Practical Methodology of Predictive Maintenance For Pipelines
Practical Methodology of Predictive Maintenance For Pipelines
Practical Methodology of Predictive Maintenance For Pipelines
net/publication/267599041
CITATIONS READS
4 621
2 authors:
Some of the authors of this publication are also working on these related projects:
All content following this page was uploaded by Anna Viktorovna Bushinskaya on 29 July 2015.
IPC2010-31197
1
Copyright © 2010 by ASME
1. INTRODUCTION the remaining lifetime (RL) for each individual defective pipe
The ultimate goal of pipeline maintenance, including integrity run [6], [12].
and risk control is to minimize overall maintenance/monitoring/ Historically predictive maintenance, when introduced, did not
inspection costs and breakdown consequences. In order to stand up to the expectations because, in the first place, there
achieve this goal, some maintenance principles have to be used were no means created to define the levels of pipeline quality
[16], [12]. degradation, which should trigger the maintenance action.The
There exist three main principles of systems maintenance ([19], comprehensive theory of how to assign the levels that trigger
[15]). One is based on the notion of inherent resource, which is inspection/monitoring and maintenance was developed in [19],
a random variable or function. The other two use reliability [16]. For more details see [18]-[21], [15], [16], [13], [12]. In
level or current condition of the system. this paper a real life case is described using some of the ideas
Historically pipelines were first used with respect to their from of the above sources.
design lifetime (the least economical method), as being highly
critical infrastructures, but poorly monitored due to lack of 2. CRITERIA FOR PIPELINE RESIDUAL LIFE
diagnostic instruments at that time [7]. Pipelines failures lead to ASSESSMENT
great economic losses (including pollution of environment) In this paper assessment of the pipeline residual life was
and, sometimes, to injuries and/or loss of human lives. The performed using several pipeline limit states (LS) criteria. They
design life of pipelines is chosen based on the size of the are presented in Table 1.
deposit, and on the forecast rate of oil/gas extraction. Pipelines
should not be allowed to fail in a manner that causes spill or Table 1. Criteria used for assessing pipeline remaining life
pollution of the environment, injuries or casualties. Therefore, LS Criteria Description
reactive maintenance is the worst choice of maintenance for Pipeline probability of failure POF is equal to
POF = Qth
pipelines, although it still takes place in a number of cases. the ultimate permissible POF Qth.
The pipelines should be operated taking into account their The depth of the most dangerous defect dd is
dd = 80%wt
current condition and purpose of usage. Pipelines should be equal to 80% of the pipe wall wt.
The maximal safe operating pressure SMOP is
maintained when their condition becomes warning or a hazard SMOP = МAОР
equal to the maximal allowable operating
issue. This is proactive maintenance. There are different kinds (ERF = 1)
pressure (МAОР).
of proactive maintenance – preventive, reliability based, The depth of the most dangerous defect dd is
condition based, predictive, etc. The most suited method of dd = 100%wt
equal to 100% of the pipe wall wt.
pipeline maintenance is based on monitoring their POF The failure pressure is equal to the maximal
(reliability (POF) and remaining life [13], [15]-[17]). MAOP = Pf
allowable operational pressure
This approach can also be adopted for maintaining pipelines,
with some adjustments that take into account the main specifics 2.1. Assessment of the pipeline POF
of pipelines and the available technology for their inspection. Pipeline probability of failure (POF) assessment is performed
The main specifics of pipeline diagnostics are: high cost of using the Gram-Charlier-Edgeworth method [14]. In this
diagnostics and verification; lack of necessary and sufficient method both the leak and the rupture criteria of pipeline
numbers of accurate measurements; absence of consistent integrity loss are utilized. Several independent methods of
transparent recommendations on how to analyze results of pipeline reliability assessment are implemented while using the
measurements containing intrinsic errors. following worldwide accepted pipeline design codes: B31G,
Remaining lifetime (RL) is a conditional random time of modified B31G, Battelle, DNV, and Shell-92.
transition of a pipeline segment from its present state to a limit In order to obtain consistent and authentic assessments of
state. RL is a random variable (RV), because it depends on reliability of real life pipelines both types of possible pipeline
many parameters random by nature and on a set of restrictions, integrity loss failure (leak and rupture) are taken into account
rules and decisions imposed by the pipeline performance and (with maximal possible completeness), as well as the
ILI technology. For pipelines all limit states can be subdivided probabilistic character of pipe geometry, pipe material
into deformation type limit states (elastic-plastic deformation) properties, sizes of defects which are present in pipe wall
and integrity-loss limit states (leak, rupture). For risk analysis thickness, operating pressure, corrosion rates, etc.
