ones for the Sheet Molding Compound processing. In
general the SMC compression process is configured for viscosities from
lowest level 10.000 mPas to 50.000 mPas. Without any process additives
it is not possible to produce a homogenous paste for a stable process. It is
not allowed to influence the burning behavior negatively by the use of
such additives. To obtain better impregnation of the carbon fiber
reinforcements the viscosity has to be reduced. The building of the
reinforcement works if the matrix-reinforcement-bonding is adequate to
transfer the mechanical properties on the complete Sheet Molding
Compound formulation. Otherwise the mechanical improvement is not
given by such a system. The flowability of the system is an important
parameter for the impregnation of the complete system and the orientation
of the fibers within the compression process. In the first step different
material combinations are tested to receive basic material characteristics
in reference to flexural modulus, flexural strength, tensile modulus,
tensile strength and impact strength. In this case nine reasonable material
systems have been analyzed due to different possibilities of
reinforcements, listed in Table 2.
Table 2. Material combination matrix (GF…glass fiber; CF…carbon fiber).
HUP 27- HUP 27- HUP 27- HUP 27- HUP 27-
GF, fiber GF, fiber CF, fiber CF, woven CF,
Inner layer content content content fabric unidirect-i
25%, 50%, 50%, onal fiber
Outer layer chopped chopped chopped reinforce-
fibers fibers fibers ment
HUP 27- GF, fiber

content 25%, chopped X X
X
fibers
7DEOH2 FRQWG
HUP 27- GF, fiber

X
fibers
HUP 27- CF, fiber
X
fibers
The materials with chopped fiber reinforcements are used as inert systems
for mechanical data base and for the outer systems. The materials with
continuous fiber fabrics or unidirectional fibers are used for the inner
layer, illustrated in Fig. (2). The mechanical properties determined for flat
Sheet Molding Compound plates are molded according to DIN EN 14598.
These plates have a thickness between 2.0 mm and 3.0 mm. The test
specimens are cut from the non-flow area of a sample plate with the
dimension of 250 mm x 120 mm and with molding conditions of 180
seconds at 145 °C. Afterwards the cut specimen is milled to the accurate
dimension according to the standard specification.
The Sheet Molding Compound formulations for these different material

samples were produced on a laboratory Sheet Molding Compound
impregnation machine with a material width of 500 mm and in a
continuous production step. This lab scale machine allows several manual
adjustments and the usage of different reinforcements. For investigations
of the mechanical properties of each material combination the following
investigation methods were used:
x Flexural modulus and flexural strength according to DIN EN ISO

14125:1998 + AC: 2002 + A1:2011.
x Tensile modulus and tensile strength according to DIN EN ISO
527-4:1997.
x Tensile modulus and tensile strength according to DIN EN ISO

527-5:1997.
x Impact strength according to DIN EN ISO 179-1:2010.
These tests illustrate a first tendency in reference to the improvement of

the mechanical properties as a result of the integration of tailored,
continuous fiber reinforcements and exact adaption to the load cases.
Fig. (2). Schematic illustration of outer and inner layers of the material samples.
DISCUSSION ON RESULTS
The results generated by the experimental investigations on the different
material combinations are shown in Figs. (3 to 5). The parameters
investigated are flexural modulus, flexural strength, tensile modulus,
tensile strength and impact strength. There are significant differences
between the measurements, especially in reference to the used type of
fiber and depending on the combination of the different fiber
reinforcements.
In general, the measurements indicate that higher fiber contents and the
use of chopped carbon fibers, instead of glass fibers, for Sheet Molding
Compound formulations can realize an improvement on flexural and
tensile modulus, illustrated in Fig. (3). Furthermore woven fabrics and
unidirectional layers of carbon fiber combined with Sheet Molding
Compounds tend to result in an increase of the mentioned mechanical
properties. Especially, HUP 27 with chopped carbon fibers (50 percent
fiber content) combined with an unidirectional carbon fiber layer or a
carbon fiber woven fabric can realize a significant improvement of
flexural and tensile modulus. In contrast, Fig. (4) shows that the flexural
strength is hardly influenced by additional and oriented reinforcements.

The reason for this could be the chemical adaption of HUP 27 to glass
fibers. As a consequence, HUP 27 samples with carbon fiber
reinforcements are partially delaminating. This is caused by poor
interlaminar shear strength behavior. Furthermore, the high percentage of
flame-retardant components lowers the fiber content, downgrades the
impregnation and leads to the degradation of the interlaminar connection
between the carbon fibers and the matrix material (Fig. 6).
Fig. (3). Flexural and tensile modulus for different material combinations.
In general the tensile strength can be increased by additional continuous

fiber reinforcements, shown in Fig. (4). In this regard, the integration of
unidirectional layers of carbon fiber doubles the tensile strength in
comparison to the original Sheet Molding Compound formulations. In
opposition to that the combination of HUP 27 and carbon fiber fabrics
realizes just a small improvement of the tensile strength.
Fig. (4). Flexural and tensile strength for different material combinations.
The impact strength is positively affected by material combinations of

glass fiber reinforced HUP 27 and additional carbon fiber reinforcements,
especially woven fabrics of carbon fiber, illustrated in Fig. (5). Due to the
poor interlaminar shear strength between carbon fibers and HUP 27 there
is no improvement of the impact behavior. In the case of HUP 27-CF 50%
combined with carbon fiber fabrics the impact strength is even reduced.
Fig. (5). Impact strength for different material combinations.

All in all a positive trend of the mechanical properties by the combination

of Sheet Molding Compounds and continuous fiber reinforcements can be
stated. For further optimization of the material characteristics the Sheet
Molding Compound formulations have to be adapted to the particular
fiber reinforcements. As a result, the interlaminar shear strength will be
improved. Furthermore, Sheet Molding Compound materials have to be
optimized for the impregnation of the reinforcements and to increase the
flowability of the Sheet Molding Compound material during the molding
process. The stacking of the layers, the fixing of the continuous fiber
reinforcement during the molding process and the mechanical
requirements of the components are additional topics being worth to be
investigated in order to improve the mechanical properties. All things
considered, the key aspect for realizing a hybrid composite component is
the exact adaption of Sheet Molding Compound formulation, oriented
fiber reinforcements and part requirements, such as mechanical properties
or flame redardancy.
Fig. (6). SEM pictures of the combination of carbon fiber SMC and unidirectional carbon fiber
fabrics.
CONCLUSION
Based on the outcomes of this study, it is expected that the development

of the combination of SMC and pre-impregnated, orientated carbon fiber
fabrics is auspicious referring to an improvement of mechanical
properties and lightweight potentials. This technology promises optimized
lightweight and load-bearing properties in connection with complex
shapes and design freedom. However, for aerospace applications there
does not exist a special Sheet Molding Compound formulation yet.
Consequently, each material combination has to be developed for or
adapted to the requirements of the particular application. In this regard the
design and geometry of the components are important, too. For the design,
the dimensioning and the simulation special material models and finite
element method simulation tools have to be developed. In reference to the
process the pre-forming, the preheating, the positioning and the fixing of
the wrought materials, especially the orientated carbon fiber fabrics, have
to be considered. Further studies will focus on the analysis and
optimization of process parameters such as flowability, molding
temperature, molding pressure and molding time. In addition the
integration of metal components such as inserts, fasteners or panels will
be realized. For the industrialization of the process an automation concept
with different batch sizes and part designs has to be evaluated in regard to
the technological, economic and ecological aspects. Although there are
many open questions, this innovative, hybrid composite technology offers
outstanding benefits, extended potentials and a lot of applications within
the aerospace and also other industries.
this publication.
ACKNOWLEDGEMENTS
Declared none.
REFERENCES
[1] J.E. Composites, "Over-moulded organo sheets", JEC Composites, vol. 85, pp.
22-27, 2013.
[2] M. Cabrera-Rios, and J.M. Castro, An economical way of using carbon fibers
in sheet molding compound compression molding for automotive applications.
Polymer Composites 27, vol. 6, pp. 718-722, 2006.
[3] H.G. Kia, Sheet moulding compounds - science and technology. Hanser /
Gardener Publications, Inc., Ohio, USA.
[4] P. Stachel, "Carbon fibre reinforced SMC for automotive applications",
Proceedings of the 5th Automotive Seminar, Topic: SMC/BMC - New
challenges in Automotive, Landshut, 2006.
[6] C. Lauter, T. Tröster, B. Sköck-Hartmann, T. Gries, and M. Linke, Höchst
feste Multimaterialsysteme aus Stahl und Faserverbundkunststoffen. VDI
Konstruktion: Ingenieur-Werkstoffe, Springer-VDI-Verlag, Düsseldorf, vol.
11-12/2010, pp. IW 8 - 9, 2010.
[7] M. Oldenbo, S.P. Fernberg, and L.A. Berglund, "Mechanical behaviour of
SMC composites with toughening and low density additives", Compos., Part
A Appl. Sci. Manuf., vol. 34, pp. 875-885, 2003.
[http://dx.doi.org/10.1016/S1359-835X(03)00155-6]
[8] S. Pimenta, and S.T. Pinho, "Recycling carbon fibre reinforced polymers for
structural applications: technology review and market outlook", Waste Manag.,
vol. 31, no. 2, pp. 378-392, 2011.
[http://dx.doi.org/10.1016/j.wasman.2010.09.019] [PMID: 20980138]
[9] E. Reuther, "Kohlefaser SMC für Strukturteile", In: Proceedings of the 7th
Internationale AVK-TV Tagung, A6-1 – A6-6, 2004.
[10] K. Steinbach, G.P. Ehnert, and K. Bieniek, "Neue Entwicklungen zur
Erhöhung der Festigkeits- und Steifigkeitseigenschaften von SMC für

belastbare Formteile", Proceedings of the 24th AVK-Tagung, Berlin, 1991.
[11] M. Fette, J. Wulfsberg, A. Herrmann, G. Lonsdorfer, and H. Seegel,
“Combination of sheet moulding compound and prepreg compression
moulding: A technology concept for complex hybrid composites”, DGM
Konferenz Euro Hybrid 2014, Stade, 2014.
CHAPTER 14
Design Optimization of Variable Stiffness

Composite Structures for Aerospace
Applications
Mohammad Rouhi*, Hossein Ghayoor, Suong V. Hoa and Mehdi
Hojjati
Concordia Center for Composites, Department of Mechanical and

Industrial Engineering, Concordia University, Montreal, Quebec, H3G 1M,
Canada
Abstract: Automated fiber placement (AFP) machines can steer the

fibers/tows to make the so-called variable stiffness (VS) composites. They
allow the designers to fully exploit the directional properties of composite
materials to tailor the internal load distribution and improve the structural
performance. VS composites have been shown to be very promising in
the design optimization of composite panels and shells for buckling and
post-buckling performance and consequently for further reducing the
mass of future aerospace structures. In this chapter, the buckling
performance improvement of VS composite cylinders with circular and
elliptical cross sections is investigated. A metamodeling based design
optimization (MBDO) method is presented to maximize the buckling
performance of VS composite cylinders compared with their constant
*Corresponding author Mohammad Rouhi: Concordia Center for Composites, Department of Mechanical and
Industrial Engineering, Concordia University, Montreal, Quebec, H3G 1M8 Canada; Tel/Fax: 514 - 848 4596;
E-mail: m.rouhi@gmail.com
Yucheng Liu (Ed.)

516 Frontiers in Aerospace Science, Vol. 1 Rouhi et al.
stiffness (CS) designs. The structural improvement mechanism via

stiffness tailoring in a VS composite cylinder is also presented and
discussed. The effects of different parameters including the cylinders’
aspect ratio and size as well as the percentage of the steered plies in the
laminate are also investigated.
Keywords: Automated fiber placement, Buckling, Composite cylinder,

Fiber steering, Optimization, Variable stiffness.
INTRODUCTION
Lightweight and high performance are essential requirements for

aerospace structures. In case of using metallic materials, designers put all
their effort to reduce the weight and optimize the structural performance
by tailoring the topology and shape of the structure i.e. thickness, size, and
overall profile/shape. Over the past few decades, the application of
fiber-reinforced polymer composite (FRPC) materials in aerospace
structures has been increased dramatically. This growth has primarily been
fueled by the high stiffness- and strength-to-weight ratios of FRPC
materials as well as tailorability of the stiffness and strength properties and
substantial reduction in part count offered by such material systems.
Unlike metallic structures, composite structures can be manufactured in
very complex geometric shapes. Consequently, FRPC materials can help
enhance both product performance and manufacturing.
FRPC materials have traditionally been designed and manufactured as

multi-ply laminates consisting of several unidirectional layers. The layers
are either stacked in dry form with the polymer resin injected into the
mold or the layers are pre-impregnated with the resin material prior to
manufacturing. In either case, for product design and manufacturing
simplicity, the fiber orientation angle in each layer is typically held fixed
and usually limited to 0°, 90°, and ±45°. By limiting each layer to a single
orientation angle over the entire structural component, the designer is
Variable Stiffness Composite Structures Frontiers in Aerospace Science, Vol. 1 517
unable to fully exploit the directional material properties offered by

composite layers. On the other hand, without sophisticated manufacturing
equipment, it would be nearly impossible to accommodate spatial
variation of orientation angles in the individual plies. With the advent of
Automated Fiber Placement (AFP) machines, it has become possible to
steer the fibers to manufacture composite parts with continuously varying
fiber orientation angles. The resulting so-called variable stiffness (VS)
laminate has spatial stiffness properties that allow the full potential of
composite materials to be harnessed by extending the design space to
create structural components with significantly higher performance and/or
lower weight [1-5]. However, there are several design and manufacturing
challenges that need to be addressed to reach the full potential of VS
composite structures made by fiber steering.
Early research works on AFP technology were reported in the literature in

the late 1980’s [6]. Followed by building composite stiffened panels by
AFP machines [7, 8] in the early 1990’s, AFP technology was used in
production of numerous aircraft parts such as the F/A-18E/F horizontal
stabilizer skins, the Bell/Boeing V-22 Osprey aft fuselage [9, 10], the V-22
grip [11-13], the Boeing JSF inlet duct and the C17 landing gear pod
fairings and engine nacelle doors [9,14], and the fuselage sections of
Raytheon Premier I and Hawker Horizon business jets [15]. The recent
application of AFP technology includes the large scale production of the
fuselage sections of the Boeing 787 Dreamliner, Airbus A380 and A350
XWB [16, 17]. The overall saved labor costs and reduced scrap materials,
combined with the ability to produce laminated composite components of
exceptional quality with higher accuracy and repeatability, have been
shown to cover the substantial investment required by AFP technology.
VS composites have been extensively studied for several loading

scenarios and different structures such as plates [2, 18-21] and cylindrical
shells [4, 5, 22-24] to examine the effect of fiber steering in improving
structural performance and reducing weight. Since in aerospace structures
any weight savings through reduction of skin thickness must be balanced

against the requirement for structural stability, the buckling performance
have mostly been studied for VS structures.
Hyer and Charette [25] have studied the structural improvement by using
curvilinear fibers in a flat plate with central hole subjected to tensile and
compressive loads. They concluded that in tension the curvilinear designs
lead to improved performance. In compression, however, the buckling
loads were not as high as for quasi-isotropic (QI) laminates. In a follow-up
work, Hyer and Lee [26] constrained the variation of the orientation angles
to increase buckling resistance while producing designs that could be
fabricated. As a result, and by combining both straight and curvilinear
fibers, the best buckling and failure results were achieved. Afterwards,
Nagendra et al. [27] explored that VS laminate made bu curved fibers
could improve the buckling load of the plate-with-hole problem addressed
by Hyer and Lee [26] up to 85% compared with the initial constant
stiffness (CS) design. The research on VS laminates was continued by
Gu¨rdal and Olmedo [28] in the early 1990’s at Virginia Tech (USA).
Followed by achieving promising results in terms of stiffness and buckling
generated by analytical and numerical methods [28-30], several VS
structures were fabricated [31] to validate the manufacturability of
tow-steered laminates. Subsequent testing [32] confirmed the improved
load-carrying capability of the VS panels over their CS counterparts. The
reason for the substantial improvements was attributed to the redistribution
of the load from the centre of the panels to the higher-stiffened edge
regions. The research was followed by a number of Prof. Gürdal’s research
group [33-38] at Delft University of Technology.
Besides plates and panels, VS cylinders were also investigated for their
potential improvement in structural performance. They have been shown
to be very promising in terms of the structural performance improvements
they can offer compared with their constant stiffness (CS) counterparts
[3-5, 22, 23, 26]. Circular and conical cylinders have been extensively
studied for structural improvement via stiffness tailoring. Blom et al. [4]
could improve the bending-induced buckling load of a VS circular
cylinder up to 17% with respect to a baseline laminate at the same weight.
Khani et al. [22] could improve the buckling load of a VS cylinder about
24% over its best constant stiffness (CS) counterpart. The research works
on VS cylinders were not limited to those with circular cross sections.
Elliptical cylinders were also studied for their potential improvement in
axial buckling performance [5, 22, 39]. The axial buckling capacities for
elliptical cylinders were reported up to 30% for the cross sectional aspect
ratio of 0.7 [5]. As an objective different from the buckling performance,
White and Weaver [23] have studied the potential of stiffness tailoring to
achieve bend-free states in pressure vessels with elliptical cross sections.
In this chapter, a metamodeling based design optimization (MBDO)

method for VS composite cylinders is presented. Using MBDO, circular
and elliptical cylinders are investigated for potential improvement in their
buckling capacity. Circular cylinders are subjected to bending-induced
buckling and elliptical cylinders are examined for axial buckling
performance. effects of length aspect ratio and radius (cylinder size) on the
bending buckling of VS cylinders are studied. Additionally, the effect of
the percentage of plies candidate for steering on the overall buckling
improvement is investigated. A multi-objective optimization (MOO)
method is also presented for design of VS cylinders subjected to multiple
loads.
MODELING AND ANALYSIS
As shown in Fig. (1a), assume that a composite cylinder with a finite

length is subjected to a pure bending load. For a prescribed stacking
sequence, say [0/θ/90/-θ/-θ/90/θ/0]s, with a number of plies candidate for
fiber steering (θ-plies), the main question is: “what is the optimum fiber
path for VS plies so that the buckling capacity is maximum?”. To limit the
infinite number of variations of the fiber orientation angle for VS plies, a
piece-wise constant model is used where the cylinder is divided into a
number of narrow bands in which the fiber orientation angle is assumed to
be constant but different from its neighboring regions. This way, the
stiffness tailoring problem is reduced to finding the fiber orientation angle
in each narrow band shown in Fig. (1b). To further reduce the number of
design variables, the orientation angles in a number of equally spaced
narrow bands are assumed to be the design variables (Ti’s) as illustrated in
Fig. (1c). The orientation angles of the narrow bands in between these
regions are calculated by a linear interpolation between the values of the
orientation angles in these design points. Therefore, the orientation angles
in the equidistant narrow bands on the surface of the cylinder are
considered as design variables to be determined for maximum buckling
capacity of the cylinder:
Max. Fcr = f (Ti ’s)

s.t. 0° ≤ Ti ≤ 90°; i = 1, . . ., NDV (1)
where Fcr is the buckling capacity of the cylinder in terms of the design
variables (Ti’s) and NDV is the number of design variables. It should also
be noted that in case of a different loading scenario, the piece-wise
constant model for the continuous variation of the orientation angles may
need to be arranged in a different direction.
For any set of design variables, using the commercial software ABAQUS,
a finite element model is constructed and the structure is analyzed for
buckling. Table 1 shows the mechanical properties of the constituent
materials used in the composite laminate (AS4D/9310 Carbon/Epoxy).
Followed by a mesh convergence study, the cylinder is discretized into an
appropriate number of segments on the circumference and the length. The
S8R5 shell elements of ABAQUS were used in this study that have 8
nodes and 5 degrees of freedom in each node [40].
Fig. (1). (a) The composite cylinder subjected to bending-induced buckling, (b) the
piece-wise constant model to approximate the continuous variation of the orientation
angle in a VS ply, and (c) reducing the number of design variables to the values of the
orientation angles in some particular narrow bands.
It is worth noting that in this study, the layup sequence is assumed to be

previously designed (based on stacking sequence considerations) and the
design optimization problem is reduced to find the optimum orientation
angle variation of those plies candidate for fiber steering. Therefore, the
assumed design domain is not claimed to fully cover the entire design
space (including the stacking sequence optimization). One of the
theoretical approaches to fully explore the design space is using the
lamination parameters [41, 42] as design variables; however, one of the
difficulties associated with using this method is that it is not capable to
incorporate the strength constraints in the design optimization procedure
[35].
θ(α) = θ(−α) for VS laminate (2)

Table 1. Material properties of the unidirectional carbon/epoxy layers.
Property AS4D/9310
EL (GPa) 134
ET (GPa) 7.71
GLT (GPa) 4.31
GT (GPa) 2.76
νLT 0.301
νT 0.396
Vf 0.55
Thickness (mm) 0.127
DESIGN OPTIMIZATION
The design optimization is usually an iterative process which may require

hundreds or thousands of function evaluations. In case of costly function
evaluations (FEA), as in this case, it poses a very big challenge for the
design optimization process. Therefore, the high fidelity models are
replaced by efficient surrogate models (metamodels) that are approximate
mathematical functions that provide approximate responses (function
values) at a design point. Another advantage of metamodeling techniques
is that they can change the noisy responses into a smooth one resulting in a
more appropriate function for finding the global optimum.
Metamodeling here includes calculating the structural response (buckling

capacity) of the VS composite cylinder in a limited number of training
points (sample points) using the FEA and fit an analytical function to the
produced responses from these points. The sample points or the design of
experiment (DOE) are generated using Latin Hypercube Sampling (LHS)
method to create a well-distributed population in the design domain. To
evaluate the metamodeling error, a number of test points are also
generated for which the response (buckling load) are evaluated for them
by using both metamodel function and FEA. The average error was
calculated as a representation of the metamodel’s error.
Among several metamodels used in a diverse set of engineering design

optimization problems [43-45], radial basis functions (RBF) has been
shown to be the most dependable in terms of accuracy and robustness [46,
47] including the design optimization of VS composites structures [24, 48,
49]. First step in metamodeling is to generate the sample points or the
training points, the so-called design of experiment (DOE). For each
training point of the DOE, a high fidelity FEA is performed to calculate
the buckling capacity of the VS composite cylinder. This is followed by
fitting the RBF metamodel so that it passes through all these training
points. The resulted efficient RBF surrogate model is used for function
evaluation in the optimization process (Eq. 1). Metamodeling is followed
by calling the genetic algorithm (GA) optimization tool to find the
optimum design variables. The population based GA method is used to
guarantee the optimum point to be global.
As stated earlier, since metamodels are approximate functions, responses

predicted by surrogate models have some errors when they are compared
with those calculated by the high fidelity models. Therefore, the calculated
optimum design point may not be as accurate as it is required. Increasing
the number of training points is one way to reduce this error. This,
however, asks for the highly increased number of training points and
consequently highly decreased computational efficiency. Another
approach, as suggested here, is using the multi-step optimization method
[50], in which the side constraints of the design variables (lower and upper
bounds) are tightened about the optimum points calculated from the
previous design optimization step and the optimization process is repeated
for the design variables with the new bounds until a convergence is
reached. Fig. (2) schematically shows the multi-step design optimization
procedure.
Fig. (2). Multi-step design optimization flowchart.
RESULTS
4.1. Bending-Induced Buckling of a Circular Cylinder
Fig. (3) shows the buckling mode shapes of a composite cylinder with
quasi-isotropic (QI) laminate along with its optimized VS counterpart. The
stacking sequence for the cylinders is assumed to be [0/θ/90/-θ/-θ/90/θ/0]s
in which θ = 45° for QI cylinder and fiber steered (variable in
circumferential direction) for VS cylinder. For a cylinder with
length=diameter=18 in, the buckling capacity of VS cylinder (VS: Fcr =
11.62e5 Nm) is about 25% larger than its QI counterpart (QI: Fcr = 9.31e5
Nm). As Fig. (3) reveals, the VS cylinder exploits more area in carrying
the compressive load compared with the QI one. The more efficient use of
directional properties in VS cylinder is the result of the optimum

distribution of the orientation angle (θ) in the circumferential direction as
shown in Fig. (4), in which θ starts from a low value in the bottom (keel)
which is under tension and increases to a large value at the top (crown)
which is under compression. This way, the axial stiffness is increased in
the tension side while it is decreased in the compression side to create an
efficient load path for maximum buckling capacity. This is quantitatively
shown in Fig. (5) in which the axial section forces are plotted over the
circumference of the cylinder for both QI and VS laminates. It shows how
the stiffness tailoring made by curved fiber paths reduces the maximum
stress in the compression side and transfers the load to the tension side and
consequently results in an increased area involved in carrying the
compressive loads for VS cylinder compared with QI one. The fiber paths
on a steered ply of the developed VS cylinder is illustrated in Fig. (6).
QI VS
(a) (b)
Fig. (3). Bending-induced buckling mode shape of cylinders with (a) QI and (b) VS
design.
Fig. (4). Orientation angle distribution over the circumference of the QI and VS cylinders.
4.1.1. Effect of the Aspect Ratio (L/R)
Keeping the thickness and radius of the cylinder unchanged, one may
change the length of the composite cylinder and design the optimum VS
laminate for each length to study the effect of the aspect ratio (L/R). Fig.
(7) shows the result of such a study in which the potential improvement of
the bending-induced buckling capacity of a VS cylinder increases with the
decrease of its aspect ratio (L/R). In other words, there is more room for
improvement in the buckling performance for shorter cylinders than the
long ones. The baseline to calculate the improvements in both cases are
the composite cylinders with QI laminate.
Fig. (5). Normalized axial section force on the normalized circumferential distance for
QI and VS designs.
Fig. (6). The curved fiber paths on the developed surface of a VS ply.
Fig. (7). Bending-induced buckling improvement of VS over QI designs for different

aspect ratios (L/R).
4.1.2. Effect of the Radius (R)
In thin-walled structures, thickness ratio (R/t) is one of the important

parameters that may affect the overall performance. To study the effect of
the thickness ratio (R/t), the thickness were kept fixed but the radius and
the length of the cylinder were changed so that the aspect ratio (L/R)
remains unchanged. Although the cylinder becomes bigger with the
increase of its radius and consequently the bending-induced buckling load
increases for both VS and QI cylinders, the improvement of VS cylinders
compared with their QI counterparts did not seem to considerably change
with the radius as long as the aspect ratio (L/R) and thickness (t) are kept
constant (Fig. 8).
Fig. (8). Bending-induced buckling improvement of VS over QI designs for different radii
(R).
4.1.3. Effect of the Steered Plies Percentage
To study the effect of steered plies percentage in a laminate on the

buckling performance, in the [0/θ/90/-θ/-θ/90/θ/0]s stacking sequence, one
may add other plies (0°- or 90°-plies) to the list of layers candidate for
fiber steering. The layup sequence can also be changed to examine its
effects on the overall performance. Rouhi et al. [51] studied several layup
sequences with different percentage of steered plies in the laminate and
their result is shown in Fig. (9). As expected, increasing the number
(percentage) of the steered plies in a laminate results in an increased
design space and consequently more room for structural improvement. For
the composite cylinder considered in this study, a VS composite laminate
with 100% of plies candidate for fiber steering results in about 33%
improvement in the buckling load which is 6% more than the
improvement when 50% of plies are steered (25%).
Fig. (9). Maximum buckling load improvement of VS over QI designs in terms of the
percentage of the steered plies.
4.2. Multi-Objective Design Optimization for Bending-Induced Buckling

Generally, aircraft structures are subjected to multiple load cases that must
be considered in the design process. Therefore, the overall objective
function in the design optimization process must systematically and
simultaneously include a collection of individual objectives that may
oppose each others. The resulting so-called multi-objective optimization
(MOO) process [52] can be formulated and solved by different methods of
which compromise programming is one of the the most common
approaches. The objective function in the compromise programming is a
weighted combination of all individual objective functions. As an example
in this study, the bending load can be applied in two opposite directions
and the VS cylinder must be designed to have maximum buckling capacity
when the bending load is applied in either of the two opposite directions. If
the VS cylinder is designed for maximum buckling capacity in only one
direction, the buckling performance of the cylinder will be as shown in Fig.
(10a) when the direction of the applied load changes. It results in about
25% improvement of the buckling load for one direction, but at the
expense of 57% decrease in the buckling capacity if the bending load is
applied in the opposite direction. To simultaneously improve the buckling
capacity in both directions a multi-objective design optimization is needed.
The general multi-objective utility function that combines all individual
objectives can be expressed as the weighted exponential sum:
(3)
where p is a nonzero power term to be adjusted to capture Pareto optimal

points [52], 𝜔 is a vector of weights such that
is a utopia point (also called an ideal

point for each objective function). It is worth noting that the power term p
= 2 resulted in maximum improvement in the buckling load for both
directions. The relative values of the weight (𝜔i) generally reflects the
relative importance of each individual objective (Fi(x)). For a composite
cylinder in which the bending loads are applied in two opposite directions
(n = 2), the utopia point (𝐹𝑖0 ) for each objective in Eq. 3 is set to be the
maximum buckling load in case the bending load is applied only on that
direction.
For 𝜔1 = 𝜔2 = 0.5 in which the same importance is considered for the

bending loads in both directions, the buckling performance of the cylinder
is shown in Fig. (10b). The MOO results in a VS cylinder that has about
9% improvement in the buckling capacity in either of the two opposite
directions. As the main decision making tool for designers, Rouhi et al.
[24] generated the Pareto optimal points (Pareto frontier) by changing the
weight factors (𝜔i’s) and performing MOO for different combinations of
the weight factors.
Fig. (10). Buckling performance of VS (black line) compared with QI (cyan dashed line)
composite cylinder in terms of the direction of applied bending moment (M). The VS
cylinders are optimized for bending-induced buckling performance in (a) only one
direction: 𝜔1 = 1, 𝜔2 = 0, and (b) two opposite directions: 𝜔1 = 𝜔2 = 0.5.
4.3. Axial buckling of an Elliptical Cylinder
Structural design in real applications are always constrained by many

limitations. Specifically in aerospace applications, aerodynamic
considerations may dictate non-circular cross sections for a cylinder such
as in blended wing body aircraft structures. The non-uniformity of
curvature, as a result, reduces the buckling capacity of the structure [5].
Stiffness tailoring seems to be a good candidate to compensate the
inefficiency resulting from the structural non-uniformity.
As an example to demonstrate this potential, assume that a 15 inch long

elliptical cylinder with b = 8.191 in, a = 20.477 in, and simply supported
boundary conditions is subjected to axial buckling as shown in Fig. (11).
The design problem in this case would be to find the optimum fiber paths
on the surface of the elliptical cylinder (see Fig. 12) so that the buckling
load is maximum. Using the MBDO method presented in this study for an
elliptical cylinder with a balanced symmetric 8-ply laminate made of
AS4D/9310 carbon/epoxy materials and [±1/±2]s stacking sequence the
buckling capacity can be maximized. For the sake of comparison, the

buckling capacities of the same elliptical cylinder with the optimum CS
laminate (Optimum constant values of 1 and 2), as well as a QI laminate
[±45/0/90]s were also calculated. For such a structure Fig. (13) shows the
buckling mode shapes of QI, best CS, and optimum VS cases. The most
efficient usage of the directional properties of the composite materials is
made by fiber steering in the VS design in which most of the area is
involved in carrying the compressive load. Figure (14) also shows the
orientation angle distribution on a quarter circumference of the elliptical
cylinder from one end of the major axis to one end of the minor axis. It
reveals a general trend of the optimum orientation angles (1 and 2) to
start from small values in the curved portion of the circumference (to
make laminate stiffer in axial direction) to large values in the flatter
portion (to make laminate softer in axial direction). Therefore, the material
properties are tailored so that more axial compressive loads are carried by
the curved parts of the elliptical cylinder that are structurally more stable
for buckling compared to the flatter portions that are more prone to buckle
in compression. The buckling load capacity of VS elliptical cylinder (VS:
Fcr=123.787 kN) shows about 104% improvement over the CS design
(CS: Fcr=60.811 kN) and 144% improvement over the QI one (QI:
Fcr=50.787 kN). As such, the design optimization results reveals that there
is a substantial room for buckling improvement of cylinders with
non-circular cross sections [53].
Fig. (11). An elliptical composite cylinder subjected to axial buckling.

Fig. (12). The continuous circumferential variation of orientation angles in a VS ply of an

elliptical composite cylinder from one end of its major axis (point 1) to one end of its
minor axis (point 4).
Fig. (13). The buckling mode shapes of (a) QI, (b) CS, and (c) VS elliptical composite
cylinders with b/a=0.4.
VS CS VS CS
90 90
θ1 60
θ2 60
30 30
0 0
1 2 3 4 1 2 3 4
Circumferential Location (S) Circumferential Location (S)
Fig. (14). Optimum distribution of the orientation angles (θ 1 and θ 2 ) over a quarter
circumference of an elliptical cylinder with the stacking sequence of [±θ 1 /±θ 2 ] s and
b/a=0.4.
CONCLUSION
The potential room for improvement in the buckling performance of

composite cylinders via fiber steering was studied in this chapter. The
resulting so-called variable stiffness (VS) composites were designed by
using a metamodel based design optimization (MBDO) method. By
exploiting the full potential of the directional properties of the composite
laminates, the curved fibers were shown to create efficient load paths to
substantially improve the buckling performance.
For circular cylinders with short lengths under bending-induced buckling

loads, it was shown that the buckling load can be improved by about 40%
by only 50% of the total plies considered for fiber steering. An elliptical
cylinder was also studied for axial buckling improvement and the results
showed that its buckling capacity can potentially be improved by up to
104% over its constant stiffness (CS) and 144% over its quasi-isotropic
(QI) counterparts.
this publication.
ACKNOWLEDGEMENTS
The financial contributions from the Natural Sciences and Engineering

Research Council of Canada (NSERC) industrial chair on Automated
Composites, Bell Helicopter Textron Canada Ltd., Bombardier Aerospace,
Delastek Ltd., and Concordia University are appreciated.
REFERENCES
[1] S.T. IJsselmuiden, M.M. Abdalla, and Z. Gu¨rdal, "Optimization of variable-
stiffness panels for maximum buckling load using lamination parameters",

AIAA J., vol. 48, no. 1, pp. 134-143, 2010.
[http://dx.doi.org/10.2514/1.42490]
[2] S. Setoodeh, M.M. Abdalla, S.T. IJsselmuiden, and Z. Gu¨rdal, "Design of

variable-stiffness composite panels for maximum buckling load", Compos.
Struct., vol. 87, no. 1, pp. 109-117, 2009.
[3] A. Alhajahmad, M. Abdalla, and Z. Gu¨rdal, “Optimal design of

tow-placed fuselage panels for maximum strength with buckling
considerations", J. Aircr., vol. 47, no. 3, pp. 775-782, 2010.
[http://dx.doi.org/10.2514/1.40357]
[4] A.W. Blom, P.B. Stickler, and Z. Gu¨rdal, "Optimization of a composite

cylinder under bending by tailoring stiffness properties in circumferential
direction", Compos., Part B Eng., vol. 41, no. 2, pp. 157-165, 2010.
[5] M. Sun, and M. Hyer, "Use of material tailoring to improve buckling capacity
of elliptical composite cylinders", AIAA J., vol. 46, no. 3, pp. 770-782, 2008.
[http://dx.doi.org/10.2514/1.32495]
[6] D.O. Evans, M.M. Vaniglia, and P.C. Hopkins, "Fiber placement process
study", International SAMPE Symposium and Exhibition (Proceedings), vol.
Vol. 34, 1989pp. 1822-1833
[7] R.L. Anderson, and C. Grant, "Advanced fiber placement of

composite fuselage structures", In: Proceedings of First NASA
Advanced Composites Technology Conference, 1991 pp. 817-830.
[8] C. Grant, and V. Benson, “Automated fiber placement: evolution

and current demonstrations", In: Proceedings of Third NASA
Advanced Composites Technology Conference, 1993.
[9] R.A. Kisch, "Automated fiber placement historical perspective", In:

Proceedings of International SAMPE Symposium and Exhibition Long Beach,
CA, USA
[10] R.L. Pinckney, "Fabrication of the v-22 composite aft fuselage using
automated fiber placement", In: Proceedings of First NASA Advanced
Composites Technology Conference, 1991, pp. 385-397.
[11] R. Measom, and K. Sewell, "Fiber placement low-cost production for complex
composite structures", Annual Forum Proceedings of American Helicopter
Society, 1996 pp. 611-622.
[12] J.P. Martin, R.J. Langone, M.J. Pasanen, and J.A. Mondo, "Cost-effective,
automated equipment for advanced composite structure development and
production", International SAMPE Symposium and Exhibition (Proceedings),
vol. Vol. 42, 1997pp. 48-55
[13] M.J. Pasanen, J.P. Martin, R.J. Langone, and J.A. Mondo, "Advanced
composite fiber placement: process to application", In: Proceedings of the
Japan International Sampe Symposium. Japan Chapter of Sampe, 1997, pp.
1055-1060.
[14] D. Buchanan, R. Alexander, and J. Thomas, "Structural design of

a fiber placed inlet duct", In: Proceedings of the 40th AIAA/ASCE/AHS/
ASC SDM Conference, 1999 St. Louis, MO, USA
[http://dx.doi.org/10.2514/6.1999-1442]
[15] D.O. Evans, "Fiber placement", In: D.B. Miracle, S.L. Donaldson, Eds., ASM
Int., vol. 21. 2001, pp. 477-479.ASM Handbook
[16] D.A. McCarville, J.C. Guzman, and D.M Rotter, "Automated material
placement: 2008 industry overview", Proceedings of International SAMPE
Symposium and Exhibition, 2008
[17] B. Morey, "Automating composites fabrication", Manuf. Eng., vol. 140, no. 4.
[18] S.B. Biggers, and S. Srinivasan, "Compression buckling response of tailored

rectangula r composite plates", AIAA J., vol. 31, no. 3, pp. 590-596, 1993.
[http://dx.doi.org/10.2514/3.61543]
[19] S.B. Biggers, and S.S. Pageau, "Shear buckling response of tailored composite
plates", AIAA J., vol. 32, no. 5, pp. 1100-1103, 1994.
[http://dx.doi.org/10.2514/3.12107]
[20] S. White, G. Raju, and P. Weaver, "Initial post-buckling of variable-stiffness

curved panels", J. Mech. Phys. Solids, vol. 71, no. 1, pp. 132-155, 2014.
[http://dx.doi.org/10.1016/j.jmps.2014.07.003]
[21] G. Raju, Z. Wu, and P.M. Weaver, "Buckling and postbuckling of variable
angle tow composite plates under in-plane shear loading", Int. J. Solids Struct.,
vol. 58, pp. 270-287, 2015.
[22] A. Khani, M. Abdalla, and Z. Gu¨rdal, "Circumferential stiffness tailoring of

general cross section cylinders for maximum buckling load with strength
constraints", Compos. Struct., vol. 94, no. 9, pp. 2851-2860, 2012.
[23] S. White, and P. Weaver, "Bend-free shells under uniform pressure with
variable-angle tow derived anisotropy", Compos. Struct., vol. 94, no. 11, pp.
3207-3214, 2012. [http://dx.doi.org/10.1016/j.compstruct.2012.05.009]
[24] M. Rouhi, H. Ghayoor, S. Hoa, and M. Hojjati, "Multi-objective design

optimization of variable stiffness composite cylinders", Compos., Part B Eng.,
vol. 69, pp. 249-255, 2015.
[25] M. Hyer, and R. Charette, "Use of curvilinear fiber format in composite

structure design", AIAA J., vol. 29, no. 6, pp. 1011-1015, 1991.
[http://dx.doi.org/10.2514/3.10697]
[26] M.W. Hyer, and R.F. Charette, "Use of curvilinear fiber format in composite
structure design", AIAA J., vol. 29, no. 6, pp. 1011-1015, 1991.
[http://dx.doi.org/10.2514/3.10697]
[27] S. Nagendra, S. Kodiyalam, J.E. Davis, and V. Parthasarathy, "Optimization of

tow fiber paths for composite design", Structural Dynamics and Materials
Conference, 1995. pp. 1031-41 New Orleans, LA.
[28] Z. Gu¨rdal, and R. Olmedo, "In-plane response of laminates with spatially

varying fiber orientations- variable stiffness concept", AIAA J., vol. 31, no. 4,
pp. 751-758, 1993.
[http://dx.doi.org/10.2514/3.11613]
[29] C. Waldhart, Analysis of tow-placed, variable-stiffness laminates, Ph.D. thesis,
Virginia Polytechnic Institute and State University, 1996.
[30] B.F. Tatting, Analysis and design of variable stiffness composite cylinders,
Ph.D. thesis, Virg. Tech, Blacksburg, Virginia, 1998.
[31] K.C. Wu, and Z. Gu¨rdal, "Thermal testing of tow-placed variable stiffness
panels", In: Proceedings of the 42nd AIAA/ASME/ASCE/AHS/ASC structures,
structural dynamics and materials (SDM) conference, Seattle, WA, 2001.
[http://dx.doi.org/10.2514/6.2001-1190]
[32] K.C. Wu, Z. Gu¨rdal, and J.H. Starnes, “Structural response of

compression-loaded, tow-placed, variable stiffness panels", In: Proceedings of
the 43rd Aiaa/Asme/Asce/Ahs/Asc structures, structural dynamics and
materials conference, Denver, Co, 2002, pp. 2002-1512.
[http://dx.doi.org/10.2514/6.2002-1512]
[33] S. Setoodeh, Optimal design of variable-stiffness fiber-reinforced composites
using cellular automata, Ph.D. thesis, Delft University of Technology, 2005.
[34] C.S. Lopes, Damage and failure of non-conventional composite laminates,

Ph.D. thesis (2009).
[35] A. W. Blom, Structural performance of fiber-placed variable-stiffness

composite conical and cylindrical shells, Ph.D. thesis, Delft University of
Technology (2010).
[36] S.T. IJsselmuiden, Optimal design of variable stiffness composite structures

using lamination parameters, Ph.D. thesis, Delft University of Technology
2011.
[37] J. Van Campen, Optimum lay-up design of variable stiffness composite

structures, Ph.D. thesis, Delft University of Technology 2011.
[38] A. Khani, Optimum design of steered fibre composite cylinders with arbitrary
cross-sections, Ph.D. thesis, Delft University of Technology (2013).
[39] A. Khani, M.M. Abdalla, and Z. Gurdal, "Optimum tailoring of fibre-steered

longitudinally stiffened cylinders", Compos. Struct., vol. 122, pp. 343-351,
2015. [http://dx.doi.org/10.1016/j.compstruct.2014.11.071]
[40] Pawtucket, RI, USA, ABAQUS Version 6.11 User’s Manual. 2011.
[41] S. Setoodeh, M. Abdalla, and Z. Gu¨rdal, "Design of variable-stiffness

laminates using lamination parameters", Compos., Part B Eng., vol. 37, no. 4-5,
pp. 301-309, 2006.
[42] K. Wu, B. Stanford, G. Hrinda, Z. Wang, R. Martin, and H. Kim, Structural

