Vous êtes sur la page 1sur 14

ARTICLE IN PRESS

International Journal of Mechanical Sciences 49 (2007) 13991412


www.elsevier.com/locate/ijmecsci

A phenomenological model of ductile fracture for API X65 steel


Chang-Kyun Oha, Yun-Jae Kima,, Jong-Hyun Baekb, Young-Pyo Kimb, Woosik Kimb
a

Department of Mechanical Engineering, Korea University, 5 Ka, Anam-dong, Sungbuk-Ku, Seoul 136-701, South Korea
Korea Gas Corporation (KOGAS) Research and Development Center, 638-1 Il-dong, Ansan, Kyunggi-do 425-150, South Korea

Received 8 February 2007; accepted 20 March 2007


Available online 24 March 2007

Abstract
This paper presents a phenomenological model of ductile fracture for the API X65 steel using the GursonTvergaardNeedleman
(GTN) model. Experimental tests and FE damage simulations using the GTN model are performed for smooth and notched tensile bars,
from which the parameters in the GTN model are calibrated. Comparison of experimental data of pre-strained, notched tensile and
fracture toughness tests with nite element (FE) damage analyses show good agreements, suggesting the validity of the calibrated
parameters. As application, the developed GTN model is applied to predict the pre-strain effect on deformation and fracture and the
results are compared with experimental data.
r 2007 Elsevier Ltd. All rights reserved.
Keywords: GTN model; Ductile fracture; API X65; Pre-strain

1. Introduction
Crack-like defect assessment is an important element in
tness-for-service analyses based on which engineering
decisions can be made whether such defect can be tolerated
for continued operation. For the last two decades,
signicant progress has been made in development of
engineering methods for crack-like defect assessment (see
for example Refs. [111]). These methods can be roughly
classied into the following three categories. The rst
category is the net-section limit load approach [3,4] based
on solid mechanics concept, where a defective structure is
assumed to fail at the load when the net section reaches
fully plastic stress state. Such approach is valid for
structures made of materials with sufcient ductility. It is
simple and could be efcient in application, as it requires
only tensile properties of the material of interest. However,
validation is rather expensive, requiring a large number of
full-scale pipe test data. The second category is the
approach based on elasticplastic (non-linear) fracture
mechanics [12,13]. In this approach, information on
Corresponding author. Tel.: +82 2 3290 3372; fax: +82 31 290 5276.

E-mail address: kimy0308@korea.ac.kr (Y.-J. Kim).


0020-7403/$ - see front matter r 2007 Elsevier Ltd. All rights reserved.
doi:10.1016/j.ijmecsci.2007.03.008

fracture toughness of the material of interest is required


in addition to tensile properties. This approach is increasingly popular and accordingly many tness-for-service
codes have been developed based on this approach(see for
example Refs. [2,49]). Although existing codes are
designed to provide conservative failure predictions of
defective structures, it is well known that structural
geometry and the mode of loading can affect fracture
toughness, which is often termed to as the crack-tip
constraint effect. To accommodate the constraint effect, a
second fracture mechanics parameter has been proposed
[1417]. Although some codes include a method to
incorporate such constraint effects, they are still in a
developing stage. A more comprehensive approach is based
on phenomenological models of fracture, which is the last
category. For ductile fracture, for instance, the micromechanical processes of void nucleation, growth and
coalescence have been modelled extensively for the last
three decades (see for example Refs. [1829]). Among
them, one notable model is due to Gurson [18] who has
developed the constitutive model of a porous material by
incorporating the void volume fraction as an extra internal
variable.The Gurson model has been improved signicantly (see for instance Refs. [2429]), to account for void

ARTICLE IN PRESS
C.-K. Oh et al. / International Journal of Mechanical Sciences 49 (2007) 13991412

1400

nucleation, coalescence and strain hardening. More recently efforts have been made on application of such
micro-mechanical models to predict failure of cracked
specimens and defective structures, via detailed computational schemes [3039]. In fact, the application of the
phenomenological models to predict failure of component
can provide signicant benets, particularly when relevant
tests are either very difcult or impossible.
This paper rstly develops a GTN (Gurson, Tvergaard
and Needleman) model of ductile fracture for the API
(American Petroleum Institute) X65 steel, popularly used
for gas and oil pipelines within Korea. The parameters in
the GTN model are calibrated by combining notched
tensile bar tests and detailed nite element (FE) analyses.
For application, the calibrated model is applied to quantify
the pre-strain effect on deformation and fracture, and the
results are compared with experimental data of prestrained smooth and notched tensile bars and single-edge
cracked bend bars. Section 2 briey introduces the GTN
model as the phenomenological model used in the present
work. Section 3 discusses calibration of the parameters in
the phenomenological model for the API X65 steel. Section
4 applies the proposed model to quantify the pre-strain
effect on deformation and fracture. The present work is
concluded in Section 5.
2. GTN model

where se denotes the macroscopic effective Mises stress; sh


represents the macroscopic hydrostatic stress; s is the
current ow stress of the matrix material; and f  denes the
effective void volume fraction. The matrix ow stress, s,
can be obtained by enforcing equality between the rates of
macroscopic and matrix plastic work:

e_pij

f pf c ;
f 4f c ;

