Sie sind auf Seite 1von 11

Paper No.

05159
CORROSION 2005
Progressive Integrity Assessment Solutions using Quantitative Methods

Shahani Kariyawasam
GE Pipeline Solutions, GE Energy,
1003 11th Street,
Calgary, Alberta
Canada T2R 1G2

Iain Colquhoun
GE Pipeline Solutions, GE Energy,
1003 11th Street,
Calgary, Alberta
Canada T2R 1G2

ABSTRACT

The paper presents a phased approach to the quantification of pipeline risk due to corrosion, SCC
and other defect- based hazards. Public concern and pressure from regulatory bodies are accelerating the need
for pipeline operators to formalize and intensify their approach to integrity management, and there is a growing
acceptance of a risk-based approach to the prioritization, planning, and evaluation of integrity assessments.
Varying amounts and quality of data are required to do risk assessments depending on the methodology used
and the required outcome. For pipelines not currently on a formal assessment plan, the data available tend to be
sparse and qualitative in nature. This level of data quality and availability seldom lends itself to a fully quantitative
risk assessment, but is adequate for qualitative and/or semi-quantitative methods, and these methods, in turn, are
adequate to address the initial prioritization of integrity assessments. As assessment data become available, the
operator can take advantage of risk methodologies that progressively incorporate quantitative data. The phased
quantitative methodologies described in this paper incorporate quantitative assessment data that caters to the
developing needs of the operators.
Most operators need a phased approach to progressively adopt the most relevant and needed assessments.
These organized and relevant steps need to be introduced according to data available and maturity of the integrity
management program. Each refinement to the quantification offers distinct advantages to the user. The paper
explains these advantages. Corrosion is used as an example, but the approach is applicable to any time-
dependent defect based hazard.

Key Words: Threat Analysis, Hazard Analysis, Quantitative Risk Assessment, Qualitative Risk Assessment,
Reliability Based Assessment, Probability of Failure, Probability of Exceedance, Consequences, Threat Mitigation

Copyright
©2005 by NACE International. Requests for permission to publish this manuscript in any form, in part or in whole must be in writing to NACE International,
Publications Division, 1440 South Creek Drive, Houston, Texas 77084 77084-4906. The material presented and the views expressed in this paper are solely
those of the author(s) and not necessarily endorsed by the Association. Printed in U.S.A.

1
Introduction
There are a wide variety of methodologies available to perform risk assessment on pipeline systems to manage
corrosion and other common threats to integrity [1,2,3,4], such as qualitative, semi quantitative, probability of
exceedance and quantitative methodologies. The selection of the most appropriate method depends to a large
extent on the quantity and quality of data available, and on the nature and extent of previous risk assessments
and integrity assessments performed on the system. The more reliable quantitative assessment data that are
incorporated into the assessment, the more defensible and consistent the results.
If reliable assessment data are not available and/or no previous, formal risk assessment has been performed on
the system, then qualitative risk assessments based on input from Subject Matter Experts (SME’s) may be
appropriate. This might be a case, for example, where due diligence is required in a short time frame on the
acquisition of older pipeline assets. Detailed cathodic protection data, for example, might not be available.
If, on the other hand, some level of risk assessment has been carried out on the system and adequately
documented, and/or if appropriate data are available and of a higher quality, these data can be incorporated into
empirical and/or mechanistic threat likelihood models to improve the accuracy of the risk assessment. Since such
methods often rely on a mixture of mechanistic quantitative models, that might not be definitively calibrated, and
qualitative/index-based algorithms, in which the precise parametric interaction is not rigorously represented, these
methods are commonly referred to as semi-quantitative methods.
If these preliminary models are refined on the basis of better mechanistic algorithms and more reliable
assessment data is described and analyzed statistically and input quantitatively and probabilistically, the risk
assessment can move towards a quantitative approach (QRA – quantitative risk assessment). However most
operators need a phased approach to progressively adopt the most relevant and needed assessments. These
organized and relevant steps need to be introduced according to data available and maturity of the integrity
management program. Each refinement to the quantification offers distinct advantages to the user. The paper
explains these advantages. Using external corrosion as an example, the data that would support QRA includes
reliable assessment data preferably consecutive runs that assess the corrosion growth rates.
The broad state of the art currently is to use semi-quantitative methods. The advantages and limitations of these
are discussed below. The body of the paper then describes a phased approach to the progressive quantification
of pipeline risk according to the needs of the operator.
Qualitative and Semi-Quantitative Risk Assessments
A basic qualitative method incorporates subject matter expert opinion organized systematically to rank pipeline
segments [1]. A wide variety of semi-quantitative methods are available [6, 7, 8]. The following example has been
described previously in more detail [8]. It is one of the more flexible formats used, but is also representative of the
general semi-quantitative methodology.
In the following, the risk in pipeline segment ‘j’ (Rj) is calculated as the product of the likelihood of failure from
threat ‘i’ (Lij x lj) and the consequences of failure (Q) by severity factored by the conditional probabilities of the
severity (leak or rupture) given failure.

