You are on page 1of 10

Copyright 2012, Brazilian Petroleum, Gas and Biofuels Institute - IBP

This Technical Paper was prepared for presentation at the Rio Oil & Gas Expo and Conference 2012, held between September, 17-
20, 2012, in Rio de Janeiro. This Technical Paper was selected for presentation by the Technical Committee of the event according
to the information contained in the final paper submitted by the author(s). The organizers are not supposed to translate or correct the
submitted papers. The material as it is presented, does not necessarily represent Brazilian Petroleum, Gas and Biofuels Institute
opinion, or that of its Members or Representatives. Authors consent to the publication of this Technical Paper in the Rio Oil & Gas
Expo and Conference 2012 Proceedings.
______________________________
1
Ph.D., Mechanical Engineer, Dept. of Structural Engineering, University of So Paulo
2
Ph.D., Civil Engineer, Dept. of Structural Engineering, University of So Paulo
3
M.Sc., Mechanical Engineer, PETROBRAS
4
M.Sc., Mechanical Engineer, CENPES, PETROBRAS
5
B.Sc., Asset Management Consultant, GL Noble Denton, Loughborough, UK


IBP1179_12
SAFETY MANAGEMENT OF PIPELINES
BASED ON STRUCTURAL RELIABILITY: HISTORICAL
PERSPECTIVE AND PROGRESS
Andr T. Beck
1
, Felipe A.V. Bazn
2
, Renato Mendes
3
,
Guilherme Donato
4
, Michael Gardiner
5

Abstract
Since the early beginnings in the sixties and seventies, Structural Reliability theory has reached a mature stage
encompassing solid theoretical developments and increasing practical applications. Structural reliability methods have
permeated the engineering profession, finding applications in code calibration, structural optimization, life extension of
existing structures, life-cycle management of infrastructure risks and costs, and so on. This review paper shows that the
ground work for Reliability Based Design and Assessment (RBDA) of onshore pipeline systems is already developed.
Hence, this allows the economic management of the risks involved in operation of pipeline systems. It is shown that
RBDA is a rational tool to safely manage the operational life of pipeline systems, optimizing initial design and the
expenditures in inspection and maintenance operations.

1. Introduction
Reliability-Based Design and Assessment (RBDA) is a rational tool for the safety management of engineering
structures. Uncertainty is inherent to engineering enterprises and it arises due to imperfect knowledge of novel
engineering systems, due to imprecise physical and mathematical models of structural behavior, due to the inherent
variability of structural materials strengths and due to natural and random fluctuations of structural loads. Uncertainty
leads to risk, and risk management must be put in place from the design to operation phase.
Risk management can be viewed as a tool to balance the apparently competing goals of economy and safety in
the design and operation of pipeline systems. When the expected costs of failure are considered and included in the cost
equation, one realizes that safety and economy are actually not competing goals: there is one optimum level of safety
which leads to minimum total expected costs. This balance between safety and costs can be addressed subjectively via
risk management, but it can also be addressed quantitatively using methods of RBDA.
This article builds the case for the readiness of design methods for RBDA management of buried pipelines.
Practical use of RBDA in the safety management of real pipelines requires the following elements to be available:
1. Statistical data to quantify past failure rates under different operational conditions and failure causes;
2. Well established criteria for risk acceptance (target reliabilities);
3. Accurate models for the most relevant failure modes: leak, burst or rupture due to corrosion, manufacturing
defects, operational accidents, mechanical damage due to third party interference;
4. Statistical data to quantify uncertainties in strength of materials, random loads, uncertain operational
conditions;
5. Well established methods to evaluate failure probabilities, given failure models and parameter statistics.
Following a historical perspective on past developments in RBDA, this paper addresses each of these five
points in a review of the state of the art for RBDA of buried pipelines.

2. Historical perspective
Historically, uncertainty in engineering design has been dealt with by conservative assumptions and the use of safety
margins against undesirable behaviors. Although conservatism has allowed the engineering profession to progress,
existing structures, life-cycle management of infrastructure risks and costs, and so on. This review paper shows that the
Rio Oil & Gas Expo and Conference 2012

