International Journal of Pressure Vessels and Piping 104 (2013) 76e82

Contents lists available at SciVerse ScienceDirect

International Journal of Pressure Vessels and Piping

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

Review

Maintenance planning under imperfect inspections of corroded pipelines

Yacine Sahraoui a, Rabia Khelif a, *, Alaa Chateauneuf b, c
a
LR3MI, Mechanical Engineering Dept., Badji Mokhtar University, BP 12, Annaba 23000, Algeria
b
Clermont Université, Université Blaise Pascal, Institut Pascal, BP 10448, F-63000 Clermont-Ferrand, France
c
CNRS, UMR 6602, Institut Pascal, F-63171 Aubière, France

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

Article history: In this paper, a maintenance policy is proposed for pipelines subject to corrosion, by taking into account
Received 9 June 2012 imperfect inspection results. The degradation of the pipeline is induced by uniform and pitting corrosion,
Received in revised form leading to losses in the pipe wall thickness. The inspection is applied to detect the corrosion defects,
15 January 2013
namely the corrosion depth and width. The inspection has a detection threshold under which no cor-
Accepted 30 January 2013
rosion can be measured. Due to uncertainties, each inspection is affected by the probability of detecting
small defects and the probability of wrong assessment in terms of defect existence and size. The present
Keywords:
work aims at integrating imperfect inspection results in the cost model for corroded pipelines, where the
Inspection
Maintenance
failure probabilities are computed by reliability methods. A numerical application on a gas pipe shows
Reliability the influence of inspection quality and cost on the choice of the optimal maintenance planning.
Corrosion Ó 2013 Elsevier Ltd. All rights reserved.
Pipelines

1. Introduction Despite the large progress in cathodic protection techniques,

and internal and external coating methods, the corrosion of pipe-
Due to degradation, the pipelines undergo various maintenance lines remains among the major difficulties facing the management
operations, such as: regular inspections to assess degradation, of gas and oil networks.
repair actions to recover existing damage, overall repair to bring the Two main methodologies are usually applied to deal with
structure close to its initial state, and finally complete replacement pipeline failure. The first one is a traditional code-based deter-
when the structure becomes technically inappropriate for oper- ministic approach, such as the codes AMSE B31G [5] and the
ation. From the safety point of view, the optimal maintenance modified B31G [6], which are applied to determine the failure
program should be defined on the basis of the minimum acceptable pressure. The second methodology is known as the probabilistic
level of failure probability [1]. From the cost-benefit point of view, approach, where the stochastic character of structural and envi-
the optimal maintenance planning should be defined on the basis ronmental variables is considered and the failure probability of the
of minimum expected cost [2,3]. pipeline is evaluated at various stages of operation [7e10].
During service, the remaining strength of pipeline depends of As the real degradation state can only be known at the
a number of factors, including the operational conditions, defects inspection time, the optimization of the intervals between
introduced by construction, corrosion and ground movement. In inspections or non-destructive testing (NDT) becomes a funda-
particular, corrosion constitutes an important cause resulting in mental problem for maintenance quality and safe operation of
pipeline failure [4]. The corrosion is active in all facilities, especially pipelines. In practice, the inspection of existing pipelines cannot
in gas and oil industries where severe effects are observed. While be perfect and their performance can be defined in terms of
ensuring production, transport and distribution, the gas and oil probability of detection (PoD) and probability of false alarm (PFA)
pipelines are often located in aggressive environments which [11e14]. These probabilities are generally sufficient to describe the
cannot be easily reproduced in laboratory to offer precise predictive inspection results and can therefore be considered in the decision
models. In addition, the pipelines are often embedded and un- process [15]. The application of Risk Based Inspection methods
available for observations, leading to high inspection costs. (RBI), on the basis of structural reliability analysis, failure con-
sequence assessment and probabilistic modeling of inspection
results, allows us to establish and to optimize the maintenance
* Corresponding author. Tel./fax: þ213 550105216. policies of aging installations by satisfying safety and availability
E-mail address: [email protected] (R. Khelif). requirements [16e22].

0308-0161/$ e see front matter Ó 2013 Elsevier Ltd. All rights reserved.
http://dx.doi.org/10.1016/j.ijpvp.2013.01.009
 Y. Sahraoui et al. / International Journal of Pressure Vessels and Piping 104 (2013) 76e82 77

standard deviation are, respectively, 0.066 and 0.037 for the mul-
tiplier kUC, and 0.53 and 0.14 for the power n [23].