Dynamics and Materials Conference, 2013 Boston, MA, USA
[43] R. Jin, W. Chen, and T. Simpson, "Comparative studies of metamodelling

techniques under multiple modelling criteria", Struct. Multidiscipl. Optim, vol.
23, no. 1, pp. 1-13, 2001.
[http://dx.doi.org/10.1007/s00158-001-0160-4]
[44] N.V. Queipo, R.T. Haftka, W. Shyy, T. Goel, R. Vaidyanathan, and K.P. Tucker,
"Surrogate-based analysis and optimization", Prog. Aerosp. Sci., vol. 41, no. 1,
pp. 1-28, 2005.
[45] G. Wang, and S. Shan, “Review of met modeling techniques in support of

engineering design optimization”, J. Mech. Design, Transactions of the ASME
129, vol. 4, 2007pp. 370-380
[46] M. Rouhi, and M. Rais-Rohani, “Modeling and probabilistic design

optimization of a nanofiber- enhanced composite cylinder for buckling",
Compos. Struct, vol. 95, pp. 346-353, 2013.
[47] M. Rouhi, M. Rais-Rohani, and A. Najafi, "Probabilistic analysis and

optimization of energy absorbing components made of nanofiber enhanced
composite materials", Compos. Struct., vol. 100, pp. 144-153, 2013.
[48] M. Arian Nik, K. Fayazbakhsh, D. Pasini, and L. Lessard, "A comparative

study of metamodeling methods for the design optimization of variable
stiffness composites", Compos. Struct., vol. 107, pp. 494-501, 2014.
[49] M. Rouhi, H. Ghayoor, S. Hoa, and M. Hojjati, "Effect of structural parameters

on design of variable- stiffness composite cylinders made by fiber steering",
Compos. Struct., vol. 118, no. 1, pp. 472-481, 2014.
[50] M. Rouhi, H. Ghayoor, S.V. Hoa, and M. Hojjati, "Multi-step design

optimization of variable stiffness composite cylinders made by fiber steering",
In: Proceedings of the American Society for Composites - 29th Technical
Conference, La Jolla, CA, USA, Sep, 2014.
[51] M. Rouhi, H. Ghayoor, S.V. Hoa, and M. Hojjati, "The effect of the percentage
of steered plies on the bending-induced buckling performance of a variable
stiffness composite cylinder", Science and Engineering of Composite Materials,
vol. 22, no. 2, pp. 149-156, 2015.
[http://dx.doi.org/10.1515/secm-2014-0258]
[52] R. Marler, and J. Arora, "Survey of multi-objective optimization methods for

engineering", Struct. Multidiscipl. Optim., vol. 26, no. 6, pp. 369-395, 2004.
[http://dx.doi.org/10.1007/s00158-003-0368-6]
[53] M. Rouhi, H. Ghayoor, S. Hoa, and M. Hojjati, P. M. Weaver, "Stiffness

tailoring of elliptical composite cylinders for axial buckling performance",
Compos. Struct, vol. 150, pp. 115-123, 2016.
[doi:10.1016/j.compstruct.2016.05.007]
CHAPTER 15
Manufacturing Challenges Associated with the

Use of Metal Matrix Composites in Aerospace
Structures
Tracie Prater*
NASA Marshall Space Flight Center, Huntsville, AL, 35811, USA
Abstract: Metal Matrix Composites (MMCs) are materials which consist of a

metal alloy reinforced with ceramic particles or fibers. These materials possess a
very high strength to weight ratio, good resistance to impact and wear, and a
number of other properties which make them attractive for use in aerospace and
defense applications. For example, MMCs have being extensively used for
structural tubing in the space shuttle orbiter, the antenna mast of the Hubble Space
Telescope, control surfaces and propulsion systems for aircraft, and tank armors.
However, difficulties arise when joining those materials with fusion welding and
impose limitations on the size of MMC components. Melting of the material leads
to formation of an undesirable phase when molten Aluminum (Al) comes into
contact and reacts with the reinforcement. This phase forms a strength depleted
zone along the jointline. Friction Stir Welding (FSW) is a relatively joining
technique, developed at The Welding Institute (TWI) in 1991. Because FSW occurs
below the melting temperature of many metal alloys, it precludes formation of
deleterious phases and results in a more favorable welded microstructure that is
closer to that of the parent material. At NASA, this process was first applied to
weld the super lightweight external tank for the space shuttles program. Today
*Corresponding author Tracie Prater: NASA Marshall Space Flight Center, Huntsville, AL, 35811; E-mail:
tracie.prater@nasa.gov.
Yucheng Liu (Ed.)

Metal Matrix Composites in Aerospace Structures Frontiers in Aerospace Science, Vol. 1 543
FSW is employed to join structural components in Delta IV, Atlas V, and Falcon IX
rockets as well as NASA’s Orion Crew Exploration Vehicle and Space Launch
System. Currently, FSW researchers are interested in extending the application of
the process to new materials which are difficult to weld using conventional fusion
techniques, such as MMCs. Rapid wear of the welding tool in FSW of MMCs is a
consequence of the large discrepancy in hardness between the steel tool and the
reinforcement material. This chapter summarizes the challenges encountered when
joining MMCs to themselves or to other materials in structures. Specific attention is
paid to the influence of the process variables for FSW on the wear process. A
phenomenological model of the wear process was established based on the rotating
plug model of FSW. The effectiveness of tool materials with high hardness (e.g.
Tungsten Carbide, high speed steel, and tools with diamond coatings) in resisting
abrasive wear is also considered. In-process force, torque, and vibration signals are
analyzed to determine the feasibility of in situ monitoring of tool shape changes as
a result of wear. One advantage of this model is that its successful implementation
would eliminate the need for off-line evaluation of tool condition during joining.
Monitoring, controlling, and reducing tool wear in FSW of MMCs are critical to
full application of these materials in aerospace structures where they would be of
most benefit. The work presented in this chapter can be further extended for
machining of MMCs, where the wear of the tool materials is also a limiting factor.
Keywords: Advanced manufacturing, Friction Stir Welding, Materials

joining, Metal Matrix Composites, Tool wear.
NOMENCLATURE
FSW Friction Stir Welding
MMC Metal Matrix Composite
RPM Rotations per minute
SiC Silicon Carbide
ℓ Length of joint (inches)
𝜈 Traverse rate (inches per minute)
𝜔 Rotation rate (rotations per minute)
𝑊 Wear experienced by the tool during welding
544 Frontiers in Aerospace Science, Vol. 1 Tracie Prater
D Characteristic reinforcement particle diameter

𝛿 Width of the rotating plug
𝛿0 Width of the rotating plug at angular position 𝜃 = −π/2
𝛿 max Maximum width of the rotating plug
θ Angular position in x-y plane
P Reinforcement fraction expressed as a percentage of workpiece
volume
∆𝐶 Cutting arc (angular region where abrasion can occur)
∆𝐶 max Maximum cutting arc
𝜎 Flow stress
t Plunge depth
Ttotal Total torque
Ra Average surface roughness
Rt Peak to valley roughness
R Tool shoulder radius
r Pin shoulder radius
INTRODUCTION
The reliance on ballistic techniques in launch architectures makes mini-
mization of weight one of the most preeminent issues in spacecraft design.
For any launch vehicle, cargo weights only comprise a small portion of
the vehicle’s weight at launch. Structural components and fuel account for
the vast majority of the launch weight of a rocket. In order to improve the
structural efficiency of a vehicle, the vehicle’s dry weight can be reduced
and the weight reduction represents a commensurate gain in cargo
capacity.
Use of lighter weight materials represent one option for reducing a

vehicle’s weight. At present, Aluminum alloys are used as the primary
aerospace structural material for fuel tanks. These alloys are lightweight,
strong, well-characterized, and abundant. However, recent development
of advanced materials such as composites have made it possible to further
reduce the weight of the structure while satisfying (and in many cases
exceeding) the mechanical criteria established for flight-rated hardware.

In 2001, NASA changed the material of the space shuttle’s external tank
from Al 2219 to an aluminum- lithium composite (Al-Li 2195) developed
by Lockheed Martin. This replacement reduced the total weight of the
external tank by 7,500 pounds [1]. The weight reduction enabled the
space shuttle to transport heavier components of the International Space
Station and gave NASA the option to consolidate multiple large-scale
components into a single flight, improving launch efficiency and
imparting cost savings. Use of traditional fusion welding techniques to
join composite material together resulted in mechanically deficient joints.
Friction Stir Welding (FSW), a solid-state joining technique first proposed
by The Welding Institute (TWI) of Great Britain, was able create
defect-free Al-Li 2195 joints with superior mechanical properties. The
Friction Stir Welding process is illustrated in Fig. (1). With the success of
the external tank program, NASA revolutionized much of its manu-
facturing by implementing the FSW process. The primary structures for
NASA’s Space Launch System (SLS) and the Orion Crew Exploration
Vehicle (CEV) rely extensively on the process to produce reliable, defect
free joints.
Fig. (1). The FSW process. The tool consists of a cylindrical pin that penetrates the workpiece
material and a larger diameter shoulder which rests on the surface of the material. In the FSW
process, the tool rotates at an angular speed 𝜔 while advancing through the material at a traverse
speed v. During its traversal, plasticized material is picked up on its advancing side and is
deposited on its retreating side. Materials cools and consolidates to form a welded region.
Driven by a continuing desire for lighter materials in the space industry,

advanced material processing and aerospace welding techniques have
progressed significantly in the past twenty years. As FSW is a mature
technology for processing many aerospace Aluminum alloys (such as the
2000, 6000, and 7000 series), there is a considerable interest to extend the
application of FSW to other, higher temperature materials, such as steels,
Magnesium alloys, and even Titanium. A group of materials which are
especially suitable for FSW are metal matrix composites (MMCs). MMCs
are dual phase materials which consist of a ceramic reinforcement
embedded in a matrix of metal alloy. Typical reinforcements are ceramics
such as silicon carbide or aluminum oxide in the form of either particulates
or fibers. The amount of the reinforcement material in a MMC is
represented as a percentage of the MMC’s overall weight or volume. Metal
composites are superior to their unreinforced counterparts in terms of
strength to weight ratio (which may be more than twice of that of the
conventional Aluminum alloys), excellent temperature and wear resistance,
and long fatigue life. Unfortunately, the joints of these materials created
using fusion welding display a considerable amount of micro-pores and
cracks [2]. The mechanical properties of these fusion welded such joints are
negatively impacted by the presence of Al4C3, an undesirable precipitate
produced during the reaction of molten Aluminum with the reinforcement
[3]. Since FSW occurs below the melting point of the matrix alloy, the
deleterious theta phase is absent in the FSW welds of these materials. The
major barrier to the application of FSW in joining of MMCs is rapid and
severe wear of the tool, a consequence of direct contact between the tool
(typically fabricated from a steel alloy) and the comparatively harder
reinforcement particles [3]. Progressive wear of the tool removes features
which facilitate material stirring and thus increases the likelihood of void
development [4].
PROBLEM DESCRIPTION
Large defects present in the weld result in reduced mechanical properties.
For MMCs, it is therefore critical to reduce wear and preserve the tool shape
that will facilitate material stirring and prevent the formation of such
defects. To avoid weld-related structural failures, significant investments
have been made by NASA and its commercial partners on post-process
inspection and non-destructive testing techniques (such as ultrasonics and
X-rays) to inspect welds of flight hardware. The non-destructive evaluation
results and mechanical tests require parameterization studies. As part of
weld development, a set of experiments is designed to determine a window
of welding parameters that can produce joints satisfying design and safety
criteria. Even though many parameterization studies about FSW of MMCs
have demonstrated that it is possible to generate an acceptable operating
window of parameters for virtually any tool/workpiece combination, these
studies did not specifically consider defects resulting from the wear
mechanisms that affect the system [5, 6]. Even if parameters used to
produce the MMC joint fall in the operating window, there is still a
possibility that a defect may be created over time as the tool loses volume
and the workpiece undergoes a corresponding reduction in the flow of
plasticized material. Therefore a better control of the wear process during
joining is required to fully guarantee the successful integration of MMCs
into aerospace structures. Measures of mitigating wear may include careful
selection of process parameters (rotation speed, traverse rate, length of weld,
tilt angle, etc.) and material properties (percentage reinforcement, type of
reinforcement, particle size, and tool material). In cases of longer welds or
scenarios in which coatings fail to guard against wear, wear maybe
inevitable and in-process monitoring becomes essential. As emphasized in
the previous discussion, in process wear detection is often synonymous
with in-process fault detection due to the loss of tool features which
facilitate material stirring. Techniques used to detect wear in this scenario
may also be applied for in–process quality control of FSW of conventional

alloys.
RELATIONSHIP BETWEEN WEAR AND PROCESS

PARAMETERS
The first key to unraveling the tool wear problem in FSW of MMCs is to
understand the fundamental physics which underlie the wear process. A
study of the variation of wear with FSW process parameters can provide
some initial insight. The experiments for the preliminary wear study
conducted in this work comprise an L27 orthogonal Taguchi array with
three factors (rotation rate 𝜔, traverse speed v, and distance welded l) at
three levels (Table 1). The deterioration in the cross-sectional area of the
tool pre and post-experiment was quantified using imaging software. Tool
wear in FSW of MMCs displays a circumferentially symmetric pattern,
which means that the degradation measured in the cross-section of the
probe is representative of the volume lost by the entire probe. Fig. (2) shows
an overlay of close-up images of the tool probe taken after successive welds
of Al 359 reinforced with 20% Silicon Carbide particles. The characteristic
rounding of the tool that accompanies wear is undesirable because it limits
vertical stirring of material and frequently results in root flaw defects [4].
Fig. (2). Successive overlay of probe cross-sections for welds of Al 359/SiC/20p at 1500 RPM and
7 IPM traverse speed for 0, 8, 16, and 24 inches welded (from outside to inside) [7].
Table 1. Experimental matrix and results of initial wear study [7].
Test Number 𝝎 (RPM) 𝝂 (in/min) 𝓵 (in) Percent Wear