2
excessive conservatism leads to unnecessary costs and does not allow for optimization of resources in safety
management of constructed facilities. Deterministic conservative design allows structures to be built and managed, but
does not provide a quantified understanding of structural safety. Hence, future structural risks cannot be properly
managed by deterministic design.
Structural reliability theory has been developed as a tool to quantify probabilities of undesirable structural
responses (failure), based on a probabilistic description of the uncertainties in structural materials and loads. Risk is
then the product of a failure probability with the costs of failure. Hence, the very prediction of future risks requires the
computation of future failure probabilities.
Since the early beginnings in the sixties and seventies, structural reliability analysis has grown into a mature
stage encompassing solid theoretical developments and increasing practical applications. Structural reliability methods
have permeated the engineering profession, finding applications in code calibration, structural optimization, life
extension of existing structures, life-cycle management of infrastructure costs, and so on.
Most progress in theoretical developments was made in the late seventies and during the eighties. At the end of
the 70s, the technical basis for Load and Resistance Factor Design (LRFD) and the concept of limit state design were
laid out. These concepts were introduced in the 1982 version of the ANSI A58.1 code, and later incorporated in the
ASCE 7 code of 1988. These codes already included partial safety factors that were calibrated to achieve target
reliability indexes in design. Significant parts of these developments were due to the pioneering work of Ellingwood et
al. (1980).
EUROCODES followed the trend of partial safety factor and limit state design (EUROCODE, 2001). Starting
from a uniform format, member states have been allowed to establish their own partial safety factors. Most European
countries have used calibration to target reliabilities to accomplish this task. The ground work for this reliability-based
calibration resulted from work of the Joint Committee on Structural Safety (JCSS), formed in 1971 by the combined
effort of six civil engineering international associations: CEB, CIB, fib, IABSE and RILEM. In 2001, the JCSS
released the "Probabilistic Model Code" (JCSS, 2001), which set a framework for full probabilistic design of
engineering structures. Following the same trend, in 1998 the ISO released ISO2394: General principles on reliability
for structures. This code established equivalence between full probabilistic design and the partial safety factor design,
making room for Reliability-Based Design and Assessment (RBDA).
With these developments, structural reliability has permeated into the structural engineering profession.
Today, most engineers who perform a design using limit states and partial safety factors are designing to achieve target
reliabilities, even when they dont know about it and it is due to calibration already developed on past.
For pipelines the first applications of RBDA come from the 90s with some work developed by the United
Kingdom Health & Safety Executive (HSE) and significant developments by British Gas Research and Technology
(now GL Noble Denton) in support of pipeline pressure uprating. Limit state design of pipeline systems using
reliability-calibrated partial safety factors is also currently addressed in the ISO16708:2006 code.
Specifically for the oil industry, the API 581 Risk Based Inspection code for static equipments (e.g. vessels)
allows application of structural reliability analysis considering the limit state equations describing the most relevant
damage mechanisms, with uncertainties of independent variables described by statistical distributions. The resulting
model is solved directly for the probability of failure.
Since 2007, the Canadian Standard CSA Z662 (2011) Oil and Gas Pipeline Systems allows for limit state
design (Annex C) and for reliability-based design (Annex O). Although both annexes are informative, Clause 4.1.7 of
CSA Z662-07 states: "Steel oil and gas pipelines may be designed as specified in Annex C, provided that such designs
are suitable for the conditions to which such pipelines are to be subjected". Clause 4.1.8 of the same code states:
"Onshore pipelines for non-sour service natural gas transmission may be designed as specified in Annex O. Where such
design methods are used, the requirements of Clause O.1 shall be met in their entirety". Hence, the conditions are set
for both partial safety factor and full probabilistic design of pipeline systems.
Efforts are also underway to incorporate reliability-based design in the ASME B31.8 code (Zhou et al., 2009).
One important role of such regulations is to remove arbitrariness in the selection of suitable probability models for the
relevant variables. Hence, these codes not only allow probabilistic design but also guide the selection of thoughtfully
tested limit state functions and probability distributions that have been shown to describe the relevant random variables.
In comparison to other design methods, RBDA has the following advantages:
It gives a quantitative measure of structural safety;
Pipeline performance is assessed with respect to the most-relevant failure modes;
Different target reliability goals can be set for different failure modes and mechanisms according to their
distinct consequences;
Inspection and maintenance may be allocated in an optimized order to achieve target safety levels;
The approach is less dependent on expert experience, and can be applied to innovative designs, involving
unconventional loads or new materials;
There is a unified decision making process, integrating design and operational decisions.
There is a tendency of modern design codes to move from prescriptive to goal-oriented regulations (Zhou et
Rio Oil & Gas Expo and Conference 2012

3
al., 2009). Reliability-based design and assessment allows the design and maintenance of pipelines to move in this
direction.

3. Statistical data on past failure rates
International databases systematically and yearly report pipeline related safety and product loss incidents in different
parts of the world. In some cases, the reports group data for the last 50-odd years. Such reports are extremely relevant
because they show what are the most common failure modes in practice. They also serve as a base for investigating the
relations between different pipeline system parameters and real observed failure rates. This information is important to
understand the effectiveness of risk mitigation measures, but also to guide the design process of new pipelines. Past
failure rates are used to define acceptable safety targets, and serve as a general reference for estimating the accuracy of
future, calculated expected failure rates.
Reports by UKOPA (2011) cover pipelines in the United Kingdom from 1962 until today. The database covers
around 741 thousand kmyear of pipelines. The overall failure frequency from 1962 to 2008 was found to be 0.234
incidents per 1000 kmyear. For the five years from 2006 to 2010, this overall failure frequency had dropped to 0.093
incidents per 1000 kmyear, probably as a result of better design, operational and risk management practices. External
corrosion and external third-party interference appear as the most common causes of failure of pipelines in this
database. In the past, these failure causes were associated to many others; in the last five years external corrosion and
third-party interference accounted for more than 50% of the incidents. The report shows that failure frequencies due to
external interference decrease with increasing pipeline diameter and wall thickness. Failure frequencies due to external
corrosion decrease significantly with increasing wall thickness.
The CONCAWE (2011) report covers 35000 km of European cross-country oil pipelines, with historical data
from 1971. The long-run failure frequency for these pipelines is 0.53 spillage incidents per 1000 kmyear, and the last 5
year average is 0.28 incidents per kmyear. For the CONCAWE pipelines, the most frequent failure cause is third party
interference, both in the long run and in the past five years. Corrosion and mechanical failure are listed as the second
and third most common causes of failure. A tendency of reduction of overall spillage frequencies can also be observed
with increasing pipe diameter. This is particularly true for spillages due to third party interference.
The European Gas pipeline Incident data Group publishes the EGIG (2008) report, covering 3.15 million
kmyear of gas pipelines operation in Europe since 1970. The overall incident frequency for this database is 0.37
incidents per 1000 kmyear for the whole period, and 0.14 incidents per 1000 kmyear for the last 5 years covered by
the report (2003 to 2007). The major reported cause of incidents is third-party interference, which accounted for nearly
50% of the incidents, both in the long term and in the last 5 years. Construction defects/material failures and corrosion
accounted for 16% and 15% of all incidents, being the second and third most common causes of failure. This report
also shows that the failure frequency due to external interference is reduced with increasing pipe diameter. Moreover, it
is also shown that this failure frequency decreases with increasing depth of cover and increasing wall thickness. The
failure frequency due to corrosion also decreases with increasing wall thickness. The report also presents statistics on
ignition probability given size of leak (pinhole/crack, hole, rupture).
It is interesting to note from these figures how correlations between observations and physical parameters need
not imply physical causality. For example, it is plain that the observed reductions in corrosion and external interference
failure frequencies with increasing wall thickness are direct physical effects. On the other hand there is no reason why
increasing pipe diameter should, in itself, also reduce the failure frequency due to external interference. Rather, one can
speculate that larger-diameter pipelines tend to be laid in more remote locations where excavations are less likely to
occur.

