Вы находитесь на странице: 1из 11

1 Copyright 2012 by ASME

Proceedings of the 2012 9th International Pipeline Conference


IPC2012
September 24-28, 2012, Calgary, Alberta, Canada

IPC2012-90626





STRUCTURAL INTEGRITY ASSESSMENT CRITERIA
AND SERVICE LIFE PREDICTION OF COLD WRINKLEBENDS
Xian-Kui Zhu and Brian N. Leis
Battelle Memorial Institute
505 King Avenue
Columbus, Ohio 43201



ABSTRACT

Three-dimensional elastic-plastic finite element analysis
(FEA) is performed in this paper to simulate the complicated
stresses and deformation of wrinklebends in a pipeline from its
bending formation to operation under cyclic loading. Three
plastic hardening models (isotropic, kinematic and combined
isotropic/kinematic) are discussed and used in FEA of
wrinklebend response that considers strain hardening and
Bauschinger effects. The FEA simulation is carried out first
for an elbow held at constant pressure while subject to cyclic
bending, which serves as a benchmark case. The results show
that the three hardening models lead to very different
outcomes. Comparable FEA simulations are then developed
for wrinklebends under cyclic pressure. Detailed parametric
analysis is considered, including finite-element type, element
sensitivity, computation time, and material input data. Based
on those results viable nonlinear FEA model is developed as
the basis to quantify wrinklebend response under service-like
conditions. Based on the FEA results, fatigue damage is
quantified using the Smith, Watson and Topper (SWT)
parameter, and thereafter a damage criterion is proposed to
predict the fatigue life of a wrinklebend under the pressure
cycles of 72%-10% of SMYS for typical X42 pipeline steel.
The results show that the wrinkle aspect ratio H/L is a key
parameter to control the service life of a wrinklebend.

KEYWORDS: Structural integrity, assessment criterion, finite
element analysis, plastic hardening model, Bauschinger effect,
fatigue damage, cyclic plastic loading, wrinklebend, elbow

INTRODUCTION
Cold pipe bends are frequently required in a pipeline
route in order to accommodate changes in the direction or
slope of the trench, or both the vertical and horizontal planes
of the pipeline [1-7]. While modern cold bends are made
under controlled conditions using a hydraulically driven
machine, in the early days wrinklebends were often made
absent such control. Thus, while the term cold bend reflects a
smooth local change in pipe shape, the term wrinklebend can
refer to a bend that involves a local buckle on the short side of
the bend. The wrinklebend terminology has historically
been used in the pipeline industry to identify bends made by
less controlled bending processes that intentionally create such
local buckles. The shape of a wrinkle is strongly dependent on
the locally unsupported length and pipe thickness, as well as
the presence of imperfection that are variables central to the
buckling theory. Depending on the geographic conditions and
routing, some transmission pipelines contain only a few
wrinklebends, while others through hilly terrain can contain
thousands of such bends. A wrinklebend may fail due to
periodic pressure change, and such incidents interrupt pipeline
operation, and pose risks to the public and environment, so
their integrity and its assessment are important. A detailed
review of historic development and integrity management of
wrinklebends was given by the authors in Reference [8], with
a review of the available acceptance criteria for wrinkles
recently done by de Larminat et al. [9].
Extensive work [1-7] has shown that the cold bending
process produces large plastic deformation that results in
residual stresses and strains at the wrinklebend in the




Copyright 2012 by ASME

2
longitudinal direction of the pipe. The residual stresses and
strains in wrinkles can be detrimental to pipeline integrity
under some operating conditions, particularly for cyclic
loading due to periodic pressure or temperature changes.
Cyclic loading can grow cracking introduced in severely
deformed wrinklebends when the bend was made. It also can
cause cracks to initiate, which likewise can grow to cause
failure. Accordingly, where wrinklebends remain in service
they pose a potential threat to pipeline integrity. It is therefore
important to quantify fatigue damage evolution at wrinkle
bends, and to develop criteria to assess their integrity.
The cold bending process that forms the wrinklebend
involves local buckling, and postbuckling deformation, and
can results in complicated plastic deformation at the wrinkle.
The finite element analysis (FEA) is the most effective tool to
simulate this complexity, and the residual stresses and strains
in the pipe, in order to quantify fatigue damage, with
consideration of the effects of its forming and the operational
conditions. For ductile materials subject to large cyclic
loading, the Bauschinger effect and strain hardening pose
unique effects that must be addressed to correctly quantify
cyclic plastic response. Numerical analyses [3-6] indicated
that the Bauschinger effect and work hardening occurred for
the pipeline steels in the bending process owing to the large
plastic deformation and residual stress at the wrinkle. As it has
long been known that the conventional isotropic hardening
model is inadequate for such applications, other hardening
models have been explored. The present authors reviewed
different kinematic hardening models and numerical methods
for modeling the Bauschinger effect and cyclic plastic
hardening in simulation of metal forming [10]. The combined
isotropic/ kinematic model was found best for such purposes.
This paper develops a reliable nonlinear FEA model to
simulate the complicated stress and plastic deformation of a
wrinklebend from its bending formation through operation
under cyclic pressure loading. Isotropic, kinematic and
combined isotropic/kinematic hardening models are briefly
reviewed, and then evaluated for a pipe bend and for a
wrinklebend in the FEA, to determine local stresses and
strains under cyclic loading. Parametric studies are then done
to identify a viable FEA simulation scheme, which included
element type, mesh sensitivity, computation time, and material
input data. A viable FEA model is thus evolved to simulate
wrinklebend response. Fatigue damage is quantified using the
SWT parameter, after which criteria are proposed for
predicting service life and for assessing pipeline integrity.

BRIEF REVIEW OF PLASTIC HARDENING MODELS
For a metallic material, a general plastic yield function is:
0 ) ( ) (
0
=
pl
f c o o o (1)
where o is the stress tensor, o is the backstress,
0
o is the
current equivalent stress and
pl
c is the accumulative
equivalent plastic strain. From the flow theory of plasticity
and the normalization rule, the associate flow rule with respect
to the yield function f is:

o
c
c
c
=
f
pl

(2)
where
pl
c is the incremental plastic strain tensor, and

is the
plastic flow factor. For the von Mises yield criterion, the yield
function f is the equivalent Mises stress:

( )( ) ' '
2
3
) ( o o o o = S S f
(3)
where S is the deviatoric stress tensor, and o' is the deviatoric
backstress tensor. Substitution of Equation (3) into (2) obtains:

( ) o c ' = - S

pl
(4)
This incremental plastic strain plus the incremental elastic
strain determine the incremental total strain for the material.
Different plastic hardening laws have been developed to
describe the evolution of the backstress tensor o and the
equivalent stress
0
o . Three commonly used plastic hardening
models considered herein are outlined next.

Isotropic Hardening Model
The often used isotropic hardening model assumes that
the yield surface enlarges equally in all directions. As its yield
center has no kinematic shift, the backstress tensor is zero:
0 = o (5)
This model is widely used for elastic-plastic stress analysis of
metallic materials under monotonic loading. However, it is
inappropriate for cyclic loading because it cannot simulate the
Bauschinger effect or anisotropy induced by work hardening.

Linear Kinematic Hardening Model
The linear kinematic hardening model is a simple
kinematic scheme that assumes a constant hardening modulus.
When temperature dependence is omitted, the evolution of the
backstress tensor o is defined by the Ziegler hardening law:

( )
pl
C c
o

o o o =
0
1
(6)
where
o is the backstress change rate tensor,
pl
c

is the
equivalent plastic strain rate, and C is the kinematic hardening
modulus. For the simple tension, C is the tangential modulus
of stress-strain response,
pl
p
d d E c o / =
. In this model, the
equivalent stress defining the yield surface size remains
constant, i.e.
0
0
o o =
, where
0
o
is the initial yield stress at
zero plastic strain. The linear kinematic hardening model can
simulate the Bauschinger effect and anisotropy.





Copyright 2012 by ASME

3
Combined Nonlinear Isotropic/Kinematic Hardening
Model
The combined nonlinear isotropic/kinematic hardening
model was developed based on the work of Lemaitre and
Chaboche [11]. The evolution of yield surface consists of two
components: an isotropic hardening component and a
nonlinear kinematic hardening component. The isotropic
hardening behavior defines the equivalent stress, describing
the size of the yield surface as a function of the equivalent
plastic strain. This evolution can be expressed as the simple
exponential law:

( )
pl
b
e Q
c
o o

+ = 1
0
0
(7)
where Q and b are the material parameters that must be
calibrated from the cyclic test data. Q defines the maximum
change in the size of elastic range and b is the rate at which
the size of the yield surface changes as the plastic strain
develops. When
0
0
o o =
, this model reduces to a nonlinear
kinematic hardening model.
The nonlinear kinematic hardening component of this
combined model considers the Bauschinger effect by
describing the translation of the yield surface in stress space
through the effect of the backstress, such that straining in one
direction reduces the yield stress in the opposite direction.
This law is defined as an additive combination of a purely
kinematic term and a relaxation term, which introduces the
nonlinearity:

( )
pl pl
C c c
o

o o o o =
1
0
(8)
where C and are the material constants that need to be
calibrated from cyclic test data. C is the initial kinematic
hardening modulus, and defines the rate at which the
kinematic hardening modulus decreases as plastic deformation
develops. If C and become zero, Equation (8) reduces to
Equation (5) for the isotropic hardening model. If is zero,
Equation (8) reduces to Equation (6) for the linear kinematic
hardening model.
Integration of Equation (8) over a half cycle of tensile
stress-strain curve gives backstress:

( )
pl
e
C
c

o

= 1
(9)
with the uniaxial stress then expressed as:

( ) ( )
pl pl
e
C
e Q
b c c

o o

+ + = 1 1
0
(10)
In general, the combined hardening model can account for
the Bauschinger effect, ratcheting, relaxation of the mean
stress, and cyclic hardening with plastic shakedown. This
plastic constitutive model incorporates several material
parameters that are usually determined from the cyclic stress-
strain test data. Chapter 11 of ABAQUS Standard Users
Manual [12], which covers inelastic mechanical properties,
provides detailed calibration procedures for these parameters,
and input methods for the cyclic test data for such plastic
hardening models.

NONLINEAR FEA SIMULATION OF ELBOW UNDER
CYCLIC BENDING
Differences that arise from the use of the three plastic
hardening models have been assessed via a benchmark
analysis that considered the response of a 90
o
elbow set
between two straight equal length pipe segments that were
end-capped. This scenario was evaluated via nonlinear FEA
simulation under cyclic bending. The average cross-sectional
radius of the elbow is R = 203.2 mm (8 inches), the pipe wall
thickness is t = 12.7 mm (0.5 inches), the elbow radius to
cross-sectional radius ratio is R
b
/R = 3, and the bend parameter
is
1875 . 0 /
2
= = R t R
b

. The length of each attached pipe arm


was chosen at 10R in order to eliminate the arm effect, which
was suggested by Robertson et al. [13] based on sensitivity
analysis of a bend under in-plane bending. Detailed elastic-
large deformation plastic FEA was performed using ABAQUS
Standard [12] in three dimensional conditions. Due to
symmetry, only one quarter of the elbow was modeled via the
FEA mesh shown in Figure 1. This mesh had 755 nodes and
700 four-node shell elements with reduced integration
(ABAQUS element type: S4R). This shell element is generally
used for structural stress analysis with large displacement and
large rotation. The simple support conditions and the
symmetric boundary conditions were imposed in the FEA
model. Applied loading includes a fixed internal pressure of
72% of the specified minimum yield stress (SMYS) and a
cyclic in-plane bending moment at the elbow ends generated
by a pair of tensile and compressive forces. This loading was
assumed to simulate the effect of earth movement on an
existing pipeline with elbows.
X
Y
Z
V1
L1
C1

Figure 1. Three-dimensional FEA mesh for the 90
o
elbow

R
Rb




Copyright 2012 by ASME

4
The applied cyclic bending moment over the first 2 cycles is
shown in Figure 2, where the fixed pressure is applied first,
and opening (closing) bending moment is positive (negative).
The maximum bending moment is the limit-moment given in
[14] for defect-free elbow under combined loading of in-plane
closing bending moment and internal pressure.

-5000
-4000
-3000
-2000
-1000
0
1000
2000
3000
4000
5000
0 1 2 3 4 5 6 7 8 9
Time
A
p
p
l
i
e
d

m
o
m
e
n
t

(
k
i
p
-
i
n
)
-9
-7
-5
-3
-1
1
3
5
7
9
A
p
p
l
i
e
d

m
o
m
e
n
t

(
k
N
-
m
)
Isotropic
Kinematic
Combined

Figure 2. Applied cyclic bending moment on the elbow

The three nonlinear plastic hardening models were used in
each of the FEA calculations for the elbow. The flow theory of
plasticity with large deformation and von Mises yield criterion
were used. The material considered is an API 5L X42 pipeline
steel. The monotonic true stressplastic strain curve for X42
was represented as shown in Figure 3(a), with cyclic true
stressplastic strain loops as shown in Figure 3(b). It is
observed that after few stress cycles, the stressstrain loops
become stable. Other mechanical properties for X42 are the
Youngs modulus of 207 GPa and the Poisson ratio of 0.3. For
the isotropic hardening model, only the monotonic true stress
strain curve is needed, with the points in Figure 3(a) used as
the material input data in ABAQUS. For the pure kinematic
model, the half-cycle test data of the fourth stress-strain loop
was used to define the kinematic hardening components
without cyclic hardening. For the combined model, both the
monotonic and cyclic test data were used to define the
isotropic and kinematic hardening components.

In a typical fatigue analysis, it is necessary to determine a
critical location where the highest stress is induced and crack
initiation likely occurs after certain loading cycles. Weib et al.
[15] pointed out that the internal pressure or in-plane bending
moment causes the maximum stress to occur at the intrados of
pipe bends. Similarly, our FEA simulation showed that the
maximum stress range and the maximum strain range occur on
the outside surface at the intrados of the elbow under the fixed
internal pressure and cyclic bending. Therefore, this point is
used as a critical point for the fatigue analysis. The following
analysis focuses on this critical point of the elbow.
0
20000
40000
60000
80000
100000
120000
140000
160000
0 0.2 0.4 0.6 0.8 1 1.2
True plastic strain
T
r
u
e

s
t
r
e
s
s

(
p
s
i
)
0
200
400
600
800
1000
T
r
u
e

s
t
r
e
s
s

(
M
P
a
)
X42 test data

Figure 3(a). Tensile true stress-strain responses for X42

-80
-60
-40
-20
0
20
40
60
80
-0.008 -0.006 -0.004 -0.002 0.000 0.002 0.004 0.006 0.008
T
r
u
e

s
t
r
e
s
s

(
k
s
i
)
-552
-414
-276
-138
0
138
276
414
552
T
r
u
e

s
t
r
e
s
s

(
M
P
a
)
cycle 1
cycle 2
cycle 3
cycle 4
cycle 5
cycle 6
monot
true plastic strain

Figure 3(b). Cyclic true stress plastic strain hysteresis
curve for X42

Figures 4 and 5, respectively, show the variation of the
axial stress and axial strain at the critical point of the elbow
for the three plastic hardening models. The x-axis time in
these figures corresponds to the arbitrary segmenting of the
history as shown in Figure 2. From these figures it is observed
that the kinematic and combined models lead to similar
stresses, but the strains are very different. Figure 6 presents
the axial stressaxial strain loops at the critical point of the
elbow under the fixed internal pressure and cyclic bending.
This figure shows that the isotropic model develops the
smallest strain range, the kinematic model develops the largest
strain range, with the results for the combined model falling
between these. These observations are consistent with those
for the cyclically loaded beam observed in Reference [10]. It
is clear from Figure 6 that the kinematic and combined models
can simulate the Bauschinger effect, large plastic strain, strain
ratcheting, and relaxation of the mean stress. From these and
prior results [10], one can conclude that the combined
isotropic/kinematic hardening model is most appropriate, and
that its FEA results would be most accurate for an elbow
subject to internal pressure and cyclic bending.




Copyright 2012 by ASME

5
-100
-80
-60
-40
-20
0
20
40
60
80
100
0 1 2 3 4 5 6 7 8 9
Time
A
x
i
a
l

s
t
r
e
s
s

(
k
s
i
)
-552
-414
-276
-138
0
138
276
414
552
A
x
i
a
l

s
t
r
e
s
s

(
M
P
a
)
Isotropic
Kinematic
Combined

Figure 4. Axial stress at the critical point versus time

-0.140
-0.120
-0.100
-0.080
-0.060
-0.040
-0.020
0.000
0.020
0.040
0 1 2 3 4 5 6 7 8 9
Time
A
x
i
a
l

s
t
r
a
i
n
Isotropic
Kinematic
Combined

Figure 5. Axial strain at the critical point versus time

-100
-80
-60
-40
-20
0
20
40
60
80
100
-0.13 -0.11 -0.09 -0.07 -0.05 -0.03 -0.01 0.01 0.03
Axial strain
A
x
i
a
l

s
t
r
e
s
s

(
k
s
i
)
-552
-414
-276
-138
0
138
276
414
552
A
x
i
a
l

s
t
r
e
s
s

(
M
P
a
)
Iso, inrtodos,top
Kinematic
combined

Figure 6. Axial stress - axial strain at the critical point

DEVELOPMENT OF NONLINEAR FEA MODEL FOR
WRINKLEBEND SIMULATION
The FEA simulations for the elbow provide insight into
which plastic hardening model best quantifies the outcomes of
cyclic loading. With the best model established relative to this
benchmark case, extensive FEA calculations were then
conducted for a variety of wrinklebends using ABAQUS [12].
The aim is to develop a reliable nonlinear FEA model as a
means to simulate complicated stresses and strains that
develop at wrinkles due to forming the wrinkles in a straight
pipe. The intention is to quantify large plastic deformation,
accounting for bending, large rotation, and cyclic plastic
deformation. In turn, this provides the basis to develop and
calibrate a wrinklebend criterion to assess fatigue damage and
service life.

Basic FEA Modeling
A thin-walled pipe (X42) with a diameter D = 16 inches
(406.4 mm) and a thickness t = 0.283 inches (7.2 mm), leading
to D/t = 57 was considered for all FEA simulations. The
diameter and thickness were selected so that the D/t ratio falls
into the typical range from 30 to 100 found in such early field
bends. The pipe segment used in the FEA model had a total
length about 5 times the pipe diameter, so as to minimize the
boundary effects due to the pipe adjacent to the wrinkle. A
wrinkle was formed in the middle of the pipe segment due to
external bending achieved by applying a bending rotation at
the pipe ends while enforcing the circular cross-section remote
to the mid-length of the pipe. Due to symmetry, only one
quarter of the pipe segment was modeled. An overview of a
typical FEA mesh formed of four-node shell elements is
shown in Figure 7, wherein the mesh is refined in the vicinity
of the wrinkle, but coarser remote to it.
As indicated above, the residual stresses and strains fields
formed during cold bending, and thereafter under the action of
pressure and cyclic bending to simulate service were
developed by first forming the wrinkle, and then subjecting it
to service loading. Accordingly, various boundary and loading
conditions were introduced in different analysis steps in
ABAQUS, except for the conventional displacement boundary
conditions. During cold bending, a boundary condition to
constrain ovalization of the pipe was applied first, with
rotational displacement then imposed at the center node in the
end section of the model. The center node was rigidly
connected to all element nodes in the end section by rigid
beams. When the end rotation was applied to a specific angle
as expected to generate a desired wrinkle size, the rotation
displacement and the ovalization constraint were released to
allow the pipe spring back to form the wrinkle. A protruded
(out of the pipe body) wrinklebend has been identified as the
worst case (shortest wrinkle fatigue life), and thus was
adopted in all FEA simulations. To obtain such a protruded
wrinkle and to present the ovalization constraint as used with
mandrels in bending process, the pipe body was restrained
during bending by use of the rigid body or the built-in
multipoint constraints in the MPC user subroutine in
ABAQUS. After the buckle formed, the internal pressures at
72% and 10% of SMYS of the X42 pipeline steel were
alternatively applied to the FEA model. To simulate buried
pipes, an axial force produced at the capped ends from the




Copyright 2012 by ASME

6
internal pressure was imposed at the center node in the end
section. Both the axial force and internal pressure were
applied as the follower force as prescribed in ABAQUS.

X
Y
Z
X
-6.
-4.
-2.
0.
2.
4.
6.
Y
0.
2.
4.
6.
V1

Figure 7. A typical finite element mesh model for
simulating wrinklebend

Fatigue Damage Parameter
The numerical studies indicated that the wrinkle shape
characterized by the aspect ratio between the wrinkle height H
and the wrinkle length L (H/L) is a key parameter that can be
related to fatigue damage and fatigue resistance. Thus, the
wrinklebend geometry is described only by the H/L ratio in
this study. For a wrinklebend with a specific H/L ratio, the
fatigue damage of wrinklebend under cyclic loading is
quantified by the following damage parameter:

amp f
D c o
max
=
(11)
where
max
o
is the maximum stress, and
amp
c
is the amplitude
of strain during cyclic loading. The damage parameter D
f
as
defined in Equation (11) was developed by Smith et al. [16].
It is often referred to as the SWT parameter. Because the SWT
damage parameter embeds a mean stress effect, this and other
similar parameters can be used to quantify fatigue damage and
to predict fatigue life of metals [10, 17-18]. The following
sections focus on discussion of FEA results in terms of the
SWT fatigue damage parameter, rather than detailed stress and
strain components in wrinkles.

Parametric Study and Implications for the FEA
Selection of Element Types. To identify a reliable and
cost-effective finite element type in the simulation of
wrinklebends, a numerical comparison was performed, as
follows. Three commonly-used three dimensional finite
element types: S4R, S8R5 and C3D20R were selected for this
purpose. As defined in ABAQUS [12], S4R is a 4-node
doubly curved, general-purposed stress/displacement shell
element with reduced integration, hourglass control and finite
membrane strain; S8R5 is a 8-node doubly curved stress/
displacement thin shell element with reduced integration and
five degrees of freedom per node; and C3D20R is a 20-node
quadratic stress/displacement solid element with reduced
integration. Usually, S4R is adopted for general-purpose shell
structures, S8R5 is good for thin-walled shell structures, and
S3D20R is employed for three-dimensional solid structures.
These three elements are often adopted in a FEA simulation
for pipes, pipelines, and cylinders. FEA results for wrinkle
bends using these element types follow below.
The parametric analysis used the same pipe geometry
model for the three types of elements. For the shell element
S4R, the FEA mesh was generated with 728 nodes and 675
elements, while for the shell element S8R5, the FEA mesh
contained 1977 nodes and 676 elements. Finally, for the brick
element C3D20R, the FEA mesh was involved in 20688 nodes
and 3920 elements with 4 layers through the thickness. Finer
meshes were generated in the wrinkle area and coarser meshes
elsewhere for all three element types considered. Note that the
solid element gives effectively results at eight integration
points, whereas the two shell element types use seven
integration points through the thickness via Simpsons rule in
order to obtain accurate results in the thickness direction.
Figure 8 presents FEA results typical of the displacement
for the nodes in the compressive side of the pipe after
wrinkling. The results for the S4R shell elements are used for
this illustration, where the applied rotation displacement was
3.97. This displacement profile shows the locally buckled
shape, with the wrinkle evident in the lower right of the figure,
formed at the plane of symmetry, with the undeformed pipe
body remote to the wrinkle remaining straight. This wrinkle
has an aspect ratio H/L = 0.704, and a size as evident in the
figure. Figure 9 shows a typical deformed mesh for the
wrinkled pipe, where the arrowed node is a critical point on
the wrinkle crown. The axial stress and axial strain at this
point are determined in the FEA for a cyclic pressure of 72%
to 10% of SMYS of X42, for which D
f
was 0.25 ksi. Similar
results have been parametrically obtained for this pipe
geometry and the same bending rotation angle for the S8R5
shell element and the C3D20R brick element.
0
1
2
3
4
5
6
0 5 10 15 20 25 30 35
b
u
c
k
l
e
d

s
u
r
f
a
c
e

(
i
n
c
h
)
z (inch)
Shell Elements S4R, model length=35 inches, nodes=728, elements=675

Figure 8. A typical buckled pipe shape in symmetric plane
using 4-node shell elements




Copyright 2012 by ASME

7
X
Y
Z
X
-6.
-4.
-2.
0.
2.
4.
6.
Y
0.
2.
4.
6.
140573.
125781.
110989.
96197.
81405.
66612.
51820.
37028.
22236.
7444.
-7349.
-22141.
-36933.
-51725.
-66517.
-81310.
-96102.
V1
Output Set: Step5, Inc 13
Deformed(2.151): Total Translation
Contour: PlateBot YNormal Stress


Figure 9. A typical deformed mesh for a wrinkled pipe
using 4-node shell elements


Figure 10 compares the FEA results for the same pipe
geometry model using S4R, S8R5 and C3D20R. Figure 10(a)
shows the residual stress after wrinkling and the working
stresses under the cyclic pressure of 72% to 10% of SMYS of
the X42, whereas Figure 10(b) shows the corresponding
residual strain and working strains. It is evident from these
two figures that (a) the three element types predict the similar
wrinkle size of H/L with the lowest value found for the 8-node
shell element; (b) the two shell elements predict the similar
axial residual stress and the similar working stress at the
maximum load, while the brick element predicts the highest
residual stress and working stress at the maximum load; (c)
the 8-node shell element predicts the largest residual and
working strains, while the 4-node shell element predicts the
smallest residual and working strains; and finally, (d) the 8-
node shell element obtains the highest fatigue damage, while
the 4-node shell element obtains the lowest fatigue damage.
The total computer CPU time used in the FEA calculation
is 10, 59 and 360 minutes, respectively for the elements S4R,
S8R5 and C3D20R. Apparently, S4R is the most cost-
effective element, and C3D20R is the most expensive element
for this FEA simulation.
As elaborated in ABAQUS Standard Users Manual [12],
the general-purpose 4-node shell element S4R is robust and
suitable to simulate large deformation and large rotation for
shell structures as the case for wrinklebends. Compared to
S4R, the thin-wall focused 8-node shell element S8R5 and the
general-purpose solid element C3D20R are inappropriate for
the simulation of wrinklebends because they require too much
CPU time and determine considerably different results in
terms of stresses, strain and fatigue damage. Therefore, the
shell element S4R is adopted hereafter as a reliable and cost-
effective element for the FEA simulation of wrinklebends. This
selection is consistent with use of the S4R shell elements for
wrinklebend simulation by Sen et al. [19].
(a)
-150
-100
-50
0
50
100
150
200
250
1 2 3
a
x
i
a
l

s
t
r
e
s
s

(
k
s
i
)
S4R
S8R5
C3D20R
after buckle 72% SMYS
10% SMYS


(b)
-1.5
-1.0
-0.5
0.0
0.5
1.0
1.5
1 2 3 4 5
d
a
m
a
g
e

(
k
s
i
)
,

H
/
L

o
r

s
t
r
a
i
n
S4R
S8R5
C3D20R
Damage
axial strain
72%SMYS
axial strain
10%SMYS
H/L
axial strain
after buckle

Figure 10. Comparisons of FEA results using three
indicated elements. (a) stresses after buckle
formed and at cyclic pressure, (b) Strains after
buckle formed and at cyclic pressure

Mesh Sensitivity Analysis. Mesh sensitivity was assessed
relative to three different FEA models of otherwise the same
pipe geometry. The first pipe model had a half length of 35
inches and contained a coarser mesh with 728 nodes and 675
elements. The second pipe model had the same half length of
35 inches, but used a refined mesh with 1862 nodes and 1776
elements. The third pipe model had a longer half-length of 52
inches and the mesh contains 1222 nodes and 1150 elements.
Figures 11(a) and 11(b) present numerical results for residual
and working stresses and strains, H/L ratio, fatigue damage
and CPU time, respectively. As evident in these figures, the
three FEA models predict the similar results for all these
variables, except for the computation time. The CPU time is
10, 30 and 18 minutes, respectively, for the first, second and
third FEA models. Thus, the first coarser mesh model is most
cost-effective, and while providing acceptable and reasonable
outcomes. Therefore, the first model was adopted for the
remaining FEA-based wrinklebend simulations.





Copyright 2012 by ASME

8
(a)
-40
-20
0
20
40
60
80
100
120
140
160
1 2 3 4
a
x
i
a
l

s
t
r
e
s
s

(
k
s
i
)

o
r

C
P
U

(
m
i
n
u
t
e
s
)
PL=35", nodes=728
PL=35", nodes=1776
PL=52", nodes=1222
after buckle
72% SMYS
10% SMYS
CPU, minutes


(b)
-0.8
-0.6
-0.4
-0.2
0.0
0.2
0.4
0.6
0.8
1 2 3 4 5
d
a
m
a
g
e

(
k
s
i
)
,

H
/
L

o
r

s
t
r
a
i
n
PL=35", nodes=728
PL=35", nodes=1776
PL=52", nodes=1222
Damage
axial strain
72%SMYS
axial strain
10%SMYS
H/L
axial strain
after buckle

Figure 11. Mesh sensitivity effect on FEA results for
wrinklebends. (a) residual and working stresses,
(b) residual and working strains

Effect of Material Input Data. The effect of material input
data on the outcomes of the FEA of wrinklebends was
assessed relative to three representations of stress-strain curve
for the same X42 steel. The first material input data set was
based on the full-range stress-strain curve, for which the last
data point involved a fracture stress S
f
= 150 ksi and a fracture
strain
013 . 1 =
f
c
, as was shown in Figure 2(a). The second
material input data set was comparable, except that the full-
range curve was truncated at S
f
= 130.2 ksi and
0 . 1 =
f
c
. The
third material input data set was also comparable to the first,
but it was truncated at the ultimate stress, for which S
f
= 96.1
ksi and
18 . 0 =
f
c
. Note that beyond the final input point
supplied ABAQUS assumes a flat stress-strain response.
Figure 12 shows the FEA results for these three material
input data sets. The comparison from this figure shows that:
(a) these three sets of material input data lead to the same
wrinkle aspect ratio of H/L, but different stresses and strains;
(b) the complete full-range stress-strain curve predicts the
lowest strains but the highest stresses and fatigue damage; and
(c) the stress-strain curve truncated at the ultimate develops
the highest strains but the lowest stresses and fatigue damage.
One anticipates that the results for the third set of material
input data would be inaccurate because the incomplete stress-
strain curve poorly quantifies postbuckling response. In
contrast, one anticipates that the results from the full-range
response (the first data set) would be accurate because it best
characterizes the deformation process up through postbuckling
deformation. On this basis, logic indicates the full-range
stress-strain curve should be adopted for purposes of the
material input data if viable FEA simulation of the
wrinklebend is sought.

(a)
-100
-50
0
50
100
150
200
1 2 3
a
x
i
a
l

s
t
r
e
s
s

(
k
s
i
)
Sf = 96.1 ksi
Sf = 130.2 ksi
Sf = 150 ksi
after buckle 72% SMYS
10% SMYS


(b)
-0.8
-0.6
-0.4
-0.2
0.0
0.2
0.4
0.6
0.8
1 2 3 4 5
d
a
m
a
g
e

(
k
s
i
)
,

H
/
L

o
r

s
t
r
a
i
n
Sf = 96.1 ksi
Sf = 130.2 ksi
Sf = 150 ksi
Damage
axial strain
72%SMYS
axial strain
10%SMYS
H/L
axial strain
after buckle

Figure 12. Material input data effect on FEA results for
wrinklebends. (a) residual and working stresses,
(b) residual and working strains

Plastic Hardening Models for Cyclic Loading. As noted
earlier, the effect of material hardening models on the
numerical simulation for wrinklebends has considered three
nonlinear plastic hardening models: the isotropic, kinematic,
and combined isotropic/ kinematic hardening models. In




Copyright 2012 by ASME

9
reference to the case where the simulated wrinklebend had an
H/L ratio of 0.5, the FEA results led to values of the SWT
fatigue damage parameter as 0.312, 0.103 and 0.283 ksi,
respectively for the isotropic, kinematic and combined models.
As anticipated from earlier discussion, the combined model
leads to a result between those for the isotropic and kinematic
models. This result is consistent with the observation in
Reference [7] for a beam and an elbow under large plastic
cyclic loading.
By use of the parametric study outlined above, and guided
by the results of related full-scale testing (see [8] for details), a
reliable nonlinear FEA model was developed for the
numerical simulation of wrinklebends. This FEA model used
the S4R shell element to simulate the severe distortion of
wrinkles, with the combined isotropic/kinematic hardening
model adopted to simulate cyclic plastic deformation, and the
full-range stress-strain curve required to quantify the response
up through postbuckling response. This elastic-plastic FEA
model is utilized next to calculate stresses and strains at
wrinklebneds, to quantify fatigue damage and to estimate the
fatigue life of wrinklebends in pipeline applications.

FATIGUE DAMAGE ANALYSIS OF WRINKLEBENDS

Fatigue Damage Quantification
The nonlinear FEA model as outlined above was used in
detailed elastic-plastic FEA for a pipe segment with a diameter
of 16 inches and a wall thickness of 0.283 inches in the X42
pipeline steel. The cyclic pressure was 72% to 10% of SMYS
of the X42. Different bending rotation angles were applied to
the pipe model to generate different wrinkle sizes so that the
wrinkle fatigue damage could be quantified directly in terms
of the wrinkle aspect ratio of H/L, relative to the maximum
stress and the strain range at the critical point of the wrinkle
crown for each pressure cycle. As noted earlier, fatigue
damage was assessed using the SWT damage parameter in
Equation (11). Four wrinkle sizes quantified via H/L as 0.129,
0.303, 0.475 and 0.706 were developed in these simulation
using the combined hardening model. Based on field measured
deformations, and past experience in related failure analyses,
these sizes can be considered representative of wrinklebends
from small to large.
Figure 13 shows the variation of the SWT fatigue damage
parameter and the wrinkle aspect ratio of H/L in reference to
the FEA results. From this figure, it is found that a linear
relationship exists between the SWT damage parameter and
the H/L ratio. For the combined hardening model, a linear
function is thus curve-fitted as:

) / ( 565 . 0 L H D
f
=
(12)
While this linear function is simple, it does quantify the trend
based on the FEA, as shown in Figure 13. It is evident from
Equation (12) that fatigue damage becomes increasingly
severe as the wrinkle shape defined by H/L ratio becomes
larger. In this context, wrinkle shape defined by H/L is a
viable metric of wrinkle severity, with severity tending to zero
as H/L likewise tends to zero. It follows that fatigue damage
of a wrinklebend can be quantified by Equation (12) in
reference to the aspect ratio of H/L.

0
0.2
0.4
0.6
0.8
0 0.2 0.4 0.6 0.8 1
Wrinkle height/Wavelength, H/L
D
a
m
a
g
e

-

S
W
T
,

k
s
i
72%-10% SMYS, X42, isotropic
fit line, isotropic
72% - 10% SMYS, X42, kinematic
fit line, kinematic
72%-10% SMYS, X42, combined
fit line, combined
1,000 cycles
10,000 cycles
750 cycles
1,500 cycles
5,000 cycles
100,000 cycles
Free capped end condition
2,500 cycles
D=0.625(H/L)
D=0.205(H/L)
D=0.565(H/L)

Figure 13. Variation of SWT damage parameter with
wrinkle ratio H/L.

The significant effect of the hardening model on the
predicted severity of a wrinklebend is evident in comparing
the FEA results based on isotropic and kinematic models
relative to that for the combined model, results for which are
also included in Figure 13 for the different wrinkle sizes
considered. These also can be represented by linear functions
fitted to the predicted data for these hardening models as:

) / ( 625 . 0 L H D
f
=
, for isotropic model (13)

) / ( 205 . 0 L H D
f
=
, for kinematic model (14)
It can be seen in Figure 13 that the numerical results for the
combined hardening model fall between those for the isotropic
and kinematic hardening models. It is also apparent that the
results for the isotropic model fall closer to those for the
combined model, while those for the kinematic model fall far
below those for the combined model.

Fatigue Life Estimation
With the fatigue damage quantified by Equation (12) for
wrinklebends in the X42 pipeline steel, the fatigue life for a
wrinklebend can be estimated, provided that the fatigue
resistance is available for that pipeline steel. As developed
from fatigue test data, Reference [10] indicates that fatigue
resistance curve for the pipeline steel can be expressed as:

( ) ( )
28 . 0 02 . 1
2 1 . 2 2 273

+ =
f f R
N N D
(15)
where N
f
denotes the number of cycles to failure.




Copyright 2012 by ASME

10
Combining the fatigue damage force expressed in
Equation (12) and the material fatigue resistance curve
expressed in Equation (15) leads to the relationship between
the wrinkle size H/L and the fatigue life N
f
in the form:

( ) ( )
28 . 0 02 . 1
2 72 . 3 2 2 . 483

+ =
f f
N N
L
H
, (16)
which is specific to the combined hardening model. This
wrinklebend fatigue criterion can be used for estimating the
service life of a wrinklebend with a known H/L for X42.
While this criterion has been validated by full-scale
experiments for wrinklebends in Reference [8], space
limitations preclude detailed discussion of that in this paper.
While Equations (13), (14) and (15) can be combined
leading to similar relationships between the wrinkle size H/L
and the fatigue life N
f
for the isotropic and kinematic models
in the form of:

( ) ( )
28 . 0 02 . 1
2 36 . 3 2 8 . 436

+ =
f f
N N
L
H
(17)
and

( ) ( )
28 . 0 02 . 1
2 24 . 10 2 7 . 1331

+ =
f f
N N
L
H
(18)
Their use is not recommended as the kinematic outcome is
very non-conservative, while the isotropic outcome is
marginally conservative.
Figure 14 illustrates the wrinkle aspect ratio versus
fatigue life curves from Eqs (16), (17) and (18). These fatigue
curves are simply referred to as H/L-N curves because they
are similar to the traditional S-N curves in the simple fatigue
test. For a given wrinkle size H/L, the fatigue life of the
wrinkle can be easily estimated from the H/L-N curves. As
noted above, the H/L-N curve for the isotropic model is
slightly conservative, but that for the kinematic model is
highly non-conservative.
By way of illustration, two wrinklebends with H/L = 0.2
and 0.4 noted in Reference [20] are evaluated relative to cyclic
pressures service between 72% and 10% of SMYS. If made of
the X42 represented by the properties discussed herein,
Equation (16) indicates fatigue lives the order of 20,000 and
3,000, respectively for the smaller and larger wrinkles. If such
a cycle occurred once per day, service lives in the order of 55
and 8 years, respectively, result. Relative insight into how to
best manage integrity can derive from such comparisons.

CONCLUSIONS
Elastic-plastic FEA and fatigue damage analysis were
presented for wrinklebends under combined pressure and
cyclic loading. Work hardening and Bauschinger effects were
considered in the context of forming the wrinkle, and during
cyclic plastic deformation relative to models for plastic
hardening, including isotropic, kinematic and combined
isotropic/kinematic plastic hardening. Such models were
briefly reviewed and then used in the FEA simulation for a
benchmark case involving an elbow, as well as wrinklebends
subject to cyclic bending loading. The results showed that the
hardening model has a significant effect on the stresses and
strains, which in turn impact the fatigue resistance.

0
0.2
0.4
0.6
0.8
1
1.2
0 5000 10000 15000 20000 25000
cyclic life, Nf
H

/

L
SWT, D1-isotropic
SWT, D1-kinematic
SWT, D1-combined

Figure 14. The H/L-N curves for wrinklebends.

Parametric analysis was performed for the wrinklebend,
including element type selection, mesh sensitivity analysis,
computation time, and the scope of material input data.
Relative to those outcomes and guided by full-scale test data a
reliable nonlinear FEA model was then developed to
numerically simulate wrinklebend responses under cyclic
pressure. Detailed numerical calculations were carried out for
a wide range of wrinkle sizes under the cyclic pressure of 72%
to 10% of SMYS of the X42 pipeline steel. With the FEA
results, fatigue damage was quantified in reference to the
SWT damage parameter, and fatigue criteria were proposed to
assess the integrity of wrinklebends. It was found that the
wrinkle geometry, pipeline geometry and pipeline operation
were important factors in formulating the criteria, with the
wrinkle aspect ratio H/L being a key parameter, and on that
basis a fatigue criterion was developed. The results indicated
that increasing wrinkle size can significantly reduce service
life of the wrinkle. It is expected that the proposed fatigue
criterion and H/L-N curve can serve as a viable tool in
determination of service life of wrinklebends and in evaluation
of pipeline integrity.


REFERENCES
[1]. Bilston, P. and Murray, N., The Role of Field Bend in
Pipeline Construction. EPRG/PRCI 9th Biennial Joint
Technical Meeting on Line Pipe Research, Houston,
Texas, May 11-14, 1993.




Copyright 2012 by ASME

11
[2]. Olson, R., Clark, T. and Odom, T., Evaluation of the
Structural Integrity of Cold Field-Bend Line Pipe.
EPRG/PRCI 10th Biennial Joint Technical Meeting on
Line Pipe Research, Cambridge, U.K., April 18-21,
1995.
[3]. Fukuda, N., Yatabe, H., Kawaguchi, S., Watanabe, T.
and Masuda, T., Experimental and Analytical Study of
Cold Bending Process for Pipelines. Journal of
Offshore Mechanics and Artic Engineering, Vol. 125,
2003, pp.153-157.
[4]. Dinovitzer, A., Fredj, A., Lazor, R. and Doblanko, R.,
Development and Validation of a Pipeline Buckle and
Wrinkle Assessment. Proceedings of 2004
International Pipeline Conference, Calgary, Alberta,
Canada, October 4-8, 2004.
[5]. Sen, M., Cheng, J.J.R., Murray, D.W. and Zhou, J.,
Mechanical Properties of Cold Bend Pipes. Journal of
Pressure Vessel Technology, Vol. 130, 2008: 021708.
[6]. Alexander, C. and Kulkarni, S., Evaluating the Effects
of Wrinkle Bends on Pipeline Integrity. Proceedings of
the 7th International Pipeline Conference, Calgary,
Canada, September 29-October 3, 2008: IPC2008-
64039.
[7]. Mohareb, M., Kulak, G.L., Elwi, A. and Murray, D.W.,
Testing and Analysis of Steel Segments. Journal of
Transportation Engineering, Vol. 127, 2001, pp. 408-
417.
[8]. Zhu, X.K., Leis, B.N. and Clark, E.B., Integrity
Management for Wrinklebends and Buckles, Final
Report to U.S. Department of Transportation PHMSA,
February 2007.
[9]. De Larminat, C., Lallemand, D., Bourgouin, L., and
Zarea, M., Assessment of Bending WrinklesA Review
of Available Acceptance Criteria. The 17th PRCI,
EPRG and APIA Jointed Technical Meetings on Pipeline
Research, Milan, Italy, May 11-15, 2009: Paper 12.
[10]. Zhu, X.K. and Leis, B.N., Elastic-Plastic Finite Element
Simulation and Fatigue Life Prediction for Beam and
Elbow under Cyclic Loading. Proceedings of 2007
ASME Pressure Vessel and Piping Conference, San
Antonio, Texas, July 22-26, 2007: PVP2007-26273.
[11]. Lemaitre, J. and Chaboche, J.L., Mechanics of Solid
Mechanics, Cambridge University Press, Cambridge,
UK, 1990, pp. 161-241.
[12]. ABAQUS, 2005, ABAQUS/Standard Users Manual,
Version 6.5, Hibbitt, Karlsson, and Sorensen Inc.,
Providence, RI, USA.
[13]. Robertson, A., Li, H. and Mackenzie, D., Plastic
Collapse of Pipe Bends under Combined Internal
Pressure and In-Plane Bending, International Journal
of Pressure Vessels and Piping, Vol. 82, 2005, pp. 407-
416.
[14]. Chattopadhyay, J. and Tomar, A.K.S., New Plastic
Collapse Moment Equations of Defect-Free and
Throughwall Circumferentially Cracked Elbows
Subjected to Combined Internal Pressure and In-Plane
Bending Moment, Engineering Fracture Mechanics,
Vol. 73, 2006, pp. 829-854.
[15]. Wei, E., Lietzmann, A. and Rudolph, J., Linear and
Nonlinear Finite-Element Analysis of Pipe Bends,
International Journal of Pressure Vessels and Piping,
Vol. 67, 1996, pp. 211-217.
[16]. Smith, K.N., Watson, P. and Topper, T.H., A Stress-
Strain Function for the Fatigue of Metal, Journal of
Materials, Vol. 5, 1970, pp. 767-778.
[17]. Koh, S.K., Fatigue Damage Evaluation of a High
Pressure Tube Steel Using Cyclic Strain Energy
Density, International Journal of Pressure Vessels and
Piping, Vol. 79, 2002, pp.791-798.
[18]. Dowling, N.E., Mechanical Behavior of Materials:
Engineering Methods for Deformation, Fracture and
Fatigue, Second edition, Prentice Hall, Upper Saddle
River, New Jersey, 1999.
[19]. Sen, M. and Cheng, R., Finite Element Analysis of
Cold Bend Pipes under Bending Loads. Proceedings
of the 8th International Pipeline Conference, Calgary,
Canada, September 27-October 1, 2010: IPC2010-
31487.
[20]. Zhu, X.K. and Leis, B.N., Elastic-Plastic Finite
Element Simulation and Fatigue Damage Analysis of
Wrinklebends. Proceedings of 2008 ASME Pressure
Vessel and Piping Conference, Chicago, Illinois, July
27-31, 2008: PVP2008-61220.

Вам также может понравиться