∑ [(P ]
3
That is: R j = L ij × (l j ) × R/F .Q R ,k + PL / F .Q L ,k )ij (1)
k =1
Rj = The risk values for a basic segment j
Lij = Likelihood of failure due to threat i for segment
lj = a non-dimensional factor to account for the effect of the length of segment j.
PL/F , PR/F = Probability of leak or rupture given a failure. These parameters are chosen to represent
common industry trends or operator experience as appropriate. [9]
QL,k , QR,k = Consequence of a leak or rupture given a leak or rupture failure
i = 1 to number of threat types considered
k = 1 to number of consequences considered

Limitations of Qualitative and Semi-Quantitative Methods


Although qualitative methods and semi-quantitative methods form a simple and convenient assessment approach
that is generally flexible with respect to data availability and quality, this flexibility comes with a price. These

2
methods essentially treat a probabilistic problem in a broadly classified deterministic manner, and therefore
require a conservative, blanket approach to account for inherent uncertainty in many of the analysis parameters.
Assessment data represent the actual state of the pipeline as opposed to predicted state given by indirect
measurements such as CP surveys. Qualitative methods do not make full use of known quantitative data from
inspections and dig verifications in a rigorous manner, since the framework for doing so is specific and
quantitative and therefore outside of the scope of the methods. Incorporation of assessment data is the best
practice available to assess the actual state of the pipeline.
A useful requirement of risk assessment methods is that they quantify the effect of mitigation so the effect of
mitigation on the risk can be demonstrated [10]. This requirement is outside of the capabilities of qualitative and
semi-quantitative methods and therefore such methods cannot be used to optimize maintenance decisions and
maintenance scheduling.
As the consequences of the failures and the costs of maintenance is not quantified in the qualitative methods they
also cannot relate risk-reduction benefits to program costs, and therefore cannot perform rational cost
optimizations. A justifiable cost benefit analysis needs quantification.

Incorporation of Pipeline Assessment Data


An assessment of the pipeline should be able to detect significant defects and also defects that could grow to be
a significant size before the next assessment. An assessment should also be able to size the defects to an
acceptable level of accuracy. In detection two main errors are of interest, namely Type 1 error; which is the error
of calling a defect when there is non and the Type 2 error; which is the error of not calling a defect when there is
one present.
The use of good quality assessment data enables reliable assurance of immediate and future integrity for defects
that grow with time, such as corrosion defects. To achieve this, assessment data has to possess the following
characteristics:
• All areas of the pipeline should be assessed (comprehensive coverage as opposed to sampling),
• All existing defects above a practically reasonable safe detectability limit should be reported. In other
words, probability of detecting an existing defect should be very high. This reduces the Type 2 error of
missing an existing defect
• Adequate sizing accuracy for defects (within a measurement tolerance). All defect dimensions that are
relevant to assessing the leak and burst pressure conditions should be reported with adequate accuracy.
• Low probability of false indication. This reduces the Type 1 error of a false call.