4. Risk acceptance criteria for pipelines (ultimate limit states)
The risk r associated with an event (e.g. failure) is the product of undesirable consequences (of failure) by a probability
of failure:
c p r = (1)
where p is probability of failure and c is a measure of failure consequences. For distributed systems like pipelines,
probabilities of failure (p) are generally given as a rate of failure per kilometer and year (failures per kmyear).When
failure consequences are measured in monetary terms (cost of failure), then risk is the expected cost of failure and also
has monetary units. Societal risk, for instance, is measured in terms of expected number of fatalities. Individual risk is
measured in terms of chance of death (per year). For fixed consequences c, the maximum (admissible) risk is r
max
, and
the target reliability (R
T
) becomes:

c r p R
T
/ 1 1
max max
= =
(2)
Consequences of failure for pipelines can be measured by the number of fatalities N resulting from an incident.
For gas pipelines, this can be estimated from the exposure of people to the heat resulting from gas ignition. The number
of fatalities is proportional to:
The ignition probability, given release of product from the pipeline;
Rio Oil & Gas Expo and Conference 2012

4
Size of the exposure area;
Population density () in that area
Probability of building occupation.
Zhou et al. (2007) showed that the exposure area is proportional to PD
2
, and that ignition probability is proportional to
D, where P is pipeline pressure and D is pipe diameter. For a constant occupation probability, this leads to:

3
D P c (3)
where consequences c are given by number of fatalities.
Risk criteria for general engineering facilities are nowadays largely known and accepted. For societal risk (risk
which affects a whole population), two criteria are generally used: fixed expectation and risk aversion. For fixed
expectation, consequences of failure are considered directly proportional to the number of fatalities. Hence, the risk is
the same for an event occurring with probability p and leading to N fatalities, and for another event occurring with
probability p/10 but leading to 10N fatalities. In general, however, society is increasingly averse to events leading to
larger numbers of fatalities. Risk aversion is modelled by measuring consequences as N raised to a larger-than-one
power. Figure 1 illustrates these two risk acceptability criteria in terms of accepted failure rates (per kmyear) for gas
pipelines. The more stringent criterion is used, leading to the acceptability bounds shown in Figure 1.
Societal risk is not a good measure for pipelines located in under-populated areas, since all admissible risk
would affect the few people living or working around the pipeline. For under-populated (e.g. rural) areas, individual
risk criteria are more appropriate. This is a measure of the risk that affects each individual living or working around the
pipeline. Nessim and Zhou (2005) have developed individual risk criteria for the pipeline classes of ASME B31.8
(2003): 10
-4
for Class 1; 10
-5
for Class 2 and 10
-6
for Class 3. These individual risk criteria were also mapped to Figure
1. Description of the different classes can be found in ASME B31.8 (2003).
From the results shown in Figure 1, the following population density-based reliability targets are obtained:

( )
( )
( )
0,66
3
3 7
0,66
3
7 3 9
3
10
1,6
3
1650
1 0
197
1 1,16 10 (individual risk)
49700
1 1,16 10 7,1 10 (fixed expectancy)
4, 05 10
1
T
PD
PD
PD
R
PD
PD
PD
=
s

=
< s

3 9
7,1 10 (risk aversion) PD

>

(4)
where P is pressure in MPa, D is diameter in mm and is population density in people per hectare.

1.E09
1.E08
1.E07
1.E06
1.E05
1.E04
1.E03
1.E02
1.E01
1.E+00
1.E+06 1.E+07 1.E+08 1.E+09 1.E+10 1.E+11 1.E+12 1.E+13
T
a
r
g
e
t

r
e
l
i
a
b
i
l
i
t
y

(
p
e
r

k
m

y
e
a
r
)
PD
3
(people/hecMPamm
3
)
Individualrisk
Societalriskwith fixedexpectation
Societalriskwith aversion function
1
1
1
1
1
1
1
1
1
1


Figure 1. Reliability targets for pipelines, adapted from Nessim and Zhou (2005).

