You are on page 1of 7

Proceedings of the 2012 9th International Pipeline Conference

IPC2012
September 24-28, 2012, Calgary, Alberta, Canada

IPC2012-90424

APPLICATION AND VALIDATION OF STATISTICALLY BASED CORROSION


GROWTH RATES

Clifford J. Maier Pamela J. Moreno


Det Norske Veritas (USA), Inc. Det Norske Veritas (USA), Inc.
Dublin, Ohio, USA Katy, Texas, USA

William V. Harper, PhD, PE David J. Stucki


Otterbein University, Mathematical Sciences Otterbein University, Mathematical Sciences
Westerville, Ohio, USA Westerville, Ohio, USA

Steven J. Polasik Thomas A. Bubenik, PhD, PE


Det Norske Veritas (USA), Inc. Det Norske Veritas (USA), Inc.
Dublin, Ohio, USA Dublin, Ohio, USA

David A. R. Shanks, P. Eng. Neil A. Bates, P. Eng.


Det Norske Veritas (Canada) Ltd. Det Norske Veritas (Canada) Ltd.
Calgary, Alberta, Canada Calgary, Alberta, Canada

ABSTRACT Validation of the SAC-predicted corrosion growth rates is


When it comes to managing the integrity of corroded important for establishing confidence in the process. This is
pipelines, operators are confronted with many difficult achieved through inspection signal comparisons, integrating
decisions – one of which is the level of conservatism that is close interval survey (CIS) results, and (when possible) field
used in pipeline integrity assessments. The financial verification. The means by which these methods are used for
implications associated with excavation, repair, rehabilitation, validating the SAC method are described in this paper.
and inspection programs typically balance the level of
conservatism that is adopted. More conservative approaches INTRODUCTION
translate into more spending, so it is important that repair In the interest of the public safety and regulatory
strategies developed based on the integrity assessment results compliance, minimum standards exist for the repair of critical
are effective. anomalies that are identified. So, the decision to address
As integrity assessment methodologies continue to evolve, critical anomalies is pre-determined. However, it is often more
so does the ability to account for local conditions. One difficult to make integrity management decisions for sub-critical
development in recent years has been the ability to evaluate anomalies. For these anomalies, operators must decide how
multiple MFL in-line inspections to determine areas of active much conservatism to build into integrity assessments. The
corrosion growth, through the combined use of statistics, financial implications associated with excavation, repair,
inspection signal comparisons, and engineering analysis. The rehabilitation, and inspection programs typically balance the
authors have previously outlined one approach (commonly level of conservatism that is adopted. More conservative
known as Statistically Active Corrosion (SAC)) that has been approaches translate into more spending, so it is important that
successfully used to identify areas of probable corrosion repair strategies developed based on the integrity assessment
growth, predict local corrosion growth rates, and maximize the results are effective.
effectiveness of integrity assessments.[1]

1 Copyright © 2012 by ASME

Downloaded From: http://proceedings.asmedigitalcollection.asme.org/ on 09/07/2017 Terms of Use: http://www.asme.org/about-asme/terms-of-use