In the pipeline industry the three main assessment methods are direct assessment, hydrostatic testing, and inline
inspection. Each of these methods is discussed below in light of the above desirable characteristics, for corrosion
assessment.
Direct assessment samples the pipeline at locations where indirect measurements and predictive equations
indicate that corrosion is most probable. This assessment does not have comprehensive coverage of the pipeline
but aids in discovering and sizing the defects in areas predicted to be the most critical using indirect
measurements and predictive equations. The confidence in Type 2 error limit cannot be established, as one does
not know what is missed. In the excavated defects the sizing accuracy is high but the non-excavated defects are
only estimated. The Type 1 error is dependent on the site and conditions. If the predictive equations are
calibrated to similar sites and conditions the probability of false indication maybe smaller but it is not as small as
for hydrostatic-testing or inline inspection.
Hydrostatic testing assesses all area of the pipeline. It accurately identifies all the critical defects that fail at or
below the test pressure. Therefore it assures immediate integrity but does not give assurance of future integrity,
as there is no rigorous assessment of the defects that remain in the pipe.
Inline inspection assesses the whole pipeline and detects defects above a detectability limit. In corrosion
assessment tools, the detectability limit is set significantly low so that the small defects that are below the limit will
not fail in the near future (10 to 15 years). The Type 1 and 2 errors are sufficiently low in presently available ILI
methods for use in the subject methodology. Measurement errors are uncertainties associated with sizing of
defects. Measurement errors, in corrosion related ILI, are adequate to assess the integrity of the pipeline.

3
However, the measurement error may need to be validated explicitly for runs in which preliminary validation
indicates poor correlation with measured values.

Deterministic Quantitative Methods using Assessment Data


For the past two decades, integrity assessment data have been used for fitness for purpose, immediate integrity
assessments, and future integrity assessments. Because they are based on physical measurements, these
assessments are quantitative in nature. Because of limitations in the early assessment tools, the assessments
traditionally dealt with uncertainty and variability in the assessment parameters in a deterministic way. The low to
medium resolution ILI tools originally available were appropriate for assessing only the immediate general
condition of the pipeline. As ILI technology improved and produced better quantitative assessment data with
greater resolution, these data could now be used to assess both immediate and future integrity, and also to
perform risk assessments.
Deterministic analyses use an upper bound value for the defect size to account for the measurement error. An
upper bound growth rate and strength parameters are also used to account for uncertainty in these parameters.
Use of upper bounds is a simple method to account for uncertainty, but it leads to undue conservatism and to
non-consistent risk assessments as can be shown in comparisons with fully probabilistic methods [11, 12,13].
In a post- ILI assessment, the highest uncertainties are associated with defect sizing and growth rates. Defect
sizing affects both immediate integrity and future integrity. Growth rates affect future integrity estimates - that is,
the estimation of time to next assessment and scheduling of mitigative actions. Variability in material properties
leads to uncertainty in the burst pressure calculations.

Use of Probability of Exceedance methods to improve dig program accuracy


A significant improvement can be made to deterministic methods by explicitly accounting for uncertainties in
defect sizing through the use of probability of exceedance techniques (POE). This methodology quantifies the
measurement error probabilistically, but all other input parameters are considered deterministic. The result is a
probability of failure or POE.
The POE approach is a statistical method used in the structural analysis of pipelines to evaluate the probability
of the likelihood of a leak (full-wall penetration) and/or a rupture (unstable propagating fracture).. This method is
based on analysis of inspection tool predictions and field measurements, and provides a means for
systematically predicting failures in a more rigorous, less deterministic fashion.
In POE analysis, the probability of failure is described by the probability of an actual defect size exceeding a given
critical value (for example, 80% wall loss for leak), or by the probability of rupture pressure calculated from an
actual defect being below the maximum allowable operating pressure. This POE concept is schematically
presented in Figure 1.
Pressure Calculated from Field Measurements, psi

1,400

0.6
60% pig call
1,300
70% pig call
MFL Tool Call : 40% wt 80% pig call
0.5
1,200
Frequency of Occurrence

0.4
1,100

Actual Depth (Field


Meausurement) Follows
0.3
Normal Distribution
1,000 MOP = 980 psi
with Mean = 50% Wall
Loss
0.2 900

>80% Wall Loss


800 Probability of
0.1
Exceedance
POE
700
0 700 800 900 1,000 1,100
0 10 20 30 40 50 60 70 80 90 100

Actual Depth, % Wall Loss


Pressure Calculated from Pig Call, psi

Figure 1. Schematic illustration of POE concept based on defect size and failure pressure