Similar reliability targets in terms of the pipeline classes of ASME B31.8 (2003) are presented by Nessim and
Zhou (2005). For service limit states and for small leaks, reliability targets are more related to economical than to safety
concerns. For small leaks in gas pipelines, the reliability target set by Nessim and Zhou (2005) is 10
-2
(failures per
Rio Oil & Gas Expo and Conference 2012

5
kmyear). For serviceability limit states (that is, limit states that do not involve loss of containment), the target of 10
-1

(failures per kmyear) is obtained from ISO 16708 (2006).

5. Failure models for corrosion and mechanical damage
One important point in RBDA is the identification of the most relevant failure modes for a given pipeline system.
Nessim and Zhou (2009) presented a methodology to identify the relevant failure modes for a specific pipeline system.
Before designing the pipeline, historical data can be used as a guide for the relevant failure modes to be considered in
design. In Section 3, it was shown that the two most relevant failure modes for onshore buried pipelines in non-sour
service, based on past observed failure rates, tend to be (external) corrosion and mechanical damage due to third party
interference. These two failure modes are addressed herein.

5.1 Failure models for external corrosion
Distributed systems like long pipelines can fail at a number of locations. Failure due to presence of a corrosion defect
can occur at any point along the pipeline where an existing corrosion defect is located. It is customary to treat multiple
corrosion defects in a long pipeline system in a bunch; that is, to evaluate a frequency of occurrence of corrosion
defects (e
c
(t) defects per km), which can grow in time due to appearance of new defects. The rate of failure due to
corrosion (
c
(t)

failures per kmyr) becomes:

( ) ( ). ( )
c c fc
t t P t = e
(5)
where P
fc
(t) is the conditional failure probability (per km) given a corrosion defect. P
fc
(t) is also a function of time
because of growth of the defect depths in time. Hence, the occurrence of failure due to corrosion can be modeled as
non-homogeneous Poisson process with arrival rate
c
(t). The cumulative failure probability becomes:

( )
0
1 exp ( )
t
f c
P t d
| |
= t t
|
\ .
}
(6)
Models for pipeline failure in presence of growing corrosion defects can be divided in two parts: one model is
necessary to estimate the burst pressure of a pipeline segment containing a corrosion defect; another model is necessary
to estimate how the corrosion defect grows in time. Models for burst pressure are described by Ahammed and Melchers
(1996), Ahammed (1998), Hong (1999), De Leon and Macias (2005), Nessim and Zhou (2009) and Zhou (2010). Many
of these models are based on the original burst pressure failure model presented by Kiefer and Vieth (1990). The
model described in Nessim and Zhou (2009) is the model incorporated in the Canadian code CSA Z662-11 (2011),
described below. For a small leak resulting from a corrosion defect, the limit state function is:

1
( ) ( )
max
g t w d t =
(7)
where w is the pipeline wall thickness and d
max
(t) is the maximum defect depth at time t. The limit state function for
burst failure of a longitudinal corrosion defect is given by:

2
( ) ( ) ( )
a
g t r t p t =
(8)
where p(t) is the operating internal pressure and r
a
is the estimated pressure for burst of the corrosion defect, which
must include the model error. The burst pressure (r
a
) was found to depend on the grade of the material, and model
errors were found to depend on the length L of the corrosion defect. The burst pressure, including model errors, is given
by:

( )
( )
1 1 0 2
3 3 0 4
( ) 1 , 241 Classes A and B
( )
( ) 1 , 241 Class X up to X60
c u
a
c y
e r t e r e s SMYS MPa
r t
e r t e r e S SMYS MPa
+ s

=

+ >

(9)
where e
i
are random variables describing model error, S
y
and S
u
are the yielding and ultimate stresses, respectively,
SMYS is the nominal minimum yield stress for the material, r
c
is the nominal, calculated burst pressure and r
0
is the
burst pressure for a pipeline segment with no corrosion defect. Burst pressure r
0
is given by:

0

1.8 , 241

2.3 , 241
u
y
w s
SMYS MPa
D
r
w s
SMYS MPa
D

>

(10)
where D is the nominal pipeline diameter. The nominal or calculated burst pressure (r
c
) is given by:

0
( )
1
( )
( )
1

a
c
a
d t
w
r t r
d t
M w
| |

|
=
|
|

\ .
(11)
where d
a
is the mean depth of the corrosion defect and M is the Folias factor, given by:
Rio Oil & Gas Expo and Conference 2012

6

( )
1/2
2 4
2
2 2
1 0.6275 0.003375 for / 50

L L
M L D w
D w D w
| |
= + s
|
\ .
(12)

( )
2
2
0.032 3.3 for / 50

L
M L D w
D w
| |
= + >
|
\ .
(13)
In Eqs. (11) to (13), defect length can grow with time, hence defect length and Folias factor become functions of time.
Model errors in Eq. (9) were found in studies developed by Advantica (now GL Noble Denton) and Battelle
(Nessim and Zhou, 2009). These studies involved the comparison of the nominal, calculated burst pressures (Eq. 9)
with burst pressures obtained by testing real corroded pipelines taken out of service. For steels of classes A and B, 38
pipeline segments were tested, and for class X, 25 segments were tested. Furthermore, it was found that model errors
depend on the length of corrosion defects. For long corrosion defects (L0.25 D), 50 real pipelines were tested. The
model errors found in the study are reported in Table 1.

