Page 56 - Pipeline Risk Management Manual Ideas, Techniques, and Resources

Designing a risk assessment model 2/35
should be evaluated. This includes a repair’s potential to cause dence. How does the model’s implied corrosion rate com-
unwanted stress concentrations. If complete and acceptable pare with wall loss observations? How does third-party
repairs that restored full component strength have been made, damage likelihood compare with dents and gouges on the
then risk assessment “penalties” can be removed. Regardless of top or side of pipe? Is the design index measure of land
repair, evidence still suggests the potential for repeat failures in movement potential consistent with observed support
the same area until the root cause identification and elimination condition or evidence of deformation?
process has been completed. C. If disagreement is apparent-the direct evidence says
Whether or not a root cause analysis has been completed, something is actually “good’ or “bad” while the risk
direct evidence can be compiled in various ways for use in model says the opposite-then perform an investigation.
a relative risk assessment. A count of incidences or a density Based on the investigation results, do one or more of the
of incidences (leaks per mile, for example) will be an following:
appropriate use of information in some cases, while a zone- Modify risk algorithms based on new knowledge.
ofinfluence or anomaly-specific approach might be better Modify previous condition assessments to reflect new
suited in others. knowledge. For example, “coating condition is actually
When such incidences are rather common-ccurring bad, not fair as previously thought” or “cathodicprotec-
regularly or clustering in locations-the density or count tion levels are actually inadequate, despite 3-year-old
approaches can be useful. For example, the density of ILI close interval survey results.”
anomalies of a certain type and size in a transmission pipeline 0 Monitor the situation carefully. For example, “existing
or the density ofnuisance leaks in a distribution main are useful third-party damage preventions are very protective of the
risk indications (see Chapters 5 and 1 I). pipe and this recent detection of a top side dent is a rare
When direct evidence is rare in time andor space, a more exception or old and not representative of the current sit-
compelling approach is to assign a zone qf influence around uation. Rescoring is not appropriate unless additional
each incident. For example, a transmission pipe leak incident is evidence is obtained suggesting that third-party damage
rare and often directly affects only a few square inches of pipe. potential is actually higher than assumed.” Note that this
However, it yields evidence about the susceptibility of neigh- example is a nonconservative use of information and is
boring sections of pipeline. Therefore, a zone of influence, X not generally recommended.
number of feet on either side of the leak event, can be assigned
around the leak. The length of pipeline within this zone of Role of leak history in riskassessment
influence is then conservatively treated as having leaked and
containing conditions that might suggest increased leak Pipeline failure data often come at a high cost-an accident
susceptibility in the future. happens. We can benefit from this unfortunate acquisition of
The recommended process incorporating direct evidence data by refining our model to incorporate the new information.
into a relative risk assessment is as follows: In actual practice, it is a common belief, which is sometimes
backed by statistical analysis, that pipeline sections that have
A. Use all available leak history and ILI results---even when experienced previous leaks are more likely to have additional
root cause investigations are not available-to help evaluate leaks. Intuitive reasoning suggests that conditions that promote
and score appropriate risk variables. Conservatively one leak will most likely promote additional leaks in the same
assume that damage mechanisms are still active. For exam- area.
ple, the detection of pipe wall thinning due to external Leak history should be a part of any risk assessment. It is
corrosion implies often the primary basis of risk estimations expressed in
0 The existence of a corrosive environment absolute terms (see Chapter 14). A leak is strong evidence of
0 Failure of both coating and cathodic protection systems failure-promoting conditions nearby such as soil corrosivity,
or a special mechanism at work such as AC-induced inadequate corrosion prevention, problematic pipe joints, fail-
corrosion or microbially induced corrosion ure of the one-call system, active earth movements, or any of
0 A pipe wall thickness that is not as thought-pipe many others. It is evidence of future leak potential. This evi-
strength must be recalculated dence should be incorporated into a relative risk assessment
Scores should be assigned accordingly. because, hopefully, the evaluator’s “degree of belief” has been
The detection of damaged coating, gouges, or dents sug- impacted by leaks. Each risk variable should always incorpo-
gests previous third-party damages or substandard installa- rate the best availableknowledge ofconditions andpossibilities
tion practices. This implies that for promoting failure.
0 Third-party damage activity is significant, or at least was Where past leaks have had no root cause analysis and/or cor-
at one time in the past rective action applied, risk scores for the type of failure can be
0 Errors occurred during construction adjusted to reflect the presence of higher failure probability
Pipe strength must be recalculated factors. A zone of influence around the leak site can be
Again, scores can be assigned accordingly. established (see Chapter 8) to penalize nearby portions of the
B. Use new direct evidence to directly validate or adjust system.
risk scores. Compare actual coating condition, pipe wall In some pipelines, such as distribution systems (see Chapter
thickness, pipe support condition, soil corrosivity, etc., with 11) where some leak rate is routinely seen, the determination
the corresponding risk variables’ scores. Compare the rela- as to whether a section of pipeline is experiencing a higher
tive likelihood of each failure mode with the direct evi- frequency of leaks must be made on a relative basis. This can be