the main limit state is integrity-loss limit states (leak or The following main uncertainties which influence pipeline
rupture). integrity are considered as random variables (RVs): the defects’
For a full description of a RV it is necessary and sufficient to depth and length; pipe wall thickness and diameter; pipe
construct its probability density function (PDF). In order to material SMYS; radial and longitudinal corrosion rates (needed
construct the RL PDF, it's necessary to have the following: when prognosis is made for a certain time depth) and MOP.
degradation models; algorithms for calculating the conditional The number of variables considered random can change
PDF's; a full group of assumed future performance scenarios depending on time and the issue in consideration.
(in the form of event / decision / fault trees), and the relevant The method constitutes itself as an assessment of the
database. Maximum economic value is obtained by assessing probability that the limit states function (LSF) for a pipe cross
section is a positive value at time t. The LSF is constructed as
2
Copyright © 2010 by ASME
the difference between the value of the function which corresponding parameters of the defects, as given by the first
describes the burst pressure, and the value of the operating and second ILI, divided by the time between these ILIs:
pressure. For full description of the GSE reliability assessment p pP
method Authors refer the Reader to paper [14]. CR L (4)
tL tP
2.2. Estimated Repair Factor Where pP , pL are, correspondingly, the parameters of the defect
The Estimated Repair Factor is defined as: as measured during the last and the previous ILIs; tP , tL are the
MAOP times of conducting, correspondingly, the previous and the last
ERF (1)
SMOP ILI.
Here SMOP is safe maximum operating pressure: In real life CRs are random variables (RVs). Some pipeline
SMOP = DF·Pf (2) operators [5] realizing this fact, utilize the following method of
Where Pf is the failure pressure (assessed using one of the forecasting the future state of the pipeline, based on predicting
world wide accepted codes B31G, modified B31G, Battelle, a maximal possible CR. When assessing the maximal possible
DNV, and Shell-92), DF is the design factor. rate of defect growth it is assumed that the probability density
The values of the design factor for the three design codes function (PDF) of the depths of the n defects, which are
depending on the on-shore pipeline location class are given in actually present in the pipeline transporting oil or gas
Table 2. condensate substances, is, as a rule, described by the Weibull
law. The two-parameter Weibull CDF has the form:
F ( d ) 1 e ( d / )
b
Table 2. Design factors for different location classifications
[B31.8]
Where d is the defect depth, α and b are the PDF parameters.
Design code Location class Design factor
1 0.72
The maximal defect depth, which is possessed or exceeded by
B31G, modified 2 0.60 the (1- γ) portion of the total number of defects found during
B31G, DNV 3 0.50 the ILI, is defined according to following formula (expression
4 0.40 for the Weibull PDF quantile):
dmax ln(1 ) b
1
(5)
According the DNV code (see [1]), SMOP is calculated as:
In the case when the distribution of the defect set is normal or
SMOP DFt Pf approximately normal, the depth of the defect with probability γ
Where DFt MF DF is total usage factor, MF = 0.9 is the is assessed using the formula for the quantile of the normal
modeling factor (based on a 95% confidence interval). distribution:
If for the pipe cross-section in consideration with a dmax d d (6)
defect ERF 1 , the defect is subject to immediate repair.
Where is the inverse of the standard normal cdf, d is
3. METHODS OF CORROSION RATE ASSESSMENTS the sample average of the defects depth, d is the sample
Any type of analysis of the future state of a pipeline, like standard deviation of the defects depth, n is the number of
residual strength, POF, etc., is based on the predicted sizes of defects present in the pipeline.
the defects which were detected during ILI or DA. If results of two sequential ILs are available, then the maximal
If the rate of the depth and the length of a corrosion defect CR, with a probability of , is defined by formula:
change are close to linear in time, then the defect parameters at
d max L d max P
time t can be assessed using following formula: amax
d t d 0 vd t tL tP
(3)
l t l0 vl t Here dmax P , dmax L are the maximal depths of the defects as
Where d0 , l0 are the depth and the length of the defect at the defined by formula (5) or (6), for the previous (P) and the last
(L) ILI correspondingly.