Table 1. Model error statistics for burst pressure of pipeline segment containing
longitudinal corrosion defects (adapted from Nessim and Zhou, 2009).
Model
error
Distribution
Short corrosion defects Long corrosion defects
Mean St. Dev. Mean St. Dev.
e
1
Deterministic 1.04 - 1.0 -
e
2
Normal -0.00056 0.001469 -0.00228 0.00383
e
3
Deterministic 1.17 or 1.15
(1)
- 1.34 -
e
4
Normal -0.007655 0.006506 -0.00938 0.00578
(1) Use 1.17 for steel classes A and B and 1.15 for class X.

Equation (8) describes burst of a corrosion defect which results in a through-wall rupture of length L; this
characterizes what is called a large leak. Burst failure may be followed (or not) by unstable growth of the defect,
leading to what is called a rupture. The limit state function for rupture is given by:

3
1.8
( ) , 241

( )
2.3
( ) , 241

u
y
w s
p t SMYS MPa
M D
g t
w s
p t SMYS MPa
M D

>

(14)
In Eqs. (7 to 13), maximum d
max
(t) and mean defect depths d
a
(t), as well as defect length L(t), grow in time,
due to corrosion growth. In the authors' opinion, the importance given in the published literature to corrosion growth
models has not been appropriate. Most authors simply assume the corrosion rate to be represented by a random
variable, and limited information on corrosion growth (which is often all that is available in practice) is used to
calibrate such random variable corrosion rate models (e.g. corrosion measures at two times only). These random
variable models do not describe the inherent variabilities of the corrosion process in time. Pandey (1998) and Pandey et
al. (2009) present a very simple Gamma process corrosion model, where corrosion is given by stationary and discrete
increments. The model results in discrete stepped corrosion growth, which is also not very realistic. Bazn and Beck
(2012) developed a model where the corrosion rate is modeled as a Gamma process with independent increments. In
this model, corrosion rates are represented as Poisson square-wave processes: hence, the resulting corrosion growth is
continuous and the variability of the corrosion process in time is properly represented.

5.2 Failure models for mechanical damage due to third party interference
Mechanical damage due to third party interference is often the primary cause of failure for onshore pipelines, as
discussed in Section 3. The frequency of occurrence of interference events, e
i
, is generally constant in time. Hence, the
rate of failure due to interference (
i
failures per kmyr) is also constant in time:

.
i i fi
P = e
(15)
where P
fi
is the conditional failure probability given an interference event. However, the cumulative failure probability
increases with increasing exposure time:

( ) ( ) 1 exp
f i
P t t =
(16)
Failure models to evaluate the conditional failure probability due to third party interference (P
fi
) were
developed by Francis et al. (2005) and by Nessim and Zhou (2009). The model by Francis et al. (2005) is based on
elastic-plastic fracture mechanics, and assumes in general the formation of a dent-like defect with a micro-cracked
notch at the root of the dent. As this model is governed by nine parameters, some of which are difficult to evaluate, it
will not be considered herein.
Nessim and Zhou (2009) present limit state equations for different damage events that can result from the
interference of a excavator bucket with a buried pipeline. Following the authors, nearly 75% of all interference events
Rio Oil & Gas Expo and Conference 2012

7
are caused by excavators. For modern large diameter pipelines, only excavators, which produce much greater forces
than backhoes, can cause immediate failure of a pipeline. The authors classified possible failure mechanisms in two
ways: puncture, which occurs if the excavator bucket tooth penetrates pipeline wall, and dent-gouge damage, which
occurs when the force is insufficient for wall penetration but produces a dent defect with a gouge at the root. The dent-
gouge defect may lead to failure once the excavator tooth is removed, or later in time due to a peak in internal pressure.
The following model was developed at C-FER Technologies (Nessim and Zhou, 2009) and was calibrated for
w between 4.0 and 12.5mm, D between 168 and 914 mm and steel classes up to X70. The limit state for wall puncture
by an excavator bucket tooth (indentor) is:

q r g
a
=
1
(17)
where r
a
is the estimated resistance (in kN), including model error, evaluated as:

( )
e
s w m l
w
D
r
u t t
a
+
+
(


=
1000
0029 , 0 17 , 1
(18)
and q is the normal impact force in kN, estimated as:

N D
R R m q 5 , 16
6919 , 0
=
. (19)
In Eqs. (18) and (19), s
u
is the tensile strength, l
t
is the cross-sectional indentor length, w
t
is the cross-sectional width of
indentor, m is the excavator mass (in tonnes), R
D
is the dynamical impact factor (equal to 2/3), R
N
is the normal load
factor (ratio between force component normal to the pipe and the total force, following uniform distribution between 0
and 1) and e is a model error random variable, characterized by a normal distribution with mean of 0.833 kN and
standard deviation of 26.7 kN.
The limit state function for dent-gouge failure is (Nessim and Zhou, 2009):

h c
g o o =
2
(20)
where o
c
is the calculated critical hoop stress and o
h
is the hoop stress resulting from internal pressure, given by:

w
D p
h
2

= o
(21)
The calculated critical hoop stress is:

(
(

|
|
.
|

\
|
=
a
K
b
b
arc
b
IC
c

125 exp cos

2
2
2
1
2 2
2
t
t
t
o
(22)
where K
IC
is the critical stress intensity factor, measured in MPam
0.5
. The stress intensity factor for pipeline steels is
generally evaluated from Charpy energy tests as:

95 , 0
0
3 / 2
5 , 0
0
|
|
.
|

\
|

|
|
.
|

\
|
=
v
v
c
v
IC
c
c
a
c E
K
(23)
where c
v2/3
is the Charpy energy for reduced specimens of size 2/3, c
v0
is an empirical coefficient equal to 110.3 Joules,
E is the elasticity modulus and a
c
is the cross-section area of the Charpy specimens (53.6 mm
2
). Parameters b
1
and b
2
in
Eq. (22) are given as:

0
1 2
1
5.1
;
1.15 1
m
b d
m m
y
a
S
Y d M w
b S Y b
a w
s
w
| |

|
\ .
= + =
| |

|
\ .
(24)
where M is the Folias factor, given by:

2
1
2
2 /
26 , 0 1
(
(

|
|
.
|

\
|
+ =
w D
l
M
(25)
and where a and l are the gouge depth and length, respectively, d
d0
is the dent depth at zero internal pressure,
approximated by:

( )
2.381
0 0.25
1.43
0.49 0.7 /
d
t y u
q
d
l s w w p D s
(
(
=
(
+

(26)
Parameter S
m
is given by
0
1 1.8 /
m d
S d D =
, and the geometry functions Y
m
and T
b
are given by:

( ) ( ) ( ) ( )
( ) ( ) ( ) ( )
2 3 4
2 3 4
1.12 0.23 / 10.6 / 21.7 / 30.4 /
1.12 1.39 / 7.32 / 13.1 / 14.0 /
m
b
Y a w a w a w a w
Y a w a w a w a w
= + +
= + +
(27)
Rio Oil & Gas Expo and Conference 2012

8
Both puncture and dent-gouge damages are assumed to lead to sharp through-wall crack-like defects. For
puncture, initial defect length is given by indentor length; for dent-gouge damage, initial defect length is the gouge
length. If the initial defect length is larger than the critical length for unstable growth (following the criterion by
Kiefner et al., 1973), then failure is classified as a rupture. Otherwise, failure is classified as a leak.
The limit state equation for rupture is
h cr
S g o =
3
, where:

( )
( )
1
2
2 68.95
250
cos exp

68.95
y
v
cr
y c
s
E C
S B
M
L s A

( | |
+
t
( |
=
( | t
+
\ .
(28)
and B is a model error random variable, characterized by a lognormal distribution with mean of 1.0 and standard
deviation of 0.089, R is pipe radius, c is half the defect length, C
v
is the full-size Charpy energy, A
c
is the full-size
Charpy shear area, o
h
is the hoop stress (Eq. 21) and M is the Folias factor (Eqs. 12 and 13).

6. Statistical data on random problem parameters
In RBDA, a proper model for the internal operational pressure of pipelines is a Poisson pulse process with independent
pulses, where each pulse represents the annual maximum of the internal pressure at a given segment. Although pipeline
operators keep record of pipeline operation internal pressures, such data is not available to the general public. Still,
Nessim and Zhou (2009) report the following data, obtained by C-FER Technologies: the ratio between the maximum
annual internal pressure and the design operational pressure can be described by a Beta distribution with mean of 0.993
and C.O.V. of 0.034. The lower and upper bounds of the Beta distribution are given as 80% and 110% of the design
pressure. This data assumes that the pipeline is operating at full capacity (maximum operational pressure equal to the
design pressure). Jiao et al. (1995) have found that the ratio between maximum annual pressure and design pressure can
be described by a Gumbel distribution with mean between 1.03 and 1.07, and C.O.V. between 1 and 2%. This relation
is valid for those locations which operate at nominal design pressure, that is, just after pumping stations.
Statistics on mechanical properties of pipeline steels have been presented by Nessim and Zhou (2009),
including a compilation of results by other authors. These results are summarized in Table 2. Statistics on pipeline
geometry are found in the same research report (Nessim and Zhou, 2009) and are presented in Table 3.
Corrosion rates and size and number of corrosion defects are strongly dependent on pipeline location and
configuration (soil corrosivity, soil pH, type of surface finish, existence of cathodic protection, etc.). Hence, no
statistics are presented herein with respect to these problem parameters. Corrosion data for different soil/pipeline
configurations are presented in (ISO 16708, 2006; Nessim and Zhou, 2009; Caleyo et al., 2009; Zhou, 2010).
Statistical data on parameters of the mechanical damage model are not presented herein due to space
limitations. This information can be found in Nessim and Zhou (2009).

Table 2. Statistical data for mechanical properties of pipeline grade steels (adapted from Nessim and Zhou, 2009).

