Академический Документы
Профессиональный Документы
Культура Документы
Eric N. Quillen
A Thesis
Presented to
the Faculty of the Graduate School
Tennessee Technological University
by
Eric N. Quillen
In Partial Fulfillment
of the Requirements for the Degree
MASTER OF SCIENCE
Mechanical Engineering
May 2005
Signature
Date
iii
DEDICATION
This thesis is dedicated to my wife Julie, whose encouragement has been critical
in the completion of my graduate degree and the composition of this thesis.
iv
ACKNOWLEDGMENTS
I would like to thank the following people for their help with this work: Dr.
Chris Wilson, Dr. Phillip Allen, Mike Renfro, Krishna Natarajan, and Richard
Gregory. I would also like to thank my employer, Fleetguard, Inc., and coworkers. Without their cooperation, it would not have been possible for me to perform
this research.
TABLE OF CONTENTS
Page
LIST OF TABLES . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
ix
LIST OF FIGURES . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
xiii
LIST OF SYMBOLS
xx
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Chapter
1.
2.
INTRODUCTION
. . . . . . . . . . . . . . . . . . . . . . . . . . . . .
TECHNICAL BACKGROUND
. . . . . . . . . . . . . . . . . . . . . .
. . . . . . . . . . . . . . . . . . . . . . . . . . . .
2.1 J-Integral
. . . . . . . . . . . . . . . . . . . .
17
RESEARCH PROCEDURE . . . . . . . . . . . . . . . . . . . . . . . .
19
19
. . . . . . . . . . . . . . . . . . . . . .
21
21
3.1.1.2 FEA-Crack . . . . . . . . . . . . . . . . . . . . . . .
22
. . . . . . . . . . . . . . . . . . . . . . .
24
26
vi
vii
Chapter
Page
3.3.1 Load . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
26
. . . . . . . . . . . . . . . . . . . . .
27
31
31
35
37
. . . . . . . . . . . . . . . . . . . . . . . . . . .
. . . . . . . . . . . . . . . .
38
. . . . . . . . . . . . . . . . . . . . . . . .
40
40
41
41
. . . . . . . . . . . . . . . . . . .
43
RESULTS . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
49
49
. . . . . . . . . . . .
49
50
51
. . . . . . . . . . . . . . . . . . .
67
86
4.
. . . . . . . . . . . . . . . . . . . . . . . .
91
. . . . . . . . . . . . . . . . . . . . . . . . . . .
93
viii
Chapter
5.
Page
4.5.1 Height Effects . . . . . . . . . . . . . . . . . . . . . . . .
93
95
. . . . . . . . . . . . . . . . . . . . . . . . . . . 104
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 108
APPENDICES
A: INSTRUCTIONS FOR MESH3D SCP MODIFICATIONS . . . . . . . . . 113
B: COARSE VERSUS REFINED MESHES FOR K-FACTORS
. . . . . . . 115
LIST OF TABLES
Table
Page
2.1
. . . . . . . . . . . . .
14
2.2
. . . . . . . . . . . . .
14
2.3
15
2.4
. . . . . . . . . . . . . . . . . . . .
16
2.5
16
3.1
Number of nodes and elements in the duplication of the Kirk and Dodds
[23] geometries . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
31
3.2
. . . . . .
33
3.3
. . . . . . . . . . . . . . . . . .
36
3.4
39
3.5
. . . . .
40
3.6
Stress vs. plastic strain data at n = 15, used for ABAQUS models . . .
47
3.7
Stress vs. plastic strain data at n = 10, used for ABAQUS models . . .
47
3.8
. . .
48
4.1
50
4.2
. . . . . . . . . . . . . .
56
4.3
57
82
4.4
ix
x
Table
4.5
Page
Maximum percent differences between McClung et al. [15] and Lei [17]
solutions (quarter symmetry) . . . . . . . . . . . . . . . . . . . . . .
83
4.6
. . . . .
94
4.7
. . . . . . . . . . .
95
4.8
. . . . . . . . . . .
96
4.9
. . . . . . . . . . . .
96
D.1
Model 4 (a/t=0.5 and a/c=0.2): h1 values for at different heights (Part 1) 134
D.2
Model 4 (a/t=0.5 and a/c=0.2): h1 values for at different heights (Part 2) 135
D.3
. . . 135
D.4
. . . 136
E.5
. 138
E.6
. 139
E.7
. . . . . . 140
E.8
. . . . . . 141
E.9
. 142
E.10
. 143
E.11
. . . . . . 144
E.12
. . . . . . 145
E.13
. 146
E.14
. 147
xi
Table
Page
E.15
. 148
E.16
. 149
E.17
. 150
E.18
F.19
. . . . . . . . . . 153
F.20
. . . . . . . . . . 154
F.21
. . . . . . . . . . 155
F.22
. . . . . 156
F.23
. . . . . 157
F.24
. . . . . . . . . . 157
F.25
. . . . . . . . . . 158
F.26
. . . . . 159
F.27
. . . . . 160
F.28
. . . . . . . . . . 161
F.29
. . . . . . . . . . 162
. . . . . . 151
. . . . . . . . 164
. . . . . . . . 165
xii
Table
G.35 Stress vs. plastic strain data at n = 5, used for ABAQUS models
Page
. . . 169
LIST OF FIGURES
Figure
Page
2.1
2.2
. . . . . . . . . . . . . . . . . .
2.3
12
2.4
Close up of the finite element mesh around the crack front used by
McClung et al. [15] . . . . . . . . . . . . . . . . . . . . . . . . . . .
12
3.1
. . . . . . . . . . . . . .
20
3.2
. . . . . . . . . . . . . . . . . . . . . . .
20
3.3
3.4
. . . . . . . . . . . . . .
22
. . . . . . . . . . . . . . . . . . . . . .
23
3.5
23
3.6
. . . . . . . . . . .
24
3.7
. . . . . . . . . . . .
26
3.8
Fully plastic element set consisting of the elements around the crack tip
28
3.9
29
3.10
29
3.11
30
3.12
Geometries used by Kirk and Dodds for estimating the J-Integral [23] .
32
3.13
Stress vs. strain curve for Kirk and Dodds elastic-plastic models [23] . .
34
3.14
41
xiii
. . . . . . . . . . . . . . . . . . . .
xiv
Figure
Page
3.15
42
3.16
44
3.17
Elastic, modified elastic, and Ramberg-Osgood stress vs. strain curves for
n = 10 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
45
51
52
52
53
53
54
54
55
55
4.10
. . . . . . . . . . . . . . . . . . . . . . .
58
4.11
58
4.1
4.2
4.3
4.4
4.5
4.6
4.7
4.8
4.9
xv
Figure
4.12
Page
59
4.13
Model 1 (a/t = 0.2, a/c = 0.2): Reduced vs. full integration elements
60
4.14
Model 8 (a/t = 0.8, a/c = 0.6): Reduced vs. full integration elements
61
4.15
63
4.16
63
4.17
64
4.18
Model 6 (a/t = 0.5, a/c = 1.0): K-Factor results for half symmetry model 64
4.19
Model 8 (a/t = 0.8, a/c = 0.6): K-Factor results for half symmetry model 65
4.20
67
4.21
68
4.22
68
4.23
69
4.24
69
4.25
70
4.26
70
4.27
71
4.28
71
4.29
72
4.30
72
xvi
Figure
Page
4.31
73
4.32
73
4.33
74
4.34
74
4.35
75
4.36
75
4.37
76
4.38
76
4.39
77
4.40
77
4.41
78
4.42
78
4.43
79
4.44
79
4.45
80
4.46
80
4.47
4.48
4.49
87
xvii
Figure
4.50
Page
87
88
88
89
89
Elastic,
Ramberg-Osgood,
modified elastic,
and modified
Ramberg-Osgood stress vs. strain curves for n = 10 . . . . . . . . . .
90
Model 1 (a/t = 0.2, a/c = 0.2): Normalized K-factor vs. angle along
crack front . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
91
4.57
Model 1 (a/t = 0.2, a/c = 0.2): h1 vs. angle along the crack front
92
4.58
Model 1 (a/t = 0.2, a/c = 0.2): h1 vs. c/w for Nasgro and FEM at n = 15 97
4.59
Model 1 (a/t = 0.2, a/c = 0.2): h1 vs. c/w for Nasgro and FEM at n = 10 97
4.60
Model 1 (a/t = 0.2, a/c = 0.2): h1 vs. c/w for Nasgro and FEM at n = 5
4.61
Model 3 (a/t = 0.2, a/c = 1.0): h1 vs. c/w for Nasgro and FEM at n = 15 98
4.62
Model 3 (a/t = 0.2, a/c = 1.0): h1 vs. c/w for Nasgro and FEM at n = 10 99
4.63
Model 3 (a/t = 0.2, a/c = 1.0): h1 vs. c/w for Nasgro and FEM at n = 5
4.64
Model 4 (a/t = 0.5, a/c = 0.2): h1 vs. c/w for Nasgro and FEM at n = 15 100
4.65
Model 4 (a/t = 0.5, a/c = 0.2): h1 vs. c/w for Nasgro and FEM at n = 10 100
4.51
4.52
4.53
4.54
4.55
4.56
. . .
98
99
xviii
Figure
Page
4.66
Model 4 (a/t = 0.5, a/c = 0.2): h1 vs. c/w for Nasgro and FEM at n = 5 101
4.67
Model 6 (a/t = 0.5, a/c = 1.0): h1 vs. c/w for Nasgro and FEM at n = 15 101
4.68
Model 6 (a/t = 0.5, a/c = 1.0): h1 vs. c/w for Nasgro and FEM at n = 10 102
4.69
Model 6 (a/t = 0.5, a/c = 1.0): h1 vs. c/w for Nasgro and FEM at n = 5 102
B.1
B.2
B.3
B.4
B.5
B.6
B.7
B.8
C.9
xix
Figure
Page
LIST OF SYMBOLS
Symbol
Description
a
aef f
b
c
ds
G
h1
h2
h3
n
nj
r
rc
t
w
x1
x2
y1
y2
A
CT OD
E
In
J
Jel
Jpl
Jtotal
K
Knorm
P
Po
Ti
ui
W
Crack depth
Effective crack length, includes plastic zone
Uncracked ligament length
Half crack length
Increment of length along the contour
Strain energy release rate
Dimensionless parameter used to calculate Jpl
Dimensionless parameter used to calculate CT OD
Dimensionless parameter used to calculate p
Strain hardening exponent
Unit vector components normal to
Crack tip radius
Radius of projected circle
Specimen thickness
Half specimen width
Distance along x-axis for projected circle
Distance along x-axis for projected circle
Distance along y-axis for projected circle
Distance along y-axis for projected circle
Crack area
Crack tip opening displacement
Youngs Modulus
Integration constant
Elastic-plastic fracture parameter
Elastic portion of the J-integral
Plastic portion of the J-integral
Sum of Jel and Jpl
Stress intensity factor
Normalized K-factor
Applied load
Limit load
Traction vector
Displacement vector
Specimen width
Dimensionless Ramberg-Osgood material constant
Plasticity constraint factor
Load line displacement
xx
xxi
Symbol
Description
ij
o
ref
ij
o
ref
EPFM
EPRI
FEA
FEM
LEFM
ODB
Strain
Strain tensor
Yield strain
Reference strain
Contour
Potential Energy
Reference stress factor
Poissons ratio
Strain energy density
Stress
Stress tensor
Yield stress
Reference stress
Angle of crack tip
Elastic plastic fracture mechanics
Electric Power Research Institute
Finite element analysis
Finite element model
Linear elastic fracture mechanics
Output data base
CHAPTER 1
INTRODUCTION
Fracture mechanics is the study of the effects of flaws in materials under load.
Modern fracture mechanics was originated by Griffith [1] in the 1920s when he successfully showed that fracture in glass occurs when the strain energy resulting from
crack growth is greater than the surface energy. In 1948, Irwin [2] extended Griffiths
strain energy release rate, G, to include metals by accounting for the energy absorbed
during plastic material flow around the flaw. By 1960, the fundamental principles of
linear elastic fracture mechanics (LEFM) were in place ([3, 4], for example).
LEFM is used to predict material failure when response to the load is elastic
and the fracture response is brittle. LEFM uses the strain energy release rate G or
the stress intensity factor K as a fracture criterion. K solutions for many geometries
have been calculated in the past and are widely available [5]. However, the design
parameters for some components violate the assumptions of LEFM. For example,
high temperatures and limited high stress cycles before component replacement are
factors that can cause significant plastic deformation and a ductile failure. In these
cases, where the LEFM approach is not valid, an elastic-plastic fracture mechanics
(EPFM) approach is required.
EPFM had its beginnings in 1961, when Wells [6] noticed that initially sharp
cracks in high toughness materials were blunted by plastic deformation. Wells proposed that the distance between the crack faces at the deformed tip be used as a
2
measure of fracture toughness. The stretch between the crack faces at the blunted
tip is known as the crack tip opening displacement (CT OD).