2.1.2. Localized corrosion

In the past decades, extensive researches have been performed
on localized corrosion, in order to derive empirical models for
Fig. 1. Idealized corrosion defect. different environmental conditions. The mostly used empirical
equations for the time-dependent corrosion depth dLC(t) and length
lLC(t) take the form [9,25]:
The present work aims at developing a complete procedure for
inspection optimization of pipelines, including predictive degra-
dation modeling, time-dependent reliability assessment, inspec-
dLC ðtÞ ¼ ðkUC þ aLC DkLC Þt n (3a)
tion uncertainties and expected cost minimization. The originality
of this work lies mainly in the coupling of the above models in
lLC ðtÞ ¼ gLC ðkUC þ aLC DkLC Þt n (3b)
order to allow for practical application on engineering pipes. First,
the degradation model of pipelines under corrosion is described.
where dLC(t) and lLC(t) are respectively the localized corrosion depth
Then, the maintenance model is developed according to the deci-
and length, DkLC is the specific rate of localized corrosion, aLC is the
sion tree. Finally, the developed model is applied to gas pipelines
localized corrosion fraction and gLC is the length-to-depth ratio of
under various corrosion rates.
localized corrosion.

2. Degradation model
2.1.3. Failure pressure
The failure of pipes is identified by using a semi-empirical
2.1. Pipeline corrosion
model based on fracture mechanics to determine the pressure at
which the vessel fails as a function of the size and the geometry of
The corrosion distribution in pipes can be stochastically
the corrosion defect. This approach has been developed in the early
described by spatial random fields. Although the corrosion geom-
seventies (e.g. see Maxey et al. [26] and Kiefener et al. [27], and still
etry is usually described by either uniform corrosion or localized
remain the most widely applied method. It is furthermore used in
corrosion, most of corrosion problems encountered in the real
several standards, such as the B31G [5,6]. A review of the approach
world are a combination of these two forms. Consequently the total
and different proposed modifications is given by Ahammed and
corrosion depth at any location x and time t can be described by the
Melchers [7]. The model presented here is based on the work in
sum of these two types of corrosion:
Refs. [7], but modified by considering both local and uniform cor-
rosion. The final failure can occur as either leakage or rupture,
dC ðx; tÞ ¼ dU ðtÞ þ dLC ðx; tÞ (1)
depending on the size of the through-wall defect. The failure
where dC(x,t) is the total corrosion depth at the location x and time pressure expression is written as:
t, dLC(x,t) is the depth of the localized corrosion defect and dU(t) is
the depth of the uniform corrosion. Fig. 1 shows a longitudinally- dLC
oriented surface corrosion defect in the wall of a pressurized ðd  dUC Þ 1
Pr ¼ 2Sf dd (4)
pipeline. In this Figure, d is the pipe wall thickness, dLC is the depth D dLC
1
of localized corrosion, dUC is the depth of uniform corrosion and lLC ðd  dÞM
is the length of the corroded region projected on the longitudinal
axis. where D is the pipe diameter, Sf is the flow stress, defined by
multiplying the yield stress by a factor mf: Sf ¼ mffy. Ahammed and
2.1.1. Uniform corrosion Melchers [7] proposed to model mf by a lognormal distribution
A practical engineering way to account for uniform corrosion is with mean value equal to 1.1 and coefficient of variation of 0.05. The
to use a power law to model the loss of wall thickness with the Folias factor M (also known as bulging factor) is a semi-empirical
exposure time [8,23]. The general form of the corrosion power law factor that covers the fracture mechanics aspects; it is given by
is written as: Ref. [7]:

dUC ðtÞ ¼ kUC t n (2)

sffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
l2LC l4LC l2LC
M ¼ 1 þ 0:6275  0:003375 ;  50
Dðd  dUC Þ D ðd  dUC Þ2
2 Dðd  dUC Þ
(5)
l2LC l2LC
M ¼ 0:0032 þ 3:3; > 50
Dðd  dUC Þ Dðd  dUC Þ

where dUC(t) is the thickness of the corroded layer in mm, t is the 2.2. Pipeline reliability
elapsed time (i.e. age of the pipe in years) and kUC and n are the
corrosion constants, to be evaluated by fitting Eq. (2) from field The reliability analysis allows us to understand how the un-
corrosion data [23,24]. For atmospheric pressure, the mean and certainties are propagating within the structural system, and hence
78 Y. Sahraoui et al. / International Journal of Pressure Vessels and Piping 104 (2013) 76e82

