You are on page 1of 8

M.

Bony Failure Pressure in Corroded


Institut Français de Mécanique Avancée,
Campus de Clermont-Ferrand,
France
Pipelines Based on Equivalent
Solutions for Undamaged Pipe
J. L. Alamilla1
e-mail: jalamill@imp.mx Simple and accurate approaches to predict failure pressures in corroded pipelines are
outlined in this work. It is shown that failure pressures for corroded pipelines can be
R. Vai predicted from the solution for undamaged pipelines using an equivalent wall thickness.

Downloaded from http://asmedigitalcollection.asme.org/pressurevesseltech/article-pdf/132/5/051001/5751587/051001_1.pdf by University of Edinburgh user on 22 March 2023


Three different yield criteria (Tresca, ASSY (average shear stress yield), and von Mises)
Instituto Mexicano del Petróleo, are reviewed in the light of reported experimental burst pressures. At first, failure pres-
Eje Central Lazaro Cardenas Norte No. 152, sures for cylindrical vessels with an infinitely long groove are studied by means of
07730, México, D.F., México numerical simulations. The effect of groove size (depth and width) over the pipeline
performance is quantified through a model. Finally, the scheme is extended to estimate
the failure pressure of thin walled vessels with irregular finite defects.
E. Flores 关DOI: 10.1115/1.4001801兴
Instituto Politécnico Nacional,
ESIA-Zacatenco, Keywords: failure pressure, infinitely long groove, burst test, damaged and undamaged
México pipelines

1 Introduction ure pressures associated with different groove lengths greater than
a pipe diameter do not vary. Therefore, these kinds of defects are
Pipeline systems are commonly used for oil and gas transpor-
called infinitely long grooves. Mok et al. 关4兴 and Fu and Kirk-
tation because they are economic and safe. However, their integ- wood 关12兴 performed finite element numerical analyses of pipe-
rity is affected by corrosion damages, which can originate failures. lines affected by infinitely long groove defects. They documented
These failures can imply health consequences for people in the that burst pressure values did not change significantly as a func-
pipeline neighborhood, environmental pollution, and economic tion of groove width. The failure pressure of a pipeline with an
losses. Security levels must be established on a reference frame- infinitely long groove represents a lower limit for the burst pres-
work in which a trade-off between safety and consequences has to sure of a vessel with an irregular defect 关10,5,11兴. Note that cor-
be done. To achieve this goal, effective integrity assessment is of roded pipes generally present defects with an irregular geometry.
primary concern. In this case, numerical methods, mainly finite element method, are
According to the plasticity theory 关1,2兴, it is accepted that a very useful tools to describe the pipeline performance in corroded
pipeline fails by pressure if the stress field in the wall thickness is zones 关8,10,4,13兴. Nevertheless, analytical approaches are avail-
proportional to the maximum stress of a tensile stress-strain rela- able 关8,10兴. In engineering applications, semi-empirical formulas
tion 关3–5兴. Such proportionality is described by a yield criterion 共e.g., B31G 关14兴 and RSTRENG 关15兴兲 are used to predict failure
specified in terms of effective stresses. The stress field is ex- pressures.
pressed in terms of its corresponding internal failure pressure and Simplicity and accuracy are needed to evaluate pipeline integ-
can be estimated analytically 关2–7兴. On the basis of Svensson’s rity. Numerical methods give precise results, but they require a
formulation 关3兴, using von Mises’s yield criterion, Cronin and great effort to obtain them. Practical formulas are very simple to
Pick 关8兴 assessed that the mean analytical failure pressure overes- apply; however, errors can be significant. The scheme presented in
timates the corresponding mean obtained from experimental pipe- this work is easy to use; still, results are accurate. Burst pressures
line tests. Therefore, they recommended to multiply the analytical are assessed by means of mechanical models suitable for predict-
solution by a scale factor of 0.86⬇ 冑3 / 2, which let us presume ing failure pressures of undamaged pipelines. Estimating the fail-
that Tresca’s criterion could be more realistic in estimating failure ure pressure of thin walled vessels with an irregular defect, a
pressures than other failure criteria. Recently, Zhu and Leis 关9兴 lower bound is given by the burst pressure of pipelines with an
proposed the ASSY 共average shear stress yield兲 criterion, which infinitely long groove, which in turn is bounded by the failure
corresponds to a statistical average of maximum effective shear pressure of undamaged pipes. Consequently, the first section of
stress of Tresca’s and von Mises’s. This new criterion was vali- this work is focused on the study of undamaged pipelines. Three
dated with 103 experimental tests. They reported that, in average, different yield criteria are evaluated in the light of experimental
the criterion matched up very closely with experimental results burst tests. Then, finite element method is used to study the rela-
关9兴. tionship among the failure pressure, width, and depth of infinitely
On the other hand, if the pipe wall thickness is affected by an long groove defects. A model is developed to estimate the burst
infinitely long groove or a corrosion defect with irregular geom- pressure of pipelines with an infinitely long groove. Subsequently,
etry, then other schemes have to be used 关8,10兴. By increasing the the idea is extended to predict the failure pressure of vessels with
length of a finite groove, the burst pressure asymptotically ap- finite irregular defects; in particular, the approach developed by
proaches the failure pressure of a pipe with an infinitely long Oliveros et al. 关10兴 is retrieved. The results of the developed mod-
groove. Hopkins and Jones 关11兴 experimentally showed that fail- els are also compared with reported experimental burst pressures.