In 1968 Rice [7] developed another EPFM parameter called the J-integral by
idealizing the elastic-plastic deformation around the crack tip to be nonlinear elastic.
The J-integral was shown to be equivalent to G for linear elastic deformation and to
the crack tip opening displacement for elastic-plastic deformation. During the same
year, Hutchinson [8], Rice, and Rosengren [9] showed that J was also a nonlinear
stress intensity parameter. The J-integral can be used as an elastic-plastic or fully
plastic crack growth fracture parameter, much like K is used as an elastic fracture
parameter.
The J-integral can be calculated using several experimental and analytical
techniques. The analytical techniques include the Electric Power Research Institute
(EPRI) estimation scheme, the reference stress method, and finite element methods.
It should be noted that many of the analytical techniques that do not directly require
finite element methods were established using finite element analysis.
There are three goals in this research. The primary goal is to develop threedimensional finite element analysis (FEA) J-integral results using ABAQUS. These
results will be compared to existing solutions. The second goal is to investigate
the effect of various finite element modelling parameters on the resulting J-integral.
These parameters include mesh density, element type, symmetry, and specimen size
effects. The third goal is to compare incremental plasticity FEAs that utilize a stress
vs. plastic strain table based on a power law hardening material with the deformation
plasticity solution for a power law material. This comparison will be made in an
3
attempt to see if the fully plastic results using a deformation plasticity model can be
approached by a series of increasing loads in an incremental plasticity model.
The finite element models (FEMs) used in this research were three-dimensional
flat plates with surface cracks. The plates contained various surface crack, height, and
width geometries. Because of the dual symmetry, only one quarter of each plate was
modeled. Meshes from two different mesh generation programs were used: mesh 3d
(Faleskog, 1996) and FEA-Crack from Structural Reliability Technology.
CHAPTER 2
TECHNICAL BACKGROUND
In this chapter the J-integral and different J-integral calculation methods will
be examined. The chapter begins with a discussion of the theory and mathematical
foundation of the J-integral. Next, two methods for calculating the J-integral are
discussed: the EPRI Estimation Scheme and the reference stress method. Both of
these methods can be implemented using hand calculations without an extensive
fracture mechanics background. In addition, both of these methods are incorporated
into Nasgro, a fracture mechanics and fatigue crack growth program. Finally, the
FEA method is used in this research, but a review is not included here. There are
many excellent texts on the subject of FEA (for example Cook et al. [10]).
2.1 J-Integral
ui
ds ,
dy Ti
x
(2.1)
where is the strain energy density, Ti are components of the traction vector, ui
are the displacement vector components, and ds is an increment of length along the
ij dij
(2.2)
and
Ti = ij nj ,
(2.3)
where ij is the stress tensor, ij is the strain tensor, and nj are unit vector components
normal to .
In idealizing elastic-plastic behavior to be the same as nonlinear elastic material
behavior, Rice assumed that the material stress versus strain curve followed a power
law relationship. The Ramberg-Osgood equation is commonly used to describe the
stress and total strain data for this type of material response:
=
+
o
o
n
,
(2.4)
6
where is the total material strain, o is the reference stress (normally defined as the
yield strength, but not necessarily the same as the 0.2% offset yield strength), o is
the strain at the reference stress and is defined by o = o /E. There are two other
material constants in Equation 2.4. The first of these, , is a dimensionless constant,
and the second, n, is the strain hardening exponent (n 1).
The J-dominated elastic-plastic stress field contains a singularity of order
1
r n+1 . For the elastic case (n = 1), this singularity reduces to r 2 in agreement
with the K-dominated field of LEFM. The following two equations were derived by
Hutchinson [8], Rice and Rosengren [9] and are called the HRR singularity. The HRR
singularity describes the actual stresses and strains near the crack tip and within the
plastic zone as
ij = o
EJ
o2 In r
1
n+1
ij (n, )
(2.5)
and
o
ij =
E
EJ
o2 In r
n
n+1
ij (n, ) ,
(2.6)
the angle at a point around the contour, and ij and ij are functions of n and .
Equations 2.5 and 2.6 are important because the J-integral determines the stress
amplitude within the plastic zone. This fact establishes J as a fracture parameter
under conditions of plastic deformation.
7
Rice [7] also showed that the J-integral is equivalent to the energy release rate
in a nonlinear elastic material containing a crack:
J =
d
dA
(2.7)
where is the potential energy and A is the area of the crack. For linear elastic
deformation:
K2
Jel = G = 0
E
(2.8)
E
,
(1 2 )
(2.9)
E0 =
E 0 = E.
(2.10)
Care should be taken when using the energy release rate with elastic-plastic
or fully plastic deformation. In an elastic material, the potential energy is released
as the crack grows. In an elastic-plastic material, a large amount of strain energy is
used in forming a plastically deformed region around the crack tip. This energy will
not be recovered when the crack grows, or when the specimen is unloaded [4].
8
2.2 EPRI Estimation Scheme
(2.11)
where Jtotal is the total J, Jel is the elastic portion, and Jpl is the plastic portion. For
small loads, Jel is much larger than Jpl . For large loads with significant deformation,
Jpl dominates. This situation is shown graphically in Figure 2.2. As discussed previously, elastic-plastic behavior is idealized to follow a nonlinear elastic path along the
stress versus strain curve.
Jel = G =
K 2 (aef f )
,
E0
(2.12)
aef f
1
1
=a+
2
1 + (P/Po )
n1
n+1
KI
o
2
,
(2.13)
where a is the half crack length, P is the applied load, Po is the limit load per unit
thickness, = 2 for plane stress and = 6 for plane strain, n is the strain hardening
exponent specific to the material, KI is the elastic stress intensity factor, and o is
the reference stress (typically the yield strength).
10
The fully plastic equations for Jpl , crack mouth opening displacement (CT OD),
and load line displacement (p ), applicable for most specimen geometries are
a P n+1
Jpl = o o bh1
,n
,
W
Po
(2.14)
a P n
,n
CT OD = o ah2
,
W
Po
(2.15)
a P n
,n
,
p = o ah3
W
Po
(2.16)
and
P
Po
n+1
,
(2.17)
where, for a center-cracked specimen, a is the half crack length and w is the half
width. Po is the reference or limit load, and is typically the load at which net cross
section yielding occurs. For center-cracked plate in tension,
Po = 4co
(2.18)
11
and
(2.19)
(2.20)
(2.21)
and
12
Figure 2.3 Sample of finite element mesh used by McClung et al. [15]
Figure 2.4 Close up of the finite element mesh around the crack front used by
McClung et al. [15]
13
Although the meshes were created in ANSYS, ABAQUS was used to perform
the analysis of the finite element models. The version of ABAQUS used for this work
was only capable of performing an incremental plasticity analysis. An EPRI-type
scheme was used to separate the elastic and plastic J values. The fully plastic values
for h1 were then calculated using
h1 =
Jpl
n+1 .
o o t o
(2.22)
A combination of three different a/t (0.2, 0.5, 0.8) and a/c (0.2, 0.6, and 1.0)
ratios were tabulated. The specimen geometry ratios were kept constant for all models
at h/c = 4 and c/w = 0.25. The values of h1 were calculated for strain hardening
exponents of n = 5, 10, and 15, and can be found in Tables 2.1, 2.2, and 2.3.
In 2004 Lei [17] duplicated part of the work performed by McClung et al. [15]
by performing elastic and elastic-plastic finite element analyses for plates containing
semi-elliptical surface cracks under tension. The models contained surface cracks with
the same a/t and a/c ratios used by McClung et al. [15]. For the elastic analysis,
Jel results were generated and converted into K using Equation 2.8. These K results
were then compared with Newman-Raju stress-intensity factor calculations [18]. The
elastic-plastic results for strain hardening values of n = 5 and n = 10 were presented
in terms of h1 . These h1 results are reproduced in Tables 2.4 and 2.5 and compare
well with McClung et al. for most geometries. The comparison with McClung et
al. and the current results are presented in more detail in Chapter 4.
a/c
0.20
0.60
1.00
0.20
0.60
1.00
0.20
0.60
1.00
a/c
0.20
0.60
1.00
0.20
0.60
1.00
0.20
0.60
1.00
a/t
0.20
0.20
0.20
0.50
0.50
0.50
0.80
0.80
0.80
a/t
0.20
0.20
0.20
0.50
0.50
0.50
0.80
0.80
0.80
0
0.198
0.324
0.358
2.539
2.319
2.007
17.731
9.688
7.242
0
0.223
0.356
0.389
4.085
3.336
2.774
37.609
17.660
12.667
2.1 McClung et
18
27
0.608
0.821
0.622
0.698
0.628
0.638
11.602 14.488
6.564
7.048
4.750
4.759
82.404 91.460
32.760 30.172
20.882 19.281
2.2 McClung et
18
27
0.523
0.703
0.544
0.604
0.550
0.553
6.957
8.841
4.264
4.561
3.210
3.179
39.550 43.774
16.725 15.850
11.077 10.311
Table
9
0.370
0.465
0.503
7.615
4.808
3.738
63.511
25.890
17.231
Table
9
0.320
0.416
0.450
4.512
3.205
2.599
29.512
13.685
9.472
81
1.398
0.839
0.557
14.712
5.200
2.827
45.805
14.323
7.533
81
1.623
0.971
0.646
22.212
7.922
4.243
95.447
29.095
15.003
90
1.416
0.847
0.556
14.811
5.186
2.804
47.496
14.800
7.625
90
1.644
0.981
0.646
22.309
7.881
4.198
98.941
30.806
15.533
14
a/t
0.20
0.20
0.20
0.50
0.50
0.50
0.80
0.80
0.80
a/c
0.20
0.60
1.00
0.20
0.60
1.00
0.20
0.60
1.00
0
0.164
0.286
0.321
1.325
1.502
1.377
7.224
4.983
3.910
tension, n = 5 [15]
54
63
72
0.897
0.988
1.062
0.605
0.631
0.652
0.447
0.442
0.439
7.136
7.764
8.222
2.917
2.938
2.968
1.800
1.730
1.677
20.632 20.051 18.960
7.034
6.695
6.266
4.080
3.734
3.397
81
1.103
0.666
0.435
8.428
2.973
1.637
18.993
6.114
3.285
90
1.117
0.672
0.435
8.516
2.976
1.630
19.369
6.178
3.270
15
a/c
0.2
0.6
1
0.2
0.6
1
0.2
0.6
1
a/c
0.2
0.6
1
0.2
0.6
1
0.2
0.6
1
a/t
0.2
0.2
0.2
0.5
0.5
0.5
0.8
0.8
0.8
a/t
0.2
0.2
0.2
0.5
0.5
0.5
0.8
0.8
0.8
9
0.3749
0.4508
0.5009
4.723
3.254
2.682
29.51
15.38
10.68
9
0.3003
0.3958
0.4368
2.327
2.03
1.78
11.78
6.938
5.075
0
0.179
0.3151
0.3575
1.343
1.564
1.44
6.723
5.388
4.119
0
0.2169
0.3554
0.3995
2.533
2.379
2.055
16.01
11.08
7.925
72
1.397
0.8627
0.6193
13.66
5.122
2.907
45.11
15.31
8.159
in tension, n = 5
45
54
0.8389 0.9556
0.6134 0.6517
0.4895 0.4912
6.265
7.039
2.897
2.974
1.878
1.837
21.43
21.16
7.656
7.178
4.482
4.106
72
1.132
0.7007
0.4863
8.089
2.993
1.722
19.57
6.396
3.505
h1 values
36
0.7223
0.5777
0.5004
5.455
2.829
1.97
21.2
8.185
5.014
[17]
63
1.053
0.6778
0.4858
7.616
2.98
1.762
20.35
6.633
3.701
81
1.451
0.8786
0.6172
13.91
5.07
2.818
44.93
15.48
8.23
81
1.177
0.7113
0.4825
8.338
2.972
1.676
18.92
6.263
3.396
90
1.475
0.8862
0.618
14.11
5.077
2.799
45.3
15.67
8.408
90
1.196
0.7177
0.4839
8.466
2.981
1.672
18.86
6.29
3.406
16
17
2.3 Reference Stress Method
As discussed previously, the EPRI J estimation scheme assumes that the material has a power law stress-strain curve. There are many materials that do not exhibit
this type of response. In 1984 Ainsworth [19] devised a method for calculating J that
did not depend on the materials behavior following a power law. This approach is
called the reference stress method. The reference stress is defined as
ref =
P
Po
o
(2.23)
where P is the applied load, Po is the same limit load defined previously in the EPRI
research [11], and o is the yield strength.
The reference strain, ref , is defined as the uniaxial strain corresponding to
ref . By inserting ref and ref into the Ramberg-Osgood equation 2.4, it can be
modified to the following form:
ref
ref
=
+
o
o
ref
o
n
.
(2.24)
Using Equations 2.23 and 2.24, Equation 2.14 can be altered to the form
Jpl = ref bh1 ref
ref o
.
(2.25)
Equation 2.25 still contains the variable h1 , a function of n - same h1 used in the EPRI
equations discussed in the previous section. Ainsworths approach was to choose Po
in such a way that the dependence of h1 on n was minimized. For certain values of
18
Po , he found that h1 was relatively constant for n 20. As a result,
h1
= h1
a
w
,1
(2.26)
a
,1
w
The fully plastic solution at n = 1 is identical to the elastic solution using a Poissons
ratio of = 0.5,
K 2 (a) = bh1
a
w
2
, 1 ref
(2.27)
where =1 for plane stress and =0.75 for plane strain. By substituting Equation
2.27 and using the conditions that establish Equation 2.26, the Jpl expression becomes
KI
Jpl =
E
Eref
1 .
ref
(2.28)
The previously discussed McClung et al. [15] finite element results were used
to develop another reference stress method. This reference stress algorithm is used
within Nasgro. Nasgro is a crack propagation and fracture mechanics program developed by NASA and the Southwest Research Institute.
CHAPTER 3
RESEARCH PROCEDURE
In this chapter, the technical approach used for this thesis is presented. The
chapter begins with a discussion of the finite element modeling including mesh generation. Next, the analysis procedure for the FEMs is discussed. Then, the work
duplicated by other researchers is reviewed, and any material properties or model
parameters specific to a geometry set are looked at as well. This duplication of other
researchers work was to validate the methodology used by ensuring that the J-integral
analysis could be performed properly. The chapter concludes with a discussion of the
general material properties used.
The finite element analysis program ABAQUS was used to calculate the Kfactors and J-integrals for a variety of specimen geometries. The models were created
with quarter symmetry to reduce the number of nodes and elements (hence, the
computational time) of each model.
Unless otherwise specified, the FEMs consisted of reduced integration, 20noded brick elements specified as C3D20R within ABAQUS. Reduced integration
elements are recommended in the ABAQUS User Manuals [21] for plastic and large
strain elastic models. Full integration elements tend to be overly stiff and the results
may oscillate. A reduced integration element has a softening effect on the stiffness
that improves the finite element results.
The elements around the crack tip were also of type C3D20R. However, the
elements were modified by collapsing the brick element into a wedge (Figure 3.1).
19
20
When the elements were degenerated, the mid-side nodes were not moved, and the
collapsed nodes were left untied (Figure 3.2). This allows for movement of the nodes
as the element is deformed and produces a 1/r strain singularity, which duplicates
the actual crack tip strain field in the plastic zone [4].
21
3.1.1 Mesh Generation
Two different programs were used to generate finite element meshes. The
first, called mesh3d scp [20] by Faleskog, is available as freeware. Many early finite
element meshes in this work were generated with mesh3d scp. However, this program
has serious limitations. Therefore, a second mesh generation program, FEA-Crack,
was also used. This software is commercially available from Structural Reliability
Technology, Colorado.
3.1.1.1 mesh3d scp.
a one-quarter model of a surface cracked plate. The program assumes that both the
geometry and the load possess planes of symmetry. This program divides the model
into three zones, as shown in Figure 3.3. The element density in each zone is altered by
changing variables in the mesh3d scp input file. The node and element numbering in
each zone is controlled such that the application of boundary conditions and external
loads is simplified. The meshes used to investigate the fully plastic volume and
location were created using mesh3d scp (Figures 3.8 - 3.11).
The program mesh3d scp requires an iterative approach. The set of input
variables for the program input file are changed, the program generates a mesh, the
mesh is plotted and then examined graphically. This process is repeated until a
satisfactory mesh by appearance is created. This program is capable of generating
good meshes for some geometries. However, this program does not work well for other
specimen geometries. For these geometries, mesh3d scp was found to produce a bad
mesh, no mesh, or, in the worst cases, a mesh with errors.
This program was originally written to generate meshes for an earlier version
of ABAQUS. This makes it necessary to modify the ABAQUS input files created by
22
research is called FEA-Crack. FEA-Crack is more robust than mesh3d scp and does
not require the same iterative approach on the users part. The mesh density in the
area around the crack can be controlled by adjusting the program settings. Also, the
generated model may be viewed immediately, and required changes to the ABAQUS
input file are minimal. A mesh created using FEA-Crack is shown in Figures 3.4 and
3.5.
23
Figure 3.5 Close up of mesh from Figure 3.4 created using FEA-Crack
24
3.2 Analysis Procedure
Each FEM analyzed for this research contained 5 contours around the crack
tip, as seen in Figure 3.6. The results for the first contour are generally considered
to be less accurate than the other contours because of numerical inaccuracy [21]. For
this reason, the K-factor and J-integral data from all of the contours, except the first,
were averaged [17]. These average K-factor and J-integral were used for all further
calculations and comparisons.
The FEMs contained multiple node sets along the crack front. A node set
is a group of nodes that have been associated as a group within ABAQUS. The
number of node sets depended on the physical size of the crack front. Each of these
particular node sets contain a number of nodes with the same coordinates. In the
untied condition, one node in each node set is constrained so that it can move in only
25
one or two directions (it stays on the plane of symmetry). The direction of constraint
depends on the symmetry plane. These constrained nodes are listed in another node
set called crack front nodes, which will be significant later. The other nodes in
each node set are not constrained.
ABAQUS generates values for the K-factor and J-integral at each of the node
sets along the crack front. An Excel macro was written to allow for examination of
the variation of the K-factor and J-integral values generated along the crack front.
The program was written to calculate the angle, as projected onto a circle, at each
crack front node. The macro first finds and records the constrained nodes found in
the node set crack front nodes, which is located in the ABAQUS input file. The
coordinates for each of these crack front nodes are then retrieved from the input file.
The crack coordinates are then mapped onto a circle, as shown in Figure 3.7. The
equation for the projection circle is shown below as
(3.1)
Two facts should be noted from Figure 3.7. First, y1 is equal to y2 . Second, the
circle radius, rc , is equal to the crack depth, a. Both of the previous statements are
valid as long as a/c 1, which is the case for this research. Using this information,
Equation 3.1 can now be rearranged into the form
x2 =
q
a2 y12 .
(3.2)
= tan
x2
y1
.
(3.3)
26
Two quantities were initially tested to ensure that the fully plastic FEM results
had converged. The first quantity was load. The second involved the fully plastic
zone specified for the FEMs.
3.3.1 Load
The applied load in the FEMs was adjusted until the resulting J-integral values
did not change with an increase in load. The final load step was also examined for
each model to ensure that the entire load was not applied. In cases where the entire
27
specified load was applied, the load was increased, and the FEM was analyzed again.
This ensured that the specified element set became fully plastic. The fully plastic
option in ABAQUS utilizes a Ramberg-Osgood material model and ends the analysis
when the observed strain for the selected element set exceeds the offset yield strain
by ten times, assuming the load or maximum number of increments have not been
reached. Also, to ensure sufficient steps in the model, the loads were set such that at
least 33% of the specified load was applied to the model.
The volume and location effect of the specified fully plastic element set was
examined for two reasons. First, it was necessary to determine how much of the
specimen must become fully plastic before the J-integral converged. The second
reason was to simplify the model generation. The two mesh generation programs used
in this research, mesh3d scp and FEA-Crack, established convenient, but different,
elements sets for use as fully plastic.
The fully plastic results were generated using the *FULLY PLASTIC command
within ABAQUS. This command requires the specification of an element set which
is monitored for the fully plastic condition discussed previously. Several fully plastic
element sets, or zones, were tested and the results compared. The fully plastic element
sets used in this research are defined as follows:
LayerCR - Contains elements around the crack tip, (Figure 3.8);
28
Figure 3.8 Fully plastic element set consisting of the elements around the crack
tip
Partial Layer 1 - Contains elements in the first layer of the model, but does
not contain the elements closest to the crack tip, (Figure 3.9);
Layer 1 - Contains the elements in the ligament plus the elements found in
LayerCR, (Figure 3.10);
Layer 2 - Contains elements in the first and second layers of the model, but
does not contain the elements closest to the crack tip (Figure 3.11).
29
30
Figure 3.11 Fully plastic element set consisting of partial layers 1 and 2
31
3.4 Comparison to Other Work
A series of models with different crack ratios and specimen sizes were generated.
These models contained geometric parameters (e. g. a/t, a/c, etc.) identical to those
used by other researchers. The current results were compared to previous work with
the intent of validating the FEMs and methods used for this research.
FEMs were generated with the same geometries and material properties used
by Kirk and Dodds in 1992 [23]. These geometries are shown in Figure 3.12. The
mesh generation program mesh3d scp was used to generate models for all three cracks
defined by Kirk and Dodds. The models consisted of 20-noded brick elements with
reduced integration. The number of nodes and elements in each model is listed in
Table 3.1.
Table 3.1 Number of nodes and elements in the duplication of the Kirk and Dodds
[23] geometries
Crack 1 Crack 2 Crack 3
Nodes
16,597
12,227
12,227
Elements
3562
2593
2593
32
Figure 3.12 Geometries used by Kirk and Dodds for estimating the J-Integral [23]
33
These FEMs were analyzed to find Jtotal using an elastic-plastic analysis.
ABAQUS utilizes an incremental plasticity model for this type of analysis, and requires a table of true stress versus plastic strain. The material properties for these
models were derived from Figure 3.13 and are listed below:
E = 3.00 104 kpsi
= 0.3
Tangent Modulus = 3.57 102 kpsi
Initial Yield = 80 kpsi.
These properties were used to calculate the total and elastic strains at the yield stress
and an arbitrary stress, selected to be much higher than the applied stress. This
arbitrarily large stress was used as an input because ABAQUS does not explicitly
allow the tangent modulus to be given. The plastic strains required by ABAQUS
were found by subtracting the total and elastic strains. Table 3.2 shows the calculated
strains.
Table 3.2 Incremental plasticity values for the Kirk and Dodds models
, kpsi total strain elastic strain plastic strain
80
2.67E-03
2.67E-03
0.00E+00
200
3.36E-01
6.67E-03
3.29E-01
34
Figure 3.13 Stress vs. strain curve for Kirk and Dodds elastic-plastic models [23]
35
3.4.2 McClung et al. [15]
The mesh generation program FEA-Crack was used to generate models for all
nine geometries defined in the research performed by McClung et al. (Table 3.3). Two
sets of models were generated. The first set contained a coarse mesh. The second set
utilized a more refined mesh around the crack front. The McClung et al. geometries
were analyzed as elastic, fully plastic and incrementally plastic models. The elastic
and fully plastic analyses were performed using both the coarse and refined meshes.
The incrementally plastic models were analyzed using only the coarse meshes.
In the elastic FEM analysis, the K factor was found in two ways. First,
ABAQUS was used to calculate K directly. Second, ABAQUS was used to find
the elastic J, and then Equation 2.8 was used to calculate K. These results were
compared to K factors calculated using equations from Newman and Raju [24]. The
Newman-Raju solution is given in Equations 3.4 - 3.9.
s
a 4
a 2
a
+ M3
M1 + M2
gf fw ,
KI =
Q
t
t
Q = 1 + 1.464
a 1.65
c
(3.4)
(3.5)
a/t
a/c
h/c
c/w
t
a
c
w
h
Model 1
0.2
0.2
4
0.25
1
0.2
1
4
4
Model 2
0.2
0.6
4
0.25
1
0.2
0.33
1.33
1.33
Model 8
0.8
0.6
4
0.25
1
0.8
1.33
5.33
5.33
Model 9
0.8
1
4
0.25
1
0.8
0.8
3.2
3.2
36
37
M1 = 1.13 0.09
M2 = 0.54 +
M3 = 0.5
a
c
0.89
,
0.2+( ac )
1
0.65+ ac
(3.6)
+ 14 1
a 24
c
a 2
g = 1 + 0.1 + 0.35
(1 sin )2 ,
t
f =
a 2
c
(3.7)
1/4
cos + sin
r 1/2
c a
fw = sec
,
2w t
(3.8)
(3.9)
where KI is the K factor at a given angle, is the applied stress, a is the crack depth,
Q is factor applicable for
a
c
is the angle, as previously defined in Figure 3.7, along the crack front, and w is the
half specimen width.
38
al. geometries by fixing the ratio h/w at four to one instead of one to one. Lei also
fixed c, therefore fixing w and h, and varied a and t.
Lei used ABAQUS to perform the analyses on his models. He used the *CONTOUR INTEGRAL command within ABAQUS to generate J-integral results for
fifteen contours around the crack tip. The averages of these contours, excluding the
first, were presented. Lei found that the deviation of data from any one contour is
less than 5% of the average value.
Lei used consistent material properties in his analyses. The properties for the
elastic analyses were set at E = 500 MPa and = 0.3. The elastic-plastic analyses
used the Ramberg-Osgood stress-strain relationship (Equation 2.4), where o = 1.0
MPa, = 1, and n = 5 and 10. For all analyses, Lei used the Mises yield criterion
and small strain isotropic hardening.
Current FEM results were compared with the results produced using the crack
propagation and fracture mechanics section of Nasgro. Nasgro is a fracture mechanics
and fatigue crack growth program developed by NASA and the Southwest Research
Institue. The same Ramberg-Osgood material properties used for the McClung geometries were duplicated for this comparison. The different geometries analyzed using
Nasgro are shown in Table 3.4.
39
40
3.5 Mesh Refinement
Two sets of finite element models were constructed using the McClung et
al. geometries [15] found in Table 3.3. The first set contained a coarse mesh refinement
along the crack front. The coarse mesh refinement along the crack front can be seen
in Figure 3.5. The second set of models had three times more elements around the
crack front (Figure 3.14). Table 3.5 shows the number of crack front nodes in the
coarse and refined meshes.
FEMs were generated to test the effect of specimen height and width on the
J-integral. The a/t ratios of 0.2 and 0.5, and the a/c ratios of 0.2 and 1.0 were
used in this analysis. The height effect models utilized the crack ratios for Model 1
(a/t = 0.2, a/c = 0.2), Model 4 (a/t = 0.5, a/c = 0.2), and Model 9 (a/t = 0.8, a/c =
1.0). The width effect models utilized the same model geometries used in the Nasgro
J-comparison work (Table 3.4).
Table 3.5 Number of crack front nodes in the coarse and refined meshes
Model a/t a/c Coarse Refined
1
0.2 0.2
31
91
2
0.2 0.6
17
49
3
0.2 1.0
17
49
4
0.5 0.2
45
133
5
0.5 0.6
17
49
6
0.5 1.0
17
49
7
0.8 0.2
73
265
8
0.8 0.6
31
91
9
0.8 1.0
17
49
41
The following material properties were used with the *Deformation Plasticity
command in ABAQUS:
42
E = 30.0 106 psi
= 0.3
o = 40.0 103 psi
= 0.5
n = 5, 10, and 15
where E is Youngs modulus, is Poissons ratio, o is yield or reference stress,
is a dimensionless constant as described in Equation 2.4, and n is the hardening
exponent. The effect of n on the stress vs. strain curves modelled using the RambergOsgood equation is shown in Figure 3.15. Notice that the smaller n is, the greater
the hardening slope
Figure 3.15 Effect of n on the stress vs. strain curve using a Ramberg-Osgood
model
43
3.7.2 Incremental Plasticity
The incremental plasticity models, with the exception of the Kirk and Dodds
comparison work, were generated using the Ramberg-Osgood equation,
=
+
o
0
n
,
(3.10)
shown again for convenience. The material properties listed in the previous section
were used to generate the a new Youngs modulus and a table of stress vs. plastic
strain for use in ABAQUS. The Youngs modulus, E = 30.0 106 psi, used for the
fully plastic analyses was not used to derive the stress vs. plastic strain tables for
E=
0
.
o (1 + )
(3.11)
This value of Youngs modulus was selected because it intersects the Ramberg-Osgood
curve at the fully plastic reference stress, o = 40.0 103 psi (Figure 3.16).
44
45
Using this scheme, the elastic strain, and therefore the J-integral, will be
underestimated at low stresses (Figure 3.17). But, for sufficiently high stresses, the
elastic strain becomes overwhelmed by the plastic strain, making the error negligible.
The reference strain can now be expressed as
o =
(3.12)
Figure 3.17 Elastic, modified elastic, and Ramberg-Osgood stress vs. strain curves
for n = 10
46
Multiplying both sides of Equation 3.10 by o and substituting Equation 3.12 yields:
+ o
n
.
(3.13)
Equation 3.13 can be divided into the elastic and plastic strains as
el =
(3.14)
E
and
pl = o
n
.
(3.15)
The plastic strains at different stresses were then calculated for use with the *PLASTIC command in ABAQUS for incremental plasticity analyses.
In summary, elastic-plastic material properties used in this research are based
on a modified Youngs modulus. This modification makes it possible to generate
incremental plasticity models that exhibit the same yield stress for all ns. The elastic
properties used for the incremental plasticity analyses are E = 20 106 and = 0.3.
The stress vs. plastic strain values used with the *Plastic command in ABAQUS are
shown in Tables 3.6 - 3.8.
47
Table 3.6 Stress vs. plastic strain data at n = 15, used for ABAQUS models
Stress Plastic Strain
40000
0.000667
41200
0.001039
42400
0.001598
43600
0.002428
44800
0.003649
46000
0.005425
47200
0.007982
48400
0.011633
49600
0.016797
50800
0.024042
52000
0.034124
53200
0.048049
54400
0.067142
55600
0.093139
56800
0.128302
58000
0.175561
59200
0.23869
60400
0.322525
61600
0.433231
62800
0.578633
64000
0.768614
65200
1.0156
48
Table 3.8 Stress vs. plastic strain data at n = 5, used for ABAQUS models
Stress Plastic Strain
40000
0.0
44000
0.001074
48400
0.001729
52800
0.002672
57200
0.003986
61600
0.005774
66000
0.008153
70400
0.011258
74800
0.015245
79200
0.020288
83600
0.026585
88000
0.034358
92400
0.04385
96800
0.055333
101200
0.069105
105600
0.085493
110000
0.104851
114400
0.127567
118800
0.15406
123200
0.184783
127600
0.220223
132000
0.260903
136400
0.307384
140800
0.360265
145200
0.420186
149600
0.487828
154000
0.563913
158400
0.649209
162800
0.744528
167200
0.850727
171600
0.968713
176000
1.099441
CHAPTER 4
RESULTS
This chapter begins with a discussion of the results for various fully plastic
element sets. Next, models generated for parameters used by Kirk and Dodds [23]
are compared to published results. McClung, Lei, and Newman-Raju data are then
compared to current FEM results. Finally, the effects of the specimen size on the
J-integral are examined, and the h1 values for various specimen widths are compared
with Nasgro results.
The mesh generation program mesh3d scp was used to generate FEMs for all
four of the fully plastic zones described in Chapter 3. The same mesh was used for
each model. Only the specified fully plastic element set was changed for the different
models. It was found that the J-integral was identical for all of the described zones.
Therefore, only Partial Layer 1 was used in later fully plastic models was used for
the FEA-Crack meshes, and LayerCR was used for any fully plastic meshes produced
with mesh scp.
The results for models generated per the Kirk and Dodds geometries are shown
in Table 4.1. The results compared quite well to the published data. The maximum
difference between the current results and the published data was 2.9%. It should be
noted that this excellent agreement in results was obtained even though the meshes
49
50
Table 4.1 Comparison of FEM results to Kirk and Dodds values
Crack
J (in-lb) Kirk and Dodds % Difference
from FEM
J (in-lb)
1
30.9
0.749
0.732
2.3
1
90
0.892
0.867
2.9
2
30.9
2.055
2.014
2.0
2
90
0.892
0.867
2.93
3
30.9
2.077
2.046
1.5
3
90
3.207
3.173
1.7
used by Kirk and Dodds contained approximately 25% the number of nodes and
elements used in this research.
Elastic, fully plastic, and incremental plasticity FEA results for the McClung
et al. geometries are presented in this section. The elastic results are compared to the
Newman-Raju [24] calculations, and graphical trends are noted in the comparison of
Leis [17] elastic results. The fully plastic data are compared to the tabular data of
McClung et al. [15] and Lei [17]. The effects of mesh refinement are discussed for
both the elastic and fully plastic FEMs. Finally the incremental plasticity and fully
plastic FEA results are compared.
51
Figure 4.1 Model 1 (a/t=0.2, a/c=0.2): Normalized K factor vs. angle along crack
front
4.3.1 Elastic Analysis
The K factors obtained from the FEMs with the McClung geometries were
normalized using
KI
Knorm = q ,
Qa
(4.1)
from Newman and Raju [25]. The results of the elastic FEM models and the Newman
and Raju [24] calculations are presented in Figures 4.1-4.9.
52
Figure 4.2 Model 2 (a/t=0.2, a/c=0.6): Normalized K factor vs. angle along crack
front
Figure 4.3 Model 3 (a/t=0.2, a/c=1.0): Normalized K factor vs. angle along crack
front
53
Figure 4.4 Model 4 (a/t=0.5, a/c=0.2): Normalized K factor vs. angle along crack
front
Figure 4.5 Model 5 (a/t=0.5, a/c=0.6): Normalized K factor vs. angle along crack
front
54
Figure 4.6 Model 6 (a/t=0.5, a/c=1.0): Normalized K factor vs. angle along crack
front
Figure 4.7 Model 7 (a/t=0.8, a/c=0.2): Normalized K factor vs. angle along crack
front
55
Figure 4.8 Model 8 (a/t=0.8, a/c=0.6): Normalized K factor vs. angle along crack
front
Figure 4.9 Model 9 (a/t=0.8, a/c=1.0): Normalized K factor vs. angle along crack
front
56
Significant increases in the K-factor at the surface and/or depth were observed
in all of the models. Only Models 8 (a/t = 0.8, a/c = 0.6) and 9 (a/t = 0.8, a/c = 1.0)
do not have large increases in the K-factor at the free surface. Models 1, 4, and 7
(all with a/c = 0.2) are the only FEMs that do not contain the same K-factor spike
repeated in the depth (Table 4.2).
When the surface and depth spikes are disregarded, the ABAQUS results compared very reasonably to the normalized K-factors calculated per the Newman and
Raju [24] equations. This favorable comparison occurred even though the mid-side
nodes were not moved to the quarter points, and the nodes along the crack tip were
left untied (two conditions which yield optimum accuracy in K-factor calculations
using FEMs). The largest observed error, approximately six percent, occurred with
Model 8. It should also be noted that the normalized K-factor results from the K
and elastic J models were very close. The elastic results are summarized in Table
4.3. This summary disregards the surface and depth results. There is no apparent
pattern to the differences.
The current FEA results were also compared visually to the graphical results
published by Lei [17]. Lei used a different normalizing scheme, resulting in different
57
Table 4.3 Maximum percent differences between Newman-Raju and FEM solutions (quarter symmetry)
Max FEA K Max FEA K
Max FEA K
Directly
from Jel
from tied nodes
Model a/t a/c
(% diff.)
(% diff.)
(% diff.)
1
0.2 0.2
-5.64
-5.38
-8.5
2
0.2 0.6
-2.28
-2.18
3
0.2 1.0
3.11
2.88
4
0.5 0.2
-6.40
-5.57
5
0.5 0.6
-2.15
-1.96
6
0.5 1.0
3.35
3.44
7
0.8 0.2
4.52
4.53
8
0.8 0.6
-7.07
-7.07
-7.17
9
0.8 1.0
2.8
3.03
scales on the y-axis, but the graphs had very similar shapes. Lei also showed some
models with the same spike at the surface that was experienced in this research.
However, the increase was not as significant. No sudden increases were observed at
the depth of his elastic models.
An investigation was performed to find the cause of the previously discussed
surface and depth K-factor spikes in the current FEMs. The first step was to explore
the potential error caused by not tying the crack tip nodes or moving the mid-side
nodes to the quarter points. Model 1 (a/t = 0.2, a/c = 0.2) and Model 8 (a/t = 0.8,
a/c = 0.6) meshes were recreated in FEA-Crack for elastic analysis only. These two
models were generated with an elastic singularity, 1/ r, created by tying the crack
tip nodes and moving the mid-side nodes to the quarter points (Figure 4.10). The
results of the two elastic models are shown graphically in Figures 4.11 and 4.12. The
K-factors produced using these two FEMs were almost identical to the previous
results for Model 1 and Model 8.
58
Figure 4.11 Model 1 (a/t=0.2, a/c=0.2): Normalized K-factor vs. angle along
crack front for untied and tied nodes
59
Figure 4.12 Model 8 (a/t=0.8, a/c=0.6): Normalized K-factor vs. angle along
crack front for untied and tied nodes
60
The second step in investigating the surface and depth K-factor spikes involved changing from reduced integration to full integration elements, type C3D20R
to C3D20 in ABAQUS, for the FEMs. The results for Models 1 and 8 are shown in
Figures 4.13 and 4.14. As mentioned in the ABAQUS Users Manuals [21], using a
full integration element type caused the K-factor results to oscillate. Full integration
elements did not correct the surface and depth deviations.
Figure 4.13 Model 1 (a/t = 0.2, a/c = 0.2): Reduced vs. full integration elements
61
Figure 4.14 Model 8 (a/t = 0.8, a/c = 0.6): Reduced vs. full integration elements
62
The default nonlinear solver for ABAQUS was also considered as a possible
source of error in the third attempt to reduce the surface and depth K-factor discrepencies. Since the K-factor calculation is linear, the solver used within ABAQUS
was changed to a linear perturbation. Unfortunately, the *CONTOUR INTEGRAL
command used to output the K-factors will not function within a linear perturbation
step. Another attempt to force a linear solution was made by using the *STATIC
command to force the solution to be performed in one step. The surface and depth
results were not altered by this approach.
In a fourth attempt to solve the free surface and depth spikes, the FEACrack Validation Manual [26] was examined. It was found that the FEA programs
WARP3D [27], ABAQUS, and ANSYS were used to produce K-factor calculations
for validating FEA-Crack meshes. The validation data were presented graphically as
K-factor vs. angle along the crack front (Figure 4.15 [26]). The K-factors produced
by these three FEA programs were virtually identical, except at the free surface. At
this location, the K-factor produced by ABAQUS was approximately 7.7% higher
than the other two FEA programs. It was also noted that there were no irregularities
presented in the depth.
Further examination of the FEA-Crack Validation Manual [26] revealed that
the surface crack results published in the validation manual were for full plates with no
symmetry conditions (Figure 4.16). To investigate boundary conditions in the depth
phenomenon, Model 6 (a/t = 0.5, a/c = 1.0) and Model 8 (a/t = 0.8, a/c = 0.6)
meshes were created with half symmetry (Figure 4.17). The results may be seen in
Figures 4.18 and 4.19.
63
Figure 4.16 FEM mesh for a flat plate with no symmetry exploited [26]
64
Figure 4.17 FEM mesh for a flat plate with half symmetry
Figure 4.18 Model 6 (a/t = 0.5, a/c = 1.0): K-Factor results for half symmetry
model
65
Figure 4.19 Model 8 (a/t = 0.8, a/c = 0.6): K-Factor results for half symmetry
model
66
The half symmetry models produced with FEA-Crack did not exhibit a Kfactor spike in the depth. Therefore, it was deduced that the boundary conditions
specified in the ABAQUS input file could be an issue. Perhaps the solver within
ABAQUS is very sensitive to the boundary conditions. Problems with the boundary
conditions in the depth or at the free surface could account for the observed increases.
An ABAQUS input file for an elastic model was examined to investigate the
boundary conditions of nodes along the crack front. For the elastic models, it was
found that there were no coincident nodes along the crack front. The crack tip nodes
were shared by the surrounding elements. The depth node was constrained in the
directions perpendicular and parallel to the crack front. All of the other crack tip
nodes were constrained only in the direction perpendicular to the crack front. This
constraint scheme is logical given a quarter symmetry model with tied nodes along
the crack front.
A possible explanation for the surface deviation is that it is a result of the
surface elements in the 3-D model. The lateral surface is subjected to plane stress,
but the surface elements have some thickness in the direction normal to the free
surface. This is a plausible, but unproven, explanation for the source of the surface
phenomenon.
67
4.3.2 Fully Plastic Analysis
The fully plastic results from the McClung geometries are presented in the
following section. It should be noted that ABAQUS outputs Jtotal , but it is difficult
to actually quantify Jel for a Ramberg-Osgood material due to the initial nonlinearity
of the power law curve. But, for fully plastic results, Jel is only a very small fraction
of Jtotal . For this reason, the fully plastic Jtotal is (for all practical purposes) equal to
Jpl . The Jpl results from ABAQUS were used to calculate the h1 values using
h1 =
Jpl
0 0 t (/0 )(n+1)
(4.2)
The h1 data from the current FEMs are plotted against the values from McClung et
al. [15] and Lei [17] in Figures 4.20-4.46.
Figure 4.20 Model 1 (a/t=0.2, a/c=0.2): h1 vs. angle along the crack front
68
Figure 4.21 Model 1 (a/t=0.2, a/c=0.2): h1 vs. angle along the crack front
Figure 4.22 Model 1 (a/t=0.2, a/c=0.2): h1 vs. angle along the crack front
69
Figure 4.23 Model 2 (a/t=0.2, a/c=0.6): h1 vs. angle along the crack front
Figure 4.24 Model 2 (a/t=0.2, a/c=0.6): h1 vs. angle along the crack front
70
Figure 4.25 Model 2 (a/t=0.2, a/c=0.6): h1 vs. angle along the crack front
Figure 4.26 Model 3 (a/t=0.2, a/c=1.0): h1 vs. angle along the crack front
71
Figure 4.27 Model 3 (a/t=0.2, a/c=1.0): h1 vs. angle along the crack front
Figure 4.28 Model 3 (a/t=0.2, a/c=1.0): h1 vs. angle along the crack front
72
Figure 4.29 Model 4 (a/t=0.5, a/c=0.2): h1 vs. angle along the crack front
Figure 4.30 Model 4 (a/t=0.5, a/c=0.2): h1 vs. angle along the crack front
73
Figure 4.31 Model 4 (a/t=0.5, a/c=0.2): h1 vs. angle along the crack front
Figure 4.32 Model 5 (a/t=0.5, a/c=0.6): h1 vs. angle along the crack front
74
Figure 4.33 Model 5 (a/t=0.5, a/c=0.6): h1 vs. angle along the crack front
Figure 4.34 Model 5 (a/t=0.5, a/c=0.6): h1 vs. angle along the crack front
75
Figure 4.35 Model 6 (a/t=0.5, a/c=1.0): h1 vs. angle along the crack front
Figure 4.36 Model 6 (a/t=0.5, a/c=1.0): h1 vs. angle along the crack front
76
Figure 4.37 Model 6 (a/t=0.5, a/c=1.0): h1 vs. angle along the crack front
Figure 4.38 Model 7 (a/t=0.8, a/c=0.2): h1 vs. angle along the crack front
77
Figure 4.39 Model 7 (a/t=0.8, a/c=0.2): h1 vs. angle along the crack front
Figure 4.40 Model 7 (a/t=0.8, a/c=0.2): h1 vs. angle along the crack front
78
Figure 4.41 Model 8 (a/t=0.8, a/c=0.6): h1 vs. angle along the crack front
Figure 4.42 Model 8 (a/t=0.8, a/c=0.6): h1 vs. angle along the crack front
79
Figure 4.43 Model 8 (a/t=0.8, a/c=0.6): h1 vs. angle along the crack front
Figure 4.44 Model 9 (a/t=0.8, a/c=1.0): h1 vs. angle along the crack front
80
Figure 4.45 Model 9 (a/t=0.8, a/c=1.0): h1 vs. angle along the crack front
Figure 4.46 Model 9 (a/t=0.8, a/c=1.0): h1 vs. angle along the crack front
81
The same spikes observed at the free surface and in the depth of the elastic
FEMs were also found in the fully plastic models. These discrepencies matched those
found for the elastic models (Table 4.2). As seen in the elastic results, the fully plastic
models, except Models 8 (a/t = 0.8, a/c = 0.6) and 9 (a/t = 0.8, a/c = 1.0), exhibited
noticeable increases at the free surface. Models 1, 4, and 7 (all with a/c = 0.2) were
the only models that did not also contain increases in the depth.
The fully plastic results from the current FEMs, McClung, and Lei did not
correlate as well as the elastic results compared with the Newman-Raju solutions.
The differences were estimated by visually examining the previous h1 plots (Figures
4.20-4.46). The comparison was visual because the angles for McClung et al. and Lei
results did not match the angles used for this research, and a method to force the
FEA-Crack or mesh3d scp mesh generation programs to place node sets at specific
angles was not known. As with the elastic analysis, the surface and depth results
were disregarded in determining the percent differences.
A visual examination of the h1 plots revealed some interesting findings. First,
the Lei results do not contain the same spikes in the h1 values at the surface and
depth that are observed in in the current FEMs. This is intriguing because the Lei
FEA work was performed using quarter symmetry flat plates in ABAQUS.
Second, the McClung et al. and current FEM results compared reasonably well
for some crack geometries but not for others. It was found that the differences between
the two sets of results increased with n. There were differences of approximately
thirty percent for some of the n = 15 models. Table 4.4 contains the largest percent
difference between the McClung et al. and current FEM results for each model.
It was observed that Models 2, 3, 5, 6, 8, and 9 (all with a/c = 0.6 or 1.0)
of the current FEMs, exhibited either spikes in h1 at the third angle or dips at the
82
fourth angle from the surface for for all three ns. The uncertainty of a spike at the
third angle or a dip at the fourth is due to the fact that this judgement is relative to
how the data trend is pictured. The values in Table 4.4 are highlighted with the *
symbol when the largest difference from the McClung et al. results occurred at the
third angle. The difference between results did not tend to be as significant at the
fourth angle. An examination of the FEMs for these models revealed that third angle
for each model, except Model 8, was located at 11.24 . The third angle in Model 8
occurrs at 8.09 . Both the McClung et al. and Lei data exhibited the same behavior.
These results occurred at an angle of 18 .
The third observation was that Lei [17] tended to correlate more closely with
the McClung et al. [15] results than the current FEMs. Unlike the current FEMs, the
Lei models did not exhibit a consistent increase in the difference as n increased. The
maximum percent difference between the two sets of results was twenty percent, ten
percent less than the difference of the current FEA results. Table 4.5 lists the largest
percent difference between the McClung et al. and Lei results for each model.
Table 4.4 Maximum percent differences between McClung et al. [15] and FEM
solutions (quarter symmetry)
n=5
n = 10
n = 15
Model a/t a/c (% diff.) (% diff.) (% diff.)
1
0.2 0.2
2.00
-5.66
-12.50
2
0.2 0.6
9.20
-9.26
-12.90
3
0.2 1.0
12.20
16.67
18.52
4
0.5 0.2
4.67
7.69
9.09
5
0.5 0.6
17.5
15.71
21.57
6
0.5 1.0
15.17
25.93
23.17
7
0.8 0.2
13.33
17.78
29.35
8
0.8 0.6
10.81
18.75
30.00
9
0.8 1.0
16.00
23.30
28.5
Note: Maximum differences at the third angle
from the surface are marked by *.
83
Table 4.5 Maximum percent differences between McClung et al. [15] and Lei [17]
solutions (quarter symmetry)
n=5
n = 10
Model a/t a/c (% diff.) (% diff.)
1
0.2 0.20
19.17
17.16
2
0.2 0.60
12.44
9.69
3
0.2 1.00
14.05
11.59
4
0.5 0.20
8.79
4.68
5
0.5 0.60
5.95
2.59
6
0.5 1.00
7.36
3.19
7
0.8 0.20
9.57
9.41
8
0.8 0.60
10.23
20.69
9
0.8 1.00
8.13
17.64
84
The J-integral spikes at the free surface and depth were also investigated for
the fully plastic FEMs. As discussed in the analysis of the elastic results, the free
surface phenomenon can be attributed, at least partially, to a plane stress issue.
Fully plastic models with half symmetry were also analyzed to ensure that the depth
phenomenon disappeared under conditions of severe deformation. Model 6 (a/t = 0.5,
a/c = 1.0) and Model 8 (a/t = 0.8, a/c = 0.6) geometries were again used. The half
symmetry models did not exhibit the jump in h1 in the depth (Figures 4.47 and 4.48).
Figure 4.47 Model 6 (a/t=0.5, a/c=1.0): h1 results for half symmetry model at
n = 15
85
Figure 4.48 Model 8 (a/t=0.8, a/c=0.6): h1 results for half symmetry model at
n = 15
86
An ABAQUS input file for an plastic model was also examined to investigate
the boundary conditions of nodes along the crack tip. As discussed previously, the
models with untied nodes contain multiple node sets along the crack front. The nodes
at each node set along the crack tip are coincident. In the depth node set, one node
was constrained in the directions perpendicular and parallel to the crack front. The
remaining nodes in the depth were only constrained in the direction perpendicular
to the crack front. The remaining node sets, including the surface, contained one
node that was constrained in the direction perpendicular to the crack front. All other
nodes in the remaining sets were left unconstrained. This constraint scheme is logical
given a quarter symmetry model with untied nodes along the crack front.
The elastic-plastic results are compared to the fully plastic results for Model
1 (a/t=0.2, a/c=0.2) and Model 2 (a/t=0.2, a/c=0.6) for n = 5, 10, and 15. These
results are shown graphically in Figures 4.49 - 4.54.
87
Figure 4.49 Model 1 (a/t=0.2, a/c=0.2): h1 vs. angle for fully plastic and elasticplastic models at n=5
Figure 4.50 Model 1 (a/t=0.2, a/c=0.2): h1 vs. angle for fully plastic and elasticplastic models at n=10
88
Figure 4.51 Model 1 (a/t=0.2, a/c=0.2): h1 vs. angle for fully plastic and elasticplastic models at n=15
Figure 4.52 Model 2 (a/t=0.2, a/c=0.6): h1 vs. angle for fully plastic and elasticplastic models at n=5
89
Figure 4.53 Model 2 (a/t=0.2, a/c=0.6): h1 vs. angle for fully plastic and elasticplastic models at n=10
Figure 4.54 Model 2 (a/t=0.2, a/c=0.6): h1 vs. angle for fully plastic and elasticplastic models at n=15
90
It was found that the h1 values of the elastic-plastic method varied with the
applied stress, but appeared to converge with the fully plastic results as the stress
was increased. The similarities between the elastic-plastic and fully plastic results
included the surface, depth, and interior angle anomalies discussed previously.
Some of the differences between the elastic-plastic and fully plastic results at
lower stresses may have been due to a mistake. The wrong stress versus plastic strain
tables were used in ABAQUS. The tables actually used for the elastic-plastic analyses
in this research are located in Appendix G. These plastic strain tables contain stresses
lower than the 40 ksi yield strength discussed in Chapter 3. This caused the utilized
stress-strain curve to be lower than the Ramberg-Osgood curve at lower stresses
(Figure 4.55). However, the difference due to this error is likely small for a fully
plastic condition.
Figure 4.55 Elastic, Ramberg-Osgood, modified elastic, and modified RambergOsgood stress vs. strain curves for n = 10
91
4.4 Mesh Refinement
Figure 4.56 Model 1 (a/t = 0.2, a/c = 0.2): Normalized K-factor vs. angle along
crack front
92
Figure 4.57 Model 1 (a/t = 0.2, a/c = 0.2): h1 vs. angle along the crack front
It should be noted that the refined FEM mesh for Model 7 (a/t = 0.8, a/c =
0.2) would not run in ABAQUS. The Model 7 geometry generated the largest FEM in
this research, containing 12,140 elements and 56,142 nodes. A warning was generated
about the number of history requests for the ABAQUS output database (ODB) file,
but the terminal error was not explicitly mentioned by ABAQUS. Because the other
refined mesh models generated results that closely matched the coarse mesh models,
this exact cause of this issue was not pursued.
The same anomalies observed in the coarse mesh FEMs were also present in
the refined mesh models. This includes the K-factor and J-integral spikes at the
free surface and depth. It should be noted that the mesh refinement did reduce the
magnitude of the surface and depth surges. The phenomenon at third and fourth
angles from the surface was also present. The values of the third and fourth angle in
the refined meshes were occurred at either 3.75 or 2.71 .
93
Many of the refined mesh models contain noticeable oscillations in the h1 data
around the crack tip. This oscillation was most significant in Models 3 (a/t = 0.2,
a/c = 1.0), 4 (a/t = 0.5, a/c = 0.2), and 6 (a/t = 0.5, a/c = 1.0). For each of these
models, the magnitude of the oscillation became larger as n increased.
Finite element models with four different crack ratios, each at three different
heights, were analyzed to check for a height effect. Although small differences were
noticed (maximum difference of 5%), no appreciable height effect was observed. The
results for Model 1 (a/t=0.2, a/c=0.2) can be seen in Table 4.6. The results for
Models 4 (a/t=0.5, a/c=0.2), 5 (a/t=0.2, a/c=0.6), and 9 (a/t=0.8, a/c=1.0) may
be found in Appendix D.
94
95
4.5.2 Width Effects
The width effect results were varied. As expected, it was found that h1 increased as c/w increased. However, a large amount of variation between the Nasgro
and FEM results was observed. These results are presented in Tables 4.7 - 4.9, and
graphically in Figures 4.58 - 4.69. Notice that Model 4a for a strain hardening value
of n = 15 does not have results posted. This FEM would not run in ABAQUS, but
the reason is not known.
Model
1
1a
1b
3
3a
3b
4
4a
4b
6
6a
6b
96
Model
1
1a
1b
3
3a
3b
4
4a
4b
6
6a
6b
Model
1
1a
1b
3
3a
3b
4
4a
4b
6
6a
6b
% diff(c)
-18.95
-41.75
-52.04
-4.03
4.58
13.8
22.97
12.95
5.77
4.43
-2.79
-6.38
97
Figure 4.58 Model 1 (a/t = 0.2, a/c = 0.2): h1 vs. c/w for Nasgro and FEM at
n = 15
Figure 4.59 Model 1 (a/t = 0.2, a/c = 0.2): h1 vs. c/w for Nasgro and FEM at
n = 10
98
Figure 4.60 Model 1 (a/t = 0.2, a/c = 0.2): h1 vs. c/w for Nasgro and FEM at
n=5
Figure 4.61 Model 3 (a/t = 0.2, a/c = 1.0): h1 vs. c/w for Nasgro and FEM at
n = 15
99
Figure 4.62 Model 3 (a/t = 0.2, a/c = 1.0): h1 vs. c/w for Nasgro and FEM at
n = 10
Figure 4.63 Model 3 (a/t = 0.2, a/c = 1.0): h1 vs. c/w for Nasgro and FEM at
n=5
100
Figure 4.64 Model 4 (a/t = 0.5, a/c = 0.2): h1 vs. c/w for Nasgro and FEM at
n = 15
Figure 4.65 Model 4 (a/t = 0.5, a/c = 0.2): h1 vs. c/w for Nasgro and FEM at
n = 10
101
Figure 4.66 Model 4 (a/t = 0.5, a/c = 0.2): h1 vs. c/w for Nasgro and FEM at
n=5
Figure 4.67 Model 6 (a/t = 0.5, a/c = 1.0): h1 vs. c/w for Nasgro and FEM at
n = 15
102
Figure 4.68 Model 6 (a/t = 0.5, a/c = 1.0): h1 vs. c/w for Nasgro and FEM at
n = 10
Figure 4.69 Model 6 (a/t = 0.5, a/c = 1.0): h1 vs. c/w for Nasgro and FEM at
n=5
103
The differences between the current FEM and Nasgro results were varied.
Some geometries produced relatively close comparisons (less than 5%) between the
Nasgro and FEM results. Others demonstrated significant differences (greater than
100%). The differences are most likely a result of the reference stress algorithm used
within Nasgro. This is not to say that the Nasgro results are less correct than the
current FEM results, and no final judgement is made here.
It was also found that no two comparisons of FEM and Nasgro results demonstrated exactly the same pattern. The specific observations are:
Model 1 (a/t = 0.2, a/c = 0.2): The Nasgro results were higher for every
tested c/w at all ns. The difference between the Nasgro and FEM results
increased with c/w for all ns.
Model 3 (a/t = 0.2, a/c = 1.0): The FEM results were higher for every
tested c/w at all ns. The difference between the Nasgro and FEM results
increased with c/w for all ns.
Model 4 (a/t = 0.5, a/c = 0.2): The FEM results were higher for every
tested c/w at all ns. The difference between the Nasgro and FEM results
increased with c/w for n = 10. The difference between the Nasgro and
FEM results decreased as c/w increased for n = 15 and n = 5.
Model 6 (a/t = 0.5, a/c = 1.0): The FEM results were higher for every
tested c/w at n = 5. The FEM and Nasgro results cross for n = 15 and
n = 10. The difference between the Nasgro and FEM results increased
with c/w for n = 15. The difference between the Nasgro and FEM results
decreased as c/w increased for n = 10 and n = 5.
CHAPTER 5
CONCLUSIONS AND RECOMMENDATIONS
This chapter contains the conclusions and recommendations for this research.
Conclusions are stated for each of the stated goals of this work. The recommendations
for further work address questions that were not answered during this study.
5.1 Conclusions
The following are conclusions reached in this research for the primary goal
of comparing three-dimensional finite element analysis J-integral results produced in
ABAQUS with work completed by other researchers:
1. The K-factor and elastic J-integral results produced in ABAQUS for this
research compare well with the Newman and Raju [18] calculations and Lei
data, except as noted below.
There is a spike in the K-factor at the surface for most analyzed
geometries. Models with a very deep crack (a/t = 0.8) have smaller
spikes. The elastic J-integral results presented by Lei [17] showed some
surface spikes, but not to same magnitude as found in this research.
There is a sudden increase in the K-factor in the depth for geometries that do not have highly elliptical (a/t = 0.2) cracks for quarter
symmetry models. There were no spikes observed for the depth in the
results presented by Lei.
2. The fully plastic J-integral results generated in this research compare reasonably well, for some geometries, with work produced McClung et al. [15]
104
105
and Lei [17]. However, there are some issues with using ABAQUS to calculate the J-integral.
The same surface and depth issues found in the current elastic FEMs
are also present in the fully plastic models.
The plastic FEMs from the current research, Lei, and McClung et
al. exhibit an anomaly at the third or fourth angle from the surface.
The value of the angle was dependent on the model and ranged from
2.71 to 18 .
3. The quarter symmetry FEA results presented by McClung et al. and Lei
did not contain surface or depth spikes in the plastic J-integral, as found
in the current FEMs. Possible reasons for this finding are given below.
Other researchers used a smoothing method.
There was an issue with the quarter symmetry FEMs used in this
research, probably related to the boundary conditions.
4. FEM and Nasgro results compared well for some, but not all, specimen
geometries. There was no apparent pattern to the variation.
The second goal of this research was to investigate the effects of mesh density,
element type, symmetry, and specimen size effects on the J-integral. Based on this
research, the conclusions are made.
1. The schemes used by ABAQUS to calculate K and J are robust and do
not require a significant amount of mesh refinement to produce converged
results.
2. A refined mesh can result in oscillations of the calculated J-integral, especially at larger n values.
106
3. Reduced integration elements give more consistent results for K-factor and
J-integral calculations in ABAQUS. Full integration elements tend to oscillate.
4. Quarter symmetry models experience a spike in the depth. Half symmetry
models do not have the same issue.
5. Specimen height has no significant effect on the J-integral for uniform tensile loading.
6. Specimen width has a very significant effect on the J-integral for uniform
tensile loading.
The third goal of this research was to compare elastic-plastic and fully plastic
results. For the elastic-plastic FEAs, the stress vs. plastic strain table was based
on the Ramberg-Osgood material properties. The elastic-plastic and fully plastic
results compared very well at sufficiently large stresses. The same surface and depth
phenomenon observed in the fully plastic FEMs were also present in the elastic-plastic
results.
5.2 Recommendations
107
2. Investigate the anomaly exhibited by some models at the third and fourth
angles from the surface.
3. Use the existing models to obtain results for corner cracks. This work can
be easily performed by changing the boundary conditions of the existing
models.
REFERENCES
108
109
[1] Griffith, A.A., The Phenomena of Rupture and Flow in Solids. Philosophical
Transactions, Series A, Vol. 221, 1920, pp. 163-198.
[2] Irwin, G.R., Fracture Dynamics. Fracturing of Metals, American Society for
Metals, Cleveland, 1948, pp. 147-166.
[3] Irwin, G.R., Analysis of Stresses and Strains near the End of a Crack Traversing a Plate. Journal of Applied Mechanics, Vol. 24, 1957, pp. 361-364.
[4] Anderson, T. L., Fracture Mechanics Fundamentals and Applications, Second
Edition, CRC Press, FL, 1991.
[5] Tada, H, P.C. Paris, and G.R. Irwin, The Stress Analysis of Cracks Handbook,
ASME Press, New York, 2000.
[6] Wells, A.A., Unstable Crack Propagation in Metals: Cleavage and Fast Fracture. Proceedings of the Crack Propagation Symposium, Vol. 1, Paper 84,
Cranfield UK, 1961.
[7] Rice, J.R. A Path Independent Integral and the Approximate Analysis of
Strain Concentraion by Notches and Cracks. Journal of Applied Mechanics,
Vol. 35, 1968, pp. 379-386.
[8] Hutchinson, J.W., Singular Behavior at the End of a Tensile Crack Tip in a
Hardening Material. Journal of the Mechanics and Physics of Solids, Vol. 16,
1968, pp.13-31.
[9] Rice, J.R. and G.F. Rosengren, Plane Strain Deformation Near a Crack Tip
in a Power-Law Hardening Material. Journal of the Mechanics and Physics of
Solids, Vol. 16, 1968, pp. 1-12.
[10] Cook, R.D., D.S. Malkus, M.E. Plesha, R.J. Witt, Concepts and Applications
of Finite Element Analysis, Fourth Edition, John Wiley & Sons, Inc., 2002.
[11] Kumar, V., M.D. German, and C.F. Shih, An Engineering Approach for
Elastic-Plastic Fracture Analysis, Report NP-1931, Electric Power Research
Institute, 1981.
[12] Shih, C.F., J-Integral Estimates for Strain Hardening Materials in Antiplane
Shear Using Fully Plastic Solutions, in Mechanics of Crack Growth, ASTM
Special Publication 590, 1976, pp. 3-22.
110
[13] Shih, C.F., and J.W. Hutchinson, Fully Plastic Solutions and Large-Scale
Yielding Estimates for Plane Stress Crack Problems, Transactions of ASME,
Journal of Engineering Materials and Technology, Series H, Vol.98, October
1976, pp. 289-295.
[14] Needleman, A., and C.F. Shih, A Finite Element Method for Plane Strain Deformations of Incompressible Solids. Computer Methods in Applied Mechanics
and Engineering, 1978, pp. 223-240.
[15] McClung, R.C., G.G. Chell, Y.D. Lee, D.A. Russell, and G.E. Orient, 1999,
Developement of a Practical Methodology for Elastic-Plastic and Fully Plastic Fatigue Crack Growth, National Aeronautics and Space Administration,
Marshall Space Flight Center, Alabama, Contract NAS8-37828.
[16] ABAQUS Analysis Users Manual Version 6.4, ABAQUS, Inc., 2003.
[17] Lei, Y., J-integral and limit load analysis of semi-elliptical surface cracks in
plates under tension. International Journal of Pressure Vessels and Piping,
2004, pp. 21-31.
[18] Newman, J.C. and Raju, I.S., An empirical stress-intensity factor equation for
the surface crack. Engineering Fracture Mechanics, 1981, pp. 185-192.
[19] Ainsworth, R.A., The Assessment of Defects in Structures of Strain Hardening
Materials. Engineering Fracture Mechanics, Vol. 19, 1984, p.633.
[20] Faleskog, Jonas, Users Manual: mesh3d scp, 1996.
[21] ABAQUS Users Manual, Version ???.
[22] Natarajan, Krishna R., Plastic J-Integral Solutions for Through Cracks Using Finite Element Method, Master of Science Thesis, Tennessee Technological
University, Cookeville, TN.
[23] Kirk, M.T. and R.H. Dodds, Approximate Techniques of J Estimation Applicable to Part-Though Surface Cracks, Engineering Fracture Mechanics, Vol.
43, No. 1, 1992, pp. 123-136.
111
[24] Newman, J.C. and Raju, I.S., Stress Intensity Factor Equations for Cracks in
Three-Dimensional Finite Bodies, Fracture Mechanics: Fourteenth Symposium
- Volume 1: Theory and Analysis, ASTM STP 791, J. C. Lewis and G. Sines,
Eds.,American Society for Testing and Materials, 1983, pp. I-238 - I-265.
[25] Newman, J.C. and Raju, I.S., Analysis of Surface Cracks in Finite Plates
Under Tension of Bending Loads, NASA Technical Paper 1578, Dec. 1979.
[26] 3D Finite Element Software for Cracks, Version 2.5, Benchmarks and Validation, Structural Reliability Technology, Boulder, CO, July, 2003.
[27] Computational Fracture Mechanics Group at University of Illinois UrbanaChampaign, http://cern49.ce.uiuc.edu/cfm/warp3d.html, accessed 1-31-2005.
[28] Lloyd, W.R., and W.G. Reuter, Evaluation of Elastic-Plastic Surface Flaw
Behavior and Related Parameters Using Surface Displacement Measurements,
Fracture Mechanics: Twenty-First Symposium, ASTM STP 1074, J.P. Gudas,
J.A. Joyce, and E.M. Hackett, Eds., American Society for Testing and Materials, Philadelphia, 1990, pp. 322-336.
[29] Irwin, G.R., Plastic Zone Near a Crack and Fracture Toughness, Sagamore
Research Conference Proceedings, Vol. 4, 1961.
[30] Dodds, R. H. Jr. and David T. Read, Experimental and Numerical Studies of
the J-Integral for a Surface Flaw, International Journal of Fracture, Vol. 43,
1990, pp. 47-67.
[31] Read,D.T., Applied J-Integral in HY-130 Tensile Panels and Implications for
Fitness for Service Assessment, Report NBSIR 82-1670, National Bureau of
Standards, Boulder, CO, 1982.
[32] Yagawa, G., Y. Kitajima, H. Ueda, Three-Dimensional Fully Plastic Solutions
for Semi-elliptical Surface Cracks, International Journal of Pressure Vessels
and Piping, Vol. 53, 1993, pp. 457-510.
APPENDICES
112
113
114
1. Insure that the correct element type (8, 20, or 27-node) is specified in the
mesh3d scp input file. This is done by changing the ETYP=?? field to
the appropriate value.
2. The mesh3d scp program creates several files when run. Three of these
contain node and element data. The file names are described below:
test.015 - contains node and coordinate data
test.016 - contains element connectivities
test.017 - contains element connectivities
Use the following steps to modify these files for use with ABAQUS 6.5.
(a) Combine test.016 and test.017 using cat test.016 test.017 > temp ele.inp.
(b) At the top of temp ele.inp, add *ELEMENT, TYPE=C3D20R,ELSET=
ELEMALL.
(c) Run abaqus free job=elements input=temp ele
3. Change the node file name from test.015 to nodes.inp.
4. Insert the *NODE at the top of the nodes.inp file.
5. Make the following changes to the test.inp file (input file for ABAQUS):
There will be three lines referencing test.015 thru test.017. Delete
these lines.
Add *INCLUDE, INPUT=nodes.inp
Add *INCLUDE, INPUT=elements.inp
At the bottom of the file, add commas to the spaces between the
contours in the *CONTOUR INTEGRAL command.
Many, but not all, of the commands between *MATERIAL and
*****STEP=1****** are not valid and should be replaced with
the appropriate commands.
115
116
Figure B.1 Model 1 (a/t=0.2, a/c=0.2): Normalized K factor vs. angle along
crack front
Figure B.2 Model 2 (a/t=0.2, a/c=0.6): Normalized K factor vs. angle along
crack front
117
Figure B.3 Model 3 (a/t=0.2, a/c=1.0): Normalized K factor vs. angle along
crack front
Figure B.4 Model 4 (a/t=0.5, a/c=0.2): Normalized K factor vs. angle along
crack front
118
Figure B.5 Model 5 (a/t=0.5, a/c=0.6): Normalized K factor vs. angle along
crack front
Figure B.6 Model 6 (a/t=0.5, a/c=1.0): Normalized K factor vs. angle along
crack front
119
Figure B.7 Model 8 (a/t=0.8, a/c=0.6): Normalized K factor vs. angle along
crack front
Figure B.8 Model 9 (a/t=0.8, a/c=1.0): Normalized K factor vs. angle along
crack front
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
Table D.1 Model 4 (a/t=0.5 and a/c=0.2): h1 values for at different heights (Part
1)
135
Table D.2 Model 4 (a/t=0.5 and a/c=0.2): h1 values for at different heights (Part
2)
Table D.3 Model 5 (a/t=0.5 and a/c=0.6): h1 values for at different heights
136
Table D.4 Model 9 (a/t=0.8 and a/c=1.0): h1 values for at different heights
137
Angle
(degrees)
0.00
3.04
6.10
9.00
11.93
14.60
17.30
19.73
22.19
24.25
26.35
28.23
30.15
31.88
33.64
35.24
36.87
41.60
Table E.5 Model 1 (a/t=0.2, a/c=0.2): K-Factor data from ABAQUS (Part 1)
Newman-Raju FEA K FEA K FEA K Directly
% diff
% diff
% diff
Directly from Jel
Tied Nodes
FEA Direct from Jel Tied Nodes
0.59
0.85
0.81
0.83
44.13
37.15
40.93
0.60
0.59
0.59
0.59
-1.16
-1.52
-0.74
0.62
0.58
0.59
0.56
-5.63
-4.86
-8.50
0.64
0.61
0.62
0.62
-4.64
-4.50
-4.24
0.68
0.64
0.64
0.63
-5.64
-5.38
-7.29
0.71
0.67
0.67
0.68
-5.35
-5.22
-5.17
0.75
0.71
0.71
0.70
-5.30
-5.18
-6.30
0.78
0.74
0.74
0.74
-5.15
-5.05
-5.13
0.80
0.77
0.77
0.76
-4.83
-4.75
-5.62
0.83
0.79
0.79
0.79
-4.95
-4.86
-4.93
0.85
0.81
0.81
0.81
-4.49
-4.40
-5.10
0.87
0.83
0.83
0.83
-4.58
-4.49
-4.55
0.89
0.85
0.85
0.85
-4.09
-4.00
-4.65
0.91
0.87
0.87
0.87
-4.31
-4.23
-4.25
0.92
0.89
0.89
0.89
-3.43
-3.34
-4.10
0.94
0.90
0.90
0.90
-4.42
-4.32
-4.34
0.95
0.92
0.93
0.92
-2.95
-2.90
-3.26
0.99
0.97
0.97
0.97
-2.33
-2.20
-2.35
138
Angle
(degrees)
46.71
50.74
55.15
58.76
62.80
66.12
69.95
73.03
76.79
79.57
83.44
85.36
90.00
Table E.6 Model 1 (a/t=0.2, a/c=0.2): K-Factor data from ABAQUS (Part 2)
Newman-Raju FEA K FEA K FEA K Directly
% diff
% diff
% diff
Directly from Jel
Tied Nodes
FEA Direct from Jel Tied Nodes
1.03
1.01
1.01
1.00
-2.44
-2.38
-2.60
1.06
1.04
1.04
1.04
-2.01
-1.91
-2.07
1.08
1.06
1.07
1.06
-1.85
-1.79
-2.05
1.10
1.09
1.09
1.08
-1.74
-1.66
-1.81
1.12
1.11
1.11
1.10
-1.63
-1.56
-1.82
1.14
1.12
1.12
1.12
-1.59
-1.52
-1.65
1.15
1.14
1.14
1.13
-1.51
-1.45
-1.70
1.16
1.15
1.15
1.14
-1.52
-1.45
-1.59
1.17
1.16
1.16
1.15
-1.48
-1.40
-1.63
1.18
1.16
1.16
1.16
-1.49
-1.42
-1.55
1.18
1.17
1.17
1.16
-1.46
-1.38
-1.61
1.19
1.17
1.17
1.17
-1.48
-1.41
-1.55
1.19
1.18
1.18
1.18
-0.90
-0.88
-1.05
139
140
141
Table E.9 Model 4 (a/t=0.5, a/c=0.2): K-Factor data from ABAQUS (Part 1)
Angle
Newman-Raju FEA K FEA K
% diff
% diff
(degrees)
Directly from Jel FEA Direct from Jel
0.00
0.82
1.17
1.10
42.52
35.03
3.04
0.82
0.80
0.80
-2.24
-2.56
6.10
0.84
0.78
0.79
-6.40
-5.57
9.00
0.87
0.83
0.83
-5.07
-4.91
11.93
0.91
0.86
0.86
-5.70
-5.37
14.60
0.95
0.90
0.90
-5.09
-4.95
17.30
0.99
0.94
0.94
-4.74
-4.56
19.73
1.02
0.98
0.98
-4.34
-4.21
22.19
1.05
1.01
1.02
-3.78
-3.67
24.25
1.08
1.04
1.04
-3.63
-3.51
26.35
1.11
1.08
1.08
-2.96
-2.87
28.23
1.13
1.10
1.10
-2.77
-2.67
30.15
1.15
1.13
1.13
-2.19
-2.10
31.88
1.17
1.15
1.15
-2.02
-1.92
33.64
1.19
1.18
1.18
-1.36
-1.27
35.24
1.21
1.19
1.19
-1.38
-1.28
36.87
1.23
1.22
1.22
-0.79
-0.73
39.40
1.25
1.25
1.25
-0.23
-0.12
42.02
1.28
1.28
1.28
0.02
0.11
44.32
1.30
1.30
1.31
0.53
0.62
46.71
1.32
1.33
1.33
0.87
0.95
48.83
1.34
1.35
1.35
1.18
1.25
51.06
1.35
1.37
1.37
1.49
1.56
142
Table E.10 Model 4 (a/t=0.5, a/c=0.2): K-Factor data from ABAQUS (Part 2)
Angle
Newman-Raju FEA K FEA K
% diff
% diff
(degrees)
Directly from Jel FEA Direct from Jel
53.05
1.37
1.39
1.39
1.72
1.79
55.15
1.38
1.41
1.41
1.97
2.03
57.05
1.40
1.43
1.43
2.15
2.21
59.05
1.41
1.44
1.44
2.35
2.41
60.87
1.42
1.46
1.46
2.48
2.54
62.80
1.43
1.47
1.47
2.66
2.71
64.55
1.44
1.48
1.48
2.76
2.82
66.42
1.45
1.49
1.49
2.88
2.92
68.12
1.46
1.50
1.50
2.93
2.99
69.95
1.47
1.51
1.51
3.03
3.08
71.60
1.47
1.52
1.52
3.10
3.15
73.40
1.48
1.53
1.53
3.17
3.22
75.00
1.49
1.53
1.53
3.21
3.26
76.79
1.49
1.54
1.54
3.26
3.31
78.34
1.49
1.54
1.54
3.29
3.33
80.13
1.50
1.55
1.55
3.33
3.37
81.62
1.50
1.55
1.55
3.34
3.39
83.44
1.50
1.56
1.56
3.37
3.41
84.81
1.51
1.56
1.56
3.38
3.42
86.72
1.51
1.56
1.56
3.39
3.43
87.68
1.51
1.56
1.56
3.39
3.43
90.00
1.51
1.57
1.57
3.76
3.79
143
144
145
Table E.13 Model 7 (a/t=0.8, a/c=0.2): K-Factor data from ABAQUS (Part 1)
Angle
Newman-Raju FEA K FEA K
% diff
% diff
(degrees)
Directly from Jel FEA Direct from Jel
0.00
1.19
1.49
1.46
25.75
23.29
2.44
1.17
1.23
1.22
4.52
4.07
4.88
1.18
1.19
1.20
0.62
1.29
7.20
1.21
1.22
1.22
0.90
0.98
9.53
1.25
1.24
1.25
-0.19
0.00
11.64
1.28
1.28
1.28
-0.08
-0.03
13.77
1.32
1.32
1.32
-0.06
0.01
15.68
1.35
1.35
1.35
0.01
0.03
17.61
1.39
1.39
1.39
0.36
0.39
19.22
1.41
1.42
1.42
0.32
0.32
20.84
1.44
1.45
1.45
0.94
0.95
22.31
1.46
1.48
1.48
1.02
1.00
23.79
1.49
1.51
1.51
1.57
1.55
25.12
1.51
1.53
1.53
1.62
1.62
26.48
1.53
1.56
1.56
2.26
2.26
27.71
1.54
1.58
1.58
2.21
2.20
28.96
1.56
1.60
1.60
2.76
2.73
30.69
1.58
1.63
1.63
3.15
3.14
32.46
1.61
1.66
1.66
3.30
3.27
34.05
1.63
1.69
1.69
3.66
3.64
35.66
1.65
1.71
1.71
3.88
3.86
37.13
1.66
1.73
1.73
4.04
4.01
38.62
1.68
1.75
1.75
4.22
4.18
146
Table E.14 Model 7 (a/t=0.8, a/c=0.2): K-Factor data from ABAQUS (Part 2)
Angle
Newman-Raju FEA K FEA K
% diff
% diff
(degrees)
Directly from Jel FEA Direct from Jel
40.00
1.70
1.77
1.77
4.28
4.24
41.41
1.71
1.79
1.78
4.38
4.34
42.72
1.72
1.80
1.80
4.37
4.33
44.05
1.74
1.81
1.81
4.41
4.37
45.29
1.75
1.83
1.83
4.33
4.30
46.57
1.76
1.84
1.84
4.34
4.30
47.76
1.78
1.85
1.85
4.22
4.17
48.99
1.79
1.86
1.86
4.12
4.07
50.14
1.80
1.87
1.87
3.92
3.86
51.32
1.81
1.88
1.88
3.77
3.72
52.43
1.82
1.88
1.88
3.61
3.54
53.58
1.83
1.89
1.89
3.41
3.36
54.66
1.84
1.90
1.90
3.18
3.12
55.77
1.85
1.90
1.90
2.97
2.91
56.83
1.86
1.91
1.90
2.72
2.65
57.91
1.86
1.91
1.91
2.47
2.41
58.94
1.87
1.91
1.91
2.19
2.12
60.00
1.88
1.92
1.92
1.94
1.86
61.01
1.89
1.92
1.92
1.65
1.57
62.05
1.90
1.92
1.92
1.39
1.30
63.03
1.90
1.92
1.92
1.09
1.01
64.06
1.91
1.92
1.92
0.78
0.70
65.02
1.92
1.93
1.92
0.47
0.39
66.03
1.92
1.93
1.92
0.19
0.10
66.98
1.93
1.93
1.92
-0.09
-0.19
67.98
1.93
1.93
1.93
-0.37
-0.48
147
Table E.15 Model 7 (a/t=0.8, a/c=0.2): K-Factor data from ABAQUS (Part 3)
Angle
Newman-Raju FEA K FEA K
% diff
% diff
(degrees)
Directly from Jel FEA Direct from Jel
68.91
1.94
1.93
1.93
-0.68
-0.77
69.89
1.95
1.93
1.93
-0.95
-1.05
70.82
1.95
1.93
1.92
-1.22
-1.33
71.79
1.96
1.93
1.92
-1.49
-1.59
72.70
1.96
1.93
1.92
-1.75
-1.86
73.67
1.96
1.93
1.92
-2.00
-2.11
74.57
1.97
1.92
1.92
-2.25
-2.36
75.52
1.97
1.92
1.92
-2.48
-2.59
76.41
1.98
1.92
1.92
-2.70
-2.81
77.36
1.98
1.92
1.92
-2.92
-3.03
78.24
1.98
1.92
1.92
-3.13
-3.24
79.19
1.99
1.92
1.92
-3.31
-3.43
80.06
1.99
1.92
1.92
-3.47
-3.59
81.01
1.99
1.92
1.92
-3.64
-3.75
81.86
1.99
1.92
1.92
-3.78
-3.90
82.82
1.99
1.92
1.91
-3.92
-4.04
83.66
2.00
1.92
1.91
-4.03
-4.15
84.62
2.00
1.92
1.91
-4.14
-4.26
85.43
2.00
1.92
1.91
-4.22
-4.34
86.42
2.00
1.91
1.91
-4.30
-4.42
87.17
2.00
1.91
1.91
-4.34
-4.47
88.21
2.00
1.91
1.91
-4.40
-4.51
88.73
2.00
1.91
1.91
-4.41
-4.53
90.00
2.00
1.92
1.91
-4.29
-4.42
148
Table E.16 Model 8 (a/t=0.8, a/c=0.6): K-Factor data from ABAQUS (Part 1)
Angle
Newman-Raju FEA K FEA K FEA K Directly
% diff
% diff
% diff
(degrees)
Directly from Jel
Tied Nodes
FEA Direct from Jel Tied Nodes
0.00
1.40
1.53
1.52
1.52
9.52
8.66
8.74
4.03
1.35
1.42
1.41
1.42
4.60
4.38
4.56
8.09
1.32
1.37
1.38
1.36
3.57
4.40
2.95
11.97
1.29
1.33
1.34
1.34
3.19
3.28
3.27
15.92
1.27
1.31
1.31
1.30
2.69
3.10
2.07
19.55
1.26
1.30
1.30
1.30
2.68
2.78
2.78
23.26
1.25
1.29
1.29
1.28
2.50
2.77
1.94
26.62
1.25
1.28
1.28
1.28
2.39
2.47
2.47
30.09
1.25
1.28
1.28
1.27
2.26
2.43
1.75
33.00
1.25
1.27
1.27
1.27
1.81
1.90
1.91
36.02
1.25
1.28
1.28
1.27
1.88
1.98
1.40
38.72
1.26
1.27
1.28
1.28
1.33
1.41
1.42
41.52
1.26
1.28
1.28
1.27
1.23
1.29
0.79
44.02
1.27
1.28
1.28
1.28
0.56
0.64
0.64
46.62
1.27
1.28
1.28
1.28
0.60
0.62
0.14
48.95
1.28
1.27
1.27
1.27
-0.45
-0.45
-0.48
149
Table E.17 Model 8 (a/t=0.8, a/c=0.6): K-Factor data from ABAQUS (Part 2)
Angle
Newman-Raju FEA K FEA K FEA K Directly
% diff
% diff
% diff
(degrees)
Directly from Jel
Tied Nodes
FEA Direct from Jel Tied Nodes
51.39
1.29
1.28
1.28
1.28
-0.49
-0.50
-0.86
54.40
1.29
1.28
1.28
1.28
-1.09
-1.04
-1.02
57.66
1.30
1.27
1.28
1.27
-2.20
-2.12
-2.41
60.46
1.31
1.28
1.28
1.28
-2.69
-2.66
-2.68
63.53
1.32
1.27
1.28
1.27
-3.40
-3.36
-3.66
66.16
1.33
1.27
1.27
1.27
-4.04
-4.02
-4.03
69.11
1.33
1.27
1.27
1.27
-4.64
-4.61
-4.89
71.61
1.34
1.27
1.27
1.27
-5.20
-5.18
-5.20
74.49
1.35
1.27
1.27
1.27
-5.67
-5.67
-5.92
76.85
1.35
1.27
1.27
1.27
-6.12
-6.11
-6.12
79.73
1.35
1.27
1.27
1.26
-6.45
-6.45
-6.69
81.89
1.36
1.26
1.26
1.26
-6.75
-6.75
-6.75
84.89
1.36
1.26
1.26
1.26
-6.93
-6.93
-7.17
86.38
1.36
1.26
1.26
1.26
-7.07
-7.07
-7.07
90.00
1.36
1.28
1.28
1.28
-5.74
-5.91
-5.97
150
151
152
153
154
155
156
157
158
159
160
Table F.27 Model 7 (a/t=0.8, a/c=0.2): h1 data from ABAQUS (Part 2)
Angle
h1 ,
h1 ,
h1 ,
(degrees) n = 5 n = 10 n = 15
55.77
23.80 54.35 112.99
56.83
23.55 53.55 111.12
57.91
23.43 53.35 110.84
58.94
23.18 52.65 109.23
60.00
23.05 52.42 108.88
61.01
22.80 51.82 107.85
62.05
22.67 51.72 107.54
63.03
22.43 51.18 106.86
64.06
22.33 51.16 106.78
65.02
22.11 50.81 106.32
66.03
22.02 50.81 106.22
66.98
21.83 50.45 105.56
67.98
21.76 50.52 105.78
68.91
21.59 50.15 104.89
69.89
21.54 50.25 105.25
70.82
21.39 49.85 104.21
71.79
21.36 49.97 104.64
72.70
21.22 49.54 103.50
73.67
21.21 49.66 104.00
74.57
21.07 49.20 102.85
75.52
21.07 49.33 103.28
76.41
20.94 48.86 102.37
77.36
20.94 48.99 102.53
78.24
20.83 48.57 101.94
79.19
20.83 48.73 102.44
80.06
20.72 48.28 101.45
81.01
20.73 48.48 102.11
81.86
20.63 48.07 101.26
82.82
20.65 48.26 101.69
83.66
20.55 47.92 101.17
84.62
20.58 48.12 101.41
85.43
20.49 47.80 101.05
86.42
20.53 48.04 101.28
87.17
20.45 47.73 100.97
88.21
20.50 47.99 101.21
88.73
20.43 47.70 100.93
90.00
20.56 48.15 101.57
161
162
163
164
Table G.30 Stress vs. strain data at n = 15, based on Equation 3.13
Stress Elastic Strain Plastic Strain Elastic to Plastic Ratio
34000
0.0017
5.82 105
29.19
35200
0.00176
9.8E-05
17.96
36400
0.00182
0.000162
11.23
37600
0.00188
0.000264
7.13
38800
0.00194
0.000422
4.60
40000
0.002
0.000667
3.00
41200
0.00206
0.001039
1.98
42400
0.00212
0.001598
1.33
43600
0.00218
0.002428
0.90
44800
0.00224
0.003649
0.61
46000
0.0023
0.005425
0.42
47200
0.00236
0.007982
0.30
48400
0.00242
0.011633
0.21
49600
0.00248
0.016797
0.15
50800
0.00254
0.024042
0.11
52000
0.0026
0.034124
0.08
53200
0.00266
0.048049
0.06
54400
0.00272
0.067142
0.04
55600
0.00278
0.093139
0.03
56800
0.00284
0.128302
0.02
58000
0.0029
0.175561
0.02
59200
0.00296
0.23869
0.01
60400
0.00302
0.322525
0.01
61600
0.00308
0.433231
0.01
62800
0.00314
0.578633
0.01
64000
0.0032
0.768614
0.00
65200
0.00326
1.0156
0.00
165
Table G.31 Stress vs. strain data at n = 10, based on Equation 3.13
Stress Elastic Strain Plastic Strain Elastic to Plastic Ratio
30000
0.0015
3.75 105
39.95488493
35200
0.00176
0.000185667
9.47932046
39600
0.00198
0.000602921
3.284010245
44000
0.0022
0.001729162
1.272292855
48400
0.00242
0.004485
0.53957637
52800
0.00264
0.010706513
0.246578879
57200
0.00286
0.023837962
0.119976701
61600
0.00308
0.050016805
0.061579303
66000
0.0033
0.099712174
0.033095257
70400
0.00352
0.190123333
0.018514298
74800
0.00374
0.348597932
0.010728692
79200
0.00396
0.617391497
0.006414083
83600
0.00418
1.060160459
0.003942799
166
167
Table G.33 Stress vs. plastic strain data at n = 15, used for ABAQUS models
Stress Plastic Strain
34000
0.00
35200
9.8 105
36400
0.000162
37600
0.000264
38800
0.000422
40000
0.000667
41200
0.001039
42400
0.001598
43600
0.002428
44800
0.003649
46000
0.005425
47200
0.007982
48400
0.011633
49600
0.016797
50800
0.024042
52000
0.034124
53200
0.048049
54400
0.067142
55600
0.093139
56800
0.128302
58000
0.175561
59200
0.23869
60400
0.322525
61600
0.433231
62800
0.578633
64000
0.768614
65200
1.0156
168
Table G.34 Stress vs. plastic strain data at n = 10, used for ABAQUS models
Stress Plastic Strain
30000
0.00
35200 0.000185667
39600 0.000602921
44000 0.001729162
48400
0.004485
52800 0.010706513
57200 0.023837962
61600 0.050016805
66000 0.099712174
70400 0.190123333
74800 0.348597932
79200 0.617391497
83600 1.060160459
169
Table G.35 Stress vs. plastic strain data at n = 5, used for ABAQUS models
Stress Plastic Strain
22000
0.00
26400
8.35 105
30000
0.000158
35200
0.000352
39600
0.000634
44000
0.001074
48400
0.001729
52800
0.002672
57200
0.003986
61600
0.005774
66000
0.008153
70400
0.011258
74800
0.015245
79200
0.020288
83600
0.026585
88000
0.034358
92400
0.04385
96800
0.055333
101200
0.069105
105600
0.085493
110000
0.104851
114400
0.127567
118800
0.15406
123200
0.184783
127600
0.220223
132000
0.260903
136400
0.307384
140800
0.360265
145200
0.420186
149600
0.487828
154000
0.563913
158400
0.649209
162800
0.744528
167200
0.850727
171600
0.968713
176000
1.099441
VITA
170