4
In Figures 1, there are uncertainties involved in the predicted size for each of the features reported by the tool.
These uncertainties are assumed to be normally distributed. The probability of failure is represented by a portion
of the cumulative distribution in the circled areas on the graphs, which is either above 80% of the wall thickness
for leak or below the MAOP for rupture (as shown in Figure 1 by the red line).
For simplicity, only the uncertainty of MFL reported corrosion size is considered. That is, the errors associated
with inspection tool predictions are statistically analyzed against field measurements and are then used in
probability of failure analyses as shown as in Figure 2. The representative corrosion growth rates along the
pipeline are considered as deterministic parameters. The minimum specified yield and ultimate tensile strengths
(SMYS and SMUTS) and the maximum allowable operating pressure (MAOP) are also considered to be
deterministic. The regression line shows an intercept, which shows that there is a bias in the measurement error
i.e., the measurement error defined as the actual to predicted ratio, does not have a mean of 1.0.
35.00

y = 0.488x + 9.021
2
30.00 r = 0.3717
r = 0.6096
S.E. = 5.23
25.00
RTD Measurement, wt%

20.00

15.00

10.00

5.00

0.00
0.00 5.00 10.00 15.00 20.00 25.00 30.00 35.00

Tool Predicted Depth, wt%

Figure 2. A plot of the field measured and the MFL tool predicted depths of 43 corrosion features
In the POE analysis, is described by the probability of an actual defect size exceeding a given critical value, for
example, 80% wall loss for leaks, by Equation (2) or by the probability of the rupture pressure calculated from an
actual defect being below the maximum allowable operating pressure (MAOP).
x
(2)
POE = 1 - ∫
0
f ( x ) dx

5
1.00E +00

1.00E-01

1.00E-02

1.00E-03
15 years
1.00E-04

1.00E-05
12 Years
1.00E-06
POE for leak

1.00E-07

1.00E-08

1.00E-09
10 years
1.00E-10

1.00E-11

1.00E-12

1.00E-13
5 years
1.00E-14

1.00E-15
0 10 20 30 40 50 60 70 80 90 100

T im e, year
ID of metal loss features
Figure 3. Probability of leak for the 100 worst metal loss features

Figure 3 shows an example of probability of failure by leak and rupture in five-year increments, for 100 worst
corrosion features in a 36” diameter, 38 km long natural gas pipeline.
The overall probability of leak and rupture of the entire pipeline can be calculated using Equation (3), which is
particularly important for developing a risk based pipeline management plan.

n3

POEcumulative = 1 − ∏ (1 − POE i ) (3)

i =1
where POEi = probability of failure of the ith feature, and ‘n’ = the total features in the pipeline.
The primary difference between deterministic and POE approach here is that the deterministic methodology uses
the dimensions of corrosion features reported by ILI and a safety factor to compensate for measurement errors
from the tool; whereas the POE analysis assumes a target failure probability based on (for example Location
Class) and compares this target failure probability to the failure probability of each individual feature (POEi) and
that of the entire segment or pipeline. As a result, POE establishes future repair strategies using target failure
probability. Figure 4 compares the predicted number of repairs using POE (for a target failure probability of 10-6)
and a more traditional, deterministic method (using a safety factor of 1.25). Note that the total repairs could be
significantly reduced using POE methodology (i.e., 43 vs. 203 repairs in 10 years).

6
Comparison of POE and Deterministic Approches for Future
Repairs

250
POE Approach
Determinstic Approach 203
200
Number of Repairs

155
150

105
100
79
57
50 38 43
31 30
17 22 22
15 9 15
05 2 2 2 3 4
0
2000 2001 2002 2003 2004 2005 2006 2007 2008 2009 2010
Year

Figure 4. Comparison of POE and deterministic approach for future repairs


The main advantage of the POE methodology is a more accurate dig program. That is fewer false digs and better
discovery of the critical ones through more accurate quantification of defect sizing. The more accurate dig
program leads to better safety and economy.

Refined Future Integrity Using Quantified Corrosion Growth Rates, Decision Tree Analysis, and Run
Comparison
Two key components of corrosion growth assessment in pipelines are accurate determination of corrosion growth
rate and application of relevant corrosion growth rate to significantly different segments to assess the future
integrity of a pipeline. PII has developed a corrosion growth assessment tool, Run Comparison1 software
(RunCom) that allows accurate determination of corrosion growth. This tool compares the raw signals of the
same defect present in two inspection runs to report the real active corrosion defects and their growth with less
error. By comparing the boxed data (sized data) the genuine growth cannot be assessed accurately. By
comparing signals the actual growth can be estimated much more accurately.