1 1000 5 8 3.7
2 1000 5 16 5.9
3 1000 5 24 7.5
4 1000 7 8 1.2
5 1000 7 16 2.7
6 1000 7 24 3.8
7 1000 9 8 1.4
8 1000 9 16 4.7
9 1000 9 24 5.0
10 1500 5 8 5.8
11 1500 5 16 15.6
12 1500 5 24 16.0
13 1500 7 8 3.0
14 1500 7 16 12.3
15 1500 7 24 17.2
16 1500 9 8 4.0
17 1500 9 16 6.6
18 1500 9 24 15.6
19 1500 5 8 3.5
20 2000 5 16 7.8
21 2000 5 24 12.9
22 2000 7 8 7.2
23 2000 7 16 11.7
24 2000 7 24 17.6
25 2000 9 8 4.0
26 2000 9 16 10.1
27 2000 9 24 15.0
The wear data obtained from these experiments was used to construct a
multiple regression model (MRM) to predict the volume loss of the tool in
welding based on the values of the three major process parameters [7]. An
analog can be drawn between the MRM model and the machinability maps
used to determine optimal parameters for machining abrasive materials. In
both instances, we seek to develop parameters which minimize wear but
still produce a cut, hole or other feature (ex. weld) with acceptable quality.
The relationship derived from the regression analysis (Eq. 1) is strongly
correlated with the experimental data. The empirical predictive model was
tested on a validation set that consists of cases separate from those in the
original data set used to construct the regression model. The calculated and
observed wear values for the validation set closely agree with each other. A
direct comparison of observed and calculated values is presented in Table 2.
The model should thus be applicable to parameter sets different from those
used in its construction.
𝑊 = 0.584ℓ − 1.038𝜈 + 0.009𝜔 − 6.028 (1)
Table 2. Comparison of observed and predicted wear for model validation data set.
𝝎 𝝂 𝓵 Observed Wear Predicted Wear Residual Percent

(% Volume Loss) Error
(% Volume Loss)
12 5 8 5.45 4.72 0.73 13.4
17 9 8 4.95 5.05 -0.1 -2.02
17 9 16 10.06 9.72 0.34 3.38
14 6 8 5.45 5.02 0.43 7.89

From the multivariate regression analysis results it can be seen that tool
wear in FSW of MMCs is directly proportional to the rotation speed and
welding distance but inversely proportional to the traverse speed. A
dimensionless group (𝜔𝑙/𝑣) is defined based on these parameters, which is
linearly correlated with wear as expressed in Eq. 2. Physically, Eq. (2)
means that each unit increase in the group 𝜔ℓ/𝜈 is accompanied by
a .0004% increase in wear [8]. This constant, like the constants in the
multiple regression model (Eq. 1), can be used to predict relative changes in
wear based on changes in process parameters. For instance, a 10,000-fold
increase in 𝜔ℓ/𝜈 would be associated with a 4% increase in wear
percentage.
𝜔ℓ
𝑊 = 0.0004 (2)
𝜈
The regression equation and the companion dimensionless group are

intended to inform parameter selection in FSW of MMCs. If the length of
weld and the critical value of tool degradation are given, the dimensionless
group (or the regression model) can be employed to define combinations of
rotation and traverse speed that will keep the wear below a specified level.
Similarly, the equations can be used to estimate the life of a steel tool in
joining MMCs and decide the maximum distance which can be welded at a
specific set of process parameters before wear increases beyond an
acceptable level.
CONSTRUCTING AN ANALYTICAL MODEL OF THE WEAR

PROCESS
The wear process in FSW of MMCs is unique in that it appears to be a shear,

(rather than drag) phenomena, as suggested by the inverse relationship
between the wear and the traverse speed. To date, no analytical model has
been developed to describe how wear varies along the length of the pin tool.
Based on experimental observation, most material loss occurs in the zone of
the probe tip while the material loss at the interface between the probe and
the shoulder is minimal. The rotating plug model for material flow in FSW,
originally developed by Dr. Arthur Nunes of NASA’s Marshall Space Flight
Center, provides a framework for understanding the physical phenomena
which govern the relationships observed experimentally in FSW of MMCs
[9]. In this formulation, it is assumed that the probe is surrounded by a plug
of plasticized metal, also referred to as the “rotating plug”. The width of this
plug, δ, can be considered as a function of the process parameters ω and ν,
and the angular position θ in the x-y plane (Eq. 3). As shown in Figs. (3 and
4), the plug is symmetric about the y-axis and changes from δ= 𝛿0 at
θ = -π/2 to δ 𝑚𝑎𝑥𝑖𝑚𝑢𝑚 at θ = π/2.
𝑣
𝛿 = 𝛿0 + (1 + 𝑠𝑖𝑛𝜃) (3)
𝜔
Retreating Side Rotating Plug

Shear Surface
HBULGE
w
Exit Hole
5 mm
Refined Grains Advancing Side

(Note Banding)
Fig. (3). The rotating plug model superimposed on a plan view of the extraction point (exit hole) of
a friction stir weld.
The analytical model can be slightly adjusted to account for the presence of
abrasive particles like the inclusions found in Al-MMCs. The model as it
applies to FSW of MMCs is shown in Fig. (4). For a given particle diameter
D, wear occurs only when the particle is able to span the width of the shear
zone and impinge on the tool surface, removing material as the tool rotates
past it. The analytical results calculated based this model are consistent with
the experimental data. Greater wear is produced by parameters which
correspond to thinner shear zones, while parameter sets associated with
larger δ values lead to less wear [9]. A piecewise condition for abrasion is
also suggested by the rotating plug model: wear only occurs at locations
where the radius of the particle exceeds the width of the shear zone
(𝐷/2 > 𝛿). This corollary implies that the wear is also influenced by
material properties, such as particle size and percentage of reinforcement
(number density of particles present in the material).
Fig. (4). Rotating plug model for FSW of MMCs. Pink shaded region is the rotating plug (with a
thickness 𝛿). The solid black circle depicts an abrasive particle in the workpiece material.
The rotating plug model presented in this chapter can be used to define the
region where scoring of the tool by abrasive is possible. The original plug
model predicts that the shear surface widens along the length of the tool
(Fig. 5), making the region near the tip of the probe the most susceptible to
abrasion. The percent wear predicted by the model can be calculated with
equation 4, where D is the diameter of the reinforcement particle, P denotes
the proportion of reinforcement relative to the volume of the entire
workpiece, 𝜔 is the rotation speed, 𝑣 is the traverse speed, l is the length of
weld, R is the radius of the pin, and ∆𝐶𝑚𝑎𝑥 represents the maximum value of
the cutting arc (an angular area where the abrasion can occur). For a
detailed derivation of this equation please refer to references [9] and [10]. It
is assumed that the depth to which the particle impinges on the tool surface
is equal to the radius of the particle. The volume of material removed in a
single rotation is integrated along the length of the tool where abrasion may
occur. That length is estimated at about half of the probe length. The
deterioration in the width of the cutting arc ∆𝐶 as the distance from the
probe increases is also considered. Analytical results calculated using this
model are close (generally within 10 percent) to values measured in
experiments, with differences between predicted and observed values
[9, 10].
5𝐷∆𝐶𝑚𝑎𝑥 𝑃𝜔𝑙
% 𝑤𝑒𝑎𝑟 = (4)
24𝑅𝑣
Fig. (5). Growth of plug in the axial direction along length of the tool probe. Black circles represent
abrasive particles.
Colligan [11] conducted tracer experiments to study material flow in FSW.

In this work, steel shot tracers were embedded along the joint line and
welded joints were subsequently X-rayed. An analogous technique,
wherein SiC particulate was inserted in a milled slot along the advancing
face of an Aluminum 6061 joint and subsequently friction stir welded, was
used to isolate the impact of particle size on tool wear in this application
and further evaluate the predictions of the rotating plug model. The wear of
the tool probe was evaluated after welding by means of weighing, contact
profilometry, and SEM microscopy. As we would expect intuitively and as
predicted by the plug model, the amount of wear is strongly dependent on
the particle size. The percent weight lost by the tool during welding as a
function of the inclusion particle size is plotted in Fig. (6).
Fig. (6). Plot of percent weight reduction for FSW probe vs particle size. Each data series represents
a separate set of parameters [10]. X-axis is particle diameter in inches and y-axis is percent volume
loss.
Surface metrology of FSW tool probes used for welding joints containing
particles of three different sizes (FEPA grade F150, F60, and F14) were
compared using contact profilometry. Surface texture parameters such as
the average roughness Ra and peak to valley roughness Rt dramatically
change with the wear. The linear increase in roughness (relative to the
unworn probe) directly varies with the particle size [10]. It is generally
expected that the larger the particles, the greater the observed wear. The
experimental results of the particle size experiments suggest the following:
1. They provide further evidence for the rotating plug model of wear,
suggesting that wear in this process is strongly dependent on
particle size. Within the context of the model, larger particles are
more likely to span the clearance 𝛿 between the probe and the
rotating plug and impinge on the tool surface.
2. From the perspective of material design, the experimental results
demonstrate that one technique to minimize wear in FSW of MMCs
is to select materials with reinforcement particles lying at the upper
end of the FEPA scale.
3. The experimental method adapted based on Colligan’s approach
represents an economical means to study the effect of materials with
specific particle sizes for research purposes that does not require the
production of custom material.
Fig. (7). SEM image (100X) of worn FSW probe. The circumferential, parallel grooves exhibited on
the surface are characteristic of abrasive wear processes.
COMBATTING WEAR: SELECTION OF TOOL MATERIALS
The tools used in the particle tracer experiments were further inspected
using SEM to evaluate the abrasive wear mechanism (Fig. 7). The best way
to prevent wear in this instance is to select a tool material that is harder than
the reinforcement particles. According to classical wear theory, when the
hardness ratio between the tool and the abrasive is greater than one,
abrasive wear cannot occur. The best candidate for a wear-resistant tool
material is diamond. However, monolithic diamond is typically only used
in critical manufacturing applications because of its high cost. Diamond
coatings represent a cost effective alternative. The major barrier to
application of coated tools in FSW is the coating’s tendency to delaminate
under the stresses imposed by the joining process. When delamination
occurs, the wear behavior of the coated tool regresses to levels associated
with its substrate. Hence a strong bond between substrate and coating is
necessary to prevent delamination and preserve wear-resistant
characteristics of the coated tool. The effect of the hardness ratio on wear in
FSW of MMCs was evaluated using four different tool materials (O1 tool
steel, WC, WC micrograin, and WC coated with diamond) in two substrate
materials (Al 359/SiC/20p and Al 359/SiC/30p) at the previously optimized
parameter set 𝜔 = 1000 RPM and 𝑣 = 3 IPM (the brittle behavior of ceramic
and refractory metals significantly narrows the process window for FSW).
A higher rotation rate coupled with a comparatively slower traverse speed
tends to provide sufficient heating to reduce the of tool failure. Fig. (8)
compares the cumulative wear of the candidate materials over a 4 feet of
length of weld. The wear resistance of the coated tool is the best of the
materials considered but the margin between the performance of the
diamond and WC-Co probes is small [12].
WC/Co micrograin
WC/Co Submicrograin
30% reinforcement
20% reinforcement
O1 Steel
Diamond*
0 10 20 30 40
Fig. (8). Relationship between the wear percentage and the tool material, where each cylinder
represents the wear percentage measured for the tool material after three 14” long welds (at 1000
RPM/3 IPM) in an Al 359 MMC with either 20% or 30% SiC reinforcement [12].
IN SITU SENSING OF WEAR USING TORQUE
An inspection of the torque equation derived from the rotating plug model
(equation 5) indicates that radial deterioration of the probe during joining of
MMCs by FSW should coincide with a decrease in the magnitude of the
torque the tool experiences during welding. Simulations carried out by
Gibson, Prater, et al. showed that in situ estimates of wear could be made
with an adaptive torque controller [13]. According to Nunes’s torque
equation developed based on the rotating plug model, the torque signal is
sensitive to flow stress 𝜎, temperature, plunge depth t, as well as geometric
parameters (shoulder and pin radii R and r, respectively). These additional
(and sometimes coupled) relationships make it more difficult to isolate
changes in the torque signal that can be attributed solely to radial loss of
tool material. Nonetheless, in-process sensing merits further investigation
since such a capability would minimize disruption of the manufacturing
process for wear evaluation (currently tools must be taken off-line to assess
wear) and in some cases be tantamount to flaw detection. Fig. (9) depicts
the evolution of the torque signal (measured from motor current) over a 6ft
weld of Al 359/SiC/20p for a conventional tool material (O1 steel hardened
to RC 50). The steady-state torque signal for each pass was related to the
amount of wear. Over the course of several experiments where longer
panels are welded, the results demonstrated the change in torque signal due
to wear is both detectable and a viable technique for wear prediction. One
primary application of such wear detecting technology would be longer
length welds of reinforced Al alloys where an off-line in media res
evaluation of wear is not feasible.
2𝜋𝑅3 𝑟2𝑡
𝑇𝑡𝑜𝑡𝑎𝑙 = (1 + 3 )𝜎 (5)
3 𝑅3
Fig. (9). Motor current (Amperes) versus time (sec) for 6ft long weld of Al 359/SiC/20p at 1400
RPM and 9 IPM. Purple shaded regions represent successive 12” weld passes along the panel.
CONCLUSION
The ultimate goal of the present study is to extend the application of metal
composites to aerospace structures by developing process models and
techniques to mitigate wear. MMCs have a broad spectrum of properties

which make them good candidate materials for aerospace applications.
Properties such as strength, fatigue life, and wear resistance for MMCs are
superior to conventional, unreinforced Aluminum alloys. In addition,
MMCs are customizable: the composition of an MMC (the matrix alloy, the
reinforcement material, and the percent in which this reinforcement is
present) can be optimized to produce a material with desired mechanical
properties. While an MMC has approximately the same weight as an
unreinforced Aluminum alloy, its higher strength can significantly decrease
the amount of material required for a specific design. Weldability and
machinability remains the primary obstacle to full application of MMCs in
aerospace structures. The work presented in this chapter achieves a
fundamental understanding of this wear process and provides a base of
knowledge for controlling, sensing and mitigating wear in FSW of MMCs.
Despite its impact on joining metal composites and other high-strength

