Professional Documents
Culture Documents
Abstract
This paper presents the effects of boundary conditions of failure pressure models for buried
pipelines on failure prediction by using a failure probability model and a finite element
analysis. The first order Taylor series expansion of the limit state function (LSF) is used in
order to estimate the probability of failure associated with corroded defects under varying
boundary conditions for long exposure periods in years. Two failure pressure models based
on a failure function composed of failure and operation pressures are adopted for the
assessment of pipeline failure. The effects of random variables such as varying boundary
conditions on the failure probability of the buried pipelines are systematically analyzed by
using a failure probability model for the corroded pipeline. Furthermore, the engineering
results estimated by a failure probability model are compared with those obtained by a finite
element analysis. The influences of statistical distributions, such as normal and lognormal, of
random variables on the failure probabilities are also investigated in detail.
Introduction
The buried pipelines transporting gas and oil are usually laid underground and exposed to
varying boundary conditions. It has been often reported as industrial examples that many
catastrophic disastrous pipeline accidents have been caused by the defect such as corrosion
arisen by aging and/or environmental effects on pipelines. It is thus very important to
investigate the effect of defect on the safety of buried pipelines for preventing heavy
economical and social losses in advance, Choi[1].
The buried pipelines usually have the various types of defects such as corrosion and
environment-assist-cracking. If these were operated under an excessive operating pressure in
the corroded pipelines, it could produce larger stresses than safety designed one. It may be
emphasized that the failure analysis should be carried out with the help of the probability
method than the conventional deterministic approach (CDA) because the CDA leads to
uncertainties in the failure analysis with random variables imposed on varying boundary
conditions, Lee et al.[2], Hopkins and Jones[3].
In this paper, two failure pressure models based on strength assessment methods were
used to evaluate the failure probabilities of the defect region in pipeline with internal
operating pressure. For each model, using the FORM (first-order reliability method) which is
one of the probability analysis method with reliability index, the effects of varying boundary
conditions on the failure probability of pipeline are systemically investigated. Furthermore,
the engineering results estimated by a failure probability model are compared with those
ECF15
estimated by the finite element analysis. At the end, the failure probabilities corresponding to
both normal and lognormal distributions of varying random variables are compared each
other.
(a)
(b)
FIGURE 2. Section through an idealized corrosion defect. (a) Parabolic, (b) Rectangular
ANSI/ASME B31G
A failure equation for the corroded pipelines is proposed by means of the data of bursting
experiment and expressed with consideration of two conditions below. First, the maximum
hoop stress cant exceed the yield strength of material. Second, relatively short corrosion is
projected on the shape of parabola and long corrosion is projected on the shape of
rectangular. The failure pressure equation for the corroded pipeline is classified by the shape
of parabola and rectangular as shown below, Ahammed[4].
2 t 1 (2 / 3)(d / t )
Pf = 1.1 yield
D 1 ( 2 / 3)(d / t ) / M
Pf = 1.1
2 yield t
D
[1 (d / t )]
for
L D
0.8 4 (Parabola)
D t
(1)
for
L D
0.8 > 4 (Rectangular)
D t
(2)
where Pf is the failure pressure, D is the outer diameter, M is the bulging factor, t is the
thickness of pipelines, d is the maximum depth of corrosion region, L is the defect length of
corrosion region and yield is the yield strength.
ECF15
for
L D
0.8 4
D t
(3)
M =
for
L D
0.8 > 4
D t
(4)
(5)
L D
L D
M = 1 + 0.6275 0.003375
D t
D t
2
L D
M = 3.3 + 0.032
D t
for
L D
50
D t
(6)
for
L D
> 50
D t
(7)
Failure Probability
We initially assume that every variable is normal distribution and the probability distribution
is determined by a mean and standard deviation. The failure probability is calculated using
FORM (first order reliability method) that is one of the methods utilizing reliability index [2].
The FORM method is denoted from the fact that it is based on a first-order Taylor series
approximation of the performance functional or limit state function (LSF). A limit state
function is defined as below.