Pipeline integrity assessments are useful in guiding pipeline The SAC process (as implemented by the authors) is based
operators to areas of corrosion that pose an integrity threat to on the reported features from the pipe listings, which are
the pipeline. However, the assessments are often based on coupled with sensitivity adjustments, statistical comparisons,
conditions that are assumed constant over long sections of and a manual ILI signal review. While verification through ILI
pipeline – perhaps entire pipeline systems. Such assumptions signal review is deemed critical by the authors, elements of this
generally lead to conservative but unrealistic results. Excess approach reflect additions and adaptations that have occurred
conservatism can result in unnecessary mitigative activities, since the SAC concept was introduced.
especially in areas that aren’t actually impacted by a particular Initial work on statistically active corrosion was done by
threat in the first place. Lara [2] in 1997 and Rust, Burgoon, and Lara in 1999 [3].
One development in recent years has been the ability to Follow-on applications were done by Rust and Johnson in 2001
evaluate multiple MFL in-line inspections to determine areas of [4]. There were several problems with these initial methods are
active corrosion growth, through the combined use of statistics, briefly described below:
inspection signal comparisons, and engineering analysis. The 1. Much of the ILI data was not used.
authors have previously outlined one approach (commonly 2. Assumption that an extreme-value distribution fit the
known as Statistically Active Corrosion (SAC)) that has been data.
successfully used to identify areas of probable corrosion 3. Estimated representative corrosion was much less than
growth, predict local corrosion growth rates, and maximize the mean corrosion.
effectiveness of integrity assessments.[1] Many operators using 4. Analysis only involved moving windows (generally
this approach have experienced significant cost savings because 100 feet).
integrity management activities were more effectively targeted With regard to data usage, initial approaches considered
to areas where corrosion growth is occurring. only those windows that have reported features in at least
Confidence in the SAC approach is realized through 10 feet of the 100-foot moving window. This in several ways
inspection signal comparisons, integrating close interval survey resulted in much of the data not being used to estimate
(CIS) results, and field verification. The SAC verification corrosion growth. For the second point, it was not realistic to
process is outlined in this paper, following a brief review of the assume that an extreme-value distribution would fit each
SAC methodology. moving window. Often an adjustment factor had to be used
depending on the data set (this was typically done for 25% to
REVIEW OF THE SAC METHODOLOGY 40% of the moving windows). Thirdly, the corrosion estimate
Figure 1 shows the components of the SAC approach, which used was the Estimated Pit Median Depth (EPMD) based on the
uses statistical analyses to identify areas where growth may be extreme-value distribution that grossly under-estimated the
occurring. mean depth of the reported features. In an analysis of
309,269 moving windows by two of the co-authors, the EPMD
was always less (see Figure 2) than the corresponding simple
sample mean with an average difference of 20.3 mils (EPMD
average of 42.3 mils versus sample mean average of 62.6 mils).
The EPMD ranged from 1.5 to 82 mils less than the sample
mean in these moving 100-foot windows.
Histogram of EPMD-Mean
9000

8000

7000

6000
Frequency

5000

4000

3000

2000

1000

0
-73.5 -63.0 -52.5 -42.0 -31.5 -21.0 -10.5 0.0
EPMD-Mean

Figure 2. EPMD Minus Mean for 309,269 Moving


Windows.
Figure 1. Overview of the SAC Approach.

2 Copyright © 2012 by ASME

Downloaded From: http://proceedings.asmedigitalcollection.asme.org/ on 09/07/2017 Terms of Use: http://www.asme.org/about-asme/terms-of-use


These and other concerns led to new developments in how the CP system, coating condition, corrosion criticality, and
statistically active corrosion is evaluated. These developments estimated corrosion rates can be developed.
were driven by efforts to include all data. Based on this
foundation, new routines were established to deal with a variety VALIDATION VIA INSPECTION SIGNAL
of real world issues such as having data in the most recent ILI COMPARISONS
run but none in an earlier ILI run. In some approaches, little or no attention is given to the
A variety of statistical methods are used as appropriate to manual ILI signal review. In other cases, the techniques used to
the given situation. The analysis is performed on moving analyze the ILI signals might ignore characteristics that reveal
windows (of usually 30 m or 100 ft lengths), joints, and girth important information about the corrosion anomalies. As
welds. Statistical techniques compare the means, maximums, illustrated below, a detailed review of the ILI signals is critical
and counts between the ILI runs for each item of interest to achieving the desired level of confidence (i.e., verification)
(window, joint, girth weld). Bias detection statistical algorithms and ultimately ensuring that integrity activities are directed to
are employed to aid the comparison of ILI runs. the locations where they are needed.
Depending on the operator’s needs, the output options are Some engineers may be inclined to only consider ILI vendor
numerous including potential summary information such as km spreadsheets for performing statistical comparisons of multiple
(mi) ranking, worst 25 joints, etc. These output options can be inspections. Limiting the scope of the comparison in this way
used to either establish general corrosion growth rates over discounts the additional knowledge that can be gained by
contiguous areas or to pinpoint specific joints of concern. reviewing the raw inspection signal data. Moreover, there is an
Results are based on average depth changes over time (for increased likelihood of predicting corrosion growth in areas
joints or windows), maximum pit depth change over time, or where no change is actually occurring, or vice versa. Common
items such as whether there is a statistically significant increase issues that can result in “false” hot spots are:
in the number of reported features over time along the pipeline. • The two ILI tools may have a different number of
The desired confidence level may be specified for statistical sensors, which could cause differences in grading of
significance, which can be important when identifying areas of some anomalies.
the pipeline for further investigation. • Many anomalies, especially near welds, may be
The statistical comparison is improved by reviewing the reported on one survey but not the other. The actual
alignment of box dimensions, as provided by the ILI vendor signals associated with these anomalies may not show
(typically in spreadsheet format) in the length, width and depth evidence of significant growth (see Figure 2).
measurements. Through a comparison of these anomaly
dimensions, small localized areas of corrosion growth can be
characterized for further investigation.
Inspection signal comparisons confirm where real change
has taken place. Such comparisons are used to also identify
locations that may be missed by the use of statistics alone. A
good understanding of corrosion mechanisms helps in
determining remaining life by estimating mean and maximum
corrosion growth rates. Ultimately, local corrosion rates are
determined, which are commonly incorporated into a
probability of exceedance (POE) analysis.
The POE analysis provides a long-term integrity assessment
based on ILI data by evaluating the likelihood of a leak and the
likelihood of a rupture; whereas, a deterministic analysis
evaluates the time to failure of a leak and the time to failure of a
rupture. Both methodologies assume a leak occurs when the
depth of external metal loss exceeds 80% of the nominal wall
thickness and a rupture occurs when the predicted burst
pressure is less than the maximum operating pressure (MOP).
One POE methodology has been proposed by Mora et al. [5]
and Vieth et al. [6].
Since local corrosion growth rates from the SAC approach
are used in the integrity assessment, operators can better
prioritize repairs and set re-assessment intervals as compared
with approaches where uniform corrosion growth rates are Figure 2: Comparison Between 2005 (Top) and 2010 Data
assumed. Ultimately, action plans involving knowledge about (Bottom) for Anomalies near Girth Weld.