1
RunComTM

7
Figure 5. True growth identified by Run Comparison
Since variations in corrosion growth along the pipeline can be significant, a single value of average or maximum
corrosion growth rate does not represent the corrosion condition of the pipeline and could result in a conservative
or non-conservative conclusion for future integrity. Decision Tree Analysis method [14] is used to categorize the
corroded regions along the pipeline and calculate the corrosion growth rates in these specific areas. Significant
variations in growth rates and corrosion severity are often observed from one area to another along the pipeline.
Segmentation strategies such as decision tree analysis can be used to characterize corrosion growth rates along
the entire pipeline and group them into logical segments on a statistical basis by progressively partitioning a large
population of non-homogenously distributed pipeline attribute data into relatively homogenous subgroups. This
reduces undue conservatism of applying the largest corrosion growth rate to all segments.

Refined Pressure Capacity And Safety - using Reliability Based Methods


Once the uncertainty in the measurement error and growth rate of corrosion defects is taken into account further
improvement in assessing burst pressure based analysis can be realized by accounting for uncertainty of the
pipeline strength, model error, and dimensions. Reliability based assessments take all these uncertainties into
account. This process is schematically represented in Figure 6. Reliability is the complement of probability of
failure.
Reliability = 1 – Probability of Failure (4)
If the models used to characterize failure due to leak and rupture are accurate the model error will be minimized.
However there is yet a quantifiable model error that exists due to the minor dependencies that cannot be
practically accounted for in the modeling equations. Therefore, the modeling error should be quantified and used
in the reliability-based methods of calculating the probability of failure.
Structural reliability methods or Monte Carlo simulation methods can be used to calculate the probability of failure
and consequently the reliability, which is the inverse of the probability of failure. These methods are well
established and have been used in the many industries, including the pipeline industry [11,15,16].
As all aspects of uncertainty are taken into account the reliability based processes represent probability of failure
and reliability most accurately. Consequently rupture pressure based reliability is best characterized using the
reliability-based methods. This helps identify the mitigative actions that increase reliability of the pipeline (or
reduces the probability of failure). It also helps assess the deterioration of reliability with time more accurately
than deterministic or POE methods.

8
Assessment
data

Measurement Data from


Repetitive Pipe
error Corrosion Model
inspections Properties and
and Test Results
Dimensions
Corrosion Corrosion
Characteristics Growth rates Pipeline Model
properties Uncertainties
Prob. density

Prob. density

Test results
Frequency

Operating
Feature Depth (mm)
conditions
Growth Rate (mm/yr) Yield Stress (MPa)

Model results

Failure Propability as
a function of time

Figure 6. Input and output of the Reliability Based Assessment

Cost Optimization - using Quantitative Risk


All the assessment methods mentioned above considers the probability of failure and its different aspects. Risk of
failure also includes the consequences of failure.
Risk = Probability of Failure x Consequences (5)
Quantitative risk based analyses incorporate the environmental, life safety, and financial consequences of failure.
Accounting for the consequences enables financial quantification. The balancing of the failure related costs with
the maintenance and preventative costs enables cost optimization and cost based decision-making. Practical
considerations such regulations can also be represented as constraints on the scenario represented. Figure 7
shows some of these tools that quantification of risk enables.

Figure 7. Decision making tools enabled by quantitative risk assessment

9
Conclusions
Extremely accurate, fully quantitative, and probabilistic techniques are available for the assessment of the risk
and reliability of pipelines due to threats such as corrosion. However the pipeline industry has used mostly
qualitative risk assessments due to the lack of organized data and inadequate physical assessment data.
Pipeline operators need a phased approach to progressively adopt the most relevant and needed assessments.
These organized and relevant steps need to be introduced according to data available and maturity of the integrity
management program.
Table 1 explains the benefits of progressively refining the quantitative methods used. Each new aspect included
and the advantages of each refinement are summarized in this table.