(2)

where
is the macroscopic plastic strain rate tensor,
and e_ p is the average effective plastic strain rate of the
material. It can be seen that the above yield locus recovers
that of von Mises in the limiting case of zero porosity
f  ! 0. Note that Tvergaard has originally suggested the
values of additional parameters as q1 1:5 and q2 1. The
effective void volume fraction f  has been introduced by
Needleman and Tvergaard [39] to better predict the effect

(3)

where 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 . The load
carrying capacity vanishes when f  1=q1 , and correspondingly f u 1=q1 .
To complete the constitutive model, a rate form of the
void volume fraction has to be established. The increases in
porosity may in general have contributions from two
processes; the growth of existing voids and the nucleation
of new voids. Mathematically this can be expressed as
f_ f_growth f_nucleation .

(4)

Assume that the matrix is plastically incompressible, the


growth part is directly related to the mesoscopic plastic
dilatation:
f_growth 1  f _epkk ,

(5)

where e_ pkk is the trace of the macroscopic plastic strain rate


tensor. Void nucleation based on plastic straining can be
included by
f_nucleation Ae_p .

Gurson [18] has developed a constitutive model for


porous ductile media based on a rigid-plastic material
behaviour and the upper bound theorem of plasticity.
Based on detailed phenomenological studies of the
bifurcation behaviour of materials containing periodic
distributions of cylindrical and spherical voids, Tvergaard
[24,25] has rened the constitutive model by incorporating
some additional parameters (q1 , q2 , resulting in the
following yield function:


s2
3sh
(1)
F e2 2q1 f  cosh q2
 1 q21 f 2 0,
s
2s

sij e_ pij 1  f se_p ,

of void coalescence,
8
>
<f;


f f f u  f c f  f ;
>
c
: c ff fc

(6)

The parameter A follows a normal distribution as


suggested in Ref. [28]
(

 )
fN
1 epe  eN 2
A p exp 
.
(7)
2
sN
sN 2p
In Eq. (7), f N is the volume fraction of particles available
for void nucleation, eN is the mean void nucleation
strain, and sN is the standard deviation of the distribution.
Eqs. (1)(7) constitute the GTN model to be used to
simulate ductile damage and failure in the present work.
The GTN model has eight parameters eN ; sN ; f N ; f o ; f c ; f f ;
q1 ; q2 that must be calibrated for a given material.
3. Calibration of damage parameter
In the previous section, the GTN model was described. It
was shown that the model included eight micro-mechanistic parameters to be calibrated for a given material. In
this section, these parameters are calibrated for the API
X65 steel, by comparing with notched round bar test
results with FE damage simulations.
3.1. Mechanical tests
For the present study, the API 5L X65 steel [40] is
chosen, which is popularly used for gas pipelines in Korea.
To determine mechanical properties, tensile and Charpy
tests were performed using specimens extracted from a pipe

ARTICLE IN PRESS
C.-K. Oh et al. / International Journal of Mechanical Sciences 49 (2007) 13991412

of outer diameter Do 762 mm and wall thickness t


17:5 mm (in the longitudinal direction). Chemical compositions and tensile properties of the present API X65 steel are
summarized in Tables 1 and 2, respectively. True stress
strain data at room temperature and values of the Charpy
energy as a function of temperature are shown in Fig. 1.
Even at 60  C, the Charpy impact energy was more than
250 J and all specimens failed by 100% dimple fracture,
indicating that the present material is sufciently tough.
One interesting point is that, as most of API X65 gas
pipelines in Korea have been built within the last 10 years,
mechanical properties of API X65 gas pipelines in Korea
tend to have quite uniform properties.
To investigate the effect of triaxial stress states on tensile
ductility of the API X65 steel, tensile tests were performed
using smooth and notched round bars with three different
notch radii. Schematic diagrams for smooth and notched
tensile bars, employed in the present work, are depicted in
Fig. 2. For all specimens, the minimum section has a
diameter of 6 mm. For notched bars, three different round
notch radii were machined, which were 6.0 mm 6R,
3.0 mm 3R and 1.5 mm 1:5R. Fig. 3 depicts tensile

testing of the round bar with the 3 mm notch radius. In


testing, axial displacement was monitored using extensometer with the length of 25 mm. For a given specimen
geometry, three tests were performed, giving a total of 12
tests. Engineering stressstrain data for all specimens tested

Fig. 2. Schematic models: (a) tensile test specimens and (b) notched tensile
test specimens with notch radii 1:5, 3, 6 (mm).

Table 1
Chemical composition of the API X65 steel
Elements (wt.%)
C
P

Mn

Si

Fe

Ceq

0.08

1.45

0.03

0.31

Balance

0.32

0.019

Table 2
Tensile properties of the API X65 steel
Youngs modulus

Poissons ratio

Yield strength

Tensile strength

E (GPa)

sy (MPa)

sy (MPa)

210.7

0.3

464.5

563.8

1401

Fig. 3. Conguration of notched tensile tests, notch 3R.

Fig. 1. (a) True stressstrain data at room temperature and (b) Charpy energy as a function of temperature for API X65.

ARTICLE IN PRESS
1402

C.-K. Oh et al. / International Journal of Mechanical Sciences 49 (2007) 13991412

Fig. 4. Experimental results for tensile tests and notched tensile tests.

are summarized in Fig. 4. It shows that yield and tensile


strengths increase with decreasing the notch radius, but the
engineering strain to fracture (tensile ductility) decreases.
Such trend is due to the fact that the triaxial stress increases
with decreasing the notch radius.
3.2. FE analysis
Detailed elasticplastic, axi-symmetric FE damage analyses based on the GTN model, described in Section 2, were
performed to simulate tensile tests of smooth and notched
round bars using ABAQUS [41]. The GTN model,
described in Section 2, has been implemented into
ABAQUS using the user-material subroutine [42]. Averaged true stressstrain data of the API X65 steel, resulting
from three tensile tests of the smooth tensile specimen, are
shown in Fig. 1. The experimental true stressplastic strain
data, shown in Fig. 1, were directly given in the FE
analysis. To incorporate a large geometry change effect, the
large geometry change option was chosen. The deformation boundary condition was applied to the top of the FE
model, and the resulting tensile load was determined from
nodal forces. The gauge length elongation was also
monitored from the FE displacement results.
The four-node solid element within ABAQUS (element
type CAX4) was used and in the minimum section,
elements with the length of 0.15 mm are uniformly spaced.
It should be noted that the element size could be important
in FE damage analysis. The number of elements and nodes
in FE meshes ranges from 484 elements/1557 nodes to
658 elements/2089 nodes, depending on a notch radius.
Typical FE meshes, employed in the present work, are
shown in Fig. 5.

Fig. 5. FE meshes for notched tensile specimens: (a) notch 1:5R and (b)
notch 3R.

3.3. Calibration of damage parameter


It was noted that the GTN model, described in
Eqs. (1)(7), has eight parameters: three parameters related
to void nucleation, eN , sN and f N ; three for void
growth and coalescence, f o , f c and f f ; and two for the
rened yield locus, q1 and q2 . In the present work, ve
parameters related to void nucleation and the rened
yield locus are xed to typical values suggested in literature
[24,25]: eN 0:3, sN 0:1, f N 0:0008; q1 1:5 and
q2 1:0. Furthermore, the parameter f o , related to initial
void volume fraction, is determined from the Franklins
formula [43] (based on the assumption of a spherical

ARTICLE IN PRESS
C.-K. Oh et al. / International Journal of Mechanical Sciences 49 (2007) 13991412

inclusion):

f o 0:054 S% 

0:001
,
Mn%

(8)

which leads to f o 0:000125 for the present material. Thus


only two parameters (f c and f f need to be calibrated. By
comparing smooth and notched tensile bar test results with
the present FE damage analyses, these two parameters are
Table 3
Calibrated parameters of the GTN model for the API X65 steel
Parameter set 1

Parameter set 2

Material parameter
q1
q2

1.5
1

1.5
1

Void nucleation
eN
sN
fN

0.3
0.1
0.0008

0.3
0.1
0.0

Void growth/coalescence
fo
fc
ff

0.000125
0.015
0.25

0.000125
0.003
0.18

1403

chosen so that the simulated results give best ts to test


results. Resulting values of f c and f f are found to be f c
0:015 and f f 0:25, as summarized in Table 2 (referred to
as the parameter set 1). In applying the GTN model to
simulate ductile failure, a proper choice of the parameters
in the GTN model is crucial. To validate the calibrated
values listed in Table 3, experimental notched bar tensile
test results will be compared with simulated ones using the
calibrated parameters in the next sub-section, and experimental loaddisplacement records from cracked bend bar
tests with simulated ones in Section 4. The calibrated
parameters will be further discussed later in Section 3.5,
with emphasis on sensitivity analysis results.
3.4. Comparison with notched bar test results
Fig. 6 compares experimental (engineering) stressstrain
data with the results from FE damage analyses using the
parameters given above (FE results with damage).
Experimental data include those from smooth round bar
test and notched bar tests with three different notch radii.
For each type of the specimen, three test data are shown.
For the FE results, conventional elasticplastic FE
analyses based on isotropic elasticplastic materials with

Fig. 6. Comparison of notched tensile tests with FE simulations: (a) smooth, (b) notched 6R, (c) notched 3R and (d) notched 1:5R.

ARTICLE IN PRESS
C.-K. Oh et al. / International Journal of Mechanical Sciences 49 (2007) 13991412

1404

the large geometry change option are also performed and


the results are also included (FE results w/o damage) in
Fig. 6 for comparison. Conventional elasticplastic FE
analyses can predict deformation behaviours of smooth
and notched tensile bars, even after the maximum loads
(necking points), but fail to simulate failure initiation and
subsequent failure. The FE damage analyses based on the
GTN model, on the other hand, can predict failure
initiation points reasonably well not only for smooth
round bars but also for notched round bars with different
notch radii. Good agreements between test data with
simulated FE results suggest that the calibrated parameters
for the GTN model are acceptable.

3.5. Discussion on calibrated parameters


In the proposed model, two parameters, f c and f f , were
calibrated; f c 0:015 and f f 0:25. It should be noted

Table 4
Parameters of the GTN model for different grades of API steels, reported
in literature

Material
fN
fo
fc
ff

Rivalin [32]

Chen [33]

Dotta [30]

API X70
0
0.00015
0.00074
0.18

API X70
0.00018
0.002
Variable
0.19

API X60
0
0.008
0
0.2

Fig. 8. Effects of the FE mesh size on simulation results for a notched


tensile bar with notch radius of 1.5 mm.

Fig. 7. Comparison of notched tensile tests with FE simulations: (a) smooth, (b) notched 6R, (c) notched 3R and (d) notched 1:5R.

ARTICLE IN PRESS
C.-K. Oh et al. / International Journal of Mechanical Sciences 49 (2007) 13991412

that the calibrated parameters are purely empirical, and we


believe that no unique parameter set can be found. We
have tried a number of different values for f c and f f (in fact
about hundred trials), and found the calibrated values
(listed in Table 3 as the parameter set 1) gave best results,
compared to experimental notched bar tests. Typical values
of the GTN parameters for slightly different grades (API
X60 and X70 steels), reported in literature, are listed in
Table 4. Note that no report can be found for the API X65
steel, considered in the present work. Although detailed
values of the parameters in literature are different, values
seem to be overall consistent to the present ones. Note that,
as shown in Table 4, the value of f N is often set to zero,
f N 0. As this value was set to typical one in literature

1405

[25,26], f N 0:0008,in the present work, the effect of the


choice of f N is also investigated. When f N 0 was used,
resulting values of f c and f f to give best ts to notched bar
tensile test results were found to be f c 0:003 and
f f 0:18, suggesting that calibrated values could strongly
depend on the choice of other parameters in the GTN
model. The resulting parameters are listed in Table 3
(referred to as the parameter set 2). Compared to the
values of the parameter set 1, values of f c and f f decrease.
Fig. 7 compares experimental (engineering) stressstrain
data with the results from FE damage analyses using the
parameter set 2. The results for the smooth bars and
notched bars with notch radii 6R and 3R are quite similar
to those in Fig. 2 (using the parameter set 1). However, for

Fig. 9. (a) A plate extracted from a pipe for pre-strain tests and (b) experimental set-up for pre-straining.

ARTICLE IN PRESS
1406

C.-K. Oh et al. / International Journal of Mechanical Sciences 49 (2007) 13991412

the case of the 1:5R notch radius, failure initiation is


predicted earlier using the parameter set 2 than using the
parameter set 1.
Note that the calibrated values also depend on the
element size. It was mentioned that the mesh size for the
FE damage analyses was xed to 0.15 mm in this paper.
The element could be related to the materials length scale
(such as void spacing for ductile fracture), this paper
assumes 0.15 mm for simplicity. Although this value is
believed to be realistic, considering similar values reported
in literature, mesh sensitivity analysis is also performed to
investigate the effect of the mesh size on simulated results.
Fig. 8 compares test data for the notched round bar (notch
radius 1:5R) with the results from FE damage analyses
with three different mesh sizes, 0.075, 0.15 and 0.30 mm.
Although the smaller mesh size predicts slightly earlier
failure initiation and accelerated subsequent failure, the
results for two mesh sizes, 0.075 and 0.15 mm, are overall
close. The mesh size of 0.3 mm gives quite different results,
but is a rather large number, compared to other reported
values for similar structural steels.
As conclusions, it is believed that unique determination of the parameters in the GTN model would be

extremely difcult, if not impossible. The simulated results


are affected not only by the choice of each parameter
in the GTN model but also by the element size. A
fundamental approach would be to calibrate the parameters based on thorough metallurgical examinations.
Instead, a pragmatic approach is taken in this paper,
where the parameters are calibrated simply by comparing
the simulated results with notched bar test results. Based
on many trials, we strongly feel that the parameter set 1
listed in Table 3 would be the best choice to simulate
ductile fracture of the API X65 steel, together with the
element size of 0.15 mm.
4. Application to pre-strain effects
In the previous section, the model of ductile fracture
based on the GTN model has been established for the API
X65 steel. Once established, it has many potential
application areas. For instance, it can be used to predict
failure behaviours not only of toughness testing specimens
(which will be shown later in this paper) but also of fullscale pipes with crack-like defects. It can be also used to
predict size effects of toughness testing specimens.

Fig. 10. (ac) Experimental results of pre-strain effects on engineering stressstrain curves for notched and smooth tensile bars and (d) quantication of
the pre-strain effect on tensile curves.

ARTICLE IN PRESS
C.-K. Oh et al. / International Journal of Mechanical Sciences 49 (2007) 13991412

One of the important issues in maintenance of gas


pipeline is integrity assessment of third-party mechanical
damage. Such third-party damage could generate plastic
deformation and thus plastic pre-strain that in turn can
affect deformation and fracture behaviours of pipelines. In
literature, a number of analytical and experimental works
have been published on quantication of pre-strain effects
on deformation, fatigue and fracture (see for instance Refs.
[44,45] for analytical results and Refs. [4649] for experimental results). Thus quantication of pre-strain effects on
tensile and fracture toughness properties for API X65 steels
is not a main concern of this paper, as it has been reported
extensively in literature. Our main concern is application
and validation of the developed GTN model to quantify
pre-strain effects, with particular emphasis on numerical
procedures.

1407

be quantied simply by shifting the true stressplastic


strain curve by the amount of pre-strain [44,45], which is
supported by analysing the present experimental results, as
shown in Fig. 10d.
Using three-point bend bars extracted from the prestrained plates, the crack-tip opening displacement
(CTOD) tests were performed according to the BS 5762

4.1. Experiments
Tests were performed to quantify pre-strain effects on
tensile and fracture toughness properties of the API X65
steel. The material used in tests is same as that in Section
3.1. A plate specimen (with the thickness of 11 mm, the
width of 60 mm and the net-length of 30 mm) was extracted
from the pipe in the longitudinal direction (Fig. 9a). This
plate is plastically loaded and unloaded in axial tension to
generate the plastic strain in the plate (see Fig. 9b). The
applied engineering strain to this plate was either 5% or
10%. It should be noted that the engineering strain of 10%
corresponds to the engineering ultimate tensile strain from
smooth notch bar tests (see Fig. 4). Detailed threedimensional FE (damage) analysis for the plate was also
performed to quantify variations of plastic strains and
damage left in the plate. Detailed results will be given in the
next sub-section. From FE analyses, it was found that
actual pre-strains left in the plate were almost same as the
applied strains minus elastic ones due to recovery. Round
tensile bars and three-point bend specimens were extracted
from the pre-strained plates. The FE analyses also
suggested that pre-strains are not uniformly distributed
along the plate width due to possible edge effect. Based on
the FE results, tensile bars and bend specimens were
extracted from the centre region (within 30% of the width)
ofthe plate.
For the round tensile bar tests, overall experimental
procedures are same as those in Section 3. For a given prestrain, both smooth and notched tensile bars with three
different round notch radii, 6.0 mm 6R, 3.0 mm 3R and
1.5 mm 1:5R were tested. For all specimens, the minimum
section has a diameter of 6mm, and axial displacement was
monitored using extensometer with the length of 25 mm.
Resulting engineering stressstrain data for all specimens
tested are summarized in Fig. 10. It shows that increasing
pre-strain increases the yield and tensile strengths but
decreases the strain to fracture. Note that the pre-strain
effect on tensile properties is similar to the notch effect. It is
well known that the pre-strain effect on tensile curves can

Fig. 11. (a) Schematic illustration of the crack-tip opening displacement


(CTOD) tests and (b) a picture of three-point bend bars.

Fig. 12. Experimental loadCMOD curves from CTOD tests with and
without pre-strain.

ARTICLE IN PRESS
1408

C.-K. Oh et al. / International Journal of Mechanical Sciences 49 (2007) 13991412

[50]. As the thickness of full-scale pipe was 17.5 mm, bend


bars with the thickness of 10.0 mm and the width of
10.0 mm (B  B specimens) were extracted, as shown in
Fig. 11. An initial notch was machined and the subsequent
pre-cracking was introduced by fatigue. The length of the
nal pre-crack was about half of the specimen width,
according to the standard procedure. Bend bars were
loaded using three-point bend, and the crack mouth
opening displacement (CMOD) was monitored as a
function of load. Resulting loadCMOD curves are
summarized in Fig. 12. It shows that the maximum loads
for pre-strained bend bars are lower but occurs at smaller
CMOD values than those for bend bars without pre-strain.
Note that J-resistance curves could not be determined due
to small sizes of tested specimens. Thus only loadCMOD
curves will be compared.
4.2. FE analysis
To quantify variations of pre-strains and damage left in
the plate, detailed elasticplastic, three-dimensional FE
damage analyses based on the GTN model, described in
Section 2, were performed using ABAQUS [41] with the
large geometry change option. For the GTN model, the
parameter set 1 was used. The deformation boundary
condition was applied to the top of the FE model, which
matches the experimental condition. Fig. 13 shows the FE
mesh, made of eight-node brick elements (C3D8 within
ABAQUS). To be consistent to previous damage analysis,
the element size in the region of interest was xed to

0:15 mm  0:15 mm  0:15 mm. Resulting variations of


equivalent plastic strains, left in the plate, are shown in
Figs. 14a and b. Note that equivalent plastic strain
distributions are measured in the centre of the plate both
along the width direction (Fig. 14a) and along the length
direction (Fig. 14b). The gures also include estimated prestrains considering elastic recovery (4.62% and 9.24%),
which are slightly lower than applied strains (5% and
10%). The results show that for the case of the 5% prestrain, estimated values are in fairly good agreements in
both width and length directions. On the other hand, for
the case of the 10% pre-strain, the FE values are about
20% higher than estimated ones along the width direction.
This is probably due to out-of-plane constraint effects.
However, along the length direction, the pre-strain
decreases. In an average sense, the estimated values of
pre-strains (4.62% and 9.24%) are assumed in the present
work. Figs. 14c and d show variations of void volume
fractions left in the pre-strained plate. Noting that the
maximum values are of interest, values of void volume
fraction for the 5% and 10% pre-strain are f 0:00014
and f 0:0002, respectively. Compared to f o 0:000125
for the plate without pre-strain, the void volume fraction
increases by 12% and 60%, respectively.
Using above information, FE damage analyses based on
the GTN model are performed to simulate tensile tests of
(smooth/notched) round bars and CTOD tests using
ABAQUS [41] with the large geometry change option. To
incorporate pre-strain into FE damage analyses, two
modications are made. The rst one is that true

Fig. 13. A FE mesh for the plate subject to pre-strain.

ARTICLE IN PRESS
C.-K. Oh et al. / International Journal of Mechanical Sciences 49 (2007) 13991412

1409

Fig. 14. Variations of (a, b) equivalent plastic strains; and (b, d) void volume fractin in the pre-strained plate.

stressstrain data of the API X65 steel, shown in Fig. 1, are


modied according to pre-strain, simply by shifting the
amount of pre-strain (4.62% and 9.24%, respectively), as
shown in Fig. 10d. The second modication is the value of
initial void volume fraction, f o . As noted, the value of f o
changes due to pre-strain, and values of f o 0:00014 and
f o 0:0002 for the pre-strain of 4.62% and 9.24%,
respectively. Four-node solid elements (element type
CAX4 for round tensile bar simulation and CPE4 for
CTOD test simulation) were used and elements with the
length of 0.15 mm are uniformly spaced in the minimum
section. The FE mesh for single-edge-cracked bars is shown
in Fig. 15.
Fig. 16 compares experimental (engineering) stress
strain data with the results from FE damage analyses
(FE results with damage) for selected cases. The FE
results, conventional elasticplastic FE analyses based on
isotropic elasticplastic materials with the large geometry
change option are also included (FE results w/o damage)
for comparison. Fig. 17 shows corresponding results for
the CTOD tests, i.e., experimental loadCMOD curves are
compared with the results from FE damage analyses (FE

Fig. 15. The FE mesh for damage simulations of the CTOD tests.

results with damage) for selected cases. The FE results,


conventional elasticplastic FE analyses based on isotropic
elasticplastic materials with the large geometry change
option are also included (FE results w/o damage).

ARTICLE IN PRESS
C.-K. Oh et al. / International Journal of Mechanical Sciences 49 (2007) 13991412

1410

Fig. 16. Comparisons of FE simulations with test results of smooth and notched bars with pre-strain.

The results show that the results from FE damage analyses


using the calibrated GTN model are in good agreements
with experimental data, and can capture the pre-strain
effect not only on plastic deformation but also on ductile
fracture.
5. Concluding remarks
This paper presents a phenomenological model of ductile
fracture for the API X65 steel using the GursonTvergaardNeedleman (GTN) model. Experimental tests and
FE damage simulations using the GTN model are

performed for smooth and notched tensile bars with three


different notch radii, from which the parameters in the
GTN model are calibrated. Validity of the proposed
parameters is checked by comparing simulated results with
experimental ones from notched bar tests and cracked bend
bar tests.
The proposed phenomenological model would be quite
useful to predict ductile failure of defective pipes made of
API X65, particularly when relevant tests are either very
difcult or impossible. For instance, it would be difcult to
extract a large-scale cracked specimen from a pipe to
investigate the crack growth and arrest behaviours. In such

ARTICLE IN PRESS
C.-K. Oh et al. / International Journal of Mechanical Sciences 49 (2007) 13991412

Fig. 17. Comparison of FE simulation with CTOD test results: (a)


without pre-strain and (b) with the 10% pre-strain.

a case, the proposed model can be applied to virtually


simulate crack growth and arrest. It would be also useful to
investigate the crack-tip constraint effect on ductile
fracture. As one example, the calibrated GTN model is
applied to quantify the pre-strain effect on deformation
and fracture of the API X65 steel. Comparison of
experimental data of pre-strained, notched tensile tests
and fracture toughness tests with FE damage analyses
show good agreements.
References
[1] Kumar V, German D, Shih CF. An engineering approach for
elasticplastic fracture analysis. EPRI Final Report NP 1931; 1981.

1411

[2] Zahoor A. Ductile fracture handbook, vol. 3. Novotech Corp; 1991.


[3] Kanninnen MF, Zahoor A, Wilkoski G, Abousayed I, Marschall C,
Broek D, et al. Instability predictions for circumferentially cracked
type-304 stainless pipes under dynamic loading. EPRI Report NP2347. Palo Alto, USA: Electric Power research Institute; 1982.
[4] ASME. Rule for in-service inspection of nuclear power plant
componentsASME Boiler and Pressure Vessel Code, Section XI;
1996.
[5] British Energy. R6: assessment of the integrity of structures containing defects, revision 4; 1998.
[6] API. API RP579recommended practice for tness-for-service
American Petroleum Institute; 2000.
[7] Bergman M, Brickstad B, Dahlberg L, Nilsson F, Sattari-Far I. A
procedure for safety assessment of components with crackshandbook. SA/FoU Report 91/01; ABSvensk Anlaggningsprovning,
Swedish Plant Inspection Ltd, 1991.
[8] BS 7910. Guide on methods for assessing the acceptability of aws in
fusion welded structures. London: British Standards Institution; 2000.
[9] SINTAP. Final procedure: SINTAP (structural integrity assessment
procedures for European industry). Brite-Euram Project, BE95-1426;
1999.
[10] Schwalbe KH, Zerbst U, Kim YJ, Brocks W, Cornec A, Heerens J,
et al. EFAM ETM 97: the ETM method for assessing the signicance
of crack-like defect in engineering structures GKSS 98/E/6. Germany:
GKSS Research Centre; 1998.
[11] Ainsworth RA. The assessment of defects in structures of strain
hardening materials. Engineering Fracture Mechanics 1984;19:
63342.
[12] Anderson T. Fracture mechanics fundamentals and applications.
Boca Raton: CRC Press; 1995.
[13] Kanninen MF, Popelar CH. Advanced fracture mechanics. Oxford:
Oxford University Press; 1985.
[14] Betegon C, Hancock JW. Two-parameter characterization of
elasticplastic crack-tip elds. Journal of Applied Mechanics 1991;58:
10410.
[15] Du ZZ, Hancock JW. The effect of non-singular stresses on crack-tip
constraint. Journal of the Mechanics and Physics of Solids
1991;39:55567.
[16] ODowd NP, Shih CF. Family of crack-tip elds characterised by a
triaxiality parameter. Journal of the Mechanics and Physics of Solids
1991;39:8981015.
[17] Yang S, Chao YJ, Sutton MA. Higher-order asymptotic crack tip
elds in a power law hardening material. Engineering Fracture
Mechanics 1993;45:120.
[18] Gurson AL. Continuum theory of ductile rupture by void nucleation
and growth. Part 1yield criteria and ow rules for porous ductile
media. Journal of Engineering Material and Technology 1977;99:
215.
[19] Thomason PF. Ductile fracture of metals. Oxford, UK: Pergamon
Press; 1990.
[20] McClintock FA. A criterion of ductile fracture by the growth of
holes. Journal of Applied Mechanics 1968;35:36371.
[21] Rice JR, Tracey DM. On the ductile enlargement of voids in triaxial
stress elds. Journal of the Physics and Mechanics of Solids 1969;
17:20117.
[22] Beremin FM. Cavity formation from inclusions in ductile fracture of
A 508 steel. Metallurgical Transactions 1981;12A:72331.
[23] Rousselier G. Ductile fracture models and their potential in local
approach of fracture. Nuclear Engineering and Design 1987;105:
97111.
[24] Tvergaard V. Inuence of voids on shear band instabilities under
plane strain conditions. International Journal of Fracture 1981;17:
389407.
[25] Tvergaard V. On Localization in Ductile Materials Containing
Spherical Voids. International Journal of Fracture 1982;18:23752.
[26] Koplik J, Needleman A. Void growth and coalescence in porous
plastic solids. International Journal of Solids and Structures
1988;24:83553.

ARTICLE IN PRESS
1412

C.-K. Oh et al. / International Journal of Mechanical Sciences 49 (2007) 13991412

[27] Chu C, Needleman A. Void nucleation effects in biaxially stretched sheets.


Journal of Engineering Materials and Technology 1980;102:24956.
[28] Tvergaard V, Needleman A. Analysis of the cup-cone fracture in a
round tensile bar. Acta Metallurgica 1982;32:15769.
[29] Bernauer G, Brocks W. Phenomenological modeling of ductile damage
and tearing-results of a European numerical round robin. Fatigue and
Fracture of Engineering Materials and Structures 2002;25:36384.
[30] Dotta F, Ruggieri C. Structural integrity assessments of high pressure
pipelines with axial aws using a micromechanics model. International Journal of Pressure Vessels and Piping 2004;81:76170.
[31] Rivalin F, Pineau A, Di Fant M, Besson J. Ductile tearing of
pipeline-steel wide plates Idynamic and quasi-static experiments.
Engineering Fracture Mechanics 2001;68:32945.
[32] Rivalin F, Besson J, Pineau A, Di Fant M. Ductile tearing of
pipeline-steel wide plates IImodeling of in-plane crack propagation. Engineering Fracture Mechanics 2001;68:34764.
[33] Chen Y, Lambert S. Numerical modeling of ductile tearing for semielliptical surface cracks in wide plates. International Journal of
Pressure Vessels and Piping 2005;82:41726.
[34] Burstow MC, Howard IC. Predicting the effects of crack tip
constraint on material resistance curves using ductile damage theory.
Fatigue and Fracture of Engineering Materials and Structures 1996;
19:46174.
[35] Roy Chowdhury S, Narasimhan R. A nite element analysis of
quasistatic crack growth in a pressure sensitive constrained ductile
layer. Engineering Fracture Mechanics 2000;66:55171.
[36] Skallerud B, Zhang ZL. A 3D numerical study of ductile tearing and
fatigue crack growth under nominal cyclic plasticity. International
Journal of Solids and Structures 1997;24:314161.
[37] Imad A, Wilsius J, Nait Abdelaziz M, Mesmacque G. Experiments
and numerical approaches to ductile tearing in an 2024-T351
aluminium alloy. International Journal of Mechanical Sciences
2003;45:184961.
[38] Gao X, Faleskog J, Shih CF. Ductile tearing in part-through cracks:
experiments and cell-model predictions. Engineering Fracture Mechanics 1998;59:76177.

[39] Needleman A, Tvergaard V. An analysis of ductile rupture modes at


a crack tip. Journal of the Mechanics and Physics of Solids
1987;35:15183.
[40] American Petroleum Institute. Specication for Line Pie. API 5L;
2000.
[41] ABAQUS Version 6.4 Users manual. Hibbitt, Karlson & Sorensen
Inc.; 2005.
[42] Siegmund T, Brocks W. A user-material subroutine incorporating the
GTN model of porous metal plasticity into the ABAQUS FE
program. GKSS Internal Report GKSS/WMG/97/12, GKSS Research Center, Geesthacht, Germany.
[43] Franklin AG. Comparison between a quantitative microscope and
chemical methods for assessment of non-metallic inclusions. Journal
of the Iron and Steel Institute 1969;207:1816.
[44] Ainsworth RA. An assessment of the effects of prestrain on
upper shelf fracture toughness. Journal of Strain Analysis 1986;21:
21923.
[45] Cosham A. A model of pre-strain effects on fracture toughness.
Journal of Offshore Mechanics and Arctic Engineering 2001;123:
18290.
[46] Hagiwara N, Masuda T, Oguchi N. Effects of prestrain on fracture
toughness and fatigue-crack growth of line pipe steels. Journal of
Pressure Vessel Technology 2001;123:35561.
[47] Minami F, Arimochi K. Evaluation of prestraining and dynamic
loading effects on the fracture toughness of structural steels by the
local approach. Journal of Pressure Vessel Technology 2001;123:
36272.
[48] Fukuda N, Hagiwara N, Masuda T. Effect of prestrain on tensile and
fracture toughness properties of linepipes. In: Proceedings of the
fourth international pipeline conference, IPC2002-27127, Calgary,
Alberta, Canada, 2002.
[49] Cosham A, Hopkins P. An experimental study of the effect of prestrain on the fracture toughness of line pipe steel. In: Proceedings
ICPVT-10, 2003, Vienna, Austria. p. 35765.
[50] BS 5762: Methods for crack opening displacement testing. London:
BS Institution: 1979.