3 Copyright © 2012 by ASME

Downloaded From: http://proceedings.asmedigitalcollection.asme.org/ on 09/07/2017 Terms of Use: http://www.asme.org/about-asme/terms-of-use


the verification work, attention is drawn to these areas. The
• The algorithms used on the two surveys may indicate sensitivity of the surveys influence signal amplitude, which is
different depths for signals that did not show evidence addressed before the SAC statistical work is performed (refer to
of significant growth. These algorithms are Figure 1).
proprietary and generally cannot be reviewed.
• The interaction criteria may be more conservative in
one year compared to the other. This could cause
anomalies to be represented differently in the boxed
data (i.e., ILI vendor spreadsheet) whereas they are the
same in the raw data.
Additionally, there are some important considerations when
automated approaches are used for comparing datasets. In most
cases, an automated approach alone is very challenging and
prone to misconceptions. In such cases, comparing the actual
signals recorded in each inline inspection run can improve the
automated approach. By comparing actual signals (see example
in Figure 3), low levels of corrosion growth can be verified or
Figure 4: Comparison of Same Portion of Line from Older
removed from consideration. This is often useful in
Inspection (Left) to More Recent Inspection (Right).
determining differences in tool recording capabilities versus
actual corrosion growth on a pipeline segment. Anomaly by
Raw ILI data from both the previous and the current
anomaly comparison may also be useful when a particular
inspections is also used to review previous excavation results.
anomaly needs to be reviewed over several inspections to
Previous repair locations are included in the manual review to
determine its severity and/or origin.
understand the differences in the signal data between the two
runs for locations that are not expected to have grown. This
understanding aids in verifying the survey accuracy. Once the
accuracy of the survey is verified, the sensitivity of surveys can
be matched to represent the actual condition of the pipeline as
closely as possible. This results in a good ILI representation of
the actual condition of the pipeline. For example, adjustments
are made to account for instances where corrosion areas have
appeared to “heal” themselves. These results can then be used
to match the ILI data sets in sensitivity to ensure that standard
tool differences have been removed prior to the statistical
processing.
The numerical comparisons can now be completed. Here
the potential hot spots of corrosion growth are identified
utilizing multiple statistical and non-statistical comparison
Figure 3: Anomaly by Anomaly Comparison. methods, such as:
• Statistically significant differences in the mean depth
When a manual review of the raw signal data is performed
in 2010 versus 2005.
for validating the SAC results, there are many aspects that are
considered: • Unusual differences in the numbers of anomalies
graded per joint in 2010 versus 2005.
• Differences in sensitivity between surveys • Unusual differences in the maximum depth in 2010
• Degree of signal noise versus 2005.
• Distinguishing manufacturing anomalies from • Overall highest depths in 2010
corrosion anomalies • Areas with small numbers of anomalies but high SAC
• Number of sensors used for each inspection growth rates
• Reporting thresholds • Other joints which stood out in the statistical analysis
• Interaction rules as unusual or likely to have had growth.
• Integration to other ILI
Areas are classified in during the review based on evidence
Figure 4 shows the change in raw signal traces over a 4 year suggesting the likelihood of corrosion growth. Typical
time horizon for one pipeline section. Some changes in the classifications used are shown in Table 1.
signal patterns are evident, as indicated by the ovals. During