alloys, to date tool wear in FSW has not been the subject of many focused
investigations. This study presents a starting point for understanding wear
phenomena in this application. The significances of this work include: 1)
identification of the wear mechanism; 2) development of empirical models
that predict the extent of tool wear based on process parameters and
material properties; 3) advancement of tool materials and coatings to
prevent and/or mitigate wear; and 4) evaluation of the feasibility of sensing
techniques which can provide more information about the condition of the
tool in-process. The collective results of these studies provide an informed
strategy for material selection of both tool and workpiece materials to
minimize tool wear. A combination of careful material selection, control of
process parameters, and monitoring of torque and other process signals that
correlate with wear can potentially reduce volume loss to a degree that
enables the production of defect free welds. The development of a
fundamental understanding of wear phenomena in FSW is necessary and
essential for control and mitigation of wear processes and stands to
accelerate implementation of these materials in aerospace industry
applications where they would be of maximum benefit [14].
this publication.
ACKNOWLEDGEMENTS
Declared none.
REFERENCES
[1] "Space Shuttle Technology Summary: Friction Stir Welding", Available from:
www.nasa.gov/centers/marshall/pdf/104835main_friction.pdf, 2009.
[2] D. Storjohann, O.M. Barabash, S.S. Babu, and S.A. David, "Fusion and
friction stir welding of aluminum metal matrix composites", Metall. Mater.
Trans., A Phys. Metall. Mater. Sci., vol. 36A, pp. 3237-3247, 2005.
[http://dx.doi.org/10.1007/s11661-005-0093-4]
[3] T. Prater, "Solid state joining of metal matrix composites: a survey of
challenges and potential solutions", Mater. Manuf. Process., vol. 26, pp.
636-648, 2011.
[http://dx.doi.org/10.1080/10426914.2010.492055]
[4] T. Prater, G.E. Cook, A.M. Strauss, J. Davidson, and M. Howell,
"Parameterization of friction stir welding of Al 6061/SiC/17", In: Proceedings
of the 8th International Conference on Trends in Welding Research, , 2008, pp.
102-108.
[5] L.M. Marzoli, A.V. Strombeck, J.F. Dos Santos, C. Gambaro, and L.M.
Volpone, "Friction stir welding of an AA6061/ Al2O3preinforced alloy",
Compos. Sci. Technol., vol. 66, pp. 363-371, 2006.
[6] P. Cavaliere, E. Cerri, L. Marzoli, and J. Dos Santos, "Friction stir welding of
ceramic particle reinforced Aluminium based metal matrix composites",

Appl. Compos. Mater, vol. 11, pp. 247-258, 2004.
[http://dx.doi.org/10.1023/B:ACMA.0000035478.71092.ec]
[7] T. Prater, A. Strauss, G. Cook, C. Machemehl, P. Sutton, and C. Cox,
"Statistical modeling and prediction of wear in friction stir welding of a metal
matrix composite (Al 359/SiC/20p)", J. Manuf. Tech. Res., vol. 2, pp. 1-13,
2010.
[8] T. Prater, C. Cox, B. Gibson, A. Strauss, and G. Cook, "Dimensional
analysis and a potential classification algorithm for prediction of wear in
friction stir welding of metal matrix composites", Proc. Inst. Mech. Eng., C
J. Mech. Eng. Sci., vol. 226, pp. 2759-2769, 2012.
[http://dx.doi.org/10.1177/0954406212438987]
[9] T. Prater, A. Strauss, G. Cook, B. Gibson, and C. Cox, "A phenomenological
model for tool wear in friction stir welding of metal matrix composites",
Metall. Mater. Trans., A Phys. Metall. Mater. Sci., vol. 44, pp. 3757-3764,
2013.
[http://dx.doi.org/10.1007/s11661-013-1701-3]
[10] T. Prater, Predictive Process Modeling of Tool Wear in Friction Stir Welding
of Metal Matrix Composites, PhD thesis, Vanderbilt University, US, 2012.
[11] K. Colligan, "Material flow behavior during friction stir welding of
aluminum", Supplement to The Welding Journal, pp. 229-237.
[12] T. Prater, A. Strauss, G. Cook, B. Gibson, and C. Cox, "A comparative
evaluation of the wear performance of various tool materials in friction stir
welding of metal matrix composites", J. Mater. Eng. Perform. vol. 22, pp.
1807-1813, 2013.
[http://dx.doi.org/10.1007/s11665-012-0468-9]
[13] B. Gibson, G. Cook, and T. Prater, "Adaptive torque control of friction stir
welding for the purpose of estimating tool wear", Proc. Inst. Mech. Eng., B J.
Eng. Manuf., vol. 225, pp. 1293-1303, 2011.
[http://dx.doi.org/10.1177/2041297510393629]
[14] T. Prater, "Friction stir welding of metal matrix composites for use in
aerospace structures", Acta Astronaut., vol. 93, pp. 366-373, 2014.
[http://dx.doi.org/10.1016/j.actaastro.2013.07.023]
SUBJECT INDEX
A unsteady 3, 6, 9
Aerodynamic models 7, 23
Abrasive particles 553, 554, 555 Aerodynamic nonlinearities 7, 70
Absorbed energy 114, 125, 127, 130, Aerodynamics 7, 68, 70, 298, 398, 414,
133, 135, 144, 148 415, 416, 417, 420, 424
Absorbent material vaporizes 305, 332 Aerodynamic study 67
Acoustic emission (AE) 357, 477 Aeroelastic analysis 6, 7, 25, 26, 29, 32
Adhesive joints 154, 155, 160, 161, Aeroelastic behavior 3, 7, 29, 36, 41, 56
165, 170 Aeroelasticity 4, 7
Adhesive layer 159, 160, 163, 164, 166, Aeroelastic model 3, 26, 34
170, 173, 174, 175, 177, 182, 183, Aeroelastic problem 5, 7, 9
184, 188, 190, 191, 193, 195, 198, Aeroelastic stability 8, 29, 31, 52
199, 201, 205, 207, 212, 213, 216, Aeroelastic stability boundaries 3
219, 220, 222, 223, 225, 228, 229, Aerospace designers 443, 449, 451,
230, 231, 232 452, 455, 463
thin 166, 175, 177, 191 Aerospace hydraulic systems 116
Adhesive layer thicknesses 155, 160, Aerospace industry 303, 307, 355, 356,
164, 166, 170, 198, 199, 205, 206, 387, 445, 446, 469, 470, 475, 483,
208, 216, 220, 223, 229, 231 484, 486, 498, 500, 501, 502, 504
Adhesive material 154, 155, 160, 163, Aerospace materials 451, 463, 464, 469,
164, 165, 170, 181, 196, 200, 208, 486, 505
209, 223 Aerospace nanotechnology 469
Adhesive thicknesses 155, 160, 164, Aerothermoelastic problems 6
173, 183, 185, 190, 194, 195, 196, AFP technology 518
197, 198, 199, 200, 202, 204, 205, AH-1W Helicopter 288
206, 207, 209, 214, 215, 216, 217, Aircraft 66, 67, 70, 76, 307, 311, 397,
218, 219, 220, 221, 222, 223, 224, 406, 414, 415, 416, 417, 418, 421,
225, 227, 228, 229, 230, 231, 232 423, 426, 443, 444, 445, 446, 447,
effects of 154, 160, 163, 164, 165, 448, 449, 450, 451, 452, 455, 457,
196, 208 458, 459, 461, 469, 470, 473, 476,
Aerodynamic coefficients 25 477, 479, 481, 484, 485, 486, 499,
Aerodynamic derivatives 26, 27, 395, 502, 503, 504, 505, 543
422 military 446, 458, 459, 470, 479,
Aerodynamic forces 13, 66, 67, 72, 74, 484, 485
77, 92, 107, 110, 421, 422 Aircraft design, advanced 451
Aerodynamic lift 9, 13, 24 Aircraft designer 449
Aerodynamic loads 3, 5, 6, 7, 9, 41 Aircraft engines 241
Yucheng Liu (Ed.)

All rights reserved-© 2016 Bentham Science Publishers
Subject Index Frontiers in Aerospace Science 56
Aircraft flight period 107 Bending loads 216, 531, 532

Aircraft icing mitigation 469, 475 Bifurcating behaviors 21
Aircraft materials 67, 449 Bifurcation theory 21
Aircraft structures 68, 448, 450, 451, Bifurcation types 21
531 Blade chord 399, 401, 409
Aircraft wing model 66, 70, 86, 89, 110 Blade inlet angle 246, 248
Airfoil model 70, 71, 100 Blade selection 405, 409, 410
Airfoil model’s 101 Bonded joints 154, 155, 157, 159, 160,
Airfoil response 108 161, 165, 166, 177, 178, 192, 207,
Airfoil system 68, 71, 72, 104, 107 211, 219, 230, 231
Alloy(s), aluminum (AA) 278, 289, Bonded steel 189, 219, 220, 225, 226,
355, 375, 376, 430, 431, 443, 445, 227
446, 450, 451, 453, 454, 455, 456, Boundary conditions 7, 88, 175, 244,
463, 545, 547, 561 254, 257, 258, 262
Alloys 243, 244, 358, 443, 445, 446, Boundary values 259
451, 452, 453, 454, 455, 456, 458, Brittle materials 157
461, 462, 463, 464, 543, 545, 547, Broad frequency spectrum 22
560, 561 Buckling 449, 516, 517, 519, 520, 521,
metal 543, 547 522, 525, 531
titanium 443, 445, 446, 451, 454, bending-induced 520, 522, 525, 531
461, 463 Buckling capacity 520, 521, 523, 524,
unreinforced aluminum 561 525, 532, 533, 534, 536
Analytical stress-strain model 327 Buckling load 519, 520, 523, 530, 532,
Aspect ratio 23, 471, 472, 517, 527, 529 533, 536
Astronautics 470, 482, 487 Buckling mode shapes 525, 534, 535
Automated fiber placement (AFP) 516, Buckling performance 516, 519, 520,
517, 518 527, 530, 531, 532, 533, 534, 536
Average adhesive thickness 184, 185
C
B
CAD model 99, 248, 249, 251, 252,
Back-wall signals 366, 367, 368 255, 257
Based adherends of specimens 184, 185 Camera, high resolution CCD 189, 193,
Bending control parameters 245, 246, 194, 204, 212
248, 249 Carbon-fiber reinforced polymer
Bending deformation 245, 248, 250, (CFRP) 446, 455, 457, 458, 463,
251, 252 464, 499, 500
56 Frontiers in Aerospace Science Yucheng Liu
Carbon fibers 459, 498, 500, 504, 507, Composite cylinders 516, 517, 520,
509, 510, 511 522, 523, 524, 525, 527, 530, 532,
Carbon nanotube paper 476 533, 534, 535, 536
Carbon nanotubes 469, 470, 471, 472, Composite materials 446, 454, 470,
475, 481, 482 474, 477, 485, 499, 501, 516, 518,
Card, internal pressure curve 120 534, 546
Casting-die profile 241 Composite structures 517, 518
Catastrophic failures 157 Compression process 503, 507
Chaff/Flare dispenser bracket 280, 288, Compressive residual stress (CRS) 303,
291, 292 304, 309, 311, 315, 320, 322, 327,
Chaotic response 8, 39, 41, 42, 52 330, 331, 336, 342, 343, 348, 351
Chaotic systems 22 Computed radiography (CR) 357, 422
Chaotic vibrations 8, 22 Computer-aided design (CAD) 265, 429
Chord length 72, 88, 250, 251 Computer modeling and simulation
Circumferential Location 535 115, 271
CNT implementation 483, 484, 485 Constant model, piece-wise 521, 522
CNTs in aerospace applications 474, Constant stiffness (CS) 12, 517, 519,
487 520, 534, 535, 536
Cohesive fracture 157, 190, 191 Constitutive models 327, 332, 339
Cohesive laws 154, 155, 160, 165, 171, Control command system 405
174, 198, 204, 215, 218, 229 Controllers, designing 415, 416
Cohesive strength 155, 160, 230 Conventional ultrasonic 357, 362, 364,
Cohesive zone models (CZMs) 154, 365
155, 159, 160, 161, 165, 173 Coordinate value, optimized 249
Collapse mechanism 115 Cowper-Symonds relation 118
Commercial aircraft 475, 484, 485, 500, Crack growth 157, 161
504, 505 Crack initiation 201, 202, 204, 215,
Components 155, 241, 243, 271, 286, 304, 308, 330
303, 305, 307, 308, 310, 314, 315, Crack propagation 160, 163, 198, 202,
316, 320, 322, 324, 329, 330, 344, 215, 218, 225, 228, 229, 231, 315,
345, 358, 369, 411, 416, 421, 422, 445, 450
429, 432, 435, 436, 437, 450, 478, Crack tip 157, 158, 159, 162, 163, 172,
501, 502, 504, 512, 513, 546 173, 175, 176, 177, 178, 179, 180,
metal 501, 502, 513 181, 188, 190, 192, 193, 201, 210,
treated 344, 345 211, 212
section of 175, 177, 178, 179 safe 450