Fig. 2. Idealized decision tree in RBI.

provides mandatory information for inspection planning [24,28]. In Z

reliability analysis, it is necessary to define a limit state function Pf ðtÞ ¼ P½GðXi ; tÞ  0 ¼ fXi ðxi ; tÞdxi (8)
(also known as performance function) which expresses the condi- GðXi ;tÞ0
tion of ‘system failure’. The limit state function G(xi) corresponds to
the safety margin, which is conventionally defined by the differ- where P½ is the probability operator and fXi ðxi ; tÞ is the joint
ence between the pipe resistance Pr and the applied pressure Pa: probability density function of the random variables at time t. As
the limit state function is nonlinear in terms of the input random
variables, iterative algorithms [29] are conveniently applied in or-
Gðxi ; tÞ ¼ Pr ðxi ; tÞ  Pa ðxi Þ (6) der to reduce the computation time. In the present work, the First
Order Reliability Method, known as FORM [29], is applied to
where xi are the realizations of the random variables of the pipe Xi . compute the reliability index b, which is defined as the minimum
This margin is defined such that Gðxi ; tÞ > 0 indicates safety and distance between the median point and the failure domain in the
Gðxi ; tÞ  0 corresponds to conventional failure. By introducing the standard Gaussian space. This index is evaluated by solving iter-
expression of Pr given by Eq (4), the above equation becomes: atively the constrained optimization problem:

sffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
X   2  
1
dLC ðtÞ bðtÞ ¼ minimize Ti x j subject to G xj ; t  0 (9)
ðd  dUC ðtÞÞ d  dUC ðtÞ i
Gðxi ; tÞ ¼ 2mf fy  Pa (7)
D dLC ðtÞ
1 where Ti(xj) is the probabilistic transformation of the model vari-
ðd  dUC ðtÞÞMðtÞ
ables to standard Gaussian variables. The solution of this opti-
The failure probability Pf(t) is evaluated by: mization problem can be obtained by optimization or reliability
 Y. Sahraoui et al. / International Journal of Pressure Vessels and Piping 104 (2013) 76e82 79

Table 1
Conditional probabilities in terms of the pipe conditions and inspection result.

Event Defect Inspection result Probability

S1 Exists Defect is not detected P(S1) ¼ P({x ¼ 1}|{d(x) ¼ 0})
S2 Exists Defect is observed with P(S2) ¼ P({x ¼ 1}|{d(x) ¼ 1})
acceptable size
S3 Exists Defect is observed P(S3) ¼ P({x ¼ 1}|{d(x) ¼ 1})
S4 Does not Defect is observed P(S4) ¼ P({x ¼ 0}|{d(x) ¼ 1})
exist
S5 Does not Defect is observed with P(S5) ¼ P({x ¼ 0}|{d(x) ¼ 1})
exist acceptable size
S6 Does not Defect is not detected P(S6) ¼ P({x ¼ 0}|{d(x) ¼ 0})
exist

algorithms [30,31]. The failure probability can thus be evaluated by Fig. 3. Effect of active corrosion on probability of failure of an underground pipe.
the first order approximation:

Pf ðtÞ ¼ P½GðXi ; tÞ  0 ¼ FðbðtÞÞ (10)

ð1  PoDÞg
where FðÞ is the cumulative Gaussian probability function. P ð S1 Þ ¼ (13)
ð1  PoDÞg þ ð1  PFAÞð1  gÞ

