You are on page 1of 10

International Journal of Pressure Vessels and Piping 113 (2014) 15e24

Contents lists available at ScienceDirect

International Journal of Pressure Vessels and Piping


journal homepage: www.elsevier.com/locate/ijpvp

Review

Fatigue analysis of corroded pipelines subjected to pressure and


temperature loadings
Divino J.S. Cunha a, *, Adilson C. Benjamin a, Rita C.C. Silva b, João N.C. Guerreiro c,
Patrícia R.C. Drach c
a
PETROBRAS Research and Development Center, Brazil
b
Federal University of Pará e UFPA, Brazil
c
National Laboratory for Scientific Computation e LNCC, Brazil

a r t i c l e i n f o a b s t r a c t

Article history: In this paper a methodology for the fatigue analysis of pipelines containing corrosion defects is proposed.
Received 10 January 2011 This methodology is based on the nominal stresses from a Global Analysis using a one-dimensional Finite
Received in revised form Element (FE) model of the pipeline together with the application of stress concentration factors (SCFs).
7 October 2013
As the stresses may exceed the yielding limit in the corrosion defects, the methodology also adopts a
Accepted 17 October 2013
strain-life approach (εeN method) which is capable of producing less conservative fatigue lives than the
stress-based methods. In addition the proposed methodology is applied in the assessment of the fatigue
Keywords:
life of an onshore-hot pipeline containing corrosion pits and patches. Five corrosion pits and five
Multiaxial fatigue
Biaxial fatigue
corrosion patches with different sizes are considered. The corrosion defects are situated on the external
Strain-life surface of the pipeline base material. The SCFs are calculated using solid FE models and the fatigue
Out-of-phase cyclic loadings analyses are performed for an out-of-phase/non-proportional (NP) biaxial stresses related to the com-
Heated pipelines bined loading (internal pressure and temperature) variations caused by an intermittent operation with
Corroded pipeline fatigue analysis hot heavy oil (start-up and shut-down). The results show that for buried pipelines subjected to cyclic
Corrosion defect SCFs combined loadings of internal pressure and temperature fatigue may become an important failure mode
when corroded pipeline segments are left in operation without being replaced.
Ó 2013 Elsevier Ltd. All rights reserved.

1. Introduction where very conservative corrosion assessment methods have been


replaced by more accurate methods which are capable of not only
Fatigue is not usually a failure mode that occurs in buried guaranteeing the pipeline structural integrity but also tolerating
pipelines subjected only to internal pressure. However, this is not larger corrosion defects for longer periods of time.
always the case if other types of loadings act on the pipeline, like This paper presents a methodology for the fatigue life assess-
the cyclic combined loadings of internal pressure and temperature ment of hot pipelines with corrosion defects in the base material.
resulting from intermittent operation with hot heavy oil. The proposed methodology is based on nominal stresses and stress
Furthermore, fatigue may become an important failure mode, concentration factors (SCFs) together with a BrowneMiller (BM)
even in buried pipelines subjected only to internal pressure, if any strain-life critical-plane method [3e5]. Although stress-life
type of defect (cracks, dents or corrosion) is found in periodical methods are more disseminated [6], strain-life methods are more
inspections [1,2]. adequate for pipelines which undergo plastic deformation.
In the case of hot pipelines designed under the traditional phi- The BM method handles plasticity, which may occur in the
losophy of preventing buckling by burying, fatigue becomes an corrosion defect, as well as the multiaxial stresses/strains and the
important failure mode when corroded pipeline segments are left out-of-phase/non-proportional (NP) characteristic of the applied
in operation without being replaced. This is a worldwide trend loadings or the corresponding stresses [3e5,7].
The proposed methodology was applied to an onshore corroded
* Corresponding author. Av Horácio Macedo 950, Cidade Universitária, Ilha do API-X60 pipeline with a 57.7 ratio of the diameter to the wall
Fundão, 21941-915 Rio de Janeiro, RJ, Brazil. Tel.: þ55 21 3865 4743; fax: þ55 21 thickness. The pipeline operates 3 times a week alternating be-
3865 3764.
tween hot heavy oil and light products at ambient temperature. The
E-mail addresses: divinocunha@petrobras.com.br, djs.cunha@hotmail.com (D.J.
S. Cunha), acbenjamin@petrobras.com.br (A.C. Benjamin), ritaccs@ufpa.br (R.C. maximum operating pressure is 8.2 MPa and the maximum oper-
C. Silva), joao@lncc.br (J.N.C. Guerreiro), pdrach@lncc.br (P.R.C. Drach). ating temperature of the oil is 80  C.

0308-0161/$ e see front matter Ó 2013 Elsevier Ltd. All rights reserved.
http://dx.doi.org/10.1016/j.ijpvp.2013.10.013
16 D.J.S. Cunha et al. / International Journal of Pressure Vessels and Piping 113 (2014) 15e24

Nomenclature T temperature
Tinst pipeline installation temperature
a straight portion of the corrosion defect depth Toper maximum operating temperature
b Basquin’s fatigue strength exponent w pipe weight
c CoffineManson fatigue ductility exponent w corrosion defect width (circumferential dimension)
BM BrowneMiller x, y, z local cylindrical co-ordinates
CM CoffineManson X, Y, Z global cylindrical co-ordinates
CP critical plane (highest damage plane) z distance between the soil surface and the trench
C/P Ang FE-SAFE nomenclature for the CP angle f bottom
CWP cylindrical wide pit afat fatigue usage factor
Cycle-Ampl FE-SAFE nomenclature for the fatigue parameter g shear strain (gij ¼ εi  εj on the shear plane iej, i, j ¼ 1,
amplitude (e/2) 2, 3, i s j)
d corrosion defect depth εi (i ¼ 1,2,3) principal strains: in-plane (ε1, ε2), out-of-plane/out-
dsi (i ¼ 1, 2, 3) stress datasets (FE-SAFE data line with the stress of-surface (ε3 h εz)
tensor components) εn normal strain εijn ¼ ðεi þ εj Þ=2 on the shear plane iej, i,
e BM parameter or strain (e h gmax þ εn) j ¼ 1, 2, 3, i s j)
e BM-parameter range (e ¼ e2  e1 ¼ De h Dgmax þ Dεn) ε0f fatigue ductility coefficient
eL elongation εtrue uniaxial true strain
D total damage (D ¼ SDi) n Poisson’s ratio
De pipe external diameter strue uniaxial true stress
Di damage of a single loading cycle (Di ¼ 1/N) s0f Basquin’s fatigue strength coefficient
DFF design fatigue factor sh hoop stress (sh ¼ pDi/(2t))
E elastic (Young’s) modulus si (i ¼ 1, 2, 3) principal stresses: in-plane (s1, s2), out-of-plane/
EFF environmental fatigue factor out-of-surface (s3 h sz)
h hardening exponent sL longitudinal stress
H hardening coefficient su engineering ultimate tensile stress
hc cyclic hardening exponent sy engineering yield stress
Hc cyclic hardening coefficient s* reference sample (stress-tensor/dataset, within the
hs soil cover stress history, taken by FE-SAFE to define the
L corrosion defect length (longitudinal dimension) orientation of stress principals and principal/shear
LP longitudinal patch planes)
n number of loading repeats s*i (i ¼ 1, 2, 3) reference principal stresses related to dataset s*
N number of strain cycles to failure (obtained from εeN f rotating angle of the principal/shear planes round the
curve) out-of-surface axis ε3 h εz (0  f  180 ), measured
N loading history life (N ¼ 1/D), FE-SAFE output fatigue between the plane normal εn and the stress principal
life s*1 , and identified in FE-SAFE output file as “C/P Ang”
NP non-proportional fx angle between the CP normal εn and the x-axis
p internal pressure (positive from x-axis towards y-axis), identified in
pd design pressure FE-SAFE output file as “CP/X/Ang”
poper maximum operating pressure q1 angle between s1 and s*1 (q1 ¼ 0 for constant direction
r pit radius or patch-bottom fillet radius principals)
R patch-top fillet radius DT temperature loading (DT ¼ T  Tinst)
SCF stress concentration factor (,)h hoop (circumferential)
SMTS minimum specified ultimate tensile stress (,)L longitudinal
SMYS minimum specified yielding stress (,)max maximum
t time (,)nom nominal
t pipe wall thickness

2. Methodology of analysis considered for the pipeline (Global Analysis). This approach differs
from that [3,4,8,9] where the stress analysis is normally performed
As shown in Fig. 1, the proposed methodology for the with the structure containing the corrosion defect (Local Analysis).
corroded pipeline fatigue life assessment has three main phases: In such cases, the stresses which result from the analysis are
(1) Stress Global Analysis, (2) Stress components amplification by already amplified in the defect, but the finite element model, being
the corrosion defect SCFs, with the SCFs obtained from a (20 ) a solid model, is more complex.
Local Analysis, and (3) Fatigue Analysis using a multiaxial strain- In other words, although the calculations of the SCFs require a
life method. The steps (20 )e(2) and (3) must be repeated for each solid model, neither this model nor the stresses from it are employed
corrosion defect. These three phases will be described in more in the fatigue analysis. Only the values of both SCFs are utilized.
detail later on. The SCFs also imply that the Global Analysis only needs to be
Using SCFs means that the stress analysis is carried out for the performed once irrespective of the type or the geometry of the
plain pipeline (uncorroded) and a one-dimensional model may be corrosion defects.
D.J.S. Cunha et al. / International Journal of Pressure Vessels and Piping 113 (2014) 15e24 17

Table 1
Design and operating parameters of the pipeline.

Parameter e

Pipe material API 5L X60


SMYS 413 MPa
SMTS 517 MPa
Elongation eL 0.22
Young’s modulus E 206 GPa
Poisson’s ratio n 0.3
Thermal expansion coefficient a 1.17  105 mm/mm/ C
Pipe external diameter De 457.2 mm (18 in)
Pipe wall thickness t 7.92 mm (0.312 in)
Soil cover hs 1.0 m
Pipe weight w 2.18 N/mm
Design pressure pd 10.30 MPa
Maximum operating pressure poper 8.2 MPa
Pipeline installation temperature Tinst 20  C
Pipeline operating temperature Toper 80  C
Design life 40 years
Number of operations per week 3
Number of operations during design life 6240
Fig. 1. Methodology flowchart. Design fatigue factor DFF 5
Environment fatigue factor EFF 2

3. Pipeline characteristics
A corrosion pit is sketched in Fig. 3. The length a of the cylin-
This study was carried out on a buried API-X60 steel pipeline drical part of the pit is equal to r, the pit depth d is equal to 2r, the
(see Fig. 2) designed according to the ASME B31.4 code [10]. It pit length L is equal to 2r and the pit width w is approximately equal
was assumed that the pipeline curvature is negligibly small and to 2r. Consequently, the pit length L and the pit width w are equal to
that the soil cover is large enough to prevent the pipeline global the pit depth d. Table 2 presents the dimensions of each pit
buckling. In this case, as the nominal stresses are uniform in both considered.
longitudinal and circumferential directions, the fatigue loading of The corrosion patch geometry is described by the following
the pipeline containing one corrosion defect reduces to only one parameters: the depth d, the length L, the width w, the fillet radius r
stress history. and R and the straight length a of the rectangular part. The shape of
The pipeline operates 3 times a week alternating between the corrosion patch is shown in Fig. 4 and the patch dimensions are
hot heavy oil and light products at ambient temperature. The in Table 3.
maximum operating pressure is 8.2 MPa and the maximum oper-
ating temperature of the oil is 80  C. 5. Local Analysis
The main characteristics of the pipeline are shown in Table 1. As
the operating temperature is below 120  C, no derating is applied 5.1. Solid finite element model
on the steel properties [10].
The corrosion defects (pits and patches) in Tables 2 and 3 were
4. Geometry of the corrosion defects modeled using solid (3D) Finite Elements (FE) and their corre-
sponding SCFs were calculated using these models. Each model was
It is supposed that the pipeline has been operating for several represented by a 2.6 m straight pipe with a single corrosion defect.
years and during this time corrosion has occurred on its external A cylindrical coordinate system was used with the following
surface. convention: X-axis (radial), Y-axis (circumferential) and Z-axis
Two types of corrosion defects are considered in this study: (longitudinal).
cylindrical wide pit (CWP) and longitudinal patch (LP). In both The local analyses of all the corroded pipeline models were
cases five corrosion defect sizes are evaluated. performed with the ANSYS program [11]. To take advantage of
The geometry of each pit is described by three key parameters:
the pit depth d, the radius of the pit root r and the length a of the
cylindrical portion of the pit. Other geometric parameters are: the
pit length L, which is the pit longitudinal dimension and the pit
width w, which is its circumferential dimension.

soil

hs

Fig. 2. Pipeline burial parameters. Fig. 3. Cylindrical wide pit (CWP): (a) top view, (b) longitudinal view.
18 D.J.S. Cunha et al. / International Journal of Pressure Vessels and Piping 113 (2014) 15e24

Table 2 Table 3
Dimensions of the cylindrical wide pits (CWPs). Dimensions of the longitudinal patches (LPs).

Defect d (mm) r (mm) a (mm) L (mm) w (mm) d/t Defect d (mm) r (mm) a (mm) L (mm) w (mm) R (mm) d/t

CWP1 0.792 0.396 0.396 0.792 0.792 0.10 LP1 0.792 0.634 0.158 60 20 1.268 0.10
CWP2 1.584 0.792 0.792 1.584 1.584 0.20 LP2 1.584 1.267 0.317 90 30 2.534 0.20
CWP3 2.376 1.188 1.188 2.376 2.376 0.30 LP3 2.376 1.901 0.475 120 40 3.802 0.30
CWP4 3.168 1.584 1.584 3.168 3.168 0.40 LP4 3.168 2.534 0.634 150 50 5.068 0.40
CWP5 3.960 1.980 1.980 3.960 3.960 0.50 LP5 3.960 3.168 0.792 180 60 6.336 0.50

symmetry only a quarter of each model was analyzed. The solid FE


The analysis of the plain pipe (uncorroded) was performed using
models were constructed using the non-conforming 8-node brick
a uniform mesh solid FE model with 4 elements through the
element SOLID45 available in the ANSYS FE library. Appropriate
thickness.
boundary conditions were applied to the symmetry planes (see
In total, eleven solid FE models were constructed: one for the
Fig. 5). The FE models were extended far enough beyond the
plain pipe and ten for the pipe with each type of corrosion pits and
corroded region to prevent the end conditions from affecting the
patches presented in Tables 2 and 3. For each solid FE model a linear
results.
analysis was performed. In these analyses, the pipeline was sub-
In the corroded region, 8 elements were used through its
jected to an internal pressure and a longitudinal tension. In fact, the
thickness and at some distance outside the corroded region this
longitudinal tension simulates the temperature loading effect.
number was reduced to 4 elements through the thickness (see
Fig. 6). This mesh density was selected after a convergence study in
which linear analyses were performed using an increasing degree 5.2. Stress concentration factors (SCFs)
of mesh refinement.
For each applied loading (internal pressure and longitudinal
tension) and each direction (longitudinal and circumferential or
hoop) the SCFs were determined as the ratio between the
maximum stress component in the corroded region and the cor-
responding stress in the uncorroded pipeline (nominal stress):
 
ski
k
ðSCFi Þ ¼   max i ¼ L; h and k ¼ p; DT (1)
ski
nom

In order to simplify the SCFs application, the SCF for longitudinal


stress was taken as the mean value between the two longitudinal
SCFs related to the internal pressure and the temperature loading.
Therefore, as the temperature (or longitudinal loading) has no in-
fluence in the circumferential (hoop) direction, in this work the
SCFs are given as:

ðSCFL Þp þ ðSCFL ÞDT


SCFL ¼ ; SCFh ¼ ðSCFh Þp (2)
2
The SCFs curves are shown in Fig. 7 as a function of the
normalized defect depth (d/t). Their numerical values are presented
in Tables 4 and 5. The SCFs unitary values corresponding to the
plain pipe (design condition) were included in the first line of both
tables as new corrosion defects named CWP0 and LP0 respectively.

Fig. 4. Longitudinal patch (LP): (a) top view, (b) longitudinal view, (c) cross-section. Fig. 5. Boundary conditions of the solid FE models.
D.J.S. Cunha et al. / International Journal of Pressure Vessels and Piping 113 (2014) 15e24 19

Fig. 6. Solid FE models used to calculate the SCFs: (a) plain pipe, (b) pit CWP3, and (c) patch LP2.

6. Global Analysis was composed of 4 load-steps [14] in such a way as to represent


both the start-up operation (internal pressure application followed
6.1. Pipe model by the temperature loading application) and the shut-down oper-
ation (internal pressure removal or depressurization of the pipeline
The Global Analysis was carried out using ABAQUS [12] and followed by the temperature loading removal or pipeline cooling).
PATRAN [13] programs. The pipeline was represented by a 1.0 m- Within each step, the loading was incremented using an auto-
long pipe segment. The stress and strain components are refer- matic time stepping algorithm, (only the initial/final values and
enced to a local coordinate system with the following convention: minimum/maximum increment limits were provided).
x-axis (longitudinal), y-axis (circumferential) and z-axis (radial). The internal pressure was applied considering the pipeline in-
A single PIPE21 available in ABAQUS FE library was used to ternal diameter as a reference.
model the 1.0 m long pipe segment. This element has two nodes
with 3-d.o.f. per node. 4,00
Both extreme nodes were considered clamped. Due to this re- Patch - (SCF)L
striction on the pipeline movement, and also the pipeline straight Patch - (SCF)h
geometry, the soil has no influence on the stress analysis. Pit - (SCF)L
3,00
As the Global Analysis is elastic, the material behavior is char-
Pit - (SCF)h
acterized only by the Young’s modulus and Poisson’s ratio.
SCF

Furthermore, the analysis is considered to be nonlinear geo-


metric (NLGEOM optional parameter).
2,00

6.2. Basic cyclic loadings

The basic loadings are made up of the internal pressure p and 1,00
temperature loading DT variations related to the pipeline inter- 0 0,1 0,2 0,3 0,4 0,5
mittent operation with hot heavy oil. d /t
In the numeric Global Analysis with ABAQUS, the basic loadings
were applied throughout 4 cycles as depicted in Fig. 8. Each cycle Fig. 7. Stress concentration factors of the corrosion defects.
20 D.J.S. Cunha et al. / International Journal of Pressure Vessels and Piping 113 (2014) 15e24

Table 4
Stress concentration factors of the corrosion pits CWPs.

Defect d/t (SCFL)p (SCFL)DT SCFLa SCFh

CWP0 0.00 1.000 1.000 1.000 1.000


CWP1 0.10 1.707 1.568 1.638 1.601
CWP2 0.20 2.099 1.867 1.983 1.901
CWP3 0.30 2.214 2.094 2.154 2.005
CWP4 0.40 2.284 2.258 2.271 2.086
CWP5 0.50 2.370 2.398 2.384 2.162
a
The SCFL mean value (see Eq. (2)) was applied to the longitudinal stresses due to
both basic loadings (internal pressure and temperature loading). Fig. 8. Basic cyclic loadings (out-of-phase/NP constant-amplitude pressure and tem-
perature variations due to the start-up and shut-down operations).

6.3. Nominal stresses


cycle repeats itself on all subsequent cycles (see Fig. 9), only one
In a pressurized, sufficiently buried pipeline, the longitudinal cycle needs to be analyzed, naturally, the first one (see Fig. 11).
strain is nil because of the soil imposed restriction. Under these It should be mentioned that the calculated fatigue life is the
boundary conditions, a hot pipeline develops only membrane same irrespective of whether the complete stress history is input
stresses: one circumferential sh and other longitudinal sL, the latter (as in Fig. 11) or only their inflection points (stresses at times 0, 1, 2,
resulting from the sum of the longitudinal stresses (sL)P and (sL)DT 3 and 4) are provided.
caused by the longitudinal displacement restraint. For a corrosion defect, the nominal/elastic stresses are amplified
The pipeline nominal stresses during the first three cycles are by the SCFs according to the following expression:
shown in Figs. 9 and 10. Similarly to the basic loadings, the longi-
si ¼ SCFi ,ðsi Þnom ; i ¼ L; h (3)
tudinal and circumferential stress components are out-of-phase
and have constant amplitudes. The SCFs of a corrosion defect may be applied either before the
The points B and D in Fig. 10 correspond to the operating pres- fatigue analysis, with the manual calculation of the Eq. (3), or while
sure application (p ¼ poper) and removal (p ¼ 0) respectively. The entering the fatigue loading (nominal stresses) to the fatigue soft-
points C and A correspond to the temperature loading application ware [8]. The second form was adopted in this study due to the
(DT ¼ DTmax ¼ Toper  Tinst) and removal (DT ¼ 0) respectively. simplicity of considering only one corrosion defect at a time and
Some typical nominal stress values acting on the pipeline during because it allows the application of a different SCF to each stress
a cycle are shown in Table 6. They correspond to the vertices of the component. Moreover, by taking different corrosion defects in the
lozenge in Fig. 10 and/or to the times 1, 2 and 3. Due to the same pipeline, as the nominal stresses are the same, only the SCFs
simplicity of the model, these values could also be obtained values need to be changed in the load definition file (*.ldf) for each
analytically (see Table 7). defect fatigue analysis.
As the stresses are elastic and the soil has no effect on their In this way, following the nomenclature shown in Fig. 12, the
variations with time, the first cycle is simply repeated throughout nominal stress history is given as two signals, and the corre-
the analysis. sponding stress tensor/datasets dsi h [sxx syy szz syz szx], i ¼ L, h, are
defined as unit tensor/datasets whose components are nil, except
7. Fatigue analysis that related to each signal, which is assumed to be equal to 1. Both
signals and unit tensor are provided in two different ASCII files
The fatigue analysis was carried out using FE-SAFE program [8]. (*.txt).
The fatigue analysis phase is characterized by the following steps:
(a) fatigue loading (nominal stresses) reading; (b) stress amplifi- 7.2. Stressestrain curves
cation by the SCFs; (c) plasticity correction (Neuber’s rule) using the
static and cyclic true stressestrain curves and finally (d) the fatigue 7.2.1. Elastic behavior
life/damage calculation using a multiaxial strain-life method Before the plasticity correction, using the multiaxial versions
together with a uniaxial εeN curve. These steps are described of the stressestrain relationship for elastic behavior [3,7,15] and
below.
400
7.1. Fatigue loading and application of SCFs Mises
(σL)nom
(σh)nom
In this analysis, the fatigue loading is given by the nominal
longitudinal and circumferential stress histories. As the first stress 200
Stress (MPa)

Table 5
Stress concentration factors of the corrosion patches LPs.
0
Defect d/t (SCFL)p (SCFL)DT SCFLa SCFh

LP0 0.00 1.000 1.000 1.000 1.000


LP1 0.10 1.719 1.577 1.648 1.769
LP2 0.20 2.152 1.847 2.000 2.445
LP3 0.30 2.520 2.104 2.312 2.902 -200
LP4 0.40 2.988 2.316 2.652 3.465 0 1 2 3 4 5 6 7 8 9 10 11 12
LP5 0.50 3.484 2.527 3.006 4.046
Time
a
The SCFL mean value (see Eq. (2)) was applied to the longitudinal stresses due to
both basic loadings (internal pressure and temperature loading). Fig. 9. Nominal stress history in the first three cycles.
D.J.S. Cunha et al. / International Journal of Pressure Vessels and Piping 113 (2014) 15e24 21

300 Table 7
1st cycle Equations of some typical nominal-stress values during a loading cycle in a buried
pipeline (see Figs. 10 and 11, and Table 6).
2nd cycle C (t = 2) B' B (t = 1)
Point Longitudinal stress (sL) Hoop stress Von Mises equivalent stress (seq)
Hoop stress (MPa)

200 A 0 0 0
B vsh sh vsh a
B0 0 sh sh
C EaDT þ vsh sh ½ðEaDTÞ2 þ vsh 1=2 b
D EaDT 0 EaDT
pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
100 a
v ¼ 1  v þ v2 y0:889 < 1/seq ðB0 Þ > seq ðBÞ.
b
v ¼ EaDTð1  2vÞ þ v2 sh > 0/seq ðCÞ > seq ðDÞ.

D (t = 3) A (t = 0)
7.3. Strain-life method
0
-200 -150 -100 -50 0 50 100
The uniaxial εeN curve is defined by the CoffineManson (CM)
Longitudinal stress (MPa)
equation:
Fig. 10. Nominal stresses according to the start-up and shut-down loading sequence.
Dε s0f
¼ ð2NÞb þ ε0f ð2NÞc (6)
2 E
the amplified nominal principal stresses (see Figs. 11 and 12), the
The fatigue life for a multiaxial strain parameter can be calcu-
corresponding nominal principal strains are calculated (see
lated by modifying the right-hand side of Eq. (6) appropriately [8],
Fig. 13). Note that the strains are triaxial (ε3 ¼ [v/
i.e., keeping the same general format and the same material con-
(1  v)](ε1 þ ε2)) while the stresses are biaxial (s3 ¼ sz ¼ 0). Both
stants (E, s0f ; ε0f , b, c). In this study, the fatigue lives of the corrosion
principal stresses and principal strains in the FE-SAFE output file
defects were obtained using the multiaxial BrowneMiller (BM)
(*.log) are elastic.
algorithm [3e5,8]:
7.2.2. Plasticity correction
Dgmax Dεn s0f
In the absence of experimental data, the static stressestrain þ ¼ 1:65 ð2NÞb þ 1:75ε0f ð2NÞc (7)
curve (uniaxial curve) was estimated from the SMYS and SMTS 2 2 E
stresses using the RambergeOsgood equation (see Fig. 14): It’s worth noting that, under uniaxial conditions (ε3 ¼ ε2 ¼ vε1),
the multiaxial BM equation (Eq. (7)) produces the same fatigue life
strue s 1=h as does the uniaxial CM equation (Eq. (6)) itself [8].
true
εtrue ¼ þ (4) The BrowneMiller equation proposes that the maximum fatigue
E H
damage occurs on the plane which experiences the maximum
and assuming the engineering ultimate strain to be half of the shear strain amplitude, and the damage is a function of both this
elongation. According to API Spec 5L [16], for X60 steel, the elon- shear strain gmax and the strain εn normal to this plane (see Fig. 15).
gation is 22%. As mentioned in Fig. 14, the following values were According to Refs. [3,4], this is an attractive fatigue criterion
obtained for the hardening parameters: H ¼ 690 MPa and h ¼ 0.08. because it uses standard uniaxial material properties and also gives
Similarly, the cyclic stressestrain curve was defined as: the most realistic life estimates for ductile metals [3,4].
When the principal stresses/strains are out-of-phase/non-
 1=hc proportional (NP), a critical plane (CP) technique is used. In the
strue strue
εtrue ¼ þ (5) biaxial case, the maximum shear planes are rotated round the ε3-
E Hc
axis, which is normal to the surface, through an angle f
Due to the lack of experimental data, in this study, the cyclic (0  f  180 ) varying typically in 10 increments (see Fig. 15). The
curve was taken to be the same as the static material curve, i.e., plane with the highest calculated damage is the critical plane, and
Hc ¼ H and hc ¼ h (see Fig. 14). That is, neither a hardening benefit
nor a detrimental softening was taken into account.
The amplified nominal stress/strain plasticity correction, which 400
(σL)nom
is an integral part of FE-SAFE, is based on a multiaxial approach
using Neuber’s rule [3e5,8,17,18]. In this process, the cyclic stresse (σh)nom

strain curves are modified to allow for the effect of biaxial stresses
200
Stress (MPa)

[3,15].

Table 6
Typical FE nominal stress values acting on the pipeline during a loading cycle (see
0
Figs. 10 and 11).

Point Time Stress (MPa)

B, B0 , C 1-2 Circumferential stress sh due to 228.48


the pressure poper -200
B 1 Longitudinal stress (sL)p due to 68.54 0 1 2 3 4
the pressure poper
Time
D 3 Longitudinal stress (sL)DT due to 144.61
the temperature loading DT ¼ Toper  Tinst
Fig. 11. Basic fatigue loading history (nominal stress components of the first cycle, also
C 2 Total longitudinal stress sL 76.07
identified in FE-SAFE [8] as the in-plane principal stresses (s1)nom h (sh)nom and
C 2 Von Mises equivalent stress seq 274.54
(s2)nom h (sL)nom).
22 D.J.S. Cunha et al. / International Journal of Pressure Vessels and Piping 113 (2014) 15e24

800

600

True stress (MPa)


400

200
Static (h=0.08, H=690 MPa)
Fig. 12. Amplification of the stresses during the fatigue loading history reading. In
order to apply different SCFs to the stress components, these were given as signals, and Cyclic (hc=0.08, Hc=690 MPa)
the datasets dsi h [sxx syy szz syz szx], i ¼ L, h, defined as unit datasets:
0
dsL ¼ [1 0 0 0 0 0] and dsh ¼ [0 1 0 0 0 0].
0 0,02 0,04 0,06 0,08 0,1 0,12
True strain (mm/mm)
the calculated damage on this plane determines the fatigue life of
Fig. 14. Estimated uniaxial stressestrain curves (see Eqs. (4) and (5)).
the structure being analyzed.
Moreover, under the condition of NP loadings the plasticity
correction is carried out using an incremental Neuber’s rule [3e
n
5,8,18] in terms of deviatoric stressestrain combined with a D ¼ nDi ¼  afat (10)
multiaxial cyclic plasticity model, i.e., kinematic hardening model, Ni
together with multiaxial stressestrain relations.
where the fatigue usage factor is given by:
In the absence of experimental data, the uniaxial fatigue pa-
rameters were estimated by adjusting the CM equation (Eq. (6)) to 1
an adequate existing εeN curve. In particular, the fatigue strength afat ¼ (11)
DFF,EFF
coefficient was estimated as [19]:
Alternatively, introducing D0 ¼ (afat)1D and N0 ¼ afatN with
s0f ¼ 1:5su (8) N h Ni, the criterion given by Eq. (10) can be rewritten as:

The proposed methodology uses the ASME best-fit curve [20e n n


D0 ¼ ðDFF,EFFÞ ¼ 01 (12)
22] which is, in fact, a strain-life (εeN) uniaxial curve [21e23]. N N
For the API X-60 steel (su ¼ 517 MPa) it can be estimated by Eq. (6)
with the following constant values (see Fig. 16):
8. Results
s0f ¼ 775:5 MPa; b ¼ 0:14; ε0f ¼ 0:31; c ¼ 0:48
(9) The fatigue life of the pipeline was calculated under the condi-
tions given in Table 1. The geometric characteristics and the SCFs of
The endurance limit was assumed to be 1  1015 cycles or the corrosion defects considered in this analysis (five pits and five
2  1015 reversals (half-cycles). patches) are shown in Tables 2e5. Although the metal loss due to
corrosion is a time dependent process, it was assumed that the
corrosion defects exist since the start of the pipeline operation and
7.4. Fatigue damage
their dimensions did not vary with time.
Fatigue cracks usually initiate from the body surface. In the case
In general, the fatigue damage is supposed to be calculated using
of a pipe, this can be on the outer or on the inner wall surfaces.
the PalmgreneMiner rule [3,5,8,15]. In this study, as the fatigue
However, assuming that the pipe is a thin shell submitted only to
loading amplitudes are constant the damage was simply calculated
pressure and temperature loadings, the stresses are the same at any
as:
pipe radial surface and so there is not this distinction in terms of
which surface cracks initiate.
2000
The absolute value of the elastoplastic BM-parameter amplitude
(ε1)nom
e/2, identified as the left-hand side of Eq. (7), and the corre-
(ε2)nom sponding fatigue life N are shown in Figs. 17 and 18, respectively, as
Principal strain (μ-strain)

1000
(ε3)nom functions of the angle f related to rotation of principal/shear planes
round the principal axis ε3 (ε3-axis h z-axis), mentioned in Section
0

-1000

-2000
0 1 2 3 4
Time

Fig. 13. Principal strains of the first cycle (triaxial strains) for the plain pipe (defect
CWP0 or defect LP0). In this particular case, as the SCFi ¼ 1.0, i ¼ L, h, these strains
coincide with the nominal principal strains related to the nominal stresses in Fig. 11. Fig. 15. BrowneMiller (BM) critical plane (CP) method.
D.J.S. Cunha et al. / International Journal of Pressure Vessels and Piping 113 (2014) 15e24 23

1000 5,E+07
ASME best-fit curve [20–22] plane 1-2
Coffin–Manson adjusting (Sf'=775.5 MPa, b=–0.14, ef'=0.31, c=–0.48) plane 1-3
100 4,E+07
Corresponding Brown–Miller equation with parameter e/2 (Eq. (7)) plane 2-3
Strain amplitude (%)

Life N (repeats)
10 3,E+07

1 2,E+07

0,1 1,E+07

0,01 0,E+00
1,E+01 1,E+02 1,E+03 1,E+04 1,E+05 1,E+06 1,E+07 1,E+08 0 30 60 90 120 150 180
2N (half-cycles) "C/P Ang" or angle Ф (degree)

Fig. 16. Uniaxial strain-life curve adopted in the fatigue analysis (CoffineManson Fig. 18. Fatigue life of the loading history throughout the rotation of the three shear
adjusting of the ASME best-fit curve). The multiaxial BM-curve (Eq. (7)) uses the same planes round the ε3-axis for the plain pipe (defect CWP0 or defect LP0). Lives corre-
uniaxial constants. spondent to the BM-parameters (|e|/2) shown in Fig. 17.

7.3. In fact, the elastic and elastoplastic values of the BM-parameter/ the first principals. The only interval, where the fatigue loadings
strain are available at both extremes of the range/cycle (e1 and e2) in were proportional, was the first load step related to the internal
function of the angle f, so that the user can calculate |e|/ pressure application, when sL ¼ (sL)p ¼ nsh (see Table 7 and Fig. 10).
2 h je2  e1j/2. According to both these figures, the BM-parameter In a defect free pipe subjected to internal pressure and an axial
is maximum and/or the fatigue life is minimum at f ¼ 30 on the loading or temperature, the stress/strain principal axes are always
plane 1e2. in the pipe longitudinal (axial) and circumferential (hoop) di-
These results mean that the likely cracks will originate at a plane rections regardless of the phase between these loadings [5]. In this
(critical plane) normal to the pipe surface (case A shown in Fig. 15) study, the fatigue loadings (biaxial stresses) are non-proportional
and whose normal vector εn makes a 30 angle with the reference due to the out-of-phase nature of the basic cyclic loadings.
principal axis s*1 related to the stress tensor s* taken as a reference The fatigue life and fatigue damage for the corrosion pits and the
[8] to define the surface orientation. In general, s* is defined as the corrosion patches are shown in Tables 8 and 9, respectively. The
stress tensor, within the stress history, with the largest principal largest elastoplastic values of the BM-parameter/strain amplitude
stress or, if the other two principals at this sample are negligibly e/2 were also included in these tables. For both types of corrosion
small, the one with at least two significant principals [8]. In this defects the larger the defect the higher the fatigue damage. As
study, the stress tensor of time t ¼ 1 (the end of the first load step previously mentioned, the “corrosion defects” named CWP0 and
with the internal pressure totally applied), whose first principal LP0 are in fact the defect free pipe (design condition).
axis coincides with the pipe circumferential (hoop) direction, was As shown in Table 8 and Fig. 19, considering the fatigue damage
chosen as a reference. acceptance criterion D  0.1 from Eqs. (10) and (11), all corrosion
Therefore, the critical plane itself makes a 30 angle with the pits were accepted for more than 40 years, even the one which has
pipe longitudinal axis (the reference second principal axis s*2 ) and/ the maximum depth (CWP5). However, the same doesn’t apply in
or, equivalently, its normal makes a 120 angle with the pipe lon- the case of patches. As shown in Table 9 and Fig. 19, only the
gitudinal axis (x-axis) in accordance with the angle fx. corrosion patches LP1 and LP2 were accepted for more than 40
Moreover, the fatigue loading history (longitudinal and years. The corrosion patches LP3, LP4 and LP5 violate the fatigue
circumferential stresses shown in Fig. 11) were all classified as damage acceptance criterion slightly above 26 years, 13 years and 6
“Non-proportional (constant direction principals)”, that is, q1 ¼ 0 at years respectively.
any time, and the circumferential stresses/strains were identified as

9. Conclusions
0,14
plane 1-2 A methodology for the fatigue life assessment of hot pipelines
0,12 plane 1-3 containing corrosion defects in the base material was presented in
|Cycle--Ampl| or |e|/2 (%)

plane 2-3 this paper. The general procedure includes three main phases:
0,10 Global Analysis of the pipeline represented by a one-dimensional

0,08
Table 8
0,06 Fatigue damage of the corrosion pits for the pipeline design life (40 years).

Defect d/t e/2 (%)a N (repeats)b D ¼ n/Nb,c


0,04
CWP0 0.00 0.1202 1.48Eþ06 0.0042
CWP1 0.10 0.1932 2.77Eþ05 0.0225
0,02 CWP2 0.20 0.2307 1.58Eþ05 0.0395
0 30 60 90 120 150 180 CWP3 0.30 0.2450 1.31Eþ05 0.0476
CWP4 0.40 0.2562 1.14Eþ05 0.0547
"C/P Ang" or angle Ф (degree)
CWP5 0.50 0.2671 1.00Eþ04 0.0624

Fig. 17. BM-parameter amplitude throughout the rotation of the three shear planes a
Largest elastoplastic values, critical plane 1e2 with f ¼ 30 or fx ¼ 120 .
b
round the ε3-axis for the plain pipe (defect CWP0 or defect LP0), after plasticity N ¼ N h 1/Di due to loading history be consisted of a single cycle.
c
correction. n ¼ 3 cycles/week  52 weeks/year  40 years ¼ 6240 cycles.
24 D.J.S. Cunha et al. / International Journal of Pressure Vessels and Piping 113 (2014) 15e24

Table 9 together with the plasticity, which may occur in the corrosion
Fatigue damage of the corrosion patches for the pipeline design life (40 years). defect, the fatigue loading was reduced to only one nominal stress
Defect d/t e/2 (%)a N (repeats)b D ¼ n/Nb,c history, and the fatigue analysis of the various corrosion defects
LP0 0.00 0.1202 1.48Eþ06 0.0042
required only the changing of the SCFs values.
LP1 0.10 0.2103 2.11Eþ05 0.0296 Finally, it’s worth pointing out that this is a purely theoretical
LP2 0.20 0.2905 7.86Eþ04 0.0794 study and testing is required to validate the approach.
LP3 0.30 0.3615 4.23Eþ04 0.1475
LP4 0.40 0.4702 2.08Eþ04 0.3000
LP5 0.50 0.6040 1.08Eþ04 0.5778
Acknowledgments
a
Largest elastoplastic values, critical plane: 1e2 with f ¼ 30 or fx ¼ 120 .
b
N ¼ N h 1/Di due to loading history be consisted of a single cycle.
c
n ¼ 3 cycles/week  52 weeks/year  40 years ¼ 6240 cycles. The authors would like to thank PETROBRAS for the permission
to publish this paper.

0,60
Pits References
0,50 Patches [1] Eiden H, Mackeinstein P. Safe service life analysis for pipelines e an engi-
neering method with various applications. In: Proc. of international confer-
0,40 ence on pipeline inspection, Edmonton, Alberta, Canada, CANMET CDD 621
86720 631; 1983. p. 583e97.
Damage

0,30 [2] Chuahan V, Swankie TD, Espiner R, Wood I. Developments in methods for
assessing the remaining strength of corroded pipelines; 2009. NACE paper
09115, NACE corrosion.
0,20 [3] Fe-safe fatigue theory reference manualFE-SAFE user manual, vol. 2. UK: Safe
Technology Limited; 2006. version 5.2.
[4] Draper J. Metal fatigue e failure and success, Journée Scientifique, Les Méth-
0,10
odes de Dimensionnement en Fatigue; 27 octobre, 2004.
[5] Socie DF, Marquis GB. Multiaxial fatigue. Warrendale, PA, USA: Society of
0,00 Automotive Engineers; 2000.
0 1 2 3 4 5 [6] Palmer-Jones R, Turner TE. Pipeline buckling, corrosion and low cycle fatigue.
In: OMAE98-0905, 17th international conference on offshore mechanics and
Corrosion defect article engineering, Lisbon, July 5e9, 1998.
[7] Technical note e biaxial fatigue. UK: Safe Technology Limited; 2003.
Fig. 19. Fatigue damage of the corrosion pits and patches for the pipeline design life: [8] Fe-safe user manual, vol. 1. UK: Safe Technology Limited; 2006. version 5.2.
40 years (see Tables 8 and 9). [9] Maksimovic S. Fatigue life analysis of aircraft structural components. Sci Tech
Rev 2005;LV(1).
[10] Anon. Pipeline transportation systems for liquid hydrocarbons and other
liquids e a supplement to ASME B31 code for pressure piping. New York: The
plain pipe model; nominal stress amplification by SCFs obtained American Society of Mechanical Engineering; 2009.
with solid FE models, and strain-life calculation. [11] Ansys engineering analysis system: user’s manual, version 8.1. ANSYS, Inc.;
The amplified stresses are elastic and may exceed the yielding 2004.
[12] Hibbit HD, Karlson BI, Sorensen P. ABAQUS documentation, version 6.6-EF.
limit. Also, due to the out-of-phase/non-proportional (NP) nature of
Pawtucket, RI 02860-04847: Hibbit, Karlson and Sorensen Inc.; 2006.
the applied loadings and the pipe cylindrical geometry, the stresses [13] MSC.Patran user’s guide, version 2005 r2. Santa Ana, CA 92707, USA:
are out-of-phase/NP and their principal directions do not change. MSC.Software Corporation; 2005.
Under such conditions, the plasticity correction is performed by [14] Klever FJ, Palmer AC, Kyriakides S. Limit-state design of high-temperature
pipelines, OMAE 1994. Pipeline Technol 1984;V:77e92.
applying a multiaxial approach using Neuber’s rule, and the fatigue [15] Dowling NE. Mechanical behavior of materials. 2nd ed. New Jersey: Prentice-
life is calculated by using an εeN method and the critical plane Hall; 1999.
technique. [16] API specification 5L e specification for line pipe. 42nd ed. USA: American
Petroleum Institute; January 2000. Effective Date July 2000.
The multiaxial BrowneMiller algorithm and the uniaxial εeN [17] Lemaitre J, Chaboche J-L. Mechanics of solid materials. Cambridge University
ASME best-fit curve were chosen for the fatigue life calculations. Press; 1990.
The proposed methodology was applied to an onshore pipeline [18] Buczynski A, Glinka G. An analysis of elasto-plastic strains and stresses in
notched bodies subjected to cyclic non-proportional loading paths. In:
containing corrosion defects on its external surface. Five corrosion Carpinteri A, de Freitas M, Spagnoli A, editors. 6th International conference on
pits and five corrosion patches, with different sizes, were consid- biaxial/multiaxial fatigue and fracture. Lisbon, Portugal: ESIS Publication 31,
ered in this analysis. The fatigue results (life and/or damage of all Elsevier; 2003. p. 265e83. 2001.
[19] Meggiolaro MA, Castro JTP. Statistical evaluation of strain-life fatigue crack
corrosion defects) showed that all pits and only the two smaller
initiation predictions. Int J Fatigue 2004;26:463e76.
patches could be accepted for more than 40 years (6240 cycles or [20] ASME boiler and pressure vessel code, section VII, division II, appendix R:
start-up/shut-down operations). The other three corrosion patches mandatory design based on fatigue analysis. American Society of Mechanical
Engineers; 2004.
would be approved up to just over 26, 13 and 6 years respectively.
[21] Langer BF. Design of pressure vessels for low-cycle fatigue. J Basic Eng Trans
This means that fatigue becomes an important failure mode when ASME September, 1962:389e402.
corroded pipeline segments are left in operation without being [22] Rahka K. Review of strain state effects on low-cycle fatigue of notched com-
replaced. ponents, vol. 263. PVP; 1993. p. 185e95. High pressure e codes, analysis, and
applications, ASME.
It should be noted that despite the complexity related to the [23] Branco CM, Fernandes AA, Castro PMST. Fatigue of welded structures. Lisbon:
multiaxial stress/strain and its out-of-phase/NP characteristics Fundação Calouste Gulbekian; 1987 [in Portuguese].

You might also like