Z = RL
(8)
where R is the resistance normal variable, and L is the load normal variable. Assuming
that R and L are statistically independent normally distributed random variables, the variable
Z is also normally distributed. The event of failure occurs when R<L, that is Z<0. The
probability of failure (PF) is given as below
PF = P[ Z < 0] =
1 Z
Z
exp
2
1
U2
dZ
exp
=
2 2 dU = ( )
(9)
where Z and Z are the mean and standard deviation of variable Z, respectively, new
variable U is U = ( Z Z ) / Z , is the cumulative distribution function for a standard
normal variable and is the safety index or reliability index denoted as below
Z
L
= R
Z
R2 + L2
(10)
ECF15
Rackwitz and Fiessler proposed a method to estimate the reliability index using the
procedure as shown in Fig. 3. In this paper, we iterated the loop to estimate a reliable
reliability index until it is converging to desire value ( 0.001 ), Mahadevan and Haldar[6].
Where the coefficient of variation (C.O.V) is denoted as below with the standard
deviation, X and the mean, X .
C.O.V =
X
X
(11)
equivalent normal variables should be equal at the checking point ( x ) on the failure surface.
Considering each statistically independent non-normal variable individually and equating its
cumulative distribution function with an equivalent normal variable at the checking point, we
obtain as below[6].
x* XN
FX ( x* ) =
N
X
XN = x * 1 (FX ( x * ) ) XN
(12)
where is the cumulative distribution function of the standard normal variable, XN and
XN are the mean and the standard deviation of the equivalent normal variable at the checking
point, and FX ( x * ) is the cumulative distribution function of the original non-normal random
variables. Equating probability density functions of the original variable and equivalent
normal variable at the checking point, we obtain as below.
f X ( x* ) =
x * XN
N
X
XN =
[ 1 (FX ( x* ) )]
f X ( x* )
(13)
where is the probability density function of the standard normal variable and f X ( x* ) is
the probability density function of the original non-normal random variables. The non-normal
variables can be treated as normal distributed variables through the transformation of Eqs.
(12) and (13).
ECF15
Z = yield f
(14)
Case Study
The random variables listed in Table 1 have been utilized to investigate the failure probability
and finite element analysis of the corroded pipeline.
TABLE 1. Random variables and their parameters used in the example.
Variable
Mean
C.O.V
Variable
Mean
C.O.V
Pa
8MPa
0.1
10mm
0.05
423MPa
0.067
3mm
0.1
200mm
0.05
600mm
0.03
yield
ECF15
FEA
MB31G
B31G
Failure Probability
0.8
0.6
0.4
0.2
0.0
0
10
15
20
25
30
35
40
3.5
1.0
3.5
B31G
Failure Probability
Reliability Index
Failure Probability
Reliability Index
3.0
0.8
2.0
1.5
0.4
1.0
0.2
3.0
2.5
0.6
2.0
1.5
0.4
1.0
0.2
0.5
0.0
0.5
0.0
10
15
20
25
30
35
Reliability Index
0.6
Reliability Index
2.5
Failure Probability
0.8
0.0
40
0.0
10
15
20
25
30
35
40
(a)
(b)
1.0
3.5
3.0
0.8
2.5
0.6
2.0
1.5
0.4
FEA
1.0
Failure Probability
Reliability Index
0.2
Reliability Index
Failure Probability
Failure Probability
MB31G
0.5
0.0
0.0
10
15
20
25
30
35
40
(c)
FIGURE 7. Relationship between failure probability and reliability index for several
failure pressure models. (a) B31G, (b) MB31G, (c) FEA.
ECF15
1 .2
G ro o v e W id th = T h ro u g h
G ro o v e W id th = 8 0 m m
G ro o v e W id th = 6 0 m m
F E A (P = 1 0 M P a )
F E A (P = 8 M P a )
M B 3 1 G (P = 1 0 M P a )
M B 3 1 G (P = 8 M P a )
B 3 1 G (P = 1 0 M P a )
B 3 1 G (P = 8 M P a )
1 .2
1 .0
Failure Probability
Failure Probability
1 .0
0 .8
0 .6
0 .4
0 .2
0 .8
0 .6
0 .4
0 .2
0 .0
0 .0
10
15
20
25
30
35
40
10
15
20
25
30
35
40
E x p o s u r e p e r io d fr o m la s t in s p e c tio n (y e a r )
(a)
(b)
FIGURE 8. Relationship between failure probability and exposure period. (a) varying
groove widths, (b) varying internal operation pressures.
Fig. 8(a) shows the variation of failure probability with varying groove width of corroded
defects in the finite element analysis. We analyse the failure probability of buried pipelines
with varying groove width of corroded defects as 60mm, 80mm and flat groove as shown in
Fig. 8(a). It is noted that the failure probability becomes larger with narrow groove width like
60mm. And it can be recognized that the failure probability of groove width, which is flat
groove and 80mm, is larger about 0.38 than those of groove width, which is 60mm.