3. Maintenance model PoDg

P ð S2 Þ ¼ (14)
PoDg þ PFAð1  gÞ
3.1. Bayesian modeling of inspection results
PoDg
P ðS3 Þ ¼ (15)
The imperfect inspection results may lead to wrong decisions PoDg þ PFAð1  gÞ
concerning cost and safety. The corrosion defect may not be
observed accurately, depending on the applied NDT technique, PFAð1  gÞ
P ðS4 Þ ¼ (16)
operator performance and external noise affecting measured sig- PoDg þ PFAð1  gÞ
nals [11]. The decision tree (Fig. 2) can be set to describe the
Bayesian formulation of the inspection process. The following PFAð1  gÞ
conditional probabilities P(Si) (Table 1) are given in terms of the P ðS5 Þ ¼ (17)
PoDg þ PFAð1  gÞ
pipe conditions (x) and inspection result d(x):
Naturally, some of these probabilities are complementary. The ð1  PFAÞð1  gÞ
P ðS6 Þ ¼ (18)
inspection results can be entirely characterized by the probability of ð1  PoDÞg þ ð1  PFAÞð1  gÞ
detection (PoD) and by the probability of false alarm (PFA):
with g ¼ P({X ¼ 1}) the probability of defect existence at the in-
PoD ¼ P ðfdðxÞ ¼ 1gjfx ¼ 1gÞ spection time.
¼ P ðfðdðt Þ þ N Þ  dmin gjfx ¼ 1gÞ (11)
3.2. Cost model

PFA ¼ P ðfdðxÞ ¼ 1gjfx ¼ 0gÞ ¼ P ðfN  dmin gjfx ¼ 0gÞ

The optimization of inspection plan is based on the mini-
(12) mization of the total expected cost. This total cost is given by the
expected costs related to each branch of the decision tree over
where d(t) is the defect size at time t, N is the measurement noise
a given time span [17,20].
and dmin is the detection threshold, specified by the inspection
technique. The probabilities P(Si) can now be expressed in terms of X
min E½CT  ¼ min Ci Pi (19)
these two probabilities, as following:

where E[CT] is the mathematical expectation of the total cost CT, Pi is

Table 2 the probability of occurrence of the i-th scenario and Ci is the total
Statistical data of uncertain variables. cost of the i-th scenario. The expected total cost E[CT], Eq. (19), is
Variable Symbol Corrosion Mean Standard decomposed as following:
case value deviation
Diameter D (mm) 600 18 E½CT  ¼ E½CIN  þ E½CR  þ E½CF  (20)
Wall thickness t0 (mm) 10 0.5
Yield stress fy (MPa) 423 28.3
Internal pressure p (MPa) 5 0.5
Length-to-depth ratio gLC 10 2
of localized corrosion Table 3
Uniform corrosion rate kUC Low 0.14 0.014 Cost data and maintenance thresholds.
Moderate 0.37 0.037
High 0.62 0.062 Item Symbol Value
Localized corrosion rate DkLC Low 0.14 0.028 Failure cost CF 230000 V
Moderate 0.37 0.074 Inspection cost Ci 3500 V
High 0.62 0.124 Repair cost CR 10400 V
Corrosion parameter n 0.53 e Detection threshold dmin 1.5 mm
Localized corrosion fraction aLC 1.00 e Size of critical defect dC 2.8 mm
80 Y. Sahraoui et al. / International Journal of Pressure Vessels and Piping 104 (2013) 76e82

Fig. 5. Influence of inspection cost on the total expected cost.

Fig. 4. Various costs in terms of time between inspections.

where E[CIN] is the expected inspection cost, E[CR] is the expected

repair cost, and E[CF] is the expected failure cost. Referring to Fig. 2,
these expected costs can be calculated as following:

 Expected failure cost:

(a)

E½CF  ¼ CF PF ðTÞ1 þ ð1  PF ðT1 ÞÞ

2 3
C1:1 PF1:1:Act PðS1 Þ þ C2:1 PF2:1:Act PðS2 Þ
 4 þC3:1 PF3:1:Act PðS3 Þ þ C4:1 PF4:1:Act PðS4 Þ 5 (21)
þC5:1 PF5:1:Act PðS5 Þ þ C6:1 PF6:1:Act PðS6 Þ

where C1.1 ¼ C2.1 ¼ C3.1 ¼ C4.1 ¼ C5.1 ¼ C6.1 ¼ CF are the direct failure
costs, CR is the repair cost, PF(T1) is the failure probability at time T1
and PFi.1.Act is the updated failure probability at time T1 for the i-th
scenario. The updating is performed because inspection results
improve the knowledge about the system degradation.
(b)
 Expected repair costs