4 Copyright © 2012 by ASME

Downloaded From: http://proceedings.asmedigitalcollection.asme.org/ on 09/07/2017 Terms of Use: http://www.asme.org/about-asme/terms-of-use


Figure 5a: 2005 Signal Data (Run 1).
Table 1. Classifications Assigned to ILI Signals During
Manual Signal Review.

Classification Description
1-Probable Significant The ILI signals appear to
Growth demonstrate a large difference
between each tool run for depth,
length, or width
2-Possible Growth The ILI signals appear to
demonstrate a difference between
each tool run, but this difference is
not as pronounced as “Probable
Significant Growth”
3-Possible Growth on The ILI signals indicate that a new
New Pipe Installed pipe joint has been installed with
some evidence that metal loss may
have occurred
4-Unlikely Growth The ILI signals do not appear to
demonstrate a difference between
each tool run
5-New Pipeline The ILI signals indicate that a new
Installed pipe joint has been installed
6-Inconclusive The ILI signals are difficult to
interpret to determine if growth has Figure 5b: 2010 Signal Data (Run 2) – Probable Significant
or has not occurred. An example Growth Areas Indicated.
would be metal loss within the girth
weld.

Figures 5-8 show various ILI vendor software


representations of areas identified in this type of procedure.
The areas contained within the blue ovals were identified as
active corrosion through the comparison of the multiple ILI raw
data sets.

Figure 6: Comparison Between 2005 and 2010 Data


Representing Possible Significant Growth.

5 Copyright © 2012 by ASME

Downloaded From: http://proceedings.asmedigitalcollection.asme.org/ on 09/07/2017 Terms of Use: http://www.asme.org/about-asme/terms-of-use


INTEGRATING CIS RESULTS AND FIELD
VERIFICATION
Additional verifications can be provided through an analysis
of the cathodic protection (CP) system and by performing
excavations. Excavations and direct examination offer the most
conclusive evidence as to whether or not corrosion growth is
occurring.
Operators routinely use CIS and rectifier data to evaluate
the performance of their CP system, so it makes sense to also
use these readily-available data for verifying the SAC results.
CIS and rectifier data provide valuable insight regarding the
corrosion activity on the pipeline. The following activities are
typically performed with the data, in conjunction with the SAC
verification:
Figure 7: Comparison Between 2005 and 2010 Data
Representing Possible Growth. • Evaluation of the adequacy of the CP protection with
respect to the -0.85 V copper sulfate electrode (CSE)
polarized potential criteria
• Evaluation of trends in the DC current output
• Evaluation of coating condition according to current
density and CP design current requirements
• Identification of possible stray current interference
locations
• Establishing the health of anode ground beds
Further, areas of corrosion activity identified by the
CIS/rectifier data review are compared with those areas
predicted by the SAC analysis to have active corrosion or fast
growth rates. Active corrosion areas predicted by both methods
can be prioritized for verification excavations. Excavations that
are identified can be prioritized using one or more of the
following criteria:

• Manual Review Results – those joints identified during


the manual signal review to contain “Probable
Significant Growth”
• CIS integration results - areas with insufficient
cathodic protection, or where “hot spots” have been
identified
• Deterministic Remaining Life – the estimated time for
the depth to reach 80% WT or the predicted failure
pressure to reach MAOP
• ILI Vendor Review Results – those joints identified
during the ILI vendor’s manual signal review to
contain “new ML”

ADDITIONAL VALIDATION USING FIELD DATA


Figure 9 [7] illustrates some of the aspects used in blending
Figure 8: Comparison Between 2005 and 2010 Data
field data (or the lack thereof) into a strategic comparison to the
Representing Probable Significant Growth.
ILI information. This is based on well-known epidemiology
constructs [8] that provides a foundation that meets both
As viewable in the above figures, the ILI raw data is an
engineering risk management protocols as well as demanding
important data set for validation of the identified statistically
forensic medical requirements for the Center for Disease
active areas.
Control. Many things are put into perspective above and
beyond the typical true positives, false positives, etc.