1
Corresponding author.
Contributed by the Pressure Vessel and Piping Division of ASME for publication 2 Burst Pressure of Undamaged Pipeline
in the JOURNAL OF PRESSURE VESSEL TECHNOLOGY. Manuscript received October 20,
2008; final manuscript received June 10, 2009; published online August 20, 2010. On the basis of Svensson’s 关3兴 work, the failure pressure of a
Assoc. Editor: Shawn Kenny. thin walled undamaged pipeline can be obtained analytically as

Journal of Pressure Vessel Technology Copyright © 2010 by ASME OCTOBER 2010, Vol. 132 / 051001-1
Downloaded from http://asmedigitalcollection.asme.org/pressurevesseltech/article-pdf/132/5/051001/5751587/051001_1.pdf by University of Edinburgh user on 22 March 2023
Fig. 2 Normalized analytical burst pressure versus strain
Fig. 1 Experimental failure pressures versus predicted failure hardening coefficient
pressures for Tresca, ASSY, and von Mises yield criteria

t0 mean errors are 1.126, 1.044, and 0.963 for von Mises, ASSY, and
pplain = k␴U 共1兲 Tresca, respectively, whereas the coefficient of variation is equal
r0 to 0.042, regardless of the considered criterion. The statistical
A pipeline is considered thin walled if the ratio of wall thickness variability of the error is small. A continuous line and two couples
to diameter is approximately less than 0.05 关3兴. Radial stresses can of lines are shown in Fig. 1. The continuous line represents the
be neglected since they are much lower than tangential ones. In ideal match, i.e., model prediction equal to the experimental re-
Eq. 共1兲, t0 is the initial wall thickness 共original, undeformed wall sult. The first couple of lines showed as dotted lines defines the
thickness兲, and r0 is the inner initial radius. As shown in Appendix ideal pressure value of ⫾1 standard deviation 共0.042兲. The second
A, the true stress-strain relation is defined by the parameters ␴0, n, couple denoted as dot-dashed lines corresponds to the ideal pres-
and ␴U = ␴0nn. Here, k is a constant, and its value depends on the sure value of ⫾2 standard deviations. As shown in Fig. 1, the
yield criterion: failure pressure results related to both ASSY and Tresca’s criteria

冦 冧
are between the mean ⫾2 standard deviations. Burst pressures
2 using von Mises’s yield criterion are above the upper dot-dashed
von Mises
e 共冑3兲n+1
n line, which means that pressure values obtained from von Mises’s

k=
1
共2e兲 n 冉 冊
2 + 冑3
2 冑3
n+1
ASSY 共2兲
criterion are inaccurate.
Herein, the predicted pressure values are associated with deter-
ministic mechanical properties, which were obtained from pipe
1 measurements and from experimental tests. These mechanical
Tresca properties are expected to be slightly different from those of the
共2e兲n defect zone where the pipe failed. A detailed analysis about the
In order to compare failure pressures obtained from Eq. 共1兲 and influence of the statistical variability of each mechanical property
experimental burst pressures, a sample of 46 tests of undamaged over the failure pressure is beyond this work.
pipelines was used. This sample corresponds to 8 tests from Cro- Figure 2 shows the ratio of the predicted to experimental failure
nin 关5兴, 2 from Mok et al. 关4兴, 9 from Papka et al. 关16兴, 8 from pressures as a function of the strain hardening exponent n. If this
Liessem et al. 关17兴, 12 from Amano et al. 关18兴, 2 from Vieth and ratio is less than 1, then the model underpredicts the real burst
Kiefner 关19兴, and 5 from Law 关20兴. Since the tests come from pressure. On the contrary, if the ratio is more than 1, then the
different authors, the nature of the data is heterogeneous, and no model overpredicts the real burst pressure. In Fig. 2, a clear trend
set dominates or can statistically bias the results. Moreover, the can be identified; ratios increase with n. For n values less than
sample is heterogeneous with respect to mechanical and geometri- 0.10, the ASSY criterion may lead to more accurate predictions.
cal properties. Steel grades range from API 5L X-42 to API 5L For n values greater than 0.10, Tresca’s criterion may lead to more
X-120, and outer diameters range from 273.0 mm to 1219.2 mm. accurate results. The same conclusions can be drawn from the
Information of five tests were incomplete as the n value was not results presented in the work of Christopher et al. 关21兴; however,
available. In these cases, Eq. 共A3兲 of Appendix A was used to they worked with thick walled cylinders. This issue requires fur-
estimate the missing parameter. ther research.
Figure 1 shows the experimental burst pressure values versus From previous results, von Mises’s yield criterion may lead to
the corresponding predicted pressures obtained from Eq. 共1兲. Dif- inaccurate failure pressure predictions of undamaged pipes.
ferent symbols were used to distinguish yield criteria. In agree- Therefore, it will not be considered in the following sections. On
ment with Refs. 关8,9兴, failure pressures associated with von Mis- the other hand, ASSY’s and Tresca’s criteria are statistically
es’s criterion overestimate the corresponding experimental equivalent. Both will be used and compared in the following.
pressures. The burst pressures related to the ASSY criterion are
among Tresca and von Mises results, coherently with the defini- 3 Burst Pressure of Pipeline With Infinitely Long
tion of the ASSY criterion. No clear tendency can be identified to
establish whether the Tresca or ASSY yield criterion is more suit- Groove Defect
able to fit experimental results. The error, defined as the ratio of In this section, the relationship among failure pressures, me-
the predicted failure pressure to the corresponding experimental chanical characteristics, and long groove size is discussed.
burst pressure, was computed for each test and criterion. The Grooves are defined in terms of groove depth dmax and width w;

051001-2 / Vol. 132, OCTOBER 2010 Transactions of the ASME


Table 1 Material mechanical properties Table 2 Summary of finite element analyses

Ludwik parameters Depth, dmax


␴Y ␴U ␴0 1 / 8t0 1 / 4t0 1 / 2t0 3 / 4t0
Material 共MPa兲 共MPa兲 共MPa兲 N Numerical failure pressure
W 共MPa兲
API 5L X-42 350.90 506.20 641.33 0.105
API 5L X-52 389.00 575.80 757.80 0.139 Material API 5L X-42 steel
API 5L X-60 540.00 672.51 843.30 0.097 0 15.99 15.99 15.99 15.99
t0 / 2 14.97 13.02 8.70 4.32
t0 14.90 12.83 8.61 4.31
2t0 14.86 12.83 8.59 4.31
mechanical characteristics are expressed in terms of the pipeline 4t0 14.84 12.82 8.58 4.31
wall thickness t0 and the steel stress-strain relation. At the end of 8t0 14.79 12.78 8.55 4.30