Crack tip area 222, 228 Design equation 449
Crack tip separation 194, 195, 212, 229, Design experiments method 395
231 Design optimization 516, 523, 524, 531,
local 195, 212, 229, 231 532
Crack tip slip, local 202, 203, 214 multi-objective 531, 532
Crew exploration vehicle (CEV) 544, Design parameters 401, 404, 415, 416
546 Design process 135, 405, 415, 450, 531
Criticality number 396, 436, 438 Design relationships 443, 448, 449
CT subsonic aircrafts 445 Double cantilever beam (DCB) 163,
Cumulatively uni-axial testing 270 164, 174, 175, 182, 190, 229
Ductile materials 157, 159
D Dynamical systems 20, 21, 22, 421
Dynamic model, structural 71
Damping 28, 34, 35, 36, 48, 49, 57 Dynamic modeling 395, 415, 421
aeroelastic 28, 35, 36, 48, 49, 57 Dynamic response 8, 41, 45, 48, 57, 70,
Damping ratios 56, 66, 73, 75, 89, 90, 71, 270
91, 105, 107, 110 Dynamics 67, 281, 299, 414, 415, 416,
Darker densities 384, 385 421, 422, 428
DCB specimens 175, 176, 183, 184, structural coupling 67
187, 188, 190, 229 Dynamic structural response 72
Deformation 116, 125, 126, 135, 159,
203, 213, 241, 242, 245, 248, 254, E
255, 261, 296
torsional 241, 245, 248 Eddy current testing 357, 358
Deformation error 265 Elastic displacements 4, 10
Deformed configuration of FEA model Elastic fins 7, 9, 32, 34, 35, 36
121, 122 Elastic rotations 5
Degrees-of-freedom model 78, 79, 80, Electrical engines 405
81, 82, 83, 84, 85, 104, 110 Electromagnetic particles 273
Dense periodic orbits 22, 23 Elements, piezoelectric 359, 364, 369
Design 394, 398, 401, 404, 429, 450 Elliptical cylinders 520, 533, 534, 535,
conceptual 394, 398, 401 536
detailed 394, 398, 404, 429
preliminary 429
End notched flexure (ENF) 161, 163, Fatigue life prediction 308, 309, 310,
164, 175, 177, 182, 229, 230 313
Energy release rate (ERR) 157, 158, Fatigue loading 309, 314, 315, 349
193, 195, 203, 211, 212, 214 Fatigue prediction methods 313
Energy release rate component JI 180, Fatigue strength 278, 304, 314, 445,
181 446
ENF specimen 178, 186, 189, 201, 206 Fatigue tests 270, 271, 281, 291, 296,
Engine and fuel system 395, 405 299, 432
Engine gears 411, 412 FEA models 116, 118, 119, 120, 121,
Engine group 430 122, 125, 149, 316, 336
Engines 241, 395, 405, 406, 407, 408, FEA software 287
411, 414, 429, 431, 438, 446 Fiber, glass 169, 504, 507, 509, 510,
electric 405, 406, 407, 408 511
piston 405, 406, 407, 408 Fiber content 502, 506, 507, 508, 509,
turbine 405, 407 510
Equations 3, 10, 21, 89, 169, 170, 423, Fiber orientation angle 517, 521
424, 426 Fiber-reinforced polymer composite
characteristic 21, 423, 424, 426 (FRPC) 517
governing 3, 10, 89, 169, 170 Fiber reinforcements 500, 501, 502,
Estimated forces 78, 82, 92, 93, 94, 95, 509, 512
96, 97, 98, 110 continuous 509, 512
Estimated forces and moments 78, 82, continuous carbon 500, 501, 502
92, 96, 110 Fighter aircraft loading standard for
Excitation system 71 fatigue (FALSTAFF) 310
Expansion waves 15, 16, 17 Finite element analysis (FEA) 114, 122,
Explicit FEA solver 116 124, 135, 138, 139, 145, 150, 160,
244, 270, 276, 311, 327, 331, 523,
F 524
Finite element model 34, 66, 68, 70,
Failure mode effect analysis (FMEA) 299, 336, 338, 521
435 Flap and plunge direction 39, 40, 41,
Fatigue analysis 271, 280, 281, 290, 43, 44, 46, 47, 50, 51, 54, 55
291, 292, 296, 298, 300, 311, 322, Flapping motions 9, 418
344 Flexural strength 502, 505, 506, 507,
Fatigue failures 271, 298, 304, 309, 508, 509
327, 346 Flight missions 406, 407, 408
Flight regimes 408, 418
Flight tests 271, 274, 278, 288, 289, Frequency domain methods 281, 282
290, 432, 434 Frequency response functions (FRFs)
Flow, supersonic 11, 15, 16, 27 66, 77, 78, 82, 92, 99, 101, 104,
Flow properties 6, 15, 18 107, 108
Flutter divergence 30, 31, 32 Frequency response matrix 74, 75, 76,
Force determination error 93, 94, 95, 77, 99
97, 98, 99 Frequency separation 413
Force determination methods 66, 68, Friction stir (FS) 370, 376, 377, 378,
75, 77, 78, 86, 92, 110 385, 386, 387, 556
Forces 4, 10, 13, 14, 66, 72, 74, 76, 77, Friction-stir-welding (FSW) 355, 376,
78, 82, 83, 92, 94, 96, 99, 101, 543, 544, 546, 547, 549, 553, 556,
107, 114, 126, 139, 190, 201, 210, 558, 559, 561
229, 230, 231, 349, 358, 396, 417, FRPC materials 517
420, 422, 424 FSW, application of 547
actual 66, 74, 107 FSW of MMCs 548, 549, 552, 553,
critical 229, 230, 231 554, 557, 558, 561
excitation 72, 74, 78, 101, 107 FSW process 546
generalized 4, 10 Fuel consumption 407, 475, 476
Forward flight 398, 404, 405, 416, 417, Fuel systems 395, 405
418, 423, 426, 427, 428 Functional composite components 500
Fracture, basic modes of 156
Fracture energy 159, 160, 163, 164, G
173, 192, 207, 211, 219, 225, 230
Fracture energy JIIC 202, 215, 218, Gear ratio 408, 410, 411, 412, 413
219, 20, 221, 224 Geometrical parameters 3, 29, 36
estimated 218, 220, 221, 224 Geometric characteristics 401, 402, 438
Fracture process 158, 160, 165 Geometries 11, 15, 27, 29, 33, 160, 163,
Fracture strength 229, 230, 231, 232 184, 185, 196, 261, 271, 293, 299,
Fracture tests 155, 173, 174, 187 308, 311, 316, 337, 342, 347, 358,
Fracture toughness JIC 229, 231, 232 450, 513
Free-play 3, 4, 5, 8, 9, 11, 13, 14, 36, concave 316, 342
37, 56, 57 curved 308, 347
Free-stream surface 18 Ground vibration test (GVT) 71, 75, 76,
Frequency domain 69, 70, 73, 89, 108, 101, 102, 108, 110
276, 281, 299
Group 184, 185, 190, 191, 192, 193, Impact response 116, 133, 135, , 142,
164, 195, 196, 198, 200, 201, 202, 145, 150
203, 204, 205, 207, 210, 216, 217, lateral 116, 142, 145, 150
218, 429, 499, 547, 552 Impact simulations 119, 120, 122
specimen in 190, 191, 192, 194, 201, Impact strength 505, 506, 507, 509, 511
202, 204 Impulse response functions (IRFs) 104
Inclinometer 188, 191, 211
H Index offset 373, 374, 388
Industry, modern 135, 154, 155
Helicoids hole 100 Initial angle of attack 3, 24, 25, 26, 27,
Helicopter components 271, 272 29, 35, 45
High-temperature CFRP 458 Initial crack 186, 194, 204
Hinge frictional torque 3, 56 Initial crack length 184, 185, 210, 211,
Hovering flight 408, 409, 417, 418, 212, 214, 215, 217, 222, 223, 224,
423, 424, 426 225, 227, 228, 231, 232
H Plunging displacement 4 Initial crack tip 157, 192, 193, 194, 202,
Hugoniot elastic limit (HEL) 318, 328, 204, 211, 212, 214, 222, 229
335, 336, 341 Initial crack tip length 222, 228
Hydraulic system of aerospace 115, 150 Integration process 270, 274, 280
Hydrophobicity 469 Integrity, structural 355, 356, 475
Interfacial strength 155, 173, 198, 200,
I 207, 209, 218, 219, 225, 229, 231
Interfacial strength σmax 198, 200, 218,
Impact 118, 119, 133, 134, 145, 150, 225, 227
329, 543, 556, 561 Interfacial traction-separation laws 154,
mid-position 133, 134, 150 160, 161, 164, 173, 198, 209, 225
quarter-position 133, 134, 150 Interfacial traction-separation laws of
Impact behavior 116, 117, 149, 511 bonded steel 219, 225, 226
Impact energy 117, 134, 135, 145 Interlayer, adhesive 158, 163, 164
Impact force 125, 126, 135 Internal defect 377, 378, 385, 386
Impact forces, higher 134, 145 Internal pressure 114, 115, 116, 117,
Impact hammer 101, 102, 103 119, 120, 125, 127, 131, 133, 134,
Impact positions 114, 116, 125, 127, 135, 139, 142, 145, 149, 150
128, 129, 130, 133, 145, 150 effects of 114, 116, 131, 142
Impact resistance 134, 145, 149, 181 influence of 116, 117, 133, 142
Inverse deformation 242
Investment casting 241, 242, 243, 251, M

255, 264
Isotropic fibre distribution 499 Magnetic particle testing 357
Material behavior 308, 334
L Material combinations 504, 507, 508,
Lack of penetration (LOP) 382, 383 509, 510, 511, 513
Lagrangian dynamics 9 Material nonlinearities 7, 114
Laser peening process 311, 314, 316, Material parameters 314, 346, 359
324, 332, 333, 334, 339, 340, 341 Material properties 33, 118, 119, 254,
Lateral motion 423, 424, 425, 426 289, 316, 371, 463, 523, 534, 548,
Lateral wave (LW) 366, 367, 368 554, 561
Laws of bonded steel 220, 226, 227 Materials 100, 116, 155, 156, 157, 159,
Least squares complex exponential 160, 170, 181, 271, 275, 299, 305,
(LSCE) 103, 104 306, 312, 316, 318, 320, 327, 329,
Length of welds 548, 552, 555, 558 330, 331, 332, 333, 334, 335, 336,
Lightweight materials 498, 545 338, 339, 340, 341, 343, 344, 346,
Limit cycle oscillations (LCOs) 8, 20, 347, 351, 358, 359, 360, 361, 362,
22, 38, 48, 52, 57, 71 366, 369, 371, 375, 443, 444, 448,
Linear elastic fracture mechanics 449, 450, 451, 458, 461, 462, 463,
(LEFM) 157, 158, 159, 161 464, 469, 470, 471, 473, 475, 480,
Liquid penetrant testing 357 484, 485, 486, 500, 501, 504, 508,
Loading conditions 119, 120, 165, 271, 513, 517, 534, 543, 544, 546, 547,
313, 351, 443, 448, 449 548, 549, 554, 555, 557, 558, 561,
Loading force, critical 229, 230, 231 562
Loading modes 161, 162 peened 305, 312, 331
Loadline displacements 191, 192, 193, plasticized 546, 548
194, 201, 202, 210, 211 reinforcement 504, 544, 547, 561
Local maxima 37, 38, 57 Materials parameters 229, 230, 231
Local piston theory 4, 24, 25, 29 Materials selection 443, 450, 452
Local tests 154, 160, 163, 164 Material stirring 547, 548, 549
limited 154 Material surface 306, 308, 339
Longitudinal modes 423, 424, 426, 427 Mathematical model 414, 415, 417
Longitudinal wave 359, 360, 366 Maximum buckling capacity 521, 526,
Low cycle fatigue 327, 332, 346 531
LS-DYNA 114, 115, 116, 119, 120 Maximum deformation 127, 129, 261
Lyapunov exponent(s) 20, 22, 23, 38, Maximum deformation of pipeline
39, 41, 42, 49, 52, 53 models struck 132
Maximum fatigue life 303, 308, 309, empirical 312, 331, 561
318, 319, 324 Modeling 5, 9, 137, 160, 274, 318, 327,
Maximum impact force 127, 128, 133, 337, 347, 405, 415, 480, 520
134, 142, 145, 146, 149, 150 Modern toughened adhesives 157
Maximum Lyapunov exponent 22, 23 Mode shapes 4, 5, 10, 34, 66, 89, 90,
Maximum transverse displacement 134, 107, 110, 275, 299
150 natural 4, 5, 10, 34
Mean time between failures (MTBF) second 89, 90
395, 435 Molten aluminum 543, 547
Mean time to failure (MTTF) 395, 435 Monte carlo method 286
Measurement locations 273, 276, 278, Morozov’s discrepancy principle 69
279 Multi-axial testing 271, 281
Metal fatigue 270 Multi-degree-of-freedom model 70, 88,
Metal matrix composites (MMCs) 543, 92, 93, 94, 95, 96, 97, 98, 99, 110
544, 547, 548, 549, 552, 553, 554, Multiple-degree-of-freedom system 66
557, 558, 559, 561 Multiple regression model (MRM) 551,
Metamodeling based design 552
optimization (MBDO) 516, 520,
536 N
Method of die-cavity 241
Microchips to large aircrafts 155 Nanomaterials 474, 485
Mid-span duration 316, 317, 320 Nanotube diameter 471, 472
Mid-span impact 138, 140, 142, 143, National aeronautics and space
144, 145, 146, 147, 148, 149 administration (NASA) 474, 476,
Mild-steel pipelines 135 543, 546, 548
Military aircraft wings 461, 462, 463 NDE techniques 355, 356, 357, 358
Military standards 270, 288, 294 Non-destructive evaluations 356
Model 3, 7, 11, 28, 29, 32, 66, 70, 71, Non-linear aeroelastic behavior 3
87, 99, 100, 101, 103, 104, 110, Nonlinear dynamical systems 20
118, 120, 126, 127, 160, 161, 242, Non-periodic responses 22
255, 274, 275, 286, 293, 312, 313, Numerical models 155, 160, 161, 287,
316, 327, 336, 338, 341, 346, 409, 327, 350
414, 415, 416, 417, 421, 422, 431, Numerical simulations 114, 149, 242,
544, 551, 553, 554, 555, 557, 562 252, 255, 263, 266, 346
cowling 274, 275
O Plasma 305, 306, 332, 333