6 Copyright © 2012 by ASME

Downloaded From: http://proceedings.asmedigitalcollection.asme.org/ on 09/07/2017 Terms of Use: http://www.asme.org/about-asme/terms-of-use


[4] RUST, S. W., and JOHNSON, E. R., Statistical Method
for Identifying Active Corrosion Based on Data from
Multiple In-Line Inspection Runs, Paper 01622, Corrosion
2001, NACE, March 11-16, 2001, Houston, Texas.
[5] MORA, R. G., PARKER, C., VEITH, P. H., and
DELANTY, B. Probability of Exceedance (POE)
Methodology for Developing Integrity Programs Based on
Pipeline Operator-Specific Technical and Economic
Factors. Proceedings of the 4th International Pipeline
Conference, ASME, 2002.
[6] VEITH, P. H., RUST, S. W., and ASHWORTH, B. P.,
Figure 9: An Epidemiological Approach to Validation “Use of In-line Inspection Data for Integrity
Comparing Field and ILI Measurements. Management”, NACE International, Proceedings of
Corrosion 99 Conference, 1999.
Complementing the comparison of ILI and field [7] http://en.wikipedia.org/wiki/Sensitivity_and_specificity
measurements (both with measurement errors) is the use of a [8] GORDIS, L., Epidemiology, 4th ed., Saunders,
more robust RMA regression over the theoretically unjustified Philadelphia.
least squares methods [9]. [9] HARPER, W. V., STUCKI, D. J., BUBENIK, T. A.,
MAIER, C. J., SHANKS, D. A. R., and BATES, N. A.,
CONCLUSIONS Improved Comparison of ILI Data and Field Excavations,
The SAC methodology is one approach that can be used for Proceedings of the 2012 9th International Pipeline
predicting corrosion activity and corrosion growth rates in Conference, Paper 90440, Calgary, 2012.
pipeline sections that have been inspected multiple times by
MFL tools. The ability to predict local corrosion growth rates
and improve long-term failure predictions benefits operators
that use the SAC approach because mitigative activities can be
more effectively targeted to specific pipeline sections where
corrosion is actively growing.
This paper presents methodologies that can be used for
validating active corrosion and growth rate predictions,
including a review of the inspection signal data, integration of
CIS data, and field verification. The authors contend that a
detailed review of the ILI signals is critical to achieving the
desired level of confidence (i.e., verification) and ultimately
ensuring that integrity activities are directed to the locations
where they are needed.

ACKNOWLEDGMENTS
The authors acknowledge the support of Det Norske Veritas
(USA), Inc., DNV Energy Canada Ltd., and selected clients in
carrying out the work needed to prepare this paper.

REFERENCES
[1] MORENO, P. J., MAIER, C. J., BATES, N. A., SHANKS,
D. A., HARPER, W. V., STUCKI, D. J., and BUBENIK,
T. A., “Development and Application of Local Corrosion
Growth Rates for Pipeline Integrity Assessments”, NACE
International, Proceedings of Corrosion 2012 Conference,
2012.
[2] LARA, P. F., Pipeline Corrosion Rate Estimation from
Smart Pig Data, RR 97-0008, ARCO Exploration and
Production Technology, Plano, Texas, November, 1997.
[3] RUST, S. W., BURGOON, D. A., and LARA, P. F.,
Identifying Active Corrosion Sites on the Trans Alaska
Pipeline System, Report for Alyeska, January 11, 1999.

7 Copyright © 2012 by ASME

Downloaded From: http://proceedings.asmedigitalcollection.asme.org/ on 09/07/2017 Terms of Use: http://www.asme.org/about-asme/terms-of-use

You might also like