E½CR  ¼ CR PðS3 Þð1  PF ðT1 ÞÞ þ CR PðS4 Þð1  PF ðT1 ÞÞ

(22)
¼ CR ð1  PF ðT1 ÞÞ½PðS3 Þ þ PðS4 Þ

 Expected inspection costs

 
E½CIN  ¼ 1  PF ðTÞ1
2 3 (c)
Ci þ C1:2 ð1  PF1:1:Act ÞPðS1 Þ þ C2:2 ðð1  PF2:1:Act ÞPðS2 Þ
 4 þC3:2 ð1  PF3:1:Act ÞPðS3 Þ þ C4:2 ð1  PF4:1:Act ÞPðS4 Þ 5
þC5:2 ð1  PF5:1:Act ÞPðS5 Þ þ C6:2 ð1  PF6:1:Act ÞPðS6 Þ
(23)

with Ci the inspection cost and C1.2 ¼ C2.2 ¼ C3.2 ¼ C4.2 ¼ C5.2 ¼
C6.2 ¼ Ci.

Table 4
Influence of inspection cost.

Inspection cost Optimal inspection interval Minimum expected cost

0.1Ci 30 yr 406.8 V/yr
Ci 34 yr 822.6 V/yr
10Ci 40 yr 4226.0 V/yr Fig. 6. Influence of the quality of detection of the expected total cost: (a) 0.1Ci, (b) 1Ci
and (c) 10Ci.
 Y. Sahraoui et al. / International Journal of Pressure Vessels and Piping 104 (2013) 76e82 81

Fig. 7. Expected costs as a function of perfect inspection time.

Fig. 8. Expected total costs for various corrosion rates.

4. Application to corroded pipe

4.1. Optimal time between inspections
The above model is now applied to steel pipe of grade X52, with
diameter D and wall thickness t, and subjected to internal pressure In the case of moderate corrosion rate, Fig. 4 depicts various
p. This pipe is designed for a service time of 50 years. The proba- expected costs (failure E[CF], inspection E[CIN] and repair E[CR]) in
bilistic data are provided in Table 2, where all variables are log- terms of time interval between inspections. Naturally, by increase
normally distributed. time between inspections, the inspection and repair costs decrease,
By using the corrosion model defined in Section 2, it can be while failure cost increase. The optimal time interval between in-
possible to compute the failure probability as a function of the pipe spections is obtained at the minimum expected total cost E[CT]
age. For the three adopted corrosion conditions in Table 2, Fig. 3 which is equal to 822.6 V/yr at 34 years.
plots the failure probability Pf(t) computed by the First Order
Reliability Method [29]. For low corrosion rate, the failure proba-
4.2. Effect of inspection cost
bility is very low all over the service time of 50 years. When the
corrosion rate is high, the failure probability strongly increases to
In order to show the effect of inspection cost on the time in-
reach almost 1.00 after 25 years of service. Moderate corrosion rate
terval, three values are considered: 0.1Ci, Ci and 10Ci. Fig. 4 shows
leads to failure probability of 0.05 at the end of the service life.
that the expected total cost and the optimal interval between in-
These curves provide a clear picture of the level of nominal risk of
spections increase with the inspection cost, whatever the chosen
pipeline failure at various stages of service life.
time interval. Table 4 provides the optimal inspection times and
In addition to variables in Table 2, the maintenance policy re-
costs. It can be seen that the cost is reduced to 406.8 V/yr when the
quires the estimation of various costs for failure, inspection and
inspection cost becomes 0.1Ci. On the opposite, a very high increase
repair, as well as the minimum detectable defect dmin and the
of the total cost (up to 4226 V/yr) is observed when the inspection
defect size at conventional failure dC; these parameters are given in
cost is high (i.e.10Ci), which can be explained by the fact that
Table 3. The critical defect size dC corresponds to the failure prob-
delaying the costly inspection/repair leads to disproportional
ability of 102, with respect to the limit state function in Eq. (7).
increase of the failure cost (Fig. 5).
In the following sections, the maintenance planning opti-
mization is carried out with respect to the three following
4.3. Inspection quality
considerations:

The inspection quality is now considered in terms of cost and

- Effect of time span between inspections;
detection level. Four detection precisions are considered: low
- Influence of the applied inspection technique cost and quality;
dmin ¼ 2.0 mm, medium dmin ¼ 1.5 mm, high dmin ¼ 1.0 mm, perfect
- Effect of the corrosion rate on the inspection planning.
dmin ¼ 0.0 mm. For a given inspection cost, a better precision leads
to reduce the time interval between inspections. For the three costs,
Table 5 Fig. 6aec shows that the expected total cost increases with the
Results for different inspection costs. detection quality. It can be observed that the curves resulting from
low qualities are more scattered. This can be explained by the lack
Inspection Inspection Optimal Minimum
cost quality inspection expected cost of precision leading to high overall uncertainties. In order to allow
interval for a better understanding, Fig. 7 is plotted for the case of perfect
0.1Ci Low precision 30 yr 401.5 V/yr inspections where the curves are less spread than in Figs. 4 and 5.
Medium precision 30 yr 406.8 V/yr
High precision 30 yr 468.9 V/yr
Perfect inspection 25 yr 474.9 V/yr Table 6
Ci Low precision 35 yr 775.8 V/yr Inspection times and total cost for different corrosion rates.
Medium precision 34 yr 822.6 V/yr
High precision 30 yr 887.8 V/yr Inspection Optimal Minimum
Perfect inspection 30 yr 895.3 V/yr cost inspection expected cost
10Ci Low precision 40 yr 4064 V/yr interval
Medium precision 40 yr 4226 V/yr Lowcorrosion rate e e
High precision 35 yr 4712 V/yr Moderatecorrosion rate 34 yr 822.6 V/yr
Perfect inspection 36 yr 4826 V/yr High corrosion rate 10 yr 2445 V/yr
82 Y. Sahraoui et al. / International Journal of Pressure Vessels and Piping 104 (2013) 76e82

