A nite element ductile failure simulation method using

stress-modied fracture strain model

Chang-Sik Oh
a
, Nak-Hyun Kim
a
, Yun-Jae Kim
a,
, Jong-Hyun Baek
b
, Young-Pyo Kim
b
,
Woo-Sik Kim
b
a
Korea University, Department of Mechanical Engineering, Anam-Dong, Sungbuk-Ku, Seoul 136-701, Republic of Korea
b
Korea Gas Corporation Research and Development Center, 638-1 Ansan, Kyonggi-do 425-150, Republic of Korea
a r t i c l e i n f o
Article history:
Received 14 January 2010
Received in revised form 4 September 2010
Accepted 13 October 2010
Available online 20 October 2010
Keywords:
Ductile fracture simulation
Finite element analysis
Stress-modied fracture strain
a b s t r a c t
This paper proposes a new method to simulate ductile failure using nite element analysis
based on the stress-modied fracture strain model. A procedure is given to determine the
stress-modied fracture strain as a function of the stress triaxiality from smooth and
notched bar tensile tests with FE analyses. For validation, simulated results using the pro-
posed method are compared with experimental data for cracked bar (tensile and bend)
tests, extracted from API X65 pipes, and for full-scale burst test of gouged pipes, showing
overall good agreements. Advantages in the use of the proposed method for practical struc-
tural integrity assessment are discussed.
2010 Elsevier Ltd. All rights reserved.
1. Introduction
For structural integrity analyses, performing full-scale tests is important, but is in general quite expensive and time-
consuming. Furthermore, it is often very difcult to perform full-scale tests reecting complex geometries and loading
conditions in practical assessment. For these reasons, an efcient tool may be needed not only to design complex full-scale
tests but also possibly to minimize expensive and time-consuming full-scale tests. Finite element ductile failure simulations
based on the local approach are quite useful in this respect and thus increasingly important.
A number of works have been reported in the literature up to present on nite element ductile failure simulations.
Depending on the model employed for simulating damage, existing works can be broadly classied into two categories.
The rst one is using a micro-mechanical model for ductile fracture, incorporating void nucleation, growth and coalescence,
for instance, the GursonTvergaardNeedleman model [116] and the Rousellier model [1720]. The second category is
using a phenomenological model for ductile fracture. One popular model is, for instance, the cohesive zone model [21
27]. As many researchers have already published a number of papers using these methods, applicability and validity of these
methods have been well discussed in the literature. From the authors point of view, it is felt that a few issues need to be
resolved in practical application of these methods. The rst one is how to nd parameters embedded in these models. For
instance, the GTN model has eight parameters related to micro-mechanism of ductile fracture. Determination of these
parameters are not an easy task, and often not robust. Although other models tend to have less parameters, robust parameter
determination is a common problem in nite element ductile failure simulations. The second issue is that, when these
methods are to be implemented into commercial nite element programs, special subroutines or elements often need to
be developed.
0013-7944/$ - see front matter 2010 Elsevier Ltd. All rights reserved.
doi:10.1016/j.engfracmech.2010.10.004