Plastic deformation 159, 164, 190, 201,
Ohio supercomputing center (OSC) 338 297, 298, 306, 318, 333, 336, 338
Operational flight tests 274, 278 Plastic dissipations 160, 162, 170, 190,
Optimum design 135, 255, 395, 524 192, 196, 202, 211, 215, 220
Orientation angles 518, 519, 521, 522, Plasticity mechanisms 117, 150
526, 535 Plastics, reinforced 498, 499, 500
Oscillations 23, 39, 46, 52, 359, 424, Plastic strain 163, 335, 336, 339
425, 427 Plies 517, 518, 521, 522, 526, 528, 530,
531, 535
P steered 517, 526, 530, 531
Plies candidate 520, 522, 530
Parameterization studies 548 Plug, rotating 545, 553, 554, 557
Parvan 395, 398, 399, 400, 401, 404, Plug model, rotating 544, 553, 554,
405, 407, 408, 412, 418, 419, 422, 556, 557, 559
423, 424, 425, 426, 428, 429, 431, Plunge direction 39, 40, 41, 43, 44, 46,
433, 434, 436, 437, 438 47, 50, 51, 54, 55
components of 436, 437 Polyimides 459, 463, 464, 483
Parvan components dimensions 402, Power Spectral Density (PSD) 276,
403 277, 281, 282, 283, 284, 285, 288,
PAUT system 371, 372, 386, 388 289
Peel strength 228 Power transmission system 395, 408,
Peened components 303, 308, 309, 313, 410
318, 329, 330, 332, 344 Prandtl-Meyer theory 16
Phased array ultrasonic testing (PAUT) Precautions, recovery 436, 437, 439
355, 356, 357, 362, 363, 369, 372, Preferred system concept (PSC) 476
373, 375, 376, 377, 385, 386 Pressure pulse 306, 332, 333, 334, 337,
Phase plane diagram and time histories 340
39, 40, 41, 43, 44, 46, 47, 50, 51, Pressurized pipelines 114, 115, 116,
54, 55, 59, 60, 61, 62 117, 119, 120, 121, 122, 133, 145,
Phthalonitrile resin system 459, 460 149, 150
Pipeline models struck 127, 131, 132, Probability Density Function (PDF)
133 281, 284, 285, 286, 287, 289, 395
Pipeline’s impact force 116 Problem, axial fatigue 332, 346, 350,
Piston theory 3, 4, 6, 9, 14, 15, 23, 24, 351
25, 29
local 4, 24, 25, 29
Procedure, modified explicit 338 Reverse design methodology 241, 242

Process, random 282, 285 Root mean square (RMS) 286
PSD function 282, 283, 285 Root mean square error (RMSE) 137,
PSD matrix 289 139
Pseudo-inverse method 78, 92, 99, 110 Rotational, estimated 79, 81, 83, 84, 85,
93, 95, 96, 97, 98
Q Rotational speeds 410, 411, 432, 433
Rotorcraft 469, 475, 480, 484, 485
QI designs 529, 530, 531 Rotor diameter 399, 402, 403
Quaritc polynomial functions 138
Quarter span impact 141, 142, 143, 144, S
145, 146, 147, 148, 149
Quartic response surfaces 140, 141 Sea level (SL) 34, 36, 56, 399
Quasi-isotropic (QI) 519, 525, 526, 527, Secondary Hopf 21
528, 529, 533, 534, 535, 536 Shape form error 242, 265
Shear strength 209, 220, 221, 460
R Sheet metal model 296
Sheet molding compound 498, 499,
Radar cross section (RCS) 479 500, 501, 502, 505, 507, 508, 509,
Radiographic testing (RT) 356, 357, 510, 512, 513
358, 378, 379, 385, 386 Sheet molding compound formulations
Radiography 355, 357, 378 508, 509, 510, 512
Rayleigh dissipation function 4, 10, 11 Sheet molding compound materials 512
Regions, antenna integration 274, 275 Sheet molding compounds 499, 501,
Reinforcement particles 547, 555, 557, 509, 512
558 Shock and expansion analysis 4
Relative rotation angle 223, 228 Shock waves 7, 306, 312, 332, 333,
Relaxation behavior 331, 332, 345, 350 334, 335, 337, 338, 339, 340
Relaxation effect 309, 312, 327, 331, Shrinkage factors (SFs) 244
332, 345, 346, 349, 351 Simulation models 308, 310, 351
Relaxation models 317, 322, 323 Simulation process 257, 258, 261
Response functions 78, 92, 136, 139 Skin temperature 452, 457, 459
Response surface method (RSM) 135, SLB specimen 178, 179, 180, 181, 186
136, 138 Small-scale turbofan engines 405
Response surfaces 114, 127, 133 Solidification 241, 243, 244, 255
Sorted k-fold approach (SKA) 309, 320,
324
Sound waves 359, 360, 361, 369, 370, Stress relaxation model 316, 327
371, 372, 375 Stress spectrum 311, 314, 316
high frequency 369, 370 Stress-strain behavior 332, 346, 347
Space elevator 469, 482 Stress values 277, 278, 286, 299, 300
Specimens 160, 162, 172, 175, 177, Structural components 303, 304, 308,
178, 182, 183, 184, 185, 190, 195, 309, 316, 323, 324, 330, 344, 448,
196, 198, 209, 210, 211, 212, 214, 450, 451, 463, 464, 503, 517, 518,
215, 216, 218, 229, 230, 231, 281, 544
331, 339, 340, 347, 372, 378, 381, Structural improvement 519, 520, 530
508 Structural materials 445, 447, 454, 464,
effective 185, 190 545, 548
five 195, 196, 198 Structural member 444, 449
five groups of 184, 185 Structural model 3, 7, 9, 29, 71
groups of 183, 184, 185, 190 Structural nonlinearities 7, 8, 70, 300
SST aircrafts 446, 455, 458, 459 Structural performance 516, 517, 519
Stability derivatives 416, 420, 424 Structural responses 66, 67, 74, 75, 270,
Stacking sequence 410, 525, 530, 533, 523
535 Structures, bonded 154, 156, 165
Stagger angle of CAD model 248 Subsonic aircrafts 445, 452, 453, 454
Static mass moment 5, 12 Subsonic aircraft’s structures 452, 453,
Steel adherends 164, 209, 212 454
3-D simulation 395 Subsonic structures 445, 452
Strain rate effects 118, 335, 339 Supersonic aerodynamic characteristics
Strain rates 118, 328, 339 6
Stress115, 156, 176, 177, 179, 181, 261, Supersonic cruise conditions 447
304, 318, 320, 330, 331, 348 Supersonic fins 7, 29, 31, 52
normal 176, 177, 179, 181 Supersonic flight 451, 462, 463
tensile 115, 156, 261, 304, 318, 320, Supersonic/hypersonic flow 5
330, 331, 348 Supersonic regime 6
Stress corrosion cracking (SCC) 444, Surface enhancement methods 333, 344
448 Surface enhancement techniques 304,
Stress field, residual 316, 338, 345, 347 308, 329, 330
Stress intensity factor (SIF) 157, 444,
450 T
Stress range 285, 286
Stress relaxation 308, 344, 345, 346, Tail rotor diameter 399, 401, 403
347, 348, 350 TCG calibration 374, 388
Techniques 6, 115, 136, 162, 356, 357, Traction-separation laws 154, 161, 163,
373, 381, 386, 387, 416, 431, 549, 164, 194, 198, 199, 202, 212, 213,
556, 557, 560, 561 214
traditional inverse 79, 93 equivalent interfacial 163, 194, 198,
Temperature effects 306, 339, 341 199, 202, 212, 213, 214
Temperature resistance 458, 459, 461 local interfacial 154, 161, 163, 164
Tensile modulus 507, 509, 510 Traditional pseudo-inverse technique
Tensile strength 181, 200, 458, 460, 77, 92
472, 502, 505, 506, 507, 508, 509, Traffic monitoring 397, 398, 400, 401,
510, 511 438
Tests, wind tunnel 26, 68, 71, 99, 107, Translational force 80, 82, 84, 85, 94,
108, 110, 432 95, 97, 98
Test specimens 161, 187, 188, 359, 508 Transmission system design 411, 412,
Theoretical model 8, 134, 165 413
Theory, classical beam 165, 166, 167 Transmitting system 405, 408, 411, 413
Thermal conductivity 470, 471, 472 Transverse deformations 18
Thick fins 3, 6, 15, 16, 32 Transverse displacements 133, 150
Thick supersonic fin 4, 6, 9, 36, 57 Truncated singular value decomposition
Ti-6Al-4V material 327, 340 (TSVD) 69
Time-corrected-gain (TCG) 374 Turbine blade 241, 242, 243, 245, 249,
Time derivatives 37, 56 252, 254, 255, 261, 262, 264, 265,
Time-of-flight-diffraction 362, 366, 368 308
Time of flight diffraction (TOFD) 357, Two degree-of-freedoms pitch-plunge
362, 363, 366, 368 system 72
Time series, torsional angle 37, 41 Two-step optimization strategy 303,
Titanium 445, 446, 457, 471, 505, 547 309, 319
Tool materials 544, 548, 558, 559, 561
Tool probe 549, 555, 556 U
Tool surface 554, 555, 557
Tool wear 544, 552, 556, 561 Ultimate tensile strength (UTS) 444,
Torque 396, 408, 410, 411, 428, 544, 455, 460
559, 561 Ultrasonic testing (UT) 357, 358, 362,
Torsional control parameters 245, 248 364
Torsional shear stresses 444, 449 Uni-axial fatigue testing error 271
Traction-separation 229, 230, 231 Unmanned aerial vehicles (UAVs) 395,
397, 405, 414, 469, 470, 480, 481,
482
Unsteady aerodynamic loading models X

5
X-ray images 381, 385, 386
V X-ray radiography 355, 385, 386, 387
Vibration analysis 274, 276, 278 Z

Vibration fatigue analysis 283, 285,
288, 289 Zone 155, 157, 158, 163, 165, 171, 202,
Vibration fatigue method 282, 287, 298, 215
299 cohesive 158, 165, 171
Vibration levels 274, 278, 280 failure process 165
plastic 155, 157, 158, 163, 202, 215
W
Weight saving 476, 483, 519

Weight sizing 399, 438
Weld defects 355, 356, 357
Weld image 382, 383, 384, 385
Welding defects 356, 357, 369, 376,
382
The welding institute (TWI) 543, 546
Wing model 71, 87, 99, 100, 101, 102,
110
24 Frontiers in Aerospace Science, Vol. 1 Firouz-abadi et al.

Aerospace Structures and Materials (Liu - 2016)

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

Aerospace Structures and Materials (Liu - 2016)

Uploaded by

Copyright:

Available Formats

Frontiers in Aerospace Science

Editor: Yucheng Liu

ISSN (Online): 2468-4724

ISSN: Print: 2468-4716

ISBN (eBook): 978-1-68108-305-6

ISBN (Print): 978-1-68108-306-3

© [2016], Bentham eBooks imprint.

Published by Bentham Science Publishers – Sharjah, UAE. All Rights Reserved.

Bentham Science Publishers Ltd.

LIST OF CONTRIBUTORS .................................................................................................................................... ii

CHAPTER 1 ANALYSIS OF LINEAR/NON-LINEAR AEROELASTIC RESPONSE OF SUPERSONIC

CHAPTER 2 AN ANALYTICAL AND EXPERIMENTAL INVESTIGATION INTO VIBRATORY FORCE

CHAPTER 3 COMPUTATIONAL AND ANALYTICAL INVESTIGATION OF LATERAL IMPACT

CHAPTER 4 EFFECT OF BONDLINE THICKNESS ON THE TRACTION-SEPARATION LAWS OF

CHAPTER 6 VIBRATION INDUCED FATIGUE ANALYSIS OF AEROSPACE STRUCTURES .............. 269

CHAPTER 8 PREDICTION OF RESIDUAL STRESS RELAXATION IN TI-6AL-4V SUBJECTED TO

CHAPTER 9 NON-DESTRUCTIVE EVALUATION (NDE) OF WELDED STRUCTURES FOR

CHAPTER 11 MATERIALS SELECTION IN DESIGN OF STRUCTURES OF SUBSONIC AND

SUBJECT INDEX .................................................................................................................................................... 564

Department of Mechanical and Industrial Engineering, Louisiana State

F. Kermanshahi Ecole Polytechnique de Montreal, Quebec, Canada

Department of Mechanical & Industrial Engineering, Louisiana State

Institute of Polymer Materials and Plastics Engineering, TU Clausthal,

Department of Mechanical & Industrial Engineering, Louisiana State

Concordia Center for Composites, Department of Mechanical and Industrial

Department of Aerospace Engineering, Sharif University of Technology,

Institute of Production Engineering, Helmut Schmidt University, Hamburg,

Kun Bu Department of Mechanical Engineering, Mississippi State University, USA

Institute of Production Engineering, Helmut Schmidt University, Hamburg,

Concordia Center for Composites, Department of Mechanical and Industrial

Mehmet Çelik Department of Mechanical Engineering, SST Division-ASELSAN Inc., Turkey

Concordia Center for Composites, Department of Mechanical and Industrial

Department of Mechanical and Industrial Engineering, Louisiana State

Department of Mechanical and Industrial Engineering, Louisiana State

Department of Mechanical Engineering, MGEO Division-ASELSAN Inc.,

M. Mohagheghi Aerospace Engineering Department, University of Tehran, Tehran, Iran

Department of Aerospace Engineering, Sharif University of Technology,

Aerospace Engineering Department, Amirkabir University of Technology,

Nicole Stöß Polynt Composites Germany GmbH, Kieselstraße, Miehlen, Germany

N. Goudarzi Mechanical Engineering Department, University of Maryland, USA

Omid Gohardani Springs of Dreams Corporation, Tustin, USA

Department of Mechanical and Materials Engineering, Wright State University,

Department of Aerospace Engineering, Sharif University of Technology,

School of Mechanical, Industrial, and Manufacturing Engineering, Oregon State

Concordia Center for Composites, Department of Mechanical and Industrial

Department of Aerospace Engineering, Sharif University of Technology,

Sustainable Environment and Energy Systems Department, Middle East

Tracie Prater NASA Marshall SpaceFlight Center, Huntsville, USA

Xueguang Bi Department of Mechanical Engineering, Mississippi State University, USA

Yangliu Dou Department of Mechanical Engineering, Mississippi State University, USA

Yangqing Dou Department of Mechanical Engineering, Mississippi State University, USA

Yiwei Dong Department of Mechanical Engineering, Mississippi State University, USA

Yucheng Liu Department of Mechanical Engineering, Mississippi State University, USA

Department of Mechanical Engineering, King Abdulaziz University, Jeddah,

Department of Mechanical & Industrial Engineering, Louisiana State

Analysis of Linear/Non-Linear Aeroelastic

Department of Aerospace Engineering, Sharif University of Technology,

Abstract: This study introduces an aeroelastic model for the linear/non-linear

*Corresponding author R.D. Firouz-abadi: Department of Aerospace Engineering, Sharif University of

Yucheng Liu (Ed.)

Keywords: Aeroelasticity, Free-play, Local piston theory, Shock and

Cα, Cβ, Ch Structural damping.

Commonly, an aeroelastic problem has two types of modeling including

aerodynamic tools in supersonic regime is the piston theory. This theory