time of performing the ILI; vd , vl are, correspondingly, the CRs
If results of only one ILI are available, then the maximal, with
in the radial and the longitudinal directions. probability , CR is defined according formula:
The main problem in providing for the integrity and safe
operation of pipelines lies in obtaining the needed quality of the d max
amax
prediction of the future technical condition (state) of the d
pipeline. In order to achieve such quality of forecasting the
Where d is the net time of pipeline operation before the time
future sizes of defects, it is necessary to accurately assess the
corrosion rates. Traditionally it is presumed that the CRs are of conducting the ILI (years).
constant values. Consider two more approaches to assessing CRs as RVs.
For the case when results of two sequential ILI are available, According to the first approach the CR is a RV with following
the CRs are defined as the difference between the parameters:
3
Copyright © 2010 by ASME
M pL M p P pL pP When individual corrosion rate for a specific defect is not
M CR possible to define physically, i.e., when the defect was not
tL tP tL tP
(7) found during the previous ILI, or when the defects were,
D pL D p P
D CR
according to the previous ILI, larger, than according to the last
tL tP ILI, the external corrosion rate is assumed to be, according to
2
4
Copyright © 2010 by ASME
Design factor 0.72
5
Copyright © 2010 by ASME
According to Table 5 and Fig. 5 (which actually is a
visualization of Table 5)., the maximal POF for the whole
5.2 Pipeline POF pipeline is when all the pipeline parameters and the defects’
Below results of several variants of calculations are presented sizes are considered as RVs; this POF is minimal when the
using the GCE methodology [14]. They differ in what pipeline parameters are considered as deterministic values, and
parameters of the pipeline are considered random. The initial the defects parameters are RVs.
data used is presented in Table 4. The POF in this particular case for the time of the second ILI
(2008) is negligible for all three approaches considered.
Table 4. The pipeline Parameters of RVs Hence, the main question for the Operator of the pipeline is as
Сoefficient of follows:
Mathematical
Symbol Variation PDF How corrosion will affect the integrity of the pipeline in the
expectation
(COV) near future and what optimal (lean) predictive maintenance
wt, in 0.50 0.02 Normal actions should be planned beforehand and executed in the most
D, in 12.75 0.02 Normal appropriate time?
SMYS, psi 42.00·103 or 52.00·103 0.07 Normal
MAOP, psi 1037.00 0.10 Normal
6
Copyright © 2010 by ASME
2 6786.50 40.90 0.20 2.83 4.0
3 8030.80 56.10 0.28 3.18 4.61
4 9091.40 38.40 0.19 1.41 2.77
5 17015.00 45.00 0.23 3.89 8.31
6 18931.30 45.20 0.23 2.95 7.38
7
Copyright © 2010 by ASME
Pressure, psi 3 1.29·10-7 1.70·10-4 1.31·10-2 1.04·10-1 1.00
4 0.00 1.11·10-16 1.11·10-15 3.56·10-14 2.13·10-12
5 1.09·10-8 3.85·10-6 2.90·10-4 4.05·10-3 1.95·10-2
6 4.91·10-12 3.56·10-10 4.25·10-8 2.69·10-6 6.40·10-5
SMOP
2009 Table 9. Predicted POF for pipe cross sections with defect
2010 according to modified B31G code
POF of pipe cross section with defect, for moments of time t
2011 Defect #
2009 2010 2011 2012 2013
MAOP 1 0.00 4.31·10-14 3.82·10-9 8.67·10-6 7.5·10-4
2012 2 8.66·10-15 1.41·10-9 1.03·10-5 1.51·10-3 1.96·10-2
-7
3 8.02·10 3.88·10-3 1.16·10-1 3.70·10-1 1.00
4 0.00 0.00 0.00 0.00 1.44·10-15
5 2.01·10-9 6.75·10-6 8.80·10-4 1.15·10-2 4.56·10-2
l Dt
d/t 6 3.33·10-15 7.77·10-12 1.38·10-8 3.75·10-6 1.60·10-4
Figure 9. The SMOP surface, according to the B31G mod
code, the MAOP plane and the degradation trajectory for According to Table 9, on the condition that the ultimate
the most dangerous defect of the pipeline permissible POF is 10-7 1/km year and with the present
corrosion rate CR, the defect # 3 will need repair already after
Analysis conducted in this Section, related to the most 1 year after the last inspection (in 2009); the defect # 5 – after
dangerous defect, shows that this defect (#3) should be 2 years (in 2010).