Corresponding author. Tel.: +82 2 3290 3372; fax: +82 2 926 9290.
E-mail address: kimy0308@korea.ac.kr (Y.-J. Kim).
Engineering Fracture Mechanics 78 (2011) 124137
Contents lists available at ScienceDirect
Engineering Fracture Mechanics
j our nal homepage: www. el sevi er . com/ l ocat e/ engf r acmech
This paper proposes a new method to simulate ductile failure using nite element method, based on a phenomenological
stress-modied fracture strain model for ductile fracture. Section 2 briey summarizes experimental results. In Section 3, the
proposed method is introduced, together with explanations on how to simulate progressive failure. Section 4 compares sim-
ulated results with experimental ones. The present work is discussed in Section 5 and concluded in Section 6.
2. Summary of experimental programs
This paper attempts to propose a nite element based ductile failure simulation method. Tests were performed for two
objectives; rstly to show how to apply the proposed method to simulate ductile failure, and secondly to validate the pro-
posed method by comparing with experimental data. These tests include smooth and notched bar tensile tests, cracked bar
tests and full-scale burst tests of gouged pipes. The material is the API X65 steel [28] popularly used for gas transportation
within Korea. Chemical composition of the API X65 steel is given in Table 1. This section briey explains these tests and
results.
2.1. Smooth and notched round bar tensile tests
Standard round bar specimens for tensile tests were extracted (in the longitudinal direction) from a API X65 pipe having
outer diameter D
o
= 762 mm and wall thickness t = 17.5 mm. To investigate the effect of triaxial stress states on tensile prop-
erties, notched round bar specimens with three different notch radii, 6.0 mm, 3.0 mm and 1.5 mm, were also made. For all
specimens, the minimum section has a diameter of 6.0 mm. Schematic diagrams for smooth and notched round tensile spec-
imens, employed in the present work, are depicted in Fig. 1.
Tensile tests were performed for smooth and round notched bars. In testing, axial displacement was monitored using
extensometer with the gauge length of 25 mm. For a given specimen geometry, tests were repeated three times. Engineering
stressstrain data from all tests are shown in Fig. 2a. Yield and tensile strengths determined from smooth bar tensile tests
were about 465 MPa and 563 MPa, respectively, as tabulated in Table 1. True stressstrain data, obtained from smooth round
bar tests using the Bridgman correction [29], are shown in Fig. 2b. The true fracture strain is $1.2, indicating that the mate-
rial is sufciently ductile. Notched bar tests show that, as the notch radius decreases, the yield and tensile strengths increase,
but the strain to fracture decreases due to the higher stress triaxialties introduced by the notch.
Nomenclature
a crack length
D
o
outer diameter
d gouge depth
E Youngs modulus
gouge length
P, P
b
pressure and burst pressure
R radius of the minimum section of tensile bars
r radial distance from the center of the smooth and notched bars
t thickness
w width
d displacement at the gouge mouth, see Fig. 5
e
e
equivalent strain, Eq. (5)
e
p
(equivalent) plastic strain
e
f
fracture strain
e
1
, e
2
, e
3
principal strain components
m Poissons ratio
r
e
, r
m
effective stress and mean normal stress, respectively, see Eqs. (1) and (2)
r
1
, r
2
, r
3
principal stress components
r
y
, r
u
yield (0.2% proof) strength and tensile strength, respectively
x, Dx damage and incremental damage
Table 1
Chemical compositions and mechanical tensile properties at room temperature of the API X65 steel, used in the present work.
Youngs modulus Poissons ratio Yield strength Tensile strength Element (wt.%)
E (GPa) m r
y
(MPa) r
u
(MPa) C P Mn S Si Fe Ceq
210.7 0.3 464.5 563.8 0.08 0.019 1.45 0.03 0.31 Balance 0.32
C.-S. Oh et al. / Engineering Fracture Mechanics 78 (2011) 124137 125
2.2. Cracked bar tests
Single-edge-cracked bar tests were performed. To investigate constraint effects due to the loading mode on failure behav-
iours, two types of cracked specimens were extracted from API X65 pipes. One is for three-point bending loading (single-
edge cracked bend specimen, SE(B), Fig. 3a) and the other is for tensile loading (single-edge cracked tension specimen,
SE(T), Fig. 3b). Relevant dimensions are also shown in Fig. 3. The width w and thickness t of SE(B) specimens were
Fig. 1. Schematic illustrations of (a) smooth bar and (b) notched bar tensile specimens (units: mm).
Fig. 2. (a) Engineering stressstrain curves from smooth and notched round bar tests, and (b) true stressstrain data.
Fig. 3. Schematic illustrations of (a) SE(B) and (b) SE(T) specimens (units: mm).
126 C.-S. Oh et al. / Engineering Fracture Mechanics 78 (2011) 124137
10 mm. The SE(T) specimens had the width of w = 12 mm and the thickness of t = 25 mm. Note that sizes of SE(T) and SE(B)
bars were small, as they were extracted directly from a pipe piece. An initial notch was machined and the subsequent pre-
cracking was introduced by fatigue. No side-groove was made for all specimens. For SE(B) specimens, the nal crack length,
a, was about 50% of the width. Specimens were loaded using three-point bend, and the crack mouth opening displacement
(CMOD) was monitored as a function of load. Load-CMOD curves resulting fromtwo identical specimens are shown in Fig. 4a.
For SE(T) specimens, the nal crack lengths were about 25%, 50% and 75% of the width, to see the effect of the crack length on
failure behaviours. A total of seven specimens were tested, two for a/w = 0.25, three for a/w = 0.5 and two for a/w = 0.75. Ten-
sile loading was applied using the grip, and the displacement was monitored using the extensometer with the 10 mm gauge
length. Resulting loaddisplacement curves are summarized in Fig. 4b. It shows that overall behaviours for SE(T) specimens
are quite different from those for SE(B) ones.
2.3. Burst tests of full-scale pipes with gouge defect
Test pipes having the outer diameter of D
o
= 762 mm, the thickness of t = 17.5 mm, and the total length of 2300 mm were
prepared. To simulate gouge defect, a 45 V-notch with a circular notch radius of 2 mm was made on the outer surface of the
pipe, as shown in Fig. 5. The depth of the gouge, d, was xed as d = 8.75 mm which is 50% of the pipe thickness (d/t = 0.5). The
axial length of the gouge, on the other hand, was systematically varied from = 100 mm to = 600 mm (Table 2). Both ends
were capped by circumferential welding. The pipes were pressurized by water and burst pressures were experimentally
determined at the point when the ligament failed. Fig. 6 depicts photos of the gouge before and after the test. The measured
maximum (burst) pressures, P
b
, from full-scale pipe tests are summarized in Table 2, which shows that the burst pressure
decreases with increasing gouge length.
Fig. 4. (a) Experimental load-CMOD curves for SE(B) specimens, and (b) experimental loaddisplacement curves for SE(T) specimens.
Table 2
Summary of full-scale tests of pipes with gouge defect.
Pipe no. MNA MNB MNC MND MNE
l (mm) 100 200 300 400 600
P
b
(MPa) 24.68 22.48 17.7 18.14 16.57
Fig. 5. Schematic illustrations of a full-scale pipe with gouge defect (units: mm).
C.-S. Oh et al. / Engineering Fracture Mechanics 78 (2011) 124137 127
3. Proposed ductile failure simulation method
3.1. Damage model and failure simulation
The damage model proposed in this paper is based on the concept of the stress-modied fracture strain model. It has been
well-known that (true) fracture strain e
f
for dimple fracture strongly depends on the stress triaxiality (dened by the ratio of
the mean normal stress r
m
and equivalent stress r
e
) [3036]:
r
m
r
e

