ASIP 2014

Dec 2-4, San Antonio, TX

Prediction of Fatigue Crack

Propagation from Cold-Expanded
Holes
Matt Watkins, Ricardo Actis
Engineering Software Research & Development, Inc., St. Louis, Missouri
Scott Prost-Domasky
Analytical Processes/Engineered Solutions, Inc., St. Louis, Missouri
Robert Pilarczyk
Hill Air Force Base, A10-ASIP, Layton, Utah

Approved for Public Release, Reference No. 75ABW-2014-0045

© 2014 ESRD, Inc. All Rights Reserved. StressCheck® is a registered trademark of ESRD, Inc.
Outline
 Context
• Rapid Innovation Fund
 Model of crack propagation
• Background info
• Input/output data
• Modeling assumptions
• Sources of error
 Crack propagation tool
• Input/output
• Predictions
• Effect of uncertainties in the predictions
 Summary and Conclusions

12/3/14 © 2014 ESRD, Inc. 2

Rapid Innovation Fund
 Rapid Innovation Fund
• Contract FA9453-12-C-0218
 USAF
• A-10, T-38, and F-16 ASIP Offices
• AFRL
 Team Members
• APES, Inc.
• TRI/Austin
• ESRD
• Hill Engineering
• Lextech

12/3/14 © 2014 ESRD, Inc. 3

RIF Objectives

 Establish/refine framework for modeling

fatigue crack growth through residual
stress fields within scope of ASIP.
• AFGROW for crack growth
• StressCheck for stress intensity factors (SIFs)
and crack trajectory calculations
 Validate model performance
 Provide quality assurance tools and
inspection guidance for cold-worked holes.
12/3/14 © 2014 ESRD, Inc. 4
Integrated Package

Contour Method
Data
Residual
Stress Residual
SIFs • Crack growth life
BAMF /
StressCheck • Crack growth “path”
AFGROW • Inspection Intervals
Geometry
Detected Crack Size or aNDI
AFGROW

Laser Metrology NDT for

for Degree of CX Cracks

Depot Data

12/3/14 © 2014 ESRD, Inc. 5

Simulation
Input Background Model solution Model solution
data information (data of interest) (numerical approximation)

(D, I) F Fnum
Physical Mathematical Numerical Prediction
reality model solution

Errors of
approximation

Conceptualization FEA Extraction

12/3/14 © 2014 ESRD, Inc. 6

Background Information
 Cold-expanded (CX) fastener holes have been shown to
provide substantial increase in fatigue life.
• Compressive residual stress (RS) slows down crack propagation
of corner flaws around CX holes.
 Experiments show that cracks at cold-worked holes
preferentially evolve at the mandrel entrance face.
• The mandrel entrance face tends to be more difficult to inspect;
cracks may grow to become critical before they are detectable.
 How to account for the beneficial effect of RS-CX in the
computation of fatigue crack propagation life?
• To support the development of a inspection intervals that are not
overly conservative, a mathematical model for simulating crack
growth at a cold worked hole was developed.

12/3/14 © 2014 ESRD, Inc. 7

Input Data
 Given
• Specimen with a cold-worked (CX) hole
• Initial crack size in the hole
• Constant amplitude or variable amplitude remote loading
• Material properties of the specimen (E, 𝜈)
• Residual stress distribution in the plane of the crack (RS)
• Crack propagation data (da/dN – Δ𝐾 curves)

Crack
length bore

Crack length surface

12/3/14 © 2014 ESRD, Inc. 8

Model Output
 Compute
• The crack trajectory from the initial pre-crack to
final failure
• The crack length vs. number of cycles curve
• The expected life of the specimen

12/3/14 © 2014 ESRD, Inc. 9

Conceptualization
Modeling Assumptions
 The driver of fatigue crack propagation is the
stress intensity factor amplitude DKI (LEFM)
• The principle of superposition is applicable.
• Residual stresses only affect the R (load ratio) for
the load cycle.

12/3/14 © 2014 ESRD, Inc. 10

Conceptualization
Modeling Assumptions
 The relation between the rate of crack growth and
Δ𝐾 is obtained from experimentally-derived da/dN
– Δ𝐾 curves at fixed load ratios.
• Curves derived for thru-cracks are applicable for
corner cracks.
• Interpolation is used for finding da/dN for load ratios
not covered by the experimental data.

12/3/14 © 2014 ESRD, Inc. 11

Conceptualization
Modeling Assumptions
 The direction of crack growth is normal to the crack
front
• The crack grows in the plane of the initial flaw.
• Each point along the crack front has a load ratio which
depends on the RS at the point.
• Only the residual stresses normal to the plane are needed
to compute 𝐾𝑟𝑒𝑠 .
 No assumptions regarding the shape of the crack.