Variable Distribution Mean C.O.V. (%) Source
Yielding
strenght /
SMYS
(*)

Normal 1.11 3.4 Jiao et al. (1997) X60
(1)

Normal 1.08 3.3 Jiao et al. (1997) X65
(1)

N/A 1.07 1.10 2.6 3.6 Jiao et al. (1995) X60
(2)

Lognormal 1.08 4 Sotberg & Leira (1994)
(3)

Normal ou Lognormal 1.1 3.5 Dados confidenciais
(4)

Tensile strenght
/ SMTS
(*)

Normal 1.12 3.0 Jiao et al. (1995) hoop X60
(5)

Normal 1.12 3.5 Jiao et al. (1995) hoop X65
(5)

Normal 1.07 2.6 Jiao et al. (1995) axial X60
(5)

Normal 1.08 2.9 Jiao et al. (1995) axial X65
(5)

N/A 1.12 --- GRI X60
(6)

N/A 1.14 --- GRI X65
(6)

N/A 1.13 --- GRI X70
(6)

(*) SMYS: specified minimum yield stress, SMTS: specified minimum tensile strength. (1) Based on a pool of steels from
different manufacturers (760 samples for X60 and 2.753 samples for X65). (2) Based on three sets of experimental results.
(3) No information about the source. (4) Manufacturer information for X60 and X70 steels. (5) Similar to (1) above. (6)
Based on confidential data from Gas Research Institute (GRI).
Rio Oil & Gas Expo and Conference 2012

9
Table 3. Statistical data on pipeline geometry (adapted from Nessim and Zhou, 2009).

Variable Distribution Mean C.O.V. (%) Source
Diameter / nominal
diameter
Deterministic 1.0 0 Jiao et al. (1995)
Normal 1.0 0.06 ---
Wall thickness /
nominal wall
thickness
Normal 1.0 0.25/nominal Jiao et al. (1997)
Normal 1.1 3.3 Jiao et al. (1997)
Normal 1.01 1.0 ---


7. Solution methods for structural reliability analysis
The first concepts of structural reliability theory were formulated in the late sixties. In forty years of development,
structural reliability theory has reached a mature stage, encompassing solid theoretical developments and increasing
practical applications. Structural reliability methods have permeated the engineering profession, finding applications in
code calibration, structural optimization, life extension of existing structures, life-cycle management of infrastructure
risks and costs, and so on.
Structural reliability problems can be broadly divided between time variant and time-invariant problems.
Time-invariant reliability problems are easier to solve, and involve only random variables. Time-variant reliability
problems involve stochastic (loading) processes, and also resistance degradation in time due to processes like fatigue or
corrosion. Both resistance degradation processes are relevant for pipeline systems. Under continuously varying random
process loading, such problems are very challenging to solve. However, accurate results can be obtained by
representing the load (internal pressure) as a random sequence of independent load pulses (Zhou, 2010). In this case,
the load pulses represent the annual maximum internal pressure to which the pipeline is submitted. The failure rate,
conditional to one load pulse of random magnitude, increases in time due to growth of corrosion defects. The solution
to the resulting time-variant reliability problem is straightforward, and can be found for instance in Melchers (1999).
Evaluating the failure rate due to corrosion, conditional to one load pulse, at each discrete time is a time-
invariant reliability problem. Similarly, the rate of failure of a pipeline due to third party interference, conditional to the
occurrence of an interference event, is a time-invariant reliability problem. Time-invariant reliability problems
involving any number of random variables, with any type of possibly-correlated (marginal) probability distribution
functions , are straightforward to solve. The First Order Reliability Method (FORM) is one of the most efficient, but it
involves finding the most probable failure point (design point). This involves solving an optimization problem in a
transformed space of standard Gaussian random variables, which is not always straightforward. In Bazn and Beck
(2010), it was shown that solving a problem of pipeline reliability under corrosion using the FORM method has led to
severe convergence difficulties and to inaccurate results due to high non-linearity of the limit state functions. Monte
Carlo simulation is always possible, and is not expensive when limit state functions are given in closed form. In Monte
Carlo simulation, one just generates samples of the random variables, and evaluates the limit state function for each
sample. The failure probability is simply evaluated from the statistics of an indicator function, which has a value of
unity for points in the failure domain and zero elsewhere. Monte Carlo simulation is also straightforward in solution of
time-variant reliability problems.

8. Concluding remarks
The scenario for reliability-based design and assessment (RBDA) of pipeline systems is set. Risk arises from
uncertainty, and by quantifying uncertainties one can quantitatively risk-manage the process. Although probabilistic
design is unlikely to become mandatory, the authors believe that reliability-based management of pipeline risk offers
significant possibilities for economic gain without compromising safety. This review paper has showed that the ground
work for Reliability Based Design and Assessment (RBDA) of onshore pipeline systems is already developed and
offers a route for expenditure on inspection and maintenance operations to be optimized, safely and economically
managing the operational life of pipeline systems.

9. Acknowledgements
Sponsorship of this research project by the Brazilian National Council for Higher Degree Education (CAPES) and by
PETROBRAS is gratefully acknowledged.

