Page 343 - Pipeline Risk Management Manual Ideas, Techniques, and Resources

14/320 Absolute Risk Estimates
that would better refine these estimates have been available. It is also Since the national pipeline system is not characterized in these terms,
important to note that frequencies and probabilities like these repre- the similarities cannot be confirmed. However, since the LMP speci-
sent averages expected only over long periods oftime. Short time peri- fies several state-of-the-art spill size reduction measures not typically
ods can have different experience and still be appropriately seen in other pipelines, it is reasonable to assume that the national data
represented by these frequencies. Therefore, the predictive power of will not underestimate the spill size potential and very probably will
these probabilities is limited. overestimate the potential.
As an additional evaluation step, the plausibility of the estimated
post-mitigation leak frequency was examined qualitatively. The A second assumption is that the <50 bbl spill size fraction seen
estimate is generally supported by this qualitative analysis. summa- under EPC operations is representative of LPP’s future spill size dis-
rized as follows: tribution. Since the <50 bbl size triggers few impacts and since >50
bbl spill fraction can be separated from the “all size” distribution,
1. Low leak frequencies over long periods oftime are being experi- the absolute validity of this assumption is not critical to thls
enced by US pipeline operators on hazardous liquid pipeline of analysis.
similar length to the LPP pipeline, but without the extraordinary An additional underlying assumption in these estimates is that the
level of mitigations as proposed in the LMP. This is indicated by relative probability of failure remains fairly constant over the life of
informal interviews with pipeline operators and with searches the project. This is accomplished by LPP reacting appropriately to
and analyses of OPS accident data. Analyses of these latter data changing conditions along the line, as is specified in the LMP. It also
are discussed in Attachment E [not included in this book]. requires that the integnty verifications as scheduled by ORA calcula-
Results of summary analyses of DOT and other data are pro- tions, ensure that the probability of failure does not exceed the pro-
vided. These data and analyses suggest that the estimated leak jected leak probabilities between integrity verifications. This is
frequency is possible, especially with increased mitigation. discussed in Appendix 9D [not included in this book].
2. The correlation as described in Attachment A [see page 2981,
although weak in terms of statistically valid data quantity and
quality, nonetheless offers a semi-quantitative linkage that sup- Description ofpotential impacts
ports the estimate.
3. Appendix T [not included in this book] shows leak rate estimates Nine distinct potential impacts are studied in this report. Impacts are
site-specific and sensitive to many variables, and therefore must be
for approximately 60 U.S. hydrocarbon liquid pipeline operators. somewhat generalized to present a risk picture of the entire line. For
These leak rates, presumably achieved under typical industry miti- modeling purposes, the frequency of each impact is potentially
gation levels, show the range of different leak rates that are possi- affected by variables of.
ble. This includes company-wide leak rates that are approaching
the estimated post-mitigation leak frequency estimates forthe LPP
pipeline. Index sum-representing the probability of pipeline failure;
4. The scenario-based analyses detailed in Attachment B [this Spill size; and
excerpt can be found in Chapter 3, see pages 00001 suggests that Tier designation-representing receptor vulnerability and sensitiv-
the estimated leak rate reductions can be achieved with rather ity (e.g., Tier 3 is hypersensitive).
modest assumptions regarding mitigation effectiveness, even for
the more problematic challenge of reducing third-party damage. However, not all impacts are modeled as being sensitive to all ofthese,
5. An alternative approach to estimating failure probabilities from due to data availability limitations. Below is a general description of
several common pipeline failure mechanisms has produced very the impacts modeled. These descriptions offer the reader a general
similar results. This alternative approach, shown in the preliminary sense of the rationale behind the calculation, but note that the actual
ORA [ORA = Operational Reliability Assessment, not included in results are based on more than a hundred calculated scenarios. More
this book] uses concepts from fracture mechanics, materials sci- detailed descriptions can be found in Attachment C [Appendix F of
ence, historical data. and statistics to calculate failure rates and this book].
probabilities. The fact that two separate approaches to failure prob-
ability estimation arrived at similar conclusions provides support Fatalities and injuries
for both calculations.
6. In the experience ofthe EA authors, the LMP reflects levels ofmit- While it is common to express risks of injuries and fatalities as a
igation unprecedented in the industry. This suggests that high function of “hours exposed,” this analysis uses only a calculation of
levels of leak rate reductions are possible, even if not commonly fatalities and injuries per reportable leak. All distinctions of rural
observed. versus urban; permanent residents versus temporary exposures; dis-
In addition to overall leak frequencies, spill size frequency also plays a tances to leaks; ignition probabilities; etc. are therefore aggregated in
role inmany ofthe impacts.Aspill sizedistributionforspillslargerthan these ratios. This implies that the LPP system is similar to the
50 bbl was derived from DOT hazardous liquid pipeline reportable national data in terms of these variables. The national pipeline sys-
spills from 1975 to early 2000. The fraction of spills smallerthan 50 bbl tem is not characterized to the extent that such similarities can be
was estimated from the 29 year EPC leak experience on the 450 mile confirmed. However, no compelling reasons are found to suggest
segment fromValve J- 1 to Crane. EPC leak experience contains too few that LPP is not similar, with regards to the distinctