ARTICLE IN PRESS

International Journal of Pressure Vessels and Piping 84 (2007) 512–525

www.elsevier.com/locate/ijpvp

Ductile failure analysis of API X65 pipes with notch-type defects

using a local fracture criterion
Chang-Kyun Oha, Yun-Jae Kima,, Jong-Hyun Baekb, Young-Pyo Kimb, Woo-Sik Kimb
a
Department of Mechanical Engineering, Korea University, 5 Ka, Anam-dong, Sungbuk-Ku, Seoul 136 701, Republic of Korea
b
Korea Gas Corporation Research and Development Center, 638-1, Il-dong Ansan Kyonggi-do 425 150, Republic of Korea
Received 3 January 2006; received in revised form 5 March 2007; accepted 5 March 2007

Abstract

A local failure criterion for API X65 steel is applied to predict ductile failure of full-scale API X65 pipes with simulated corrosion and
gouge defects under internal pressure. The local failure criterion is the stress-modified fracture strain as a function of the stress triaxiality
(defined by the ratio of the hydrostatic stress to the effective stress). Based on detailed finite element (FE) analyses with the proposed
local failure criterion, burst pressures of defective pipes are estimated and compared with experimental data. For pipes with simulated
corrosion defects, FE analysis with the proposed local fracture criterion indicates that predicted failure takes place after the defective
pipes attain maximum loads for all cases, possibly due to the fact that the material has sufficient ductility. For pipes with simulated gouge
defects, on the other hand, it is found that predicted failure takes place before global instability, and the predicted burst pressures are in
good agreement with experimental data, providing confidence in the present approach.
r 2007 Elsevier Ltd. All rights reserved.

Keywords: API X65 steel; Ductile fracture; Finite element (FE) analysis; Gouge; Local fracture strain criterion

1. Introduction models incorporating the stress state effect [6–18]. Recently

the authors have proposed a stress-modified fracture strain
For fitness-for-service analyses of underground gas for API X65 steel as a function of stress triaxiality (defined
pipelines, engineering assessment methods against possible by the ratio of the hydrostatic stress to the effective stress)
defects need to be developed, such as corrosion and gouge [19]. To determine the stress-modified fracture strain,
(see for example Refs. [1–3]). Noting that typical gas smooth and notched tensile bars with five different notch
pipelines are made of sufficiently ductile materials, the net- radii were tested, from which true fracture strains were
section limit load approach can be used, where a damaged determined as a function of the notch radius. Experimental
pipe is assumed to fail at the load when the net section is in results were compared with the results from detailed
the fully plastic state [4,5]. Although the net-section limit elastic-plastic, finite element (FE) analyses based on the
load approach is simple for practical application, valida- large geometry change option, which led to the local failure
tion is rather expensive, requiring a large number of full- criterion in terms of the fracture strain as a function of the
scale pipe test data. A more fundamental approach would stress triaxiality.
be based on local failure criteria for ductile fracture. It has In this paper, the local failure criterion for the API X65
been shown that ductile fracture strongly depends on the steel, proposed by the authors in Ref. [19], is applied to
stress state and can be effectively predicted using strain predict ductile failure of full-scale pipes with simulated
corrosion and gouge defects under internal pressure. Based
on detailed FE analyses with the proposed local failure
Corresponding author. Tel.: +82 2 3290 3372; fax: +82 31 290 5276. criterion, burst pressures of defective pipes are estimated
E-mail address: [email protected] (Y.-J. Kim). and compared with experimental data. Noting that

0308-0161/$ - see front matter r 2007 Elsevier Ltd. All rights reserved.
doi:10.1016/j.ijpvp.2007.03.002
 ARTICLE IN PRESS
C.-K. Oh et al. / International Journal of Pressure Vessels and Piping 84 (2007) 512–525 513

assessment equations for gouge defects are not yet X65 steel are summarized in Tables 1 and 2, respectively.
available, a parametric study is also performed. True stress–strain data at room temperature and values of
the Charpy energy as a function of temperature are also
shown in Fig. 1. Even at 60 1C, the Charpy impact energy
2. Stress-modified fracture strain criteria was more than 250 J and all specimens failed by 100%
dimple fracture, indicating that the present material is
The authors have recently proposed a local failure sufficiently tough. One interesting point is that, as most of
criterion to predict ductile fracture of API X65 steel. The the API X65 gas pipelines in Korea have been built within
criterion is in terms of true strain as a function of the stress the last 10 years, mechanical properties of API X65 gas
triaxiality, and has been determined from results of smooth pipelines in Korea tend to have quite uniform properties.
and notched round bar tests, combined with those from To investigate the effect of triaxial stress states on ductility,
detailed FE analyses. The procedures and resulting tensile tests were also performed using notched bars with
criterion are briefly summarized below. Detailed informa- four different notch radii (see Fig. 1), extracted from the
tion can be found in Ref. [19]. same pipe (outer diameter Do ¼ 762 mm and wall thickness
To determine mechanical properties, tensile and Charpy t ¼ 17.5 mm). The test results showed that, as the notch
tests were performed using specimens extracted from a pipe radius decreases, yield and tensile strengths increase, but
of outer diameter Do ¼ 762 mm and wall thickness engineering strain to fracture decreases (see Fig. 2 for
t ¼ 17.5 mm (in the longitudinal direction), made of the selected results). Such a trend is due to the fact that the
American Petroleum Institute (API) 5L X65 steel [20], triaxial stress increases with decreasing notch radius.
popularly used for gas pipelines in Korea. Chemical To simulate tensile tests of smooth and notched speci-
compositions and tensile properties of the present API mens, detailed elastic-plastic, axi-symmetric FE analyses
were performed using ABAQUS [21] with the large
Table 1 geometry change option, and variations of the triaxial
Chemical composition of the API X65 steel, used in the present work stress and strain within the specimen were determined. To
Element (wt %) avoid problems associated with incompressibility, the
reduced integration element within ABAQUS (2002)
C P Mn S Si Fe Ceq (element type CAX8R) was used. For all cases except for
0.08 0.019 1.45 0.03 0.31 Balance 0.32 the case of the 0.2 mm notch radius, elements with a length
of 0.15 mm were uniformly spaced in the minimum section,
which was shown to be sufficiently fine for the present
purpose. As the fully plastic state is of main interest, stress
Table 2 and strain gradients ahead of the notch tip were not so
Mechanical tensile properties at room temperature of the API X65 steel, steep. For the case of the 0.2 mm notch radius, finer meshes
used in the present work
with the smallest element size of 0.05 mm were used in the
Young’s modulus Poisson’s ratio Yield strength Tensile strength notch tip region. Materials were modeled as isotropic
E (GPa) n sy (MPa) su (MPa) elastic-plastic materials that obey the incremental plasticity
theory. True stress–strain experimental data from smooth
210.7 0.3 464.5 563.8
tensile bar tests were directly given in the FE analysis.

Fig. 1. (a) True stress–strain data at room temperature and (b) Charpy energy as a function of temperature for API X65 steel.
 ARTICLE IN PRESS
514 C.-K. Oh et al. / International Journal of Pressure Vessels and Piping 84 (2007) 512–525

Fig. 3 compares selected experimental engineering stress– principal stresses as

strain data from notched tensile bars with the FE results,
1  0:5
suggesting that the FE analysis can simulate tensile se ¼ pffiffiffi ðs1  s2 Þ2 þ ðs1  s3 Þ2 þ ðs3  s2 Þ2 . (2)
deformation behavior even after necking but cannot 2
simulate failure of tensile test specimens. Fig. 4 shows the The equivalent strain, ee, on the other hand, is defined by
radial variation of the stress triaxiality and equivalent pffiffiffi
strain in the minimum (necked) section of smooth and 2 0:5
e ¼ ð1  2 Þ2 þ ð1  3 Þ2 þ ð3  2 Þ2 . (3)
notched bars at the point of failure initiation (see Fig. 3 to 3
identify failure initiation points). The stress triaxiality is The distance (r) is normalized with respect to the radius of
defined by the ratio of the mean normal (hydrostatic) the minimum section (a), and the values of r/a ¼ 0 and
stress, sm, to the equivalent stress, se r/a ¼ 1 mean the center and the free surface of the
specimen, respectively. Application of the damage indica-
sm s1 þ s2 þ s3 tor based on the Rice and Tracey formula [7] showed that,
¼ , (1)
se 3se for all cases except for the case of the 0.2 mm notch radius,
failure is expected to occur from the center of the specimen.
where si (i ¼ 1–3) denote the principal stresses.
For the case of the 0.2 mm notch radius, the notch tip was
The equivalent stress, se, is expressed in terms of three
found to be the critical site for failure initiation. By
combining FE results with notched bar tensile test ones,
ductile failure criteria in terms of the equivalent strain to
failure as a function of the stress triaxiality were
established. Noting that point-wise information on stress
and strain is obtained from FE analyses, two different
approaches were taken in the previous work. One approach
was to develop a failure criterion based on stress and strain
at the location where failure is most likely to initiate. Such
a point should correspond to the location with the highest
stress triaxiality and strain, for instance. This approach was
termed the ‘‘critical location criterion.’’ The other ap-
proach was based on average stress and strain over the
minimum section, and was termed the ‘‘section average
criterion.’’ Fig. 5 shows the evolution of the stress
triaxiality in terms of the equivalent strain, measured in
the center of the minimum section of smooth and notched
tensile bars. For the case of the 0.2 mm notch radius,
failure initiation was estimated to occur at the notch tip,
and thus the data are excluded for the critical location
Fig. 2. Schematic illustrations: (a) smooth tensile bars; (b) and (c) notched criterion. The results in Fig. 5 show that the stress
tensile bars (dimensions in mm). triaxiality in the center of the specimen depends strongly

Fig. 3. Comparison of experimental engineering stress–strain data for notched tensile bars with FE results: (a) notched bar with the 3 mm notch radius
and (b) notched bar with the 6 mm notch radius.
 ARTICLE IN PRESS
C.-K. Oh et al. / International Journal of Pressure Vessels and Piping 84 (2007) 512–525 515

Fig. 4. (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. 5. Variations of the stress triaxiality with the equivalent strain for smooth and notched bar specimens: (a) based on the critical location criterion and
(b) based on the section average criterion.

on the equivalent strain. As a ductile failure criterion whereas for the section average criterion
should include the history of stress and strain, average  
sm
stress triaxiality is introduced, defined by ef ¼ 3:41 exp 2:23 þ 0:22. (6)
  Z se
sm 1 ef sm
¼ de , (4) Application of Eq. (5) or (6) to tensile test results of
se ave ef 0 se notched bars led to correct predictions of failure loads and
where eef denotes the equivalent strain to failure initiation. fracture strain for all cases considered, including the case of
Such a definition incorporates the history effect of stresses the 0.2 mm notch radius.
and strains on ductile fracture. For a given notch radius, To predict ductile failure of defective pipes made of API
the calculated average stress triaxiality is constant, and is X65 steels using the present approaches, the proposed
shown in Fig. 5. Fig. 6 summarizes the resulting equivalent equation, Eq. (5) or (6), should be combined with detailed
strain to failure initiation (true fracture strain, eef), as a elastic-plastic FE analyses from which local stresses and
function of the stress triaxiality for the API X65 steel, strains are determined. For instance, from the FE analysis,
considered in the present work. For the critical location stress and strain information can be monitored as a
criterion, fracture strain was expressed as function of load. Over the loading history, the stress
  triaxiality should be averaged according to Eq. (4), which
sm in turn gives fracture strain from Eq. (5) or (6). When the
ef ¼ 3:29 exp 1:54 þ 0:10, (5)
se (equivalent) strain from the FE analysis equals the fracture
 ARTICLE IN PRESS
516 C.-K. Oh et al. / International Journal of Pressure Vessels and Piping 84 (2007) 512–525

Fig. 7. Pipe with a simulated corrosion defect.

3.1. Full-scale pipe tests

Test pipes were prepared out of API 5L X65 pipelines

from Korea Gas Corporation (KOGAS) (same as those
Fig. 6. True fracture strains (equivalent strain to fracture) as a function of from which smooth and notched tensile bars were extracted
stress triaxiality [19]. in Section 2). A pipeline of total length 12 m was cut into
pieces with 2.3 m length, and both ends were capped by
strain estimated from Eq. (5) or (6), then ductile failure circumferential welding. The geometrical configuration of
is assumed to initiate. One notable point is that the use of the pipe tested is shown in Fig. 7, and dimensions of test
Eq. (5) will in general give a different failure point from specimens are summarized in Table 3. The mechanical
that of Eq. (6). It will be shown, however, in this paper, defect to simulate corrosion was machined in a rectangular
that both equations provide practically the same failure shape as shown in Fig. 7. The defect was machined to keep
loads. the same thickness at the bottom, and corner edges were
The proposed ductile failure criteria are physically sound rounded to avoid excessive stress concentration. For
and quite simple to use. To validate applicability of the characterizing a corrosion defect, the depth d, the width c
proposed ductile failure criteria, they are applied to predict and the length l are used, as depicted in Fig. 7.
ductile failure of defective pipes made of the API X65 steel The pipes were pressurized by water and burst pressures
and compared with experimental data from full-scale pipe were experimentally determined at the point when the
tests in the next section. Two kinds of defects are ligament failed. The burst testing equipment is shown in
considered; metal loss due to corrosion and gouge due to Fig. 8a. All specimens showed bulging deformation around
third party damage, which are typical degradation mecha- the defect area, and final failure occurred at the bottom of
nisms for underground gas pipes. the defect area with crack-like penetration in the long-
itudinal direction as shown in Fig. 8b. The defect area
shows a significant amount of thickness reduction along
3. Application to failure predictions of corroded API X65 the penetration line, probably caused by local necking
pipes prior to final failure. The resulting burst pressures are
summarized in Table 3. This shows that burst pressures are
Development of assessment methods against corrosion affected by the defect depth and length, but the effect of the
defects has been active in the last two decades. A number of defect width is much less significant.
methods have been proposed, such as ASME B31G [22],
RSTRENG [23] and PCORRC [24]. Note that these 3.2. FE analysis
methods are in principle based on limit load analysis of
pipes with corrosion defects. When the pipe materials have Three-dimensional elastic-plastic FE analyses were
medium- to high-toughness, application of the limit load performed to simulate the pipe burst tests using ABAQUS
approach is plausible and thus it has been popularly used in [21]. Fig. 9 depicts a typical FE mesh. A quarter of a full
practice due to its simplicity. In the present work, however, pipe was modeled due to symmetry conditions. The
pipes with simulated corrosion defects are assessed using machined pit was modeled as a rectangular shape with
the proposed stress-modified strain criteria and the results rounded corners in accordance with the test specimen, and
are compared with full-scale pipe test data. the model is designed with 20-node iso-parametric brick
 ARTICLE IN PRESS
C.-K. Oh et al. / International Journal of Pressure Vessels and Piping 84 (2007) 512–525 517

Table 3
Summary of full-scale tests of pipes with corrosion defect

Pipe no. Do (mm) L (mm) t (mm) l (mm) c (mm) d/t Pexp (MPa)a Ratio Pmax/Pexpb Ratio PPCORRC/Pexpc

DA 200 50 0.25 24.11 1.025 0.978

DB 200 50 0.5 21.76 1.044 0.940
DC 200 50 0.75 17.15 1.035 0.902
LA 762 2300 17.5 100 50 0.5 24.30 1.025 0.939
LC 300 50 0.5 19.80 1.042 0.943
CB 200 100 0.5 23.42 0.967 0.874
CC 200 200 0.5 22.64 0.989 0.904
a
Pexp, experimentally-measured burst pressure.
b
Pmax, predicted burst pressure from FE.
c
PPCORRC, burst pressure predicted from Eq. (7).

Fig. 8. (a) Test set-up of full-scale burst test for pipes, (b) defect, and (c) failure.

elements with the reduced integration option. The numbers A monotonically increasing hydrostatic pressure was
of elements and nodes are from 2376 elements/11,673 applied to the inner surface of the model, together with
nodes to 6150 elements/29,688 nodes and through the axial force to simulate closed cap conditions. True
thickness, three elements were used in the corroded section. stress–strain data, determined from uni-axial tension tests,
 ARTICLE IN PRESS
518 C.-K. Oh et al. / International Journal of Pressure Vessels and Piping 84 (2007) 512–525

the present approach, occur after the maximum pressure

takes place. The pipe burst test, however, is load
controlled, and thus when the pressure takes its maximum
value, unstable burst takes place. Accordingly the max-
imum pressure is independent of the use of Eq. (5) or (6). In
fact, such behavior occurs for all cases considered in the
present work. Load–displacement records for all cases are
summarized in Fig. 12. This shows that failure occurs after
maximum load, except for two cases (d/t ¼ 0.25,
l ¼ 200 mm and c ¼ 50 mm; and d/t ¼ 0.5, l ¼ 100 mm
and c ¼ 50 mm) where calculations stopped at the max-
imum load point due to numerical difficulties. Even when
Fig. 9. A FE mesh for the pipe with the simulated corrosion defect. calculations were ceased, strains did not reach critical
values. Thus the present approach suggests that failure of
the present API X65 pipes with simulated corrosion is
governed by global plastic instability (zero slope in
load–displacement curves), possibly due to sufficiently high
ductility of the material considered in the present work (see
Fig. 2b).
Although the present approach suggests that failure is
governed by global plastic instability, and thus these
experimental data are not appropriate to validate the
present approach, at least the maximum loads predicted by
the FE analysis could be compared with experimental data.
Table 3 lists ratios of the maximum pressure (Pmax),
predicted from the FE analysis, to experimental ones
(Pexp). Overall the predicted values are in good agreement
with experimental ones, and differences are no more than
5% for all cases considered. For corroded pipes with
sufficient ductility, one solution to estimate the burst
pressure of pipes with local wall thinning is the so-called
PCORRC equation [24]
" !!#
2t d l
Pf ¼ su 1  1  exp 0:157 pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi ,
Do t Ro ðt  dÞ
(7)
Fig. 10. A typical FE load–radial displacement record for pipes with
simulated corrosion defects with failure points predicted from the present where Do and Ro denote the outer diameter and radius,
approach. respectively; and su is the ultimate tensile strength. It is
known that Eq. (7) is based on limit load analysis for pipes
with constant depth thinning under internal pressure.
as shown in Fig. 1a, were directly given in the FE analysis. Table 3 lists ratios of the maximum pressure (PPCORRC),
Incremental plasticity with the large geometry change predicted from Eq. (7), to experimental ones (Pexp). Overall
option was invoked. the predicted values are less than the experimental ones,
but generally by no more than 13%.
3.3. Comparisons

Fig. 10 shows a typical load vs. radial displacement 4. Application to failure predictions of API X65 pipes with
record, determined from the FE analysis. It also includes gouges
failure points, predicted by the present approach, namely,
when the equivalent strain from the FE analysis equals the The second type of defect considered in the present work
fracture strain according to Eq. (5) or (6). This shows that is a gouge due to third party damage. The proposed
radial displacement increases initially with internal pres- method is applied to predict the maximum pressure of
sure, and then the pressure takes its maximum value and pipes with gouges, and the results are compared with full-
decreases with increasing radial displacement. A typical scale pipe test data. Having sufficient confidence, further
deformed shape in the defective region is shown in Fig. 11, parametric FE analyses are performed to provide an
clearly showing local necking. Failure points, predicted by engineering assessment equation for gouge defects.
 ARTICLE IN PRESS
C.-K. Oh et al. / International Journal of Pressure Vessels and Piping 84 (2007) 512–525 519

Fig. 11. Deformed shape in the defective region for pipes with simulated corrosion defects.

Fig. 12. Effects of defect geometries on FE load–radial displacement records for pipes with simulated corrosion defects.
 ARTICLE IN PRESS
520 C.-K. Oh et al. / International Journal of Pressure Vessels and Piping 84 (2007) 512–525

4.1. Experiments pipe thickness (d/t ¼ 0.5). The axial length of the gouge, on
the other hand, was systematically varied from ‘ ¼ 50 to
Fig. 13 schematically depicts a pipe with a simulated 600 mm. The pipes were pressurized by water and burst
gouge, and some important dimensions are summarized in pressures were experimentally determined at the point when
Table 4. The geometry of the pipe is the same as that in the ligament failed. Fig. 14 depicts photos of the gouge
Section 3, having the diameter of Do ¼ 762 mm, the before and after the test. During the test, strain values were
thickness of t ¼ 17.5 mm, and the total length of monitored using strain gauges, as shown in Fig. 14a. Note
L ¼ 2300 mm. The gouge is characterized by a 451 V-notch that two gauges, S1 and S2, were circumferentially located
with a circular notch radius of 2 mm. In all tests, the depth 24.9 and 54.9 mm, respectively, away from the center of the
of the gouge is fixed as d ¼ 8.75 mm which is 50% of the gouge. The measured maximum pressures from full-scale

Fig. 13. Schematic illustrations for pipes with gouge. Fig. 15. A typical finite element mesh for pipes with gouge.

Table 4
Summary of full-scale tests of pipes with gouge defect

Pipe no. Do (mm) L (mm) t (mm) d/t l (mm) Pexp (MPa)a Ppred/Pexp Eq. (5)b Ppred/Pexp Eq. (6)b

MNA 0.5 100 24.68 0.979 0.983

MNB 0.5 200 22.48 0.960 0.973
MNC 762 2300 17.5 0.5 300 17.70 1.103 1.117
MND 0.5 400 18.14 1.003 1.011
MNE 0.5 600 16.57 1.020 1.021
a
Pexp, experimentally-measured burst pressure.
b
Ppred, predicted burst pressure.

Fig. 14. (a) Initial gouge defect and strain measurement locations, and (b) gouge at failure.
 ARTICLE IN PRESS
C.-K. Oh et al. / International Journal of Pressure Vessels and Piping 84 (2007) 512–525 521

pipe tests are summarized in Table 4, which shows that the that the results for the MNC test are the worst, and for all
burst pressure decreases with increasing gouge length. other cases the FE results agree very well with experimental
values, as in the MNA case. From FE analyses, the
4.2. FE analysis equivalent strain and stress triaxiality were monitored as a
function of internal pressure. Fig. 17 shows typical radial
The pipes with gouges were modeled using FEs. A typical variations of stress triaxiality and equivalent strain. In
FE mesh is shown in Fig. 15, with a detailed view of the these figures, the radial distance is normalized with respect
defective region. The number of elements and nodes in FE to the minimum ligament size, and the values of 0 and 1
meshes were from 7290 elements/33,838 nodes to 11,345 denote the inner surface and the notch tip of the pipe,
elements/52,145 nodes, depending on the gouge length. For respectively. These show that the equivalent strain takes
all cases, however, the smallest element size in the defective the maximum value at the notch tip, but the maximum
area was fixed at 0.32 mm  0.67 mm  1.19 mm. It is value of stress triaxiality occurs somewhere in between the
believed that the present mesh size is appropriate for notch tip and the center. To apply the critical location
engineering purposes, as meshing using smaller mesh sizes criterion, stress and strain information at all points in the
with appropriate aspect ratios was quite difficult due to minimum ligament were monitored to find out the failure
increasing number of elements. Internal pressure was initiation location. For the present cases, failure initiation
applied to the inner surface of the pipe, together with end is found to occur always at the tip of the notch, and thus
forces to simulate the closed end condition. the failure criterion is checked using stress and strain
information extracted from the first element at the notch
4.3. Comparisons tip. For the criterion based on averaged information, stress
and strain were averaged over the minimum ligament from
Fig. 16 compares axial and hoop strains measured from the notch tip to the inner surface. Fig. 18 shows typical
tests (S1 and S2) with FE results for selected cases. Note variations of the stress triaxiality with equivalent strain for

Fig. 16. Comparison of strain variations measured from pipe tests with FE results: (a) MNA pipe test and (b) MNC pipe test.
 ARTICLE IN PRESS
522 C.-K. Oh et al. / International Journal of Pressure Vessels and Piping 84 (2007) 512–525

Fig. 17. Distributions of stress triaxiality and equivalent strain for pipes with gouge: (a) MNB pipe test and (b) MNE pipe test

Fig. 18. Variations of the stress triaxiality with the equivalent strain for pipes with gouge for the critical location criterion and the section average
criterion: (a) MNA pipe test and (b) MNE pipe test.

selected cases. This also includes constant stress triaxiality section average, although they are practically the same.
values averaged over the loading history according to The results also show that failure predicted by the present
Eq. (4). Note that for other cases, trends are quite similar approach occurs before global plastic instability points
to those in Fig. 18. It can be seen that the magnitude of the (characterized by zero slope in load–displacement curves).
stress triaxiality does not depend on the equivalent strain This is in contrast to the cases of pipes with simulated
and is almost constant, and accordingly is also close to the corrosion defects in Section 3, where failure is indeed
average stress triaxiality, defined by Eq. (4). Another governed by global plastic instability. Estimated burst
notable point is that the values of the stress triaxiality are pressures are compared with experimentally measured data
quite low: for the critical location criterion, it is about 0.55, in Table 4 and in Fig. 20. Note that the pressure is
and for the section average criterion, is about 0.85. Such normalized with respect to the theoretical limit pressure of
low values are typical for pipes under internal pressure. a plain pipe
Fig. 19 shows radial displacements, determined from FE t t sy þ su 
analysis, as a function of pressure. Note that the average Po ¼ sf ¼ , (8)
Rm Rm 2
stress triaxiality for the critical location criterion is lower
than that for the section average criterion. However, the where sf, sy and su denote the flow strength, yield strength
magnitude of strain for the critical location criterion is and ultimate tensile strength, respectively. As shown in
much higher than that for the section average criterion. As Table 4, the estimated burst pressures are in good
a result, the criterion based on the critical location provides agreement with experimentally measured ones with differ-
a slightly lower burst pressure than that based on the ences less than 3%, except the MNC case where the
 ARTICLE IN PRESS
C.-K. Oh et al. / International Journal of Pressure Vessels and Piping 84 (2007) 512–525 523

of the proposed ductile fracture criteria to defect assess-

ment of gas pipelines.
Noting that an engineering assessment equation for
gouge defects is not currently available, gouge defects
could be assessed using an expression of failure pressure of
axial surface cracked pipes under internal pressure,
assuming that the gouge is idealized as an axial crack.
One popular expression in [5] is given by
sffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
 
1  a=t ‘2
Pf ¼ Po ; M ¼ 1 þ 1:61 , (9)
1  ða=tÞð1=MÞ 4Rm t
where a and ‘ denote the crack depth and length,
respectively. As the gouge is idealized as an axial crack,
Eq. (9) can be used to estimate failure pressure, simply by
replacing the crack depth a with the gouge depth d.
Resulting predictions are compared in Fig. 20, showing
that Eq. (9) gives conservative (lower) failure loads, as
expected.

4.4. Assessment equations against gouge defects

Fig. 19. Variations of pressure with radial displacement, determined from A large number of full-scale tests might be needed to
the FE analyses, indicating predicted failure points. develop an engineering assessment equation against
gouges, which is not only expensive but also time
consuming. Having sufficient confidence in the use of the
proposed approach to assess gouge defects, systematic
parametric FE analyses are performed to provide an
engineering assessment equation for gouge defects. For
the parametric study, the dimensions of the pipe are fixed
to those in the previous cases; the diameter of
Do ¼ 762 mm, the thickness of t ¼ 17.5 mm, and the total

Fig. 20. Comparison of experimental burst pressure for pipes with gouge
with the FE results and Eq. (9).

difference is as much as 12%. Although such a difference

is still not significant, it can be seen from Fig. 16 that the
experimental data for the MNC case are somewhat lower
than expected from the trend curve. Good agreement Fig. 21. Comparison of burst pressure solutions with FE results for pipes
shown in this example provides confidence in application with gouge.
 ARTICLE IN PRESS
524 C.-K. Oh et al. / International Journal of Pressure Vessels and Piping 84 (2007) 512–525

length of L ¼ 2300 mm, and the axial length of the gouge gouge defects, on the other hand, it is found that predicted
(‘) is systematically varied from 100 to 600 mm. The gouge failure takes place before global instability, and the
depth (d), on the other hand, is varied from 4.375 to predicted burst pressures are in good agreement with
13.125 mm, leading to the relative gouge depth (d/t) from experimental data, providing confidence in the present
0.25 to 0.75. A total of 25 FE calculations are performed. approach. A parametric study is performed, from which a
The burst pressures, resulting from the present FE simple equation is proposed to predict burst pressure for
analyses, are shown in Fig. 21. In Fig. 21, failure loads, gouge defects in gas pipes made of the particular API X65
predicted using Eq. (9) assuming that the gouge is idealized steel considered in the present work.
by the crack, are shown. Predictions using Eq. (9) are quite The methodology of predicting ductile failure of
conservative, and the trend of the gouge length effect is defective structures using a local fracture strain criteria is
somewhat different. Inspired by Eq. (9), the following not a new but an old concept. The significance of the
equation is proposed to estimate the burst pressure of API present work, however, is validation of the methodology by
X65 pipes with gouge defects based on the FE results systematic comparison with full-scale pipe test data. The
sffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi results in this paper suggest that the local fracture strain
 
1  d=t d ‘2 criteria could be successfully applied to predict ductile
Pf ¼ Po ; M g ¼ 1 þ 0:64 .
1  ðd=tÞð1=M g Þ t 4Rm t failure of pipes with notch-type defects and could reduce
(10) expensive and time-consuming full-scale tests. It should be
also noted that the developed fracture strain criteria should
Resulting predictions are compared with the FE results in be applied to the specific material (API X65 steel)
Fig. 21. Eq. (10) agrees well with the FE results for larger d/ considered in the present work. As API X65 gas pipelines
t, but tends to be slightly conservative for smaller d/t and in Korea tend to have quite uniform mechanical properties,
larger ‘. It should be noted that the FE results are for the developed fracture strain criterion is at least useful to
specific values of pipe geometry (Do ¼ 762 mm and assess notch-type defects in gas pipelines within Korea. For
t ¼ 17.5 mm, giving Rm/t ¼ 21.3) and for the specific API instance, one interesting application area is to assess
X65 steel considered in the present work, and thus the use combined gouge and dent. Systematic full-scale experi-
of Eq. (10) should be limited. In particular, Eq. (10) does ments are quite difficult due to the many variables
not include the term related to the material’s toughness and involved, and the present approach could be used to
thus should only be used to assess gouges in gas pipes made develop assessment equations with limited experimental
of the particular API X65 steel considered in the present data. For another material, the same procedure can be
work. On the other hand, as noted, most of the API X65 taken to develop the local fracture strain criteria specific to
gas pipelines in Korea have been built within the last 10 that material.
years, and mechanical properties of API X65 gas pipelines In the present work, only notch-type defects, not crack-
in Korea tend to have quite uniform properties. Thus, Eq. like defects, are considered. Although the present approach
(10) would be useful to assess gouges in API X65 gas can be applied to crack-like defect problems in principle,
pipelines in Korea. micro-mechanical based damage models, such as the model
based on Gurson [10], are popularly used these days and
5. Concluding remarks seem to be better suited. The authors have also developed a
micro-mechanical model of ductile fracture for the same
In this paper, a local failure criterion for the API X65 steel using the Gurson–Tvergaard–Needleman (GTN)
steel is applied to predict ductile failure of full-scale pipes model [25].
with simulated corrosion and gouge defects under internal
pressure. The local failure criterion is the stress-modified
Acknowledgments
fracture strain for the API X65 steel as a function of the
stress triaxiality (defined by the ratio of the hydrostatic
This research is performed under the program of Basic
stress to the effective stress). Based on detailed FE analyses
Atomic Energy Research Institute (BAERI), is a part of the
with the proposed local failure criteria, burst pressures of
Nuclear R&D Programs funded by the Ministry of Science
defective pipes are estimated and compared with full-scale
& Technology (MOST) of Korea, and under the program
pipe test data.
of the Brain Korea 21 Project in 2006.
For pipes with simulated corrosion defects, FE analysis
with the proposed local fracture criterion indicates that
predicted failure takes place after the defective pipes attain References
maximum loads for all cases, and thus the present
approach suggests that pipe failure is governed by global [1] Cosham A, Hopkins P. The effect of dents in pipelines-guidance in
instability. This is because the material considered in this the pipeline defect assessment manual. Int J Pressure Vessels Pip
2004;81:127–39.
paper has sufficient ductility. Nevertheless the maximum [2] Wilkowski G, Stephens D, Krishnaswamy P, Leis B, Rudland D.
pressures, predicted from the present FE analysis, agree Progress in development of acceptance criteria for local thinned areas
well with experimental data. For pipes with simulated in pipe and piping components. Nucl Eng Des 2000;195:149–69.
 ARTICLE IN PRESS
C.-K. Oh et al. / International Journal of Pressure Vessels and Piping 84 (2007) 512–525 525

[3] API. API RP579—recommended practice for fitness-for-service. [15] Anderson TL. Fracture mechanics-fundamentals and applications.
American Petroleum Institute; 2000. Boca Raton, FL: CRC Press; 1995.
[4] Kanninnen MF, Zahoor A, Wilkowski G, Abousayed I, Marschall C, [16] Alves M, Jones N. Influence of hydrostatic stress on failure of
Broek D, et al. Instability predictions for circumferentially cracked axisymmetric notched specimens. J Phys Mech Solids 1999;47:
Type-304 stainless pipes under dynamic loading. EPRI report NP- 643–67.
2347. Palo Alto, USA: Electric Power research Institute; 1982. [17] Bao Y, Wierzbicki T. On fracture locus in the equivalent strain and
[5] ASME. ASME boiler and pressure vessel code section XI. 1992. stress triaxiality space. Int J Mech Sci 2004;46:81–98.
[6] McClintock FA. A criterion of ductile fracture by the growth of [18] Bao Y. Dependence of ductile crack formation in tensile test on
holes. J Appl Mech 1968;35:363–71. stress triaxiality, stress and strain ratios. Eng Fract Mech 2005;72:
[7] Rice JR, Tracey DM. On the ductile enlargement of voids in triaxial 505–22.
stress fields. J Phys Mech Solids 1969;17:201–17. [19] Oh CK, Kim YJ, Park JM, Baek JH, Kim WS. Development of
[8] Clausing DP. Effect of plastic strain state on ductility and toughness. stress-modified fracture strain for ductile failure of API X65 steel. Int
Int J Fract Mech 1970;6:71–85. J Fract 2007;143:119–33.
[9] Hancock JW, Mackenzie AC. On the mechanisms of ductile failure in [20] American Petroleum Institute. Specification for line pipe. 2000.
high-strength steels subject to multi-axial stress states. J Phys Mech [21] Hibbitt, Karlson & Sorensen, Inc. ABAQUS version 6.4 user’s
Solids 1976;24:147–69. manual. 2005.
[10] Gurson AL. Continuum theory of ductile rupture by void nucleation [22] American Society of Mechanical Engineer. Manual for remaining
and growth—yield criteria and flow rules for porous ductile media. strength of corroded pipelines. ANSI/ASME B31.G, 1991.
J Eng Mater Technol 1977;99:2–15. [23] Kiefner JF, Vieth PH. A modified criterion for evaluating the
[11] Mackenzie A, Hancock J, Brown D. On the influence of state of stress remaining strength of corroded pipe. American Gas Association,
on ductile failure initiation in high strength steels. Eng Fract Mech Catalog no. L51609, PR3-805. 1989.
1977;9:167–88. [24] Leis BN, Stephens DR. An alternative approach to assess the
[12] Ritchie RO, Thompson AW. On macroscopic and microscopic integrity of corroded line pipe—part I: current status. Part II:
analyses for crack initiation and crack growth toughness in ductile alternative criterion. In: Proceedings of the 7th international offshore
alloys. Metall Trans A 1985;16A:233–48. and polar engineering conference, 1997.
[13] Garrison Jr WR, Moody NR. Ductile fracture. J Phys Chem Solids [25] Oh CK, Kim YJ, Baek JH, Kim YP, Kim WS. A micro-mechanical
1987;48:1035–74. model of ductile fracture for API X65 steels and application to pre-
[14] Thomason PF. Ductile fracture of metals. Oxford, UK: Pergamon strain effects on deformation and fracture. Int J Mech Sci 2007; in
Press; 1990. press.