Downloaded from http://asmedigitalcollection.asme.org/pressurevesseltech/article-pdf/132/5/051001/5751587/051001_1.pdf by University of Edinburgh user on 22 March 2023


this section, a simple model is proposed to quantify the effect of 40t0 14.45 12.49 8.25 4.21
such characteristics over the failure pressure. ␲D / 2 14.32 12.37 8.22 4.18
There are few experimental tests reported on cylindrical thin 3␲D / 4 14.10 12.19 8.08 4.06
walled vessels with infinitely long groove defects. Therefore, nu- ␲D 14.01 12.05 8.03 4.05
merical simulations were performed by means of finite element
method using the commercial computer program ANSYS 关22兴. To Material API 5L X-52 steel
0 17.49 17.49 17.49 17.49
have a reference calibration work, some of the geometrical 共D
t0 / 2 16.52 14.46 9.66 4.79
= 500 mm, t0 = 8 mm兲, and mechanical properties 共API 5L X-42, t0 16.44 14.27 9.54 4.79
X-52, and X-60兲 of pipes reported by Mok et al. 关4兴 were used. 2t0 16.40 14.24 9.53 4.79
The geometrical and mechanical characteristics of the studied 4t0 16.34 14.18 9.53 4.77
pipes are given in Table 1. 8t0 16.26 14.12 9.48 4.76
To perform numerical simulations, rigid displacements and de- 40t0 15.76 13.67 9.17 4.61
formations out of the plane, i.e., along the axis, were restricted to ␲D / 2 15.65 13.52 9.05 4.55
zero. In plane large deformations were considered, and plane ele- 3␲D / 4 15.49 13.34 8.88 4.49
ments were defined by eight nodes 共plane 183: solid quad 8, node ␲D 15.33 13.17 8.78 4.43
183兲 were used. In order to ensure the numerical convergence,
around the groove a refined mesh was used 共Fig. 3兲. The mesh Material API 5L X-60 steel
was selected after performing several simulations with different 0 21.74 21.74 21.74 21.74
mesh schemes. Horizontal displacement was constrained along the t0 / 2 20.34 17.49 11.62 5.82
t0 20.14 17.36 11.59 5.81
symmetry axis 共see Fig. 3兲. A field of uniform internal pressure
2t0 20.09 17.35 11.58 5.79
was applied around the pipeline section, and it was stepwise in-
4t0 20.03 17.31 11.57 5.79
creased during the numerical analysis. The plasticity was repre- 8t0 19.99 17.27 11.53 5.79
sented by the multilinear kinematic hardening model, which uses 40t0 19.57 16.87 11.28 5.67
the von Mises yield criterion and the Prandtl–Reuss flow rule. It ␲D / 2 19.48 16.77 11.21 5.63
was considered that failure pressure is reached when the equiva- 3␲D / 4 19.24 16.55 11.07 5.56
lent stress at each point along a radius is greater or equal to the ␲D 19.05 16.39 10.94 5.50
true maximum stress ␴U. In order to have a reference framework,
the experimental tests and numerical results given in Refs. 关4,5兴
were reproduced. The finite element model was adopted to per-
form numerical experiments and to study the relation between yield criterion. As can be seen in Fig. 4, the normalized burst
depth and groove width. The results of numerical simulations are pressure can be treated as independent of the steel material. How-
presented in Table 2 and in Figs. 4 and 5. ever, it is a function of the depth and width of the groove. In
In Table 2, for each numerical experiment, groove geometry is general, the normalized failure pressure values tend to decrease as
given along with calculated burst pressure. Failure pressures are the depth increases; meanwhile, they slightly decrease as the
normalized by the corresponding burst pressures of undamaged groove width increases. The reduction is more sensible for shal-
pipes, so that enlisted numerical values do not depend on any low grooves than for deep grooves. Normalized burst pressures
change suddenly for very narrow defects. This agrees with the

Fig. 4 Normalized numerical pressure plong groove versus the


normalized width of an infinitely long groove for different steel
Fig. 3 Detailed mesh scheme for a pipe with a groove defect materials

Journal of Pressure Vessel Technology OCTOBER 2010, Vol. 132 / 051001-3


Downloaded from http://asmedigitalcollection.asme.org/pressurevesseltech/article-pdf/132/5/051001/5751587/051001_1.pdf by University of Edinburgh user on 22 March 2023
Fig. 6 Gamma parameter for different groove depths versus
the width of an infinitely long groove for different steel
materials

冉 冊
Fig. 5 Normalized numerical pressure plong groove versus the
normalized depth of an infinitely long groove ⴱ dmax
pplain = pplain 1 − 共3兲
t0
Note the similarity with the approach of B31G 关14兴. It turns out that
B31G considers the pressure of an undamaged pipe with reduced
experimental results performed by Hopkins and Jones 关11兴. For wall thickness as a lower limit and that it does not take into
normalized groove widths approximately greater than 0.2, the nor- account the damage width.
malized failure pressure tends to be constant. In particular, it tends On the other hand, pplain is an upper bound for plong groove be-
toward the normalized burst pressure of an undamaged pipe with cause a damaged pipe theoretically cannot support a failure pres-
wall thickness t0 − dmax. For this reason, in engineering applica- sure beyond the corresponding pplain. Rigorously pⴱplain
tions, it is considered that groove width does not contribute sig- ⱕ plong groove ⱕ pplain. According to these statements and based on
nificantly to the failure pressure. In Fig. 5, the relationship be- the ideas of Cronin and Pick 关8兴, plong groove can be estimated as
tween normalized burst pressure and normalized groove depth is ⴱ ⴱ
illustrated. The continuous line represents normalized failure pres- plong groove = pplain + 共pplain − pplain 兲␥ 共4兲
sures for defect free pipes whose wall thickness is reduced by Here, ␥ = ␥共dmax , w兲 is a function of depth dmax and width w of the
dmax / t0 percent. If shallow grooves are considered, the difference
groove. The values of ␥ are in the interval between 0 and 1. The
among the results is less than 10%. By increasing groove depth,
first term of Eq. 共4兲 denotes the failure pressure of the undamaged
normalized failure pressure tends toward that of a defect free pipe
pipe with reduced wall thickness t0 − dmax, and the second term
with reduced wall thickness. Results confirm what was previously
denotes the contribution to the failure pressure of the remaining
stated: In engineering applications, it can be considered that
material located between t0 and t0 − dmax. Substituting Eq. 共3兲 in
groove width does not contribute significantly to the failure pres-
Eq. 共4兲, it can be stated that
sure.
For pipelines with an infinitely long corrosion defect, the failure
pressure can be estimated by the product of the burst pressure of
the undamaged pipe and the damage factor 1 − dmax / t0 关23兴; this is
plong groove = 冋
pplain 1 −
dmax
t0
共1 − ␥兲 册 共5兲