As a matter of fact, the variation of the expected total cost increases [8] Ahammed M, Melchers RE. Probabilistic analysis of underground pipe-
lines subject to combined stresses and corrosion. Eng Struct 1997;19(12):
rapidly around the optimum value of 30 years. For different in-
988e94.
spection costs and qualities, Table 5 indicates the optimal times and [9] Ahammed M. Probabilistic estimation of remaining life of a pipeline in the
total costs. In this example, the impact of inspection cost is much presence of active corrosion defects. Int J Press Vessels Piping 1998;75:321e9.
higher than quality, but this observation is strongly dependent on [10] Melchers RE. The effect of corrosion on the structural reliability of steel off-
shore structures. Corros Sci 2005;4:655e64.
the provided values and data. [11] Rouhan A, Schoefs F. Probabilistic modeling of inspection results for offshore
structures. Struct Saf 2003;25:379e99.
4.4. Influence of corrosion rates [12] Straub D. Probabilistic modeling of non-destructive testing of steel structures.
In: Proceedings of the 4th international Ph.D, symposium in civil engineering,
vol. 2; 2002. p. 311e20. Munich.
In order to analyze the influence of corrosion rate on the optimal [13] Straub D, Faber MH. Modeling dependency in inspection performance. In:
interval between inspections, the expected total cost is plotted in Der Kiureghian, Madanat, Pestana, editors. Applications of statistics and
probability in civil engineering. Rotterdam: Millpress; 2003. p. 1123e30.
Fig. 8 for the three corrosion rates provided in Table 2. As expected, [14] Pakrashi V, Schoefs F, Memet JB, O’Connor A. ROC dependent event isolation
the optimal inspection time is strongly dependent on the corrosion method for image processing based assessment of corroded harbour struc-
rate, as it highly decreases with the environment aggressiveness. In tures. Struct Infrastruct Eng 2010;6(3):365e78.
[15] Sheils E, O’Connor A, Breysse D, Schoefs F, Yotte S. Development of a two-
case of high corrosion rate, the optimal interval is about 10 years stage inspection process for the assessment of deteriorating infrastructure.
and 34 years in the case of a moderately aggressive (Table 6). The Reliab Eng Syst Saf 2010;95:182e94.
optimal time is beyond the service life when the corrosion rate is [16] Madsen HO, Sørensen JD, Olesen R. Optimal inspection planning for fatigue
damage of offshore structures. In: Proceedings of the 5th ICOSSAR, vol. 3;
low, and no inspection should be scheduled in this lifetime.
1989. p. 2099e106. San Francisco.
[17] Goyet J, Faber MH, Paygnard JC, Maroini A. Optimal inspection and repair
5. Conclusions planning: case studies using IMREL software. In: Proceedings of the 13th
offshore mechanics and arctic engineering conference, vol. 2; 1994. p.
325e33.
This work presents a complete approach for Inspection-Repair [18] Straub D, Faber MH. Risk based inspection planning for structural systems.
policy of corroded pipelines allowing to take into account the er- Struct Saf 2005;27:335e55.
rors in inspection results. The proposed procedure allows us to [19] Faber MH, Engelund S, Sørensen JD, Bloch A. Simplified and generic risk based
inspection planning. In: Proceedings of the 19th offshore mechanics and arctic
compare various strategies, by comparing the effectiveness of var- engineering conference, New Orleans; 2000.
ious inspection techniques and frequencies. The formulations of the [20] Goyet J, Straub D, Faber MH. Risk based inspection planning for offshore in-
expected cost in different situations are suggested and it concludes stallations. Struct Eng Int 2002;12(3):200e8.
[21] Faber MH, Straub D, Goyet J. Unified approach to risk based inspection
with an illustration to decide on the optimal frequency of in- planning for offshore production facilities. J Offshore Mech Arctic Eng Trans
spections for different rates of corrosion. The numerical application ASME 2003;125(2):126e31.
shows the coherence of the proposed model as well as its capacity [22] Straub D, Faber MH. Computational aspects of risk based inspection planning.
Comput-Aided Civ Infrastruct Eng 2006;21(3):179e92.
to take account for practical inspection planning. [23] Kucera V, Mattsson E. Atmospheric corrosion. In: Mansfeld F, editor. Corrosion
mechanics. New York: Marcel Dekker; 1987.
References [24] Amirat A, Chateauneuf A, Chaoui K. Reliability assessment of underground
pipelines under the combined effect of active corrosion and residual stress. Int
[1] Han PH. Inspection and maintenance planning of pipeline under external J Press Vessels Piping 2006;83(2):107e17.
corrosion considering generation of new defects. Struct Saf 1999;21:203e22. [25] Qian G, Niffenegger M, Li S. Probabilistic analysis of pipelines with corrosion
[2] Laggoune R, Chateauneuf A, Aissani D. Opportunistic policy for optimal pre- defects by using FITNET FFS procedure. Corros Sci 2011;53:855e61.
ventive maintenance of a multi-component system in continuous operating [26] Maxey WA, Kiefener JF, Eiber RJ, Duffy AR. Ductile fracture initiation, propa-
units. Comput Chem Eng 2009;33:1499e510. gation and arrest in cylindrical vessels. In: Fracture toughness, proceedings
[3] Laggoune R, Chateauneuf A, Aissani D. Impact of few failure data on the national symposium on fracture mechanics, part II, ASTM STP 536. American
opportunistic replacement policy for multi-component systems. Reliab Eng Society for Testing and Materials; 1972. p. p.70e81.
Syst Saf 2010;95:108e19. [27] Kiefner JF, Maxey WA, Eiber RJ, Duffy AR. Failure stress levels of flaws in
[4] Baker M. Stress corrosion cracking studies, integrity management program pressurized cylinders, progress in flaw growth and fracture toughness testing,
DTRS56-02-D-70036. Department of Transportation, Office and Pipeline ASTM STP 536. American Society for Testing and Materials; 1973. p. 461e81.
Safety; 2004. [28] Khelif R, Chateauneuf A, Chaoui K. Reliability-based assessment of polyeth-
[5] ASME-B31G. Manual for determining the remaining strength of corroded ylene pipe creep lifetime. Int J Press Vessels Piping 2007;84:697e707.
pipelinesda supplement to ASME B31G code for pressure piping. NewYork: [29] Ditlevsen O, Madsen HO. Structural reliability methods. New York:
American Society for Mechanical Engineer; 1991. Wiley; 1996.
[6] ASME-B31G. Manual for determining the remaining strength of corroded [30] Yan-Gang Z, Tetsuro O. A general procedure for first/second-order reliability
pipelinesda supplement to ASME B31G code for pressure piping. NewYork: method (FORM/SORM). Struct Saf 1999;21:95e112.
American Society for Mechanical Engineer; 1995. [31] Castillo E, Sarabia JM, Solares C, Gomez P. Uncertainty analyses in fault trees
[7] Ahammed M, Melchers RE. Reliability estimation of pressurized pipelines subject and Bayesian networks using FORM/SORM methods. Reliab Eng Syst Saf 1999;
to localized corrosion defects. Int J Press Vessels Piping 1996;69:267e72. 65:29e40.