10. References
Ahammed M, 1998, Probabilistic estimation of remaining life of a pipeline in the presence of active corrosion
defects, International Journal of Pressure Vessels and Piping, v. 75, pp. 321-329.
Rio Oil & Gas Expo and Conference 2012

10
Ahammed M, Melchers RE, 1996, Reliability estimation of pressurised pipelines subject to localised corrosion
defects, International Journal of Pressure Vessels and Piping, v. 69, pp. 267-272.API 581, 2008: Risk-Based
Inspection Technology, American Petroleum Institute.
ASME B31.8, 2003: Gas Transmission and Distribution Systems, American Society of Mechanical Engineers, NY.
ASCE 7-05, 2006: Minimum Design Loads for Buildings and Other Structures. American Society of Civil Engineering.
Bazn FAV, Beck AT, 2011: Reliability of pipelines under corrosion and mechanical damage, Applications of Statistics
and Probability in Civil Engineering. London: Taylor & Francis Group, 2285 - 2292.
Bazn FAV, Beck AT, 2012, A stochastic model for failure assessment of corroded pipelines. Proceedings of the 1
st

International Symposium on Uncertainty Quantification and Stochastic Modeling (Uncertainties 2012), 538-548.
Caleyo F, Velzquez JC, Valor A, Hallen JM, 2009, Markov chain modeling of pitting corrosion in underground
pipelines, Corrosion Science, v. 51, pp. 2197-2207.
CONCAWE, 2011: Performance of European cross-country oil pipelines; Statistical summary of reported spillages in
2009 and since 1971.
CSA Z662, 2007: Oil and Gas Pipeline Systems, Canadian Standards Association, Mississauga, ON.
De Leon D, Macas OF, 2005, Effect of spatial correlation on the failure probability of pipelines under corrosion,
International Journal of Pressure Vessels and Piping, v. 82, pp. 123-128.
EGIG, 2008: 7th Report of the European Gas Pipeline Incident Data Group.
Ellingwood B, Galambos TV, MacGregor JM, Cornell CA, 1980: Development of a Probability Based Load Criterion
for American National Standard A58, Building Code Requirements for Minimum Design Loads in Buildings and
Other Structures, Special Publication 577, National Bureau of Standards, Washington.
EUROCODE 2001: prEN 1990: Basis of Structural Design - Annex C: Basis for Partial Factor Design and Reliability
Analysis. European Committee for Standardization, Brussels, Final Draft.
Francis A, Jandu C S, Andrews R M, Miles T J, Chauhan V, 2005, Development of a new limit state function for the
failure of pipelines due to mechanical damage, PRCI/APIA/EPRG Joint Technical Meeting, Orlando
Hong HP, 1999, Inspection and maintenance planning of pipeline under external corrosion considering generation of
new defects, Structural Safety, v. 21, pp. 203-222.
ISO 16708, 2006: Petroleum and natural gas industries - Pipeline transportation systems - Reliability-based limit state
methods, International Standards Association.
ISO 2394, 1998: General principles on reliability for structures, International Standards Association.
JCSS, 2001: Probabilistic Model Code, Joint Committee on Structural Safety.
Jiao G, Sotberg T, Igland R, 1995, SUPERB 2M Statistical data: basic uncertainty measures for reliability analysis of
offshore pipelines. Report No. STF70 F95212, Superb Project No. 700411.
Jiao G, Sotberg T, Bruschi R, Igland R, 1997, The Superb Project: Linepipe Statistical Properties and Implications in
Design of Offshore Pipelines. Proceedings of the 16th International Conference on Offshore Mechanics and Arctic
Engineering (OMAE), Yokohama, Japan.
Kiefner JF, Maxey WA, Eiber RJ, Duffy AR, 1973: Failure stress levels of flaws in pressurized cylinders. Progress in
flaw growth and fracture toughness testing. ASTM STP 536, 461-481.
Kiefner JF, Vieth PH, 1990, New method corrects criterion for evaluating corroded pipe, Oil and Gas Journal 6, 56-59.
Melchers RE, 1999: Structural reliability analysis and prediction. 2nd ed. New York, USA: John Wiley & Sons.
Nessim M, Zhou W, 2005, Target Reliability Levels for the Design and Assessment of Onshore Natural Gas Pipelines.
C-FER Technologies, GRI Report No. GRI-04/0230.
Nessim M, Zhou W, 2009, Guidelines for Reliability Based Design and Assessment of Onshore Natural Gas Pipelines.
C-FER Technologies, project n L177, Final Report submitted to Pipeline Research Council International, Inc.,
Edmonton, Alberta, Canada.
Pandey MD, 1998: Probabilistic models for condition assessment of oil and gas pipelines, NDT&E Intern. 31, 349-358.
Pandey MD, Yuan X-X, van Noortwijk JM, 2009, The influence of temporal uncertainty of deterioration on life-cycle
management of structures, Structure and Infrastructure Engineering, v. 5, pp. 145-156.
UKOPA, 2011: UKOPA Pipeline Product Loss Incidents (1962-2010) 8th Report of the UKOPA Fault Database
Management Group. United Kingdom Onshore Pipeline Operators Association.
Zhou W, 2010: System reliability of corroding pipelines. Int J Pressure Vessels & Piping 87: 58795.
Zhou J, Rothwell B, Nessim M, Zhou W, 2009: Reliability-Based Design and Assessment Standards for Onshore
Natural Gas Transmission Pipelines. Journal of Pressure Vessel Technology (ASME), Vol. 131.
Zhou J, Rothwell B, Nessim M, Zhou W, 2007: Review of rule based design and reliability based design for onshore
pipelines, 16th Joint Technical Meeting, Canberra, Australia.
Zhou W, 2010: System reliability of corroding pipelines, Int. Journal of Pressure Vessels and Piping 87, 587-595.

You might also like