the idea taken in B31G 关14兴. Besides this model, Cronin 关5兴 pro- Defining the equivalent wall thickness tElong groove = t0 − dmax共1
posed another approximated approach to evaluate failure pres- − ␥兲 and considering Eq. 共1兲, then Eq. 共5兲 can be rewritten as
sures of pipes with infinitely long grooves. The latter scheme is
obtained assuming a deformed pipe configuration, along with suit- tElong groove
able hypotheses on forces, and performing a static equilibrium of plong groove = k␴U 共6兲
r0
forces. According to this author, the approach is less conservative
than the B31G 关5兴. The failure pressure is expressed in terms of the The failure pressure of a pipe with an infinitely long groove is
unknown critical deformation, and the solution has to be guessed equal to that of an undamaged pipeline with equivalent wall thick-
iteratively. In the following, a simpler equation to estimate groove ness tElong groove. Since 0 ⱕ ␥ ⱕ 1, the equivalent wall thickness
failure pressures is proposed. It is assumed that the groove failure takes values in the interval 关t0 − dmax , t0兴. Figures 6 and 7 show the
pressure is bounded by two pplain pressures. The first one corre- relationship between ␥ and the parameters dmax and w. These
sponds to the burst pressure of the undamaged pipe. The second figures were obtained from the numerical simulations given in
one is the failure pressure of a pipe with wall thickness t0 − dmax; Table 2. In Fig. 6, it is shown that ␥ is greater for shallow defects.
i.e., the groove width is equal to the perimeter. These bounds are Whichever groove depth is considered, ␥ decreases as the groove
combined to obtain an explicit solution. width increases. Minor differences are due to the steel grade; the
Here and thereafter, the failure pressure pplain corresponds to the pattern is ruled by the geometry of the groove. In Fig. 7, ␥ is
burst pressure of an undamaged pipeline with initial wall thick- studied considering just one steel grade and varying depth and
ness t0; plong groove denotes the failure pressure of a pipeline with width of the groove. Trends shown in Fig. 6 are emphasized.
initial wall thickness t0 and an infinitely long groove defect with From a theoretical point of view, ␥ has to satisfy the following
depth dmax and width w; and pⴱplain is the failure pressure of an limits: ␥ = 1 when dmax = w = 0; ␥ = 0 when w = ␲D 共where ␲D is
undamaged pipeline with initial wall thickness t0 − dmax, called the corresponding exterior perimeter of the pipe兲; finally, ␥ → 0
reduced wall thickness. As discussed above, pⴱplain is a lower when dmax → t0 and w ⫽ 0. To represent the behavior of ␥ as a
bound for plong groove. According to Eq. 共1兲, pⴱplain and pplain can be function of dmax and w and to satisfy the described limits, the
related as follows: following function is proposed:

051001-4 / Vol. 132, OCTOBER 2010 Transactions of the ASME


pR = plong groove + 共pplain − plong groove兲gd
min
共8兲
where gmin
d takes values between 0 and 1 and depends on the
corrosion depth along the longitudinal profile of the defect.
Mainly, this factor quantifies the contribution to the burst pressure
due to the remaining metal between the surface along the defect
and dmax. Algebra related to gmin
d is discussed in detailed form by
Oliveros et al. 关10兴, and a practical procedure to compute gmin d is
given in Appendix B.
Substituting Eq. 共5兲 into Eq. 共8兲, Eq. 共8兲 can be rewritten as


pR = pplain 1 −
dmax
t0
共1 − ␥兲共1 − gmin
d 兲 册 共9兲

Downloaded from http://asmedigitalcollection.asme.org/pressurevesseltech/article-pdf/132/5/051001/5751587/051001_1.pdf by University of Edinburgh user on 22 March 2023


The equivalent thickness defined in the previous section can be
modified and generalized as

tER = t0 − dmax共1 − ␥兲共1 − gmin


d 兲 共10兲
According to Eqs. 共1兲 and 共10兲, Eq. 共9兲 can be rewritten as
Fig. 7 Gamma parameter versus the depths of an infinitely
tER
long groove for different groove widths pR = k␴U 共11兲
r0
Accordingly, the failure pressure of a pipe with a corrosion defect


␥共dmax,w兲 = 1 −
dmax
t0
冊冉
q1
1−
w
␲D
冊 q2
e−q3共dmax/t0兲 共7兲
of irregular geometry is equal to that of an undamaged pipeline
with equivalent wall thickness tER, which is defined in the interval
关t0 − dmax , t0兴. Equation 共11兲 is an extension of Eqs. 共1兲 and 共6兲; it
The constants q1 = 1.3, q2 = 1.5, and q3 = 3.5 were obtained from a can be used to estimate failure pressures of undamaged and dam-
nonlinear least-squares minimization. Data given in Table 1 were aged pipelines. Using the corresponding wall thickness, Eq. 共11兲
used. In Fig. 8, numerical results from finite element simulations allows us to estimate the burst pressure of vessels with a finite
and corresponding failure pressures obtained from Eqs. 共6兲 and 共7兲 irregular defect, with an infinitely long groove and with no defect
were compared. All pressures were normalized by their corre- at all.
sponding pplain value. Steel API 5L X-42 was the only one that Failure pressures associated with an infinitely long groove
was used as the normalized pressures can be considered indepen- change suddenly for very narrow defects, in agreement with the
dent of material characteristics. In Fig. 8, the axis associated with experimental results reported in Ref. 关11兴. For larger defects, fail-
the width groove is in logarithmic scale to appreciate differences ure pressures are almost equal to that of undamaged pipes with
for each value of w. Results obtained from Eqs. 共6兲 and 共7兲 are reduced wall thickness 共plong groove ⬇ pⴱplain, w = ␲D, and ␥ = 0兲.
very close to those obtained from finite element method. Equation This suggests that for engineering purposes of an analysis and a
共7兲 and its fitted parameters can be used to estimate plong groove. prediction of failure pressure, the groove width can be neglected
关8,14兴. Equations are simpler and do not require a datum that is
generally unavailable. If it is assumed that plong groove = pⴱplain, then
4 Burst Pressure of Pipeline With Corrosion Defect of the equivalent wall thickness given by Eq. 共10兲 can be expressed
Irregular Geometry as
According to Oliveros et al. 关10兴, the failure pressure pR of a tER = t0 − dmax共1 − gmin
d 兲 共12兲
vessel with a finite irregular defect can be obtained as
In Figs. 9共a兲 and 9共b兲, experimental failure pressures are com-
pared with those calculated using Eqs. 共11兲 and 共12兲. To this pur-
pose, 56 reported burst tests associated with corrosion defects
were used to calibrate the model: 10 from Vieth and Kiefner 关19兴,
7 from Mok et al. 关4兴, 23 from Cronin 关5兴, 5 from Hopkins and
Jones 关11兴, 4 from Mannucci et al. 关24兴, and 7 from Choi et al.
关25兴. The error was defined as the quotient of the predicted failure
pressure divided by the corresponding experimental burst pres-
sure. For each test and yield criterion, such error was computed
considering plong groove = pⴱplain. Its mean and standard deviation,
for Tresca’s criterion 共Fig. 9共a兲兲, is 0.90 and 0.13, respectively.
For the ASSY criterion 共Fig. 9共b兲兲, the mean is 0.98 and the stan-
dard deviation is 0.14. Failure pressures were denoted by filled
symbols, as shown in Figs. 9共a兲 and 9共b兲. In general, we can state
that predictions obtained using the ASSY criterion are more pre-
cise than the ones obtained by Tresca. Note that from results plot-
ted in Fig. 1, statistical equivalence between the two yield criteria
was stated. Now, differences in mean values mark that this con-
clusion has to be modified when Eqs. 共11兲 and 共12兲 are used. As
shown in Figs. 9共a兲 and 9共b兲, we distinguished between real cor-
rosion defects and machined defects; the filled circles denote real
Fig. 8 Numerical simulations of normalized pressure corrosion defects, and the filled squares denote machined defects.
plong groove „finite element… versus analytical results „Eqs. „6… There is no different trend between these two sets of defects. The
and „7…… generalized pressure model can be applied to both.

Journal of Pressure Vessel Technology OCTOBER 2010, Vol. 132 / 051001-5


calculated means and standard deviations, it can be stated that the
ASSY yield criterion should be preferred to Tresca’s.

5 Application Example
In this section, the proposed approach is exemplified. Different
types of damage defects are considered. Predicted failure pres-
sures are estimated, discussed, and, in some cases, compared with
experimental results by Mok et al. 关4兴. The pipeline corresponds
to an API 5L X-60 with an exterior diameter of 508 mm and a
wall thickness of t0 = 6.35 mm 共r0 = 247.65 mm兲. Material prop-
erties are displayed in Table 1. The ASSY yield criterion is used

Downloaded from http://asmedigitalcollection.asme.org/pressurevesseltech/article-pdf/132/5/051001/5751587/051001_1.pdf by University of Edinburgh user on 22 March 2023


共k = 0.92兲. Equations 共7兲, 共B3兲, 共10兲, and 共11兲 will be used to esti-
mate failure pressure.
5.1 Undamaged Pipe. dmax = 0 and w = 0, so that Eq. 共7兲 leads
to ␥ = 1. Equation 共B3兲 gives gmind = 1 关10兴. From Eq. 共10兲, the
equivalent wall thickness is obtained: tER = t0. Failure pressure is
estimated using Eq. 共11兲: pR = 15.88 MPa. The difference with the
mean experimental value 共15.45 MPa and 15.25 MPa, see Ref.
关4兴兲 is less than 4%.
5.2 Infinitely Long Groove, Knowing Real Width. dmax
= t0 / 2 and w = 4t0 are considered. Equation 共7兲 leads to ␥ = 0.07.
Equation 共B3兲 gives gmin d = 0. From Eq. 共10兲, the equivalent wall
thickness is obtained: tER = 3.39 mm. Failure pressure is estimated
using Eq. 共11兲: pR = 8.49 MPa. The difference with the experi-
mental value 共8.40 MPa, see 关4兴兲 is 1%.
5.3 Infinitely Long Groove, Neglecting Real Width. The
depth is still dmax = t0 / 2, while the width is considered equal to the
perimeter of the pipe w = ␲D. That is, from Eq. 共7兲, ␥ = 0. Again,
Eq. 共B3兲 gives gmind = 0. From Eq. 共10兲, the equivalent wall thick-
ness is obtained: tER = t0 − dmax = 3.175 mm. Failure pressure is es-
timated using Eq. 共11兲: pR = 7.94 MPa. The difference with the
experimental value 共8.40 MPa, see 关4兴兲 is 5%. In this case, Eq.
共10兲 or Eq. 共12兲 could be used to compute the equivalent wall
Fig. 9 „a… Experimental failure pressures versus predicted fail-
thickness. Choosing Eq. 共12兲, the evaluation of Eq. 共7兲, i.e., ␥, can
ure pressures for Tresca yield criterion. „b… Experimental failure
pressures versus predicted failure pressures for ASSY yield be skipped.
criterion. 5.4 Finite Defect, Knowing Real Width. The pipe is affected
by a corrosion defect with a maximum depth of dmax = t0 / 2, a
width of w = 4t0, and a length of l = 50 mm. This defect is repre-
sented by a parabolic geometry, given by the following expres-
For cases in which the groove widths were available 共e.g., see sion:
experiments in Mok et al. 关4兴 and Choi et al. 关25兴兲, the failures
pressures were also estimated, taking width into account 共empty
squares and triangles in Figs. 9共a兲 and 9共b兲兲. For the Tresca crite-
rion, the mean and standard deviation are 0.92 and 0.077, respec- d共x兲 =
冦 冋 冉 冊册
dmax 1 −
2x
l
2
, − l/2 ⱕ x ⱕ l/2
冧 共13兲
tively; meanwhile, for the ASSY criterion, the values are 1.00 and 0 elsewhere
0.084, respectively. In these cases, the geometry of the defect
corresponds specifically to grooves. Any filled symbol just below Equation 共7兲 leads to ␥ = 0.07. The factor gmin
d can be estimated by
an empty symbol corresponds to the burst pressure evaluated, ne- substituting the geometry, given by Eq. 共13兲, into Eq. 共B1兲 or
glecting or considering the width of the defect. Failure pressures numerically by means of Eq. 共B3兲. gmin d = 0.74 was computed.
taking into account the groove width are slightly greater than the From Eq. 共10兲, the equivalent wall thickness is obtained: tER
corresponding failure pressure obtained using a reduced wall = 5.58 mm. Failure pressure is estimated using Eq. 共11兲: pR
thickness 共w = ␲D, plong groove = pⴱplain兲. Errors obtained, ignoring or = 13.96 MPa.
considering the width of the groove, have practically equal mean
values. However, standard deviations are different probably be- 5.5 Finite Defect, Neglecting Real Width. The pipe is af-
cause of the size of the available sample. fected by a corrosion defect with a maximum depth of dmax
In Fig. 9共b兲, triangles are used to show the analytical and ex- = t0 / 2 and length of l = 50 mm. The width is considered equal to
perimental failure pressures of five pipelines with infinitely long the exterior perimeter of the pipe w = ␲D. This defect is repre-
grooves. The predicted failure pressures, using the ASSY crite- sented by a parabolic geometry given by Eq. 共13兲. From Eq. 共7兲,
rion, match up with the corresponding experimental pressures. ␥ = 0. Equation 共B3兲 gives gmind = 0.74. From Eq. 共10兲, the equiva-
Despite the accuracy obtained using Eq. 共10兲, Eq. 共12兲 should lent wall thickness is obtained: tER = 5.52 mm. Failure pressure is
be preferred. Differences between the results obtained using Eq. estimated using Eq. 共11兲: pR = 13.81 MPa. In this case, Eq. 共10兲 or
共10兲 or Eq. 共12兲 are small. The simplicity introduced with Eq. 共12兲 Eq. 共12兲 could be used to compute the equivalent wall thickness.
is significant. Moreover, in professional practice, the defect width Choosing Eq. 共12兲, the evaluation of Eq. 共7兲, i.e., ␥, can be
is often not specified in the inspection reports. Finally, comparing skipped.

051001-6 / Vol. 132, OCTOBER 2010 Transactions of the ASME


6 Conclusions x ⫽ coordinate associated with a damage depth
We introduced an accurate and simple method to compute the d共x兲
failure pressure of pipelines with any of an infinitely long groove di ⫽ ith measured corrosion depth
or a finite irregular corrosion defect. Finite element simulations Ii,j ⫽ value associated with interval 关xi , x j兴
were used to study the relationship among burst pressure, steel xi ⫽ axial coordinate associated with corrosion
properties, and groove geometry. For a pipeline with an infinitely depth di
long groove or with a corrosion defect of irregular geometry, the N ⫽ number of points where the corrosion depth di
failure pressure can be obtained from the solution of an undam- is known
aged pipeline using an equivalent wall thickness. The failure pres- 关b̂ − â兴 ⫽ interval where the corrosion defect is defined
sure of undamaged pipelines is predicted by the classic Svens-
son’s formula for thin walled cylindrical vessels. The equivalent l = b̂ − â ⫽ axial length of damage
wall thickness depends on the depth and width of the defect. How- ␤ ⫽ auxiliary variable
ever, the failure pressure barely depends on the width of the de-

Downloaded from http://asmedigitalcollection.asme.org/pressurevesseltech/article-pdf/132/5/051001/5751587/051001_1.pdf by University of Edinburgh user on 22 March 2023


fect, and it is convenient to consider the depth only. Different
Appendix A: Constitutive Function
yield criteria 共Tresca, von Mises, and ASSY兲 were compared, and To predict the failure pressure of pipelines, mathematical mod-
it was established which one is best. Results clearly show that the els based on the mechanics of continuous media 关8,9兴 must be
ASSY criterion is adequate to predict the failure pressures of un- used. These models depend on selected hypotheses. For example,
damaged and damaged pipes. a constitutive function has to be chosen to define the stress-strain
relation. In particular, the elastoplastic behavior of steel can be
Acknowledgment described by a Ramberg–Osgood or Ludwik function. The
Ramberg–Osgood function is defined by three parameters; mean-
This work was mainly supported by Instituto Mexicano del while, the Ludwik function is defined by only two, the true stress
Petróleo under Grant No. D.00433. We would like to thank Úrsula ␴ and the true strain ␧. These parameters are related through the
Iturrarán for the kind comments. Suggestions from reviewers power law ␴ = ␴0␧n, where n is the strain hardening exponent. Due
helped to improve the manuscript. to its simplicity, in this work the Ludwik function is used. The
Ramberg–Osgood function can represent adequately the stress-
strain curve; meanwhile, Ludwik’s function does not fit ad-
Nomenclature equately for elastic strains. However, it adequately represents
pplain ⫽ failure pressure of an undamaged pipeline plastic deformations, and it has the advantage that it leads simple
pⴱplain ⫽ failure pressure of an undamaged pipeline with and explicit solutions as Eqs. 共1兲, 共6兲, and 共11兲.
reduced wall thickness In literature, there are experimental burst tests in which a con-
plong groove ⫽ failure pressure of a pipeline with an infinitely stitutive equation is not given; however, the yield and the ultimate
long groove stress are specified. From experimental stress-strain relations,
pR ⫽ failure pressure of a pipeline with a corrosion Maes et al. 关26兴 and Zhu and Leis 关9兴 introduced semi-empirical
defect of irregular geometry equations to find out the value of parameter n. In this work, a
k ⫽ constant associated with a certain yield different approach is taken; n is obtained from the linear stress ␴Y
criterion and the ultimate stress ␴U as follows:
␴0 ⫽ scale parameter of the Ludwik function
n ⫽ hardening coefficient of the Ludwik function ␴Y ␴U
= n 共A1兲
e ⫽ 2.1782… ␧Yn ␧U
t0 ⫽ initial or undeformed wall thickness With Ludwik’s function, the strain at the ultimate stress ␴U is
tElong groove ⫽ equivalent undeformed wall thickness for an equal to n 共␧U = n兲. Also, the elastic strain ␧Y must satisfy, at least
infinitely long groove
approximately, that ␧Y ⬇ ␴Y / E, where E is the Young modulus.
tER ⫽ equivalent undeformed wall thickness for a
We assume that E ⬇ 207 GPa 共which is reasonable for a typical
corrosion defect
steel material兲 and ␴Y = SMYS 共where SMYS is the minimum
r0 ⫽ inner radius of a pipe 共undeformed兲
yield stress兲, and we express ␴U in terms of the corresponding
D ⫽ outer diameter 共undeformed兲
engineering stress ␴Ueng = ␴U / en 共e = 2.1782. . .兲. Then, Eq. 共A1兲
dmax ⫽ maximum corrosion depth
w ⫽ width of the groove can be expressed as
␥ ⫽ auxiliary function ln共␴Y /en␴Ueng兲 = n ln共␴Y /E兲 − n ln共n兲 共A2兲
q1 , q2 , q3 ⫽ parameters of function ␥
␲ ⫽ 3.14159… or
gmin
d ⫽ minimum value of gd共 · 兲 ln共␴Y /en␴Ueng兲
gd共 · 兲 ⫽ function that quantifies the remaining material ␰= =1 共A3兲
n ln共␴Y /E兲 − n ln共n兲
contribution at each point x0 of defect
␴U ⫽ maximum true stress To find out n, Eq. 共A2兲 must be solved numerically. The correct
␧U ⫽ true strain at ultimate true stress value of n corresponds to ␰ = 1. Note that even if n is taken as a
␧Y ⫽ true strain at yield stress random variable, e.g., Maes et al. 关26兴, its mean value would be
␴Ueng ⫽ maximum engineering stress associated with ␰ = 1. This should be accomplished no matter what
␧Y eng ⫽ engineering strain at engineering yield stress ␴Y / ␴U is.
For completeness, note that if the true yield stress and strain
E ⫽ Young modulus
共␴Y eng , ␧Y eng兲 were considered, the hypothesis on Young’s modulus
␴Y ⫽ true yield stress
␰ ⫽ auxiliary variable 共E ⬇ 207 GPa兲 should be omitted. Typically, ␧Y eng corresponds to
sech共 · 兲 ⫽ hyperbolic secant 0.5% of the total strain 关9兴 or to 0.2% of the plastic strain 关2兴.
x0 ⫽ evaluation point
xa , xb ⫽ coordinates to define geometrical bounds for
Appendix B: Computation of gmin
d
the defect According to Ref. 关10兴, the parameter gmin
d takes values in the
d共x兲 ⫽ corrosion depth at point x interval 关0,1兴. gmin
d is obtained from a function gd, which essen-

Journal of Pressure Vessel Technology OCTOBER 2010, Vol. 132 / 051001-7


et al. 关10兴. Finally, gmin
d is obtained as the minimum value that
results from evaluating gd at different evaluation points.

References
关1兴 Hill, R., 1950, The Mathematical Theory of Plasticity, Oxford University
Press, New York.
关2兴 Mendelson, A., 1968, Plasticity: Theory and Application, Macmillan, New
York, pp. 312–317.
关3兴 Svensson, N. L., 1958, “The Bursting Pressure of Cylindrical and Spherical
Vessels,” ASME J. Appl. Mech., 25, pp. 89–96.
Fig. 10 Geometry of a corrosion defect in discretized points 关4兴 Mok, D. H. B., Pick, R. J., Glover, A. G., and Hoff, R., 1991, “Bursting of
Line Pipe With Long External Corrosion,” Int. J. Pressure Vessels Piping, 46,
pp. 195–216.
关5兴 Cronin, D. S., 2000, “Assessment of Corrosion Defects in Pipelines,” Ph.D.
tially depends on the corrosion defect depth at the evaluation point thesis, Department of Mechanical Engineering, University of Waterloo, ON,

Downloaded from http://asmedigitalcollection.asme.org/pressurevesseltech/article-pdf/132/5/051001/5751587/051001_1.pdf by University of Edinburgh user on 22 March 2023


x0 and on the geometry of adjacent corrosion. The pipeline failure Canada.
pressure is associated with the minimum value of the function gd, 关6兴 Law, M., and Bowie, G., 2007, “Prediction of Failure Strain and Burst Pres-
sure in High Yield-to-Tensile Strength Ratio Linepipe,” Int. J. Pressure Vessels
d = min 兵gd共x0兲其. An appropriate Cartesian system is used
i.e., gmin Piping, 84, pp. 487–492.
x0苸关â,b̂兴 关7兴 Xue, L., Widera, G. E. O., and Sang, Z., 2008, “Burst Analysis of Cylindrical
to describe the defect geometry; the damage depth is defined, and Shells,” ASME J. Pressure Vessel Technol., 130, p. 014502.
along one of its axes, the position of each depth in the interval 关8兴 Cronin, D. S., and Pick, R. J., 2002, “Prediction of the Failure Pressure for
Complex Corrosion Defects,” Int. J. Pressure Vessels Piping, 79, pp. 279–287.
关â , b̂兴 is given along the other. According to Oliveros et al. 关10兴, 关9兴 Zhu, X. K., and Leis, B. N., 2006, “Average Shear Stress Yield Criterion and
the function gd can be written in the form Its Application to Plastic Collapse Analysis of Pipelines,” Int. J. Pressure Ves-


sels Piping, 83, pp. 663–671.
bˆ 关10兴 Oliveros, J., Alamilla, J. L., Astudillo, E., and Flores, O., 2008, “Prediction of

gd共x0兲 = 1 − sech共␤共x0 − x兲兲d共x兲dx 共B1兲 Failure Pressures in Pipelines With Corrosion Defects,” ASME J. Pressure
␲dmax â
Vessel Technol., 130, p. 021703.
关11兴 Hopkins, P., and Jones, D. G., 1992, “A Study of the Behavior of Long and
where sech共 · 兲 is the hyperbolic function, ␤ = 共D共t0 − dmax兲兲−1/2, D Complex-Shaped Corrosion in Transmission Pipelines,” Proceedings of
OMAE’S 1992 11th International Conference on Offshore Mechanics and Arc-
= 2共r0 + t0兲 is the initial exterior diameter 共in contrast with the tic Engineering, Calgary, Canada, pp. 211–217.
notation used in the body of the article兲, r0 is the inner initial 关12兴 Fu, B., and Kirkwood, M. G., 1995, “Predicting Failure Pressure of Internally
radius, t0 is the initial wall thickness, d共x兲 is the depth of the Corroded Line Pipe Using the Finite Element Method,” OMAE, 5, pp. 175–
184.
defect at x, and dmax is the maximum depth of the defect along the 关13兴 Chouchaoui, B. A., 1993, “Evaluating the Remaining Strength of Corroded
entire longitudinal profile. In engineering applications, every de- Pipelines,” Ph.D. thesis. Department of Mechanical Engineering, University of
fect profile can be approximated by a discrete set of depths di Waterloo, Canada.
关14兴 ASME B31G, 1991, “Manual for Determining the Remaining Strength of Cor-
= di共xi兲 at points xi, with i = 1 , . . . , N 共Fig. 10兲. If the function d共x兲 roded Pipelines,” ASME Paper No. B31G-1991.
is approximated to be constant between two consecutive points 关15兴 Vieth, P. H., and Kiefner, J. F., 1989, “A Modified Criterion for Evaluating the
关xi , xi+1兴, then the complete defect profile can be represented as Remaining Strength of Corroded Pipe,” American Gas Association, Final Re-
follows: port on Project PR 3-805 to the Pipeline Research Committee.
关16兴 Papka, S. D., Stevens, J. H., Macia, M. L., Fairchild, D. P., and Petersen, C.
1 W., 2004, “Full-Size Testing and Analysis of X120 Linepipe,” Int. J. Offshore
d共x兲 = 2 共di共xi兲 + di+1共xi+1兲兲, xi ⱕ x ⱕ xi+1, i = 1, . . . ,N − 1 Polar Eng., 14, pp. 42–51.
共B2兲 关17兴 Liessem, A., Graef, M. K., Knauf, G., and Marewski, U., 2004, “Influence of
Thermal Treatment on Mechanical Properties of UOE Linepipe,” Proceedings
This representation is simple and makes a close solution of the of the Fourth International Conference of Pipeline Technology, Ostend, Bel-
integral of Eq. 共B1兲 for each interval 关xi , xi+1兴, i = 1 , . . . , N − 1. Ac- gium, pp. 1263–1281.
关18兴 Amano, K., Matsuoka, M., Ishihara, T., Tanaka, K., Inoue, T., Kawaguchi, Y.,
cording to Eq. 共B2兲, the depth profile is not smooth; however, the and Tsukamoto, M., 1986, “Significance of Yield Ratio Limitation to Plastic
numerical convergence of the integral is guaranteed if ␧ ⱕ 兩xi+1 Deformation of Pipeline in High Pressure Proof Test,” Proceedings of the
− xi兩 is small and close to zero. Seventh Symposium on Line Pipe Research, Houston, TX, Oct. 14–16, pp.
8.1–8.21.
Substituting Eq. 共B2兲 in Eq. 共B1兲, the integral can be solved as

冋兺 册
关19兴 Vieth, P. H., and Kiefner, J. F., 1993, “Database of Corroded Pipe Tests,”
i=j−1 l=N−1 Pipeline Research Council International, Inc., Report No. L51689.


1 关20兴 Law, M., 2005, “Use of the Cylindrical Instability Stress for Blunt Metal Loss
gd共x0兲 = 1 − Ii,i+1 + I j,0 + I0,j+1 + Il,l+1 Defects in Linepipe,” Int. J. Pressure Vessels Piping, 82, pp. 925–928.
␲dmax i=1 l=j+1 关21兴 Christopher, T., Rama Sarma, B. S. V., Govindan Potti, P. K., Nageswara Rao,
B., and Sankarnarayanasamy, K., 2002, “A Comparative Study on Failure
共B3兲 Pressure Estimations of Unflawed Cylindrical Vessel,” Int. J. Pressure Vessels
where Ii,j is the integral of Eq. 共B1兲 scaled by 1 / ␤ and evaluated Piping, 79, pp. 53–66.
关22兴 ANSYS, 2003, Finite Element Method Software 7.1, Analysis System Inc.
between xi and x j, 关23兴 Kiefner, J. F., 1979, “Corroded Pipe: Strength and Repair Methods,” Fifth


xj Symposium on Line Pipe Research, Houston, TX.
关24兴 Mannucci, G., Demofonti, G., and Di Biagio, M., 2005, “X100-Fracture Ini-
Ii,j = ␤ sech共␤共x0 − x兲兲d共x兲dx = 共di + d j兲关tan−1共e␤共x0−xi兲兲 tiation and Propagation,” First Super High Strength Steel, Rome, Italy, Nov.
xi 2–5.
关25兴 Choi, J. B., Goo, B. K., Kim, Y. J., and Kim, W. S., 2003, “Development of
− tan−1共e␤共x0−x j兲兲兴 共B4兲 Limit Load Solutions for Corroded Gas Pipelines,” Int. J. Pressure Vessels
Piping, 80, pp. 121–128.
and function tan 共 · 兲 is the inverse tangent. In Fig. 10, subscripts
−1
关26兴 Maes, M. A., Dann, M., and Salama, M. M., 2008, “Influence of Grade on the
are listed. More details about function gd can be found in Oliveros Reliability of Corroding Pipelines,” Reliab. Eng. Syst. Saf., 93, pp. 447–455.

051001-8 / Vol. 132, OCTOBER 2010 Transactions of the ASME

You might also like