r
1
r
2
r
3
3r
e
1
r
e

1
2
fr
1
r
2

2
r
1
r
3

2
r
3
r
2

2
g
_
2
where r
i
(i = 13) denotes the principal stress. Although detailed expressions differ slightly, the dependence of e
f
on the
stress triaxiality can be modelled using an exponential function
e
f
/ exp a
r
m
r
e
_ _
3
where a is a material constant. A detailed form of the fracture strain can be found from notched bar tensile tests, as will be
described in Section 3.3. Once the form of e
f
is available as a function of the stress triaxiality, incremental damage due to
plastic deformation, Dx, is calculated (at each gauss point) using
Dx
De
p
e
f
4
where Dep is the equivalent plastic strain increment, calculated from FE analysis.
When the accumulated damage becomes unity, x= RDx= 1, ductile failure is assumed and incremental crack growth is
simulated by reducing all stress components at the gauss point sharply to a small plateau value, as schematically depicted in
Fig. 7. Decreasing stresses to zero can cause numerical problems, and thus stresses are assumed to remain a nite (cut-off)
value. The results from sensitivity analysis for the effects of the decreasing slope and cut-off stress value will be given later.
3.2. Implementation to FE program ABAQUS
Suppose the fracture strain e
f
is determined for a given material. Then damage simulation using the proposed method
requires information on the stress triaxiality and incremental plastic strain values, which can be determined from elas-
ticplastic FE analysis. Thus the proposed method can be easily implemented in commercial FE programs. To implement
the proposed damage model to ABAQUS [37], two user subroutines were developed. The rst one is the USDFLD subroutine
to re-dene eld variables. By dening as state variables, information on the mean normal (hydrostatic) stresses and equiv-
alent stresses/strains at gauss points is passed into the UHARD subroutine. In the UHARD subroutine, accumulated damage is
calculated according to Eq. (4) using information passed from the USDFLD subroutine. When the accumulated damage be-
comes critical (unity), stresses are relaxed simply by changing the yield surface.
3.3. Determination of fracture strain e
f
To apply the present damage model, the rst step is to determine the fracture strain e
f
as a function of the stress triaxiality
for a given material. By combining notched bar test results and elasticplastic FE analysis, the explicit form of e
f
can be found
Fig. 6. Gouge defect (a) before the test, and (b) after the test.
128 C.-S. Oh et al. / Engineering Fracture Mechanics 78 (2011) 124137
as follows. Elasticplastic, axi-symmetric FE analyses simulating tensile tests of smooth and notched round bar specimens
were performed to determine variations of the triaxial stresses and strains within the specimens. Symmetric conditions were
fully utilized and the second order, reduced integration elements (CAX8R within ABAQUS) were used for efcient computa-
tion. Typical FE meshes with the different notch radii are shown in Fig. 8. The number of elements and nodes in typical FE
meshes ranged from 484 elements/1557 nodes to 658 elements/2089 nodes. To incorporate the large geometry change effect
in tensile testing, the large geometry change option was chosen.
Fig. 9 compares experimental engineering stressstrain data from smooth and notched (notch radius = 3 mm) tensile
tests with the FE results. Although the FE analysis cannot reproduce failure of tensile test specimens, it can well simulate
deformation behaviour even after necking up to failure initiation points. Agreements between the test results and FE ones
are quite good up to failure initiation points which are indicated in Fig. 9 using the cross symbols.
Fig. 8. FE meshes for notched tensile bars: (a) notch = 1.5 mm and (b) notch = 3 mm.
Fig. 7. Schematic illustration of simulating ductile failure using stress relaxation.
Fig. 9. Comparison of experimental engineering stressstrain data for (a) smooth and (b) notched (radius = 3 mm) tensile bars with FE results.
C.-S. Oh et al. / Engineering Fracture Mechanics 78 (2011) 124137 129
Local stress and strain elds in the minimumsection of the tensile bars were extracted from the FE results as a function of
applied load. Fig. 10 show radial variations of the stress triaxiality r
m
/r
e
and equivalent strain e
e
in the minimum (necked)
section of the smooth and notched bars at the point of failure initiation (see Fig. 9). The equivalent strain, e
e
, is dened by
e
e

2
p
3

fe
1
e
2

2
e
1
e
3

2
e
3
e
2

2
g
_
5
where e
i
(i = 13) denotes the principal strain. In the gure, the distance (r) is normalized with respect to the radius of the
minimum section (R), and the values of r/R = 0 and r/R = 1 mean the center and the free surface of the specimen, respectively
(Fig. 8). The stress triaxiality decreases with increasing notch radius, but always attains its maximum value in the center of
the specimen, regardless of the notch radius. Furthermore, equivalent strain attains its maximum value in the center of the
bar, except for the 1.5 mm case where the maximum value of equivalent strain occurs at the notch tip. Thus for all cases
except for the 1.5 mm case, failure is expected to initiate in the center of the minimum section in the specimen. Even for
the 1.5 mm case, as the stress triaxiality in the center is much higher than that in the notch tip, failure is also expected
to initiate in the center of the minimum section in the specimen.
Fig. 11a shows the evolution of the stress triaxiality in terms of the equivalent strain for smooth and notched round ten-
sile bars in solid lines. Both the stress triaxiality and the equivalent strain are extracted in the center of the minimum section
of test specimens, where failure initiation is expected to occur. The last points in Fig. 11a correspond to the failure initiation
points. It shows that the stress triaxiality in the center of the specimen depend on the equivalent strain. As a ductile failure
criterion should include the history of stress and strain, average stress triaxiality is introduced, dened by
r
m
r
e
_ _
ave

1
e
ef
_
e
ef
0
r
m
r
e
de
e
6
Fig. 10. (a) Stress triaxiality distributions and (b) equivalent strain distributions for smooth and notched tensile bars at the failure initiation point,
determined from the FE analyses.
Fig. 11. (a) Variations of the stress triaxiality with the equivalent strain for smooth and notched bar tensile tests, and (b) fracture strain as a function of the
stress triaxiality.
130 C.-S. Oh et al. / Engineering Fracture Mechanics 78 (2011) 124137
where e
ef
denotes the equivalent strain to failure initiation. Such denition can incorporate the history effect on stresses and
strains on ductile fracture. For a given notch radius, the calculated average stress triaxiality is constant, and is shown in
Fig. 11a with dotted lines. Resulting equivalent strains to failure initiation (called the fracture strain) are shown in
Fig. 11b, as a function of the (average) stress triaxiality. Note that one point in Fig. 11b corresponds to the result for one notch
radius. It shows that the fracture strain decreases sharply with increasing the stress triaxiality. Noting that the fracture strain
is found to be exponentially dependent on the stress triaxiality [36], the following regression is proposed for the fracture
strain e
f
:
e
f
3:29exp 1:54
r
m
r
e
_ _
0:10 7
which is shown in Fig. 11b with a solid line. It shows that Eq. (7) agrees with the data and captures dependence of the stress
triaxiality on equivalent strain to fracture. Note that the history dependent ductile failure strain, similar to the proposed one,
was originally discussed by Marini et al. [38].
3.4. GursonTvergaardNeedleman (GTN) Model
For comparison, results from the proposed model are also compared with those using the GTN (GursonNeedleman
Tvergaard) model. The GTN model to simulate ductile damage and failure is given by [1,5,6].
U
r
2
e

r
2
2q
1
f

cosh q
2
3r
m
2

r
_ _
1 q
2
1
f
2
0 8
r
ij
_
e
p
ij
1 f
_

re
p
f

f f 6 f
c
f
c

u
fc
f
f
fc
f f
c
f > f
c
_
_
f
_
f
growth

_
f
nucleation
_
f
growth
1 f
_
e
p
kk
;
_
f
nucleation
A
_

e
p
; A
f
N
s
N

2p
p exp
1
2
e
p
e
e
N
s
N
_ _
2
_ _
9
In Eqs. (8) and (9), r is the current ow stress of the matrix material; f

denes the effective void volume fraction;

_
e
p
ij
is the
macroscopic plastic strain rate tensor;
_

e
p
is the average effective plastic strain rate of the material; f denes the current void
volume fraction; f
f
and f
c
denote the void volume fraction at failure and at a critical situation at which f starts to deviate from
f

, respectively; f

u
is the value f

at fracture (i.e., f

u
= f

u
(f
f
));
_
e
p
kk
is the trace of the macroscopic plastic strain rate tensor; f
N
is
the volume fraction of particles available for void nucleation, e
N
is the mean void nucleation strain; and s
N
is the standard
deviation of the distribution. Note that the GTN model has eight micro-mechanical parameters (e
N
, s
N
, f
N
; f
o
, f
c
, f
f
; q
1
, q
2
) that
must be calibrated for a given material.
For the API X65 steel, the authors have already calibrated parameters for the GTN model in Ref. [39]. Five parameters re-
lated to void nucleation and the rened yield locus were xed to typical values suggested in literature [26]: e
N
= 0.3, s
N
= 0.1,
f
N
= 0.0008; q
1
= 1.5 and q
2
= 1.0. The parameter f
o
, related to initial void volume fraction, was determined from the Franklins
formula [40] (based on the assumption of a spherical inclusion):
f
o
0:054 S%
0:001
Mn%
_ _
10
which leads to f
o
= 0.000125 for the present material. Two parameters (f
c
and f
f
) were calibrated by comparing smooth and
notched tensile bar test results with simulated results using the GTN model. They were chosen so that the simulated results
give best ts to test results. Resulting values of f
c
and f
f
were found to be f
c
= 0.015 and f
f
= 0.25. In applying the GTN model to
simulate ductile failure, a proper choice of the parameters in the GTN model is crucial. It should be noted that the calibrated
parameters are purely empirical, and we believe that no unique parameter set can be found. The calibrated values resulted
from more than hundred trials, when led to a conclusion that the above sets gave best results, compared to experimental
notched bar tests.
Finally it should be noted that the calibrated parameters may depend on the nite element size. The above parameters
were calibrated using the element size of 0.15 mm (150 lm) which will be used throughout this paper.
4. Results
4.1. Sensitivity analysis
To apply the proposed model to simulate ductile failure, several points related to the proposed FE damage analysis need
to be resolved. They include, for instance, the element type and size. For the element type, rst-order elements (the
four-node quadrilateral element for two dimensional and axi-symmetric problems and the eight-node brick element for
C.-S. Oh et al. / Engineering Fracture Mechanics 78 (2011) 124137 131
Fig. 13. Sensitivity analysis results from 2-D SE(B) and SE(T) test simulations: (a) and (b) the effect of the decreasing slope; (c) and (d) the effect of the
tolerance.
Fig. 12. FE meshes for the SE(B) specimen with a/w = 0.46: (a) 2-D mesh and (b) 3-D mesh.
132 C.-S. Oh et al. / Engineering Fracture Mechanics 78 (2011) 124137
three-dimensional problems) with full integrations are used in the present work. Although second-order elements can be
used in principle, they often exhibit uctuating hydrostatic stress values and thus can cause numerical problems when
the accumulated damage is calculated. For the element size, it is rather obvious that the element size can affect simulated
results from FE damage analyses, and thus cannot be chosen arbitrarily. It should be chosen to reect the materials length
scale (such as void spacing for ductile fracture), and this is true not only for the proposed method but also for all FE damage
analyses. In this paper, the mesh size effect on simulated results was not investigated, and the size was xed to 0.15 mm
(150 lm).
1
As the proposed method decreases stresses at the gauss point to the cut-off value when the accumulated damage be-
comes unity (as schematically depicted in Fig. 7), appropriate decreasing slope and cut-off value should be chosen. To inves-
tigate the effects of the decreasing slope and the cut-off value on simulated results, sensitivity analysis was performed for the
cracked (SE(B) and SE(T)) bars, as effects would be the most pronounced for crack problems. For efcient investigations, two-
dimensional (2-D) simulations were performed rather than three-dimensional (3-D) ones. A typical FE mesh is shown in
Fig. 12a. Four-node plane strain solid elements (element type CPE4) of the size 0.15 mm 0.15 mm were uniformly spaced
in the cracked section. Simulated loaddisplacement records for SE(B) and SE(T) specimens are compared with experimental
data in Fig. 13. It shows that simulated results do not depend on the decreasing slope and the cut-off value, as long the
decreasing slope was taken to be smaller than 1/5000 (when the strain increases by 0.1, the stress decreases more than
500 MPa) and the cut-off value is less than 50 MPa. Although the decreasing slope and cut-off value should be ideally zero,
such values can cause numerical problems. Thus, for numerical efciency, the cut-off value and decreasing slope in this work
were taken to be 10% of the yield strength (46 MPa) and 1/5000, respectively.
4.2. Smooth and notched bar tensile test results
Fig. 14 compares smooth and notched bar tensile results with simulated FE results resulting from three different analyses.
One is conventional elasticplastic analysis without any damage model (shown in dotted lines). The other two are using
Fig. 14. Comparison of smooth and notched bar tensile test results with simulated ones: (a) notch radius = 1.5 mm, (b) notch radius = 3 mm, (c) notch
radius = 6 mm, and (d) smooth bar.
1
One reason to x the element size as 0.15 mm is to compare the present results with those using the GTN model. As noted in Section 3.4, the GTN
parameters were calibrated using the element size of 0.15 mm.
C.-S. Oh et al. / Engineering Fracture Mechanics 78 (2011) 124137 133
damage models, one using the GTN model (shown in open symbols), described in Section 3.4, and the other using the pro-
posed method (shown in closed symbols). For all cases, the element size was xed to 0.15 mm. To incorporate the large
geometry change effect in tensile testing, the large geometry change option was chosen.
The conventional elasticplastic FE analysis can predict plastic behaviours of smooth and notched tensile tests up to the
failure initiations, indicated in the gures. However, after failure initiation, predictions deviate from experimental data, as
the conventional analysis cannot simulate failure. The FE results using the GTN model agree well with the experimental re-
sults for all cases. This is not surprising as the parameters associated with the GTN model are determined so as to give good
ts to smooth and notched bar tensile data, as explained in Section 3.4. For the FE analysis using the proposed method, agree-
ment is very good for smooth bar tensile test data. However, for notched bar tensile test data, the results are not as good as
those using the GTN model. Although the failure initiation can be well predicted, subsequent crack growth seems to be un-
der-predicted. Considering the fact that the present method does not involve any tting process, predictions for notched bar
tests can be regarded as good.
4.3. Cracked bar test results
For comparison with cracked bar test results, 3-D FE analyses were performed, rather than 2-D plane strain analyses. This
is because the tested cracked bars do not have any side-groove, and thus the plane strain assumption may not appropriate. A
typical FE mesh for 3-D analyses is shown in Fig. 12b. Eight-node brick elements with full integrations (element type C3D8)
of the size 0.15 mm 0.15 mm 15 mm were uniformly spaced in the cracked section. The number of elements and nodes
in FE meshes were from 8202 elements/9791 nodes to 15,390 elements/17,654 nodes. True stressstrain data were directly
given in the FE analysis, and incremental plasticity with the large geometry change option was invoked.
Fig. 15 compares cracked bar test results with simulated FE results. Again for the FE results, those from three different
analyses are compared with experimental data; the conventional analysis without any damage model (in dotted lines),
the analysis using the GTN model (in open symbols), and that using the proposed method (in closed symbols). Deviations
of the conventional elasticplastic FE results from experimental data are due to crack growth. The proposed method gives
overall good predictions for all cases considered, and agreements with experimental data are overall slightly better than
those for the GTN model.
Fig. 15. Comparison of cracked bar test results with simulated ones: (a) SE(B) with a/w = 0.46; (b), (c) and (d) SE(T) with a/w = 0.25, a/w = 0.5 and a/w = 0.75,
respectively.
134 C.-S. Oh et al. / Engineering Fracture Mechanics 78 (2011) 124137
4.4. Burst tests of full-scale pipes with gouge defect
Burst tests of full-scale pipes with gouge defects are simulated using FE analyses. A typical nite element mesh is shown
in Fig. 16, with a detailed view of the defective region. Eight-node brick elements with full integrations (element type C3D8)
of the size 0.15 mm 0.15 mm 15 mm were uniformly spaced in the defective section. The number of elements and nodes
in FE meshes were from 43,082 elements/51,088 nodes to 74,028 elements/89,569 nodes, depending on the gouge length.
Internal pressure was applied to the inner surface of the pipe, together with end forces to simulate the closed end condition.
True stressstrain data were directly given in the FE analysis, and incremental plasticity with the large geometry change op-
tion was invoked.
Fig. 17a shows simulated pressuredisplacement curves of gouged pipes. The displacement d is measured at the gouge
mouth, as schematically shown in Fig. 5. For FE results, the results from two different analyses are compared; one from con-
ventional elasticplastic analysis and the other from the proposed method. Failure points predicted from the proposed meth-
od are indicated using the cross symbols. Estimated burst pressures from the proposed method are compared with
experimentally measured data in Fig. 17b. The estimated burst pressures are in good agreement with experimentally mea-
sured ones with differences less than $3%, except the MNC case where the difference is as much as $12%. Although such a
difference is still not signicant, it can be seen from Fig. 17b that the experimental data for the MNC case are somewhat low-
er than expected from the trend curve. Estimated burst pressures using the GTN model are also compared in Fig. 17b, show-
ing that predictions are overall close to those using from the proposed method. Both experimental and estimated burst
pressures decrease with increasing gouge length, as shown in Fig. 17b. Such a trend can be readily explained by idealizing
gouge as an axial crack. Empirical-based burst pressures for axial cracked pipes are given by [41].
P
f

r
y
r
u
2
_ _
2t
D
o
1
d
t
1
d
t

1
M
_ _
; M

1 1:61

2
2D
o
tt

11
Fig. 16. Typical FE mesh for pipes with gouge.
Fig. 17. (a) FE pressuredisplacement curves including predicted failure points, and (b) comparison of predicted burst pressures with experimental ones.
C.-S. Oh et al. / Engineering Fracture Mechanics 78 (2011) 124137 135
where d and denote the crack depth and length, respectively. This equation predicts that burst pressures should decrease
with increasing crack length. Burst pressures estimated from Eq. (11) are compared with experimental and estimated ones in
Fig. 17b, showing that Eq. (11) gives conservative (lower) failure loads, as expected.
5. Concluding remarks
In this paper, a new method to simulate ductile failure based on FE analysis is proposed. The method is based on the
stress-modied fracture strain model which has been well-known phenomenological model for ductile fracture. Incremental
damage is dened by the ratio of incremental plastic strain and stress-modied fracture strain. When the accumulated dam-
age becomes unity, then stresses at the gauss point are decreased to a small value to simulate progressive failure. To validate
the proposed method, tests are performed using specimens extracted from API X65 grade pipes. The stress-modied fracture
strain as a function of the stress triaxiality is rstly obtained by combining FE analyses and experimental data of smooth and
notched bar tensile tests. Then simulated results using the proposed method are then compared with experimental data
from cracked bar (tensile and bend) tests and from full-scale burst test of gouged pipes. Good comparisons with experimen-
tal results support the condence of the proposed method.
The proposed method adopts the stress-modied fracture strain model as a failure criterion for ductile fracture, which has
been known for many years. Using notched bar tensile tests, determination of the stress-modied fracture strain model for a
given material is straightforward and robust, as described in this paper. Once the stress-modied fracture strain model is
determined, the proposed damage model does not include any parameter and thus does not require any calibration proce-
dure, which offers signicant advantage in practical application.
Acknowledgements
This research is supported by Basic Atomic Energy Research Institute and Nuclear R&D Program
(M2080608000208M060800210), funded by the Korea Science & Engineering Foundation.
References
[1] Gurson AL. Continuum theory of ductile rupture by void nucleation and growth. Part 1-yield criteria and ow rules for porous ductile media. J Engng
Mater Tech 1977;99:215.
[2] Tvergaard V. Inuence of voids on shear band instabilities under plane strain conditions. Int J Fract 1981;17:389407.
[3] Tvergaard V. On Localization in Ductile Materials Containing Spherical Voids. Int J Fract 1982;18:23752.
[4] Koplik J, Needleman A. Void growth and coalescence in porous plastic solids. Int J Solids Struct 1988;24:83553.
[5] Chu C, Needleman A. Void nucleation effects in biaxially stretched sheets. J Engng Mater Tech 1980;102:24956.
[6] Tvergaard V, Needleman A. Analysis of the cup-cone fracture in a round tensile bar. Acta Metall 1982;32:15769.
[7] Bernauer G, Brocks W. Phenomenological modeling of ductile damage and tearing-results of a European numerical round robin. Fat Fract Engng Mater
Struct 2002;25:36384.
[8] Dotta F, Ruggieri C. Structural integrity assessments of high pressure pipelines with axial aws using a micromechanics model. Int J Pres Ves Piping
2004;81:76170.
[9] Rivalin F, Pineau A, Di Fant M, Besson J. Ductile tearing of pipeline-steel wide plates I-dynamic and quasi-static experiments. Engng Fract Mech
2001;68:32945.
[10] Rivalin F, Besson J, Pineau A, Di Fant M. Ductile tearing of pipeline-steel wide plates II-modeling of in-plane crack propagation. Engng Fract Mech
2001;68:34764.
[11] Chen Y, Lambert S. Numerical modeling of ductile tearing for semi-elliptical surface cracks in wide plates. Int J Pres Ves Piping 2005;82:41726.
[12] Roy Chowdhury S, Narasimhan R. A nite element analysis of quasistatic crack growth in a pressure sensitive constrained ductile layer. Engng Fract
Mech 2000;66:55171.
[13] Skallerud B, Zhang ZL. A 3D numerical study of ductile tearing and fatigue crack growth under nominal cyclic plasticity. Int J Solids Struct
1997;24:314161.
[14] Imad A, Wilsius J, Nait Abdelaziz M, Mesmacque G. Experiments and numerical approaches to ductile tearing in an 2024-T351 aluminium alloy. Int J
Mech Sci 2003;45:184961.
[15] Gao X, Faleskog J, Shih CF. Ductile tearing in part-through cracks: experiments and cell-model predictions. Engng Fract Mech 1998;59:76177.
[16] Needleman A, Tvergaard V. An analysis of ductile rupture modes at a crack tip. J Mech Phys Solids 1987;35:15183.
[17] Rousselier G. Ductile fracture models and their potential in local approach of fracture. Nucl Engng Des 1987;105:97111.
[18] Roos E, Seidenfuss M, Kramer D, Krolop S, Eisele U, Hindenlang U. Application and evaluation of different methods for determining crack resistance
curves. Nucl Engng Des 1991;130:297308.
[19] Burstow MC, Howard IC. Predicting the effects of crack tip constraint on material resistance curves using ductile damage theory. Fat Fract Engng Mater
Struct 1996;19:46174.
[20] Bron F, Besson J. Simulation of the ductile tearing for two grades of 2024 aluminium alloy thin sheets. Engng Fract Mech 2006;73:153152.
[21] Chen CR, Kolednik O, Scheider I, Siegmund T, Tatschl A, Fischer FD. On the determination of the cohesive zone parameters for the modelling of micro-
ductile crack growth in thick specimens. Int J Fract 2003;120:51736.
[22] Chen CR, Kolednik O, Heerens J, Fischer FD. Three dimensional modelling of ductile crack growth: cohesive zone parameters and crack tip triaxiality.
Engng Fract Mech 2005;72:207294.
[23] Sorensen BF, Jacobsen TK. Determination of cohesive laws by the J integral approach. Engng Fract Mech 2003;70:184158.
[24] Scheider I, Brocks W. Simulation of cup-cone fracture using the cohesive model. Engng Fract Mech 2003;70:194361.
[25] Borst R. Numerical aspects of cohesive-zone models. Engng Fract Mech 2003;70:174357.
[26] Tvergaard V, Hutchinson JW. The relation between crack growth resistance and fracture process parameters in elasticplastic solids. J Mech Phys Solids
1992;40:137797.
[27] Cornec A, Scheider I, Schwalbe K-H. On the practical application of the cohesive model. Engng Fract Mech 2003;70:196387.
[28] American Petroleum Institute. Specication for Line Pie. API 5L; 2000.
[29] Bridgman P. Studies in large plastic ow and fracture. New York: McGraw-Hill Book Company Inc.; 1952.
[30] McClintock FA. A criterion of ductile fracture by the growth of holes. J Appl Mech 1968;35:36371.
136 C.-S. Oh et al. / Engineering Fracture Mechanics 78 (2011) 124137
[31] Hancock JW, Mackenzie AC. On the mechanisms of ductile failure in high-strength steels subject to multi-axial stress states. J Mech Phys Solids
1976;24:14769.
[32] Mackenzie A, Hancock J, Brown D. On the inuence of state of stress on ductile failure initiation in high strength steels. Engng Fract Mech
1977;9:16788.
[33] Hancock JW, Cowling MJ. Role of state of stress in crack-tip failure processes. Metal Sci 1980:293304.
[34] Thomason PF. Ductile fracture of metals. UK: Pergamon Press, Oxford; 1990.
[35] Anderson T. Fracture mechanics fundamentals and applications. CRC Press; 1995.
[36] Rice JR, Tracey DM. On the ductile enlargement of voids in triaxial stress elds. J Mech Phys Solids 1969;17:20117.
[37] ABAQUS Version 6.7. Users manual. Inc., and Dassault Systemes; 2007.
[38] Marini B, Mudry F, Pineau A. Ductile rupture of A508 steel under nonradial loading. Engng Fract Mech 1985;22:37586.
[39] Oh C-K, Kim Y-J, Baek J-H, Kim Y-P, Kim W-S. A phenomenological model of ductile fracture for API X65 steel. Int J Mech Sci 2007;49:1399412.
[40] Franklin AG. Comparison between a quantitative microscope and chemical methods for assessment of non-metallic inclusions. J Iron Steel Inst
1969;207:1816.
[41] ASME. ASME Boiler and pressure vessel Code Section XI. 1992.
C.-S. Oh et al. / Engineering Fracture Mechanics 78 (2011) 124137 137