Name of Method Phased Refinement Advantage gained


Qualitative Risk Assessment - Economic relative system ranking
Deterministic methods using Incorporates assessment of Enables dig program and mitigation
Assessment data actual state of pipeline
Probability of Exceedance Accounting for measurement Accurate dig program; less false digs
methods error and estimating probability and more confident detection of
of failure/exceedance of criteria critical defects
Decision Tree Analysis Identification of localized Segment specific corrosion analysis
subpopulations with distinct enables more accurate and
corrosion characteristics consequently economic integrity
program
Quantified Corrosion Growth Accurate corrosion growth rate Accurate time to next inspection and
Rates - Run Comparison estimation using best practices future integrity
in industry, i.e., consecutive run
comparison at signal level
Reliability Based Assessment Accounting for pipeline More accurate burst pressure
strength, dimensions and analysis leads to
model error uncertainties (1) safer and more accurate integrity
management plans and (2)
comprehensive quantification of
effects of mitigation
Quantitative Risk Assessment Accounting for quantified Optimization of costs, cost benefit
consequences analysis, rationale for decision making
Table 1 . Progressive Integrity Assessment Solutions using Quantitative Methods

REFERENCES
1. “Pipeline Risk Management Manual.” W. K. Mauhlbauer. 2nd Edition. Gulf Publishing Company. 1992
2. “Prioritization scheme used to establish maintenance program.” “ASPIRE”. A. Espinoza, D. Jones, S.
Peet, L. Espinosa. Pipeline & Gas Industry, August 2000.
3. ‘Risk Based Integrity Management’, J.Dawson and D. Jones. Risk Based & Limit State Design &
Operation of Pipelines’, IBC Conference, Oslo, Norway 4 -5 October 1999.
4. “The path forward for risk based integrity management” S. Kariyawasam, I. Colquhoun, and E. Choong.
Congreso Internacional de Ductos, Puebla, 12 – 14 Noviembre del 2003.
5. “System-Wide Risk-Based Pipeline Integrity Program at TransCanada.” R. Sutherby, L.Fenyvesi, I.
Colquhoun, and M. Rizkalla. Proceedings of International Pipeline Conference. Vol 1. 2000
6. GE PII PipeView Integrity, Technical Reference Manual – Risk Assessment. 2004.
7. “Rapid Analysis Tool for High-Level Risk Assessment of Pipelines.” Colquhoun et al. IPC 2002 – 27264.
Proceedings of IPC’02. 4th International Pipeline Conference. Calgary Sept 29 – Oct 3, 2002.

10
8. “A Validation Strategy for Semi-Quantitative Risk Assessment in Pipeline Integrity Management.”
Colquhoun, Kariyawasam, Gu, Choong, Yang, Mei, and Qingshan. IPC 2004. Proceedings of IPC-04.
5th International Pipeline Conference. Calgary Oct 4th – Oct 8th. 2004.
9. US DOT Office of Pipeline Safety, Research and Special Programs Administration. Distribution and
Transmission Accident and Incident Data. http://ops.dot.gov/IA98.htm.
10. US DOT. 49 CFR Part 195. Pipeline Safety: Pipeline Integrity Management in High Consequence Areas
(Hazardous Liquid Operators With 500 or More Miles of Pipeline); Final Rule.
11. “Resistance Factors and Companion-Action Load Factors for Reinforced Concrete Design in Canada”.
S.N.Kariyawasam. D.M. Rogowsky. and J.G. MacGregor. Canadian Journal of Civil Engineering. Vol. 24.
pp. 593-602. 1997.
12. “Statistical Considerations in fatigue”. Paul S Veers. ASM Handbook. Fracture and fatigue. Volume 19.
pp. 295-301. 1996.
13. “Guide on Methods for Assessing the Acceptability of Flaws in Metallic Structures”. BS7910:1999. British
Standard. Annex K, p.144, 1999.
14. “An Approach to Assessment of Corrosion in Pipelines”, B. Gu. R. Kania. S. Sharma and M. Gao.
International Pipeline Conference. Paper 27243. 2002.
15. “Reliability-Based Limit-State Design and Re-qualification of Pipelines” Y. Bai, and R. Song. Proceedings
of the Seventeenth International Offshore Mechanics & Arctic Engineering Conference. OMAE98-2849.
Lisbon. Portugal. 1998.
16. J. G. MacGregor, Safety and Limit States Design for Reinforced Concrete. Canadian Journal of Civil
Engineering. Vol. 3. No. 4. pp. 484-513. 1976.

11

Das könnte Ihnen auch gefallen