Fig. 8(b) shows the variation of failure probability for varying internal operation pressure.
It is noted that the failure probability increases for all models with increasing internal
operation pressure. Especially, it is noted that the amount of increasing of FEA model is the
largest about 0.42 when the exposure period is 20 year, whereas the amount of increasing of
B31G model is the smallest.
1 .0
3 .5
1 .0
3 .5
B 3 1G
F ailu re P ro ba bility(log no rm al)
0 .8
M B 3 1G
3 .0
1 .5
0 .4
1 .0
Failure Probability
2 .0
0 .2
2 .5
0 .6
2 .0
1 .5
0 .4
1 .0
0 .2
0 .5
0 .0
0 .5
0 .0
10
15
20
25
30
35
Reliability Index
2 .5
Reliability Index
Failure Probability
0 .6
3 .0
0 .8
0 .0
40
0 .0
10
15
20
25
30
35
40
(a)
(b)
1 .0
3 .5
FEA
F ailu re P ro b ab ility (lo g n o rm al)
3 .0
F ailu re P ro b ab ility (n o rm a l)
0 .8
2 .5
0 .6
2 .0
1 .5
0 .4
1 .0
Reliability Index
Failure Probability
0 .2
0 .5
0 .0
0 .0
10
15
20
25
30
35
40
(c)
FIGURE 9. Effect of normal and lognormal distribution of random variable on failure
probabilities for several failure pressure models. (a) B31G, (b) MB31G, (c) FEA.
Fig. 9 shows relationships between the failure probability and the reliability index for two
different statistical distributions of the random variables such as the normal and lognormal
ECF15
distribution. For the B31G model, it is noted that the reliability index of lognormal
distribution is larger than that for normal distribution, whereas the failure probability for
lognormal distribution is smaller than that for normal distribution. However, for the MB31G
model and FEA model, it is noted that reliability index and failure probability for lognormal
distribution is larger than that for normal distribution at some range of exposure periods,
whereas it is reversed at a certain range of exposure periods. The effects of different
statistical distributions of random variables on the reliability index are larger than those of
failure probability.
Conclusion
In this study, FORM (first order reliability method) and finite element analysis model are
utilized to extract useful technical information in carrying out the effective failure control for
the corroded pipeline. Using B31G, MB31G and FEA models, the effect of the corrosion
depth, the corrosion length, the thickness, the diameter, the inner fluid pressure and the yield
stress of pipeline on the failure probability is systematically studied and the following results
are obtained:
1. It is recognized that the failure probability increases with increasing of exposure periods
for ANSI/ASME B31G model, MB31G model and finite element analysis model. The
results by FEA model show, however, the highest increasing rate of failure probability.
2. It is noted that the increasing rate of failure probabilities for the narrower groove width
is higher than that of the wider groove width.
3. It is found that the effect of the statistical distribution of random variables such as
normal and lognormal on the reliability index and failure probability is not big. However,
the reliability index and failure probability for lognormal distribution is larger than that
for normal distribution at some range of exposure periods, whereas it is reversed at a
certain range of exposure periods.
Acknowledgment
The authors are grateful for the support provided by a grant from the KOSEF and Safety and
Structural Integrity Research Center at the SungKyunKwan University. The authors wish to
thank all the members concerned.
References
1. S. C. Choi, Gas Safety Journal, Vol.26, No.5, 25-33, 2000.
2. O. S. Lee, J. S. Pyun and D. H. Kim, Int. J. of KSPE, Vol. 4, No. 6, 12-19, 2003.
3. P. Hopkins and D. G. Jones: In Proceedings of the 11th International Conference on
Offshore Mechanics and Arctic Engineering, ASME, Vol. V, Part A, 211-217, 1992.
4. M. Ahammed, Int. J. Pressure Vessels and piping, Vol. 75, 321-329, 1998.
5. J. F. Kiefner and P. H.Vieth, Oil and Gas Journal, 88(32), 56-59, 1990.
6. S. Mahadevan and A. Haldar, Probability, Reliability and Statistical Method in
Engineering Design, John Wiley & Sons, 2000.
7. O. S. Lee and H. J. Kim, Proceeding of KSPE, Vol. 23, No. 11, pp. 2096-2101, 1999.