removed or repaired as soon as in 2012. Consider defects that are located in a one kilometer stretch of
the pipeline which includes the defect # 3 (not necessarily
5.6. Prediction of pipeline POF located at the center of the segment). These defects are nine in
Here two design cases are considered: when the corrosion rates number. Calculate now the variance based on only these ten
CRs are deterministic and when they are RVs. defects (including defect # 3): D[vrc] = 1.38·10-4 in2, D[vlc] =
For the case when the CRs are considered as normally 3.95·10-2 in2 (compare with corresponding variances calculated
distributed RVs, the means of the CRs of defects were found by based on the assessment of the ILI tool tolerance, see Table 7).
eqn. (7) and the variances of the CRs, according eqn. (9), are Using these variances, assess the predicted POF for the pipe
equal to 2SD2 / 9 7.70 104 in2 for all defects. Here cross-section which contains the defect # 3 (see Table 10).
SD 0.06 in. Other initial data are considered as RVs and their
Table 10. Predicted POF for pipe cross section with defect
parameters are given in Table 4.
#3. The local (censored) variance is used
The predicted POF was defined for the six most dangerous
POF of pipe cross section with defect, for moments of
defects which parameters are given in Table 6. The corrosion
Code time t
rates were considered as RVs, parameters of which are shown
2009 2010 2011 2012 2013
in Table 7. The results of calculus are given in Tables 8, 9.
B31G 3.37·10-8 6.12·10-6 3.90·10-4 8.88·10-3 1.00
-7 -4 -2 -1
Table 7. The PDF parameters of the CRs of the six most B31Gmod 1.51·10 2.90·10 2.53·10 2.49·10 1.00
dangerous defects
Radial CR Longitudinal-axis CR Implementation of the local variances of CRs leads, as can be
Defect #
Mean, in
Variance,
Mean, in
Variance, observed in this particular case, to a smaller radial CR variance
in2 in2 but a much larger longitudinal CR variance, as compared to the
1 0.02 0.31 general case, when the CRs are assessed using the ILI tool
2 0.02 0.39 tolerance value, as given by the Vendor. At the same time, the
3 0.05 0.47 pipeline POF is also smaller as compared to the general case.
7.70·10-4 7.70·10-4
4 0.02 0.02 It should be noted that, when considering CR of defects as
5 0.01 1.07
RVs, the difference between the POF is substantial and can be,
6 0.01 0.02
on the average, of four orders of magnitude (104), when
compared to the case when considering CRs as DVs.
Table 8. Predicted POF for pipe cross sections with defect This situation calls for drastically raising the measurement
according to B31G code accuracy of defect sizes.
POF of pipe cross section with defect, for moments of time t
Defect # 5.7. Residual time to pipeline repair/failure criteria
2009 2010 2011 2012 2013 After the last ILI the decision making person (DMP) has to
1 5.62·10-14 2.79·10-11 2.74·10-8 7.48·10-6 3.40·10-4 make up her/his mind as to when and where to conduct the
2 9.54·10-12 9.39·10-9 5.32·10-6 4.20·10-4 5.99·10-3
8
Copyright © 2010 by ASME
needed maintenance/repair actions or the next ILI. In order to conducting the ILI
figure this out, the DMP needs to know what time is left from POF of pipe cross-section
the time of the last ILI and the set of prescriptive or physical with defect = 10-6 11-Sep-2008 15 (0.5)
criteria that govern the specific pipeline operation. The (B31Gmod)*
prescriptive criteria serve as warning or alarm levels of POF POF of pipe cross-section
10-Oct-2008 29 (1.0)
with defect = 10-7 (B31G)*
that trigger maintenance action [19], [16], [12].