12/3/14 © 2014 ESRD, Inc. 12

Sources of Error

Sources of Error – Crack Propagation Model

Aleatory Epistemic Numerical
(random) (model-form) Approximation
Initial stresses before Driver of crack Computation of SIFs
CX propagation with RS
Residual stresses after Direction of crack Computation of SIFs
CX extension remote load
da/dN-DK data Interpolation and fitting Numerical Integration
of test data to find DN
Number of control
points to define the
crack shape
12/3/14 © 2014 ESRD, Inc. 13
Sources of Error
Aleatory Model-form Approximation
uncertainties uncertainties Errors

(D, I) F Fnum
Physical Mathematical Numerical
Prediction
reality model solution

Errors of
approximation

Conceptualization FEA Extraction

12/3/14 © 2014 ESRD, Inc. 14

Crack Propagation Tool
Input – Dimensions

Input
options

Basic
dimensions of
the specimen

12/3/14 © 2014 ESRD, Inc. 15

Crack Propagation Tool
Input – Dimensions

Input
options

Basic
dimensions of
the specimen

12/3/14 © 2014 ESRD, Inc. 16

Crack Propagation Tool
Input – Material

Select Linear or Harter-T

Import da/dN data interpolation (assessment
and material of model-form
properties uncertainty)
from AFGROW
*.lkp file format

12/3/14 © 2014 ESRD, Inc. 17

Crack Propagation Tool
Input – Residual Stresses

Import tabular RS
data or pull from
RS database

12/3/14 © 2014 ESRD, Inc. 18

RS Database

 Residual stress database rs_database.dll

• Accessible from any .Net or COM-enabled
software
• Includes stand alone RS Visualizer for
visualization and interpolation
• Select from measured RS distributions
• Interpolate between measured RS distributions
 Interpolation parameters: CW%, thickness, edge
distance, hole diameter
• Populate the database with your own proprietary
data in ASCII file format
12/3/14 © 2014 ESRD, Inc. 19
RS Database

Contact Bob Pilarczyk for access details

robert.pilarczyk@us.af.mil

12/3/14 © 2014 ESRD, Inc. 20

Crack Propagation Tool
Input – Load
Import load
spectrum
from AFGROW
*.sub file format

12/3/14 © 2014 ESRD, Inc. 21

Crack Propagation Tool
Input – Solver Options

Initial crack
parameters

Size of Da for integration

Multiple solution runs to

check errors in SIFs

12/3/14 © 2014 ESRD, Inc. 22

 Crack Propagation Tool
Output

Crack front display

Crack
step
selection
list Details for
each crack
step

Failure
indicators

12/3/14 © 2014 ESRD, Inc. 23

 Crack Propagation Tool
Output

Crack front display

Crack
step
selection
list Details for
each crack
step

Failure
indicators

12/3/14 © 2014 ESRD, Inc. 24

Crack Propagation Tool
Sample Results

12/3/14 © 2014 ESRD, Inc. 25

Crack Propagation Tool
Sample Results

12/3/14 © 2014 ESRD, Inc. 26

Experimental Comparisons
No RS (Not Coldworked)

* Robert Pilarczyk, Experimentally Derived Beta Corrections to Predict Fatigue Crack Growth at Cold Expanded Holes in
7075-T651, Master Thesis, Department of Mechanical Engineering, University of Utah, May 2008
12/3/14 © 2014 ESRD, Inc. 27
Experimental Comparisons
Cold-Worked & Pre-Cracked
Surface Pre- Bore Pre- Fatigue Life,
Coupon
crack, in. crack, in. Cycles
1 0.02810 0.03486 121312
2 0.02166 0.04136 65890
3 0.02200 0.03204 126587
4 0.02032 0.02936 180066

Prediction
Fatigue Life
(121,312)

12/3/14 © 2014 ESRD, Inc. 28

 Experimental Comparisons
Sensitivity to RS
Predicted Life [Cycles] Specimen
Life
90% RS 100% RS 110% RS [Cycles]

78,913 137,305 266,500 121,312

+/-10% variation in RS
3.4
+ 94%
Prediction  137,305
– 43%

12/3/14 © 2014 ESRD, Inc. 29

Summary & Conclusions
 The underlying mathematical model was clearly
defined.
• Inputs, outputs, assumptions
 For in-plane crack propagation only the component of
the residual stress tensor normal to the plane is
required.
 𝐾𝑅𝐸𝑆 is computed independently of 𝐾𝑀𝐸𝐶𝐻
• 𝐾𝑅𝐸𝑆 and 𝐾𝑀𝐸𝐶𝐻 are computed using the superconvergent
contour integral method.
• 𝐾𝑅𝐸𝑆−𝑁𝑈𝑀 and 𝐾𝑀𝐸𝐶𝐻−𝑁𝑈𝑀 are obtained by the finite
element method at each control point along the crack front
using StressCheck.

12/3/14 © 2014 ESRD, Inc. 30

Summary & Conclusions
 The proposed model accounts for the
beneficial effects of RS in the calculation of
fatigue life of cracks in CX holes.
• Good estimates of crack propagation life
• Crack propagation maps consistent with
experimental observations
 Prediction of fatigue life is strongly affected
by the statistical dispersion of the residual
stresses.
• 10% in RS  a factor of 3 or more in fatigue life
estimates.

12/3/14 © 2014 ESRD, Inc. 31