POF of pipe cross-section
There are different approaches in the pipeline industry to with defect = 10-5 8-Dec-2008 59 (2.0)
assessing the residual time to pipeline repair/failure criteria, (B31Gmod)*
i.e., 0.8wt defect depth, POF reaching a prescribed value for POF of pipe cross-section
pipelines of specific type (on-shore, off-shore, production ,etc.), 13-Feb-2009 67 (2.2)
with defect = 10-6 (B31G)*
or actual loss of integrity (leak, rupture). POF of pipe cross-section
16-Jun-2009 123 (4.1)
Each approach yields its own assessments. These assessments with defect = 10-5 (B31G)*
are given below for the most dangerous defect #3, in DEFECT DEPTH= 80%wt 25-Jul-2009 39 (1.3)
chronological order, starting with the time to reach the POF= SMOP = МAОР (B31Gmod) 12-Mar-2011 595 (9.8)
10-7, from the moment of conducting the last ILI. DEFECT DEPTH= 100%wt 12-Aug-2011 153 (5.1)
Tables 11 and 12 also give the times between the time when * - Parameters of the defects (depth, length, CRs) and of the pipeline (diameter,
wall thickness, SMYS, MAOP) are considered as RVs.
each defect reaches the 80%wt criteria and the moment of time
the pipeline loses its integrity (by leak or rupture).
The times given in the last column of Tables 11 and 12 are
Table 11. Times of failure for the most dangerous defect #3, valid only for the conditions described in detail above. For
according to different criteria safety reasons these time intervals can be divided by a safety
Time between the factor (SF > 1.0, its value determined/assessed by the
Predicted date moments of Operator), when the actual conditions differ from the ones
Used criteria analyzed in the paper, and given in monthly increments (shown
of defect failure arrival of adjacent
criteria, days (mo) in brackets).
POF of pipe cross-section The Figure 10 is a visualization of Table 11. It turns out that the
353 (11.7)
with defect = 10-7 24-Jun-2009 most sensitive criteria is, obviously, POF = 10-7, using the
(from the last ILI)
(B31Gmod)* B31G mod. All criteria related to specific values of POF using
POF of pipe cross-section different design codes serve as warning and alarm levels of
15-Aug-2009 52 (1.7)
with defect = 10-7 (B31G)* pipeline safety. They provide needed awareness for the DMP or
POF of pipe cross-section
trigger different actions depending on the engineering
with defect = 10-6 3-Sep-2009 19 (0.7)
(B31Gmod)* principles used by the Operator. The last three criteria (in
POF of pipe cross-section capital letters in Tables 11 and 12) are physical criteria because
with defect = 10-5 19-Nov-2009 77 (2.5) they describe triggered actions or actual loss of containment by
(B31Gmod)* the pipeline. The warning level leads to activity which is ruled
POF of pipe cross-section by the fact that there is a relatively large amount of time before
30-Nov-2009 11 (0.3)
with defect = 10-6 (B31G)* advancement of a leak or rupture, which may be used by the
POF of pipe cross-section Operator to plan and carefully execute necessary actions. The
22-Mar-2010 112 (2.7)
with defect = 10-5 (B31G)* alarm level triggers, in general, more fast activity because the
DEFECT DEPTH = 80%wt 7-Feb-2011 322 (6.7) time to actual failure is much shorter.
SMOP = МAОР (B31Gmod) 16-Jul-2012 519 (17.3)
DEFECT DEPTH = 100%wt 20-Feb-2013 225 (7.5)
Table 12. Times of failure for the most dangerous defect #3,
according to different criteria (ILI tool tolerance added to
all defect sizes)
Time
between the
moments of
Used criteria Date arrival of
adjacent
criteria, days
(mo)
In this case, the POF
POF of pipe cross-section of pipe cross-section Figure 10. Times for conducting prescribed maintenance
with defect = 10-7 (using with defect is 0 actions or time to failure for the most dangerous defect,
B31Gmod)* already > 10-7 at the according to different criteria
moment of
9
Copyright © 2010 by ASME
Risk Assessment, Rehabilitation and Repair Conference,
The actual spill (failure) triggers a disaster scenario, in which Houston, Texas, pp. 1-13, 1994
every single moment counts. For the three described above [9] Kulkarni, R.B., and Conroy, J.E. Development of a
cases the cost of the same activities dramatically differ Pipeline Inspection and Maintenance Optimization
depending on the time that is available to perform the needed System. GRI, 1998
actions. [10] NACE Standard RP0502-2002. Pipeline External
Corrosion Direct Assessment Methodology.
6. CONCLUSIONS [11] Nicoletti E.S.M., R.D. de Souza, Dr Sérgio da Cunha
The described above methodology permits implementation Barros. A Practical Approach in Pipeline Corrosion
of the most versatile predictive asset maintenance when modeling. Part2. Short-term integrity forecasting. In: J. of
operating pipelines. Pipeline Engineering, March, 2009
The comprehensive practical predictive pipeline [12] Timashev S.A. Risk assessment and Reliability-Based
maintenance methodology described in this paper provides the Maintenance for Large Pipelines. Chapter 40, Engineering
pipeline Operator with a powerful tool for creating and Design Reliability, Reliability Handbook, CRC Press,
implementing an optimal IMP for each specific pipeline, using 2005
results of specific pig runs. [13] Timashev S.A. Risk Based Management of Pipeline
The Operator, as a decision making person (DMP), receives Integrity and Safety. Intensive Short Course, 2002-2010
information in the needed format, which allows her/him to [14] Timashev S.A., M.G. Malyukova M.G., Poluian L.V.,
decide which of the criteria to use and how and when to Bushinskaya A.V. Internet– Oriented Method Of
conduct the necessary ILI, maintenance and repair actions. Reliability Analysis Of On-Shore Pipelines With Growing
Defects, Proceedings of IPC Conference, Paper #
7. ACKNOWLEDGEMENTS IPC2008-64545, 2008
The Authors would like to express their gratitude to the [15] Timashev, S.A. Machinery Diagnostic and Residual Life
unknown Reviewers for their comments and recommendations Monitoring System. Machinery Reliability Conference,
which helped enhancing the content of the paper. Charlotte, NC, Industrial Communications, Inc.
Knoxville, TN, pp. 176-188, 1998
8. REFERENCES [16] Timashev, S.A. Optimal System Integrity and
[1] Alkazraji D. A Quick Guide to Pipeline Engineering. Maintenance Control, COMADEM, 2000.
Series editor: Clifford Matthews, Woodhead Publishing [17] Timashev, S.A, Shalin M.G. et al. Optimal Control of Far
Limited and Matthews Engineering Training Ltd, North Pipelines Integrity and Maintenance. OMAE'2001.
Cambridge, England, 2008 [18] Timashev, S.A. On a Problem of Reliability Control.
[2] Bushinskaya A.V., Timashev S.A. Statistical analysis of Third International Vilnius Conference on Probability
ILI data for optimal pipeline integrity management. All Theory and Mathematical Statistics, Abstracts of
Russia Conference «Safety of Critical Infrastructures and Communications, pp. 186-187, 1981
Territories», Ural Branch, Russian Academy of Sciences, [19] Timashev, S.A. Reliability of Large Machine Systems,
Ekaterinburg, December, 2008, In Russian Moskva, Nauka, 184 pp, 1982; SEAG, Pavia, Italy, pp.
[3] Davis, M.J. Tenneco’s Efforts for Verifying Pipeline 102, 1984
Integrity. ASA Distribution/Transmission Conference, [20] Timashev, S.A., Kopnov, V.A. Optimal Death Process
Toronto, Canada, 1988 Control in Two-Level Policies. Fourth International
[4] DNV Recommended Practice RP-F101. Corroded Vilnius Conference on Probability Theory and
pipelines. Det Norske Veritas, Norway, 2004 Mathematical Statistics, Abstracts of Communications,
[5] Enterprise Standard #0-03-22-2008. Methodology of pp. 308-309, 1985
assessment of potential hazard and residual life of [21] Timashev, S.A., Kopnov, V.A. Optimal Resource
pipelines with corrosion defects and discontinuities in Assessment with Respect to Fatigue Crack Formation
welds and parent metal, detected during ILI. Criteria. Journal of Machine Building and Structure
"OrenburgGasprom", 2006, In Russian Reliability Problems, #1, Moscow, Nauka, pp. 65-70,
[6] Kariyawasam S., Peterson W., Revised Corrosion 1990 (in Russian).
Management with Reliability Based Excavation Criteria.
Proceedings of IPC Conference, Paper # IPC2008-64536,
2008
[7] Kiefner, J.F., Vieth, P.H., Orban, J.E., and Feder, P.E.
Methods for Prioritizing Pipeline Maintenance and
Rehabilitation. Final Report on Project PR3-919 to
American Gas Association, Battelle, 1990
[8] Kulkarni, R.B., and Conroy, J.E. Pipeline Inspection and
Maintenance Optimization System (PIMOS). Pipeline
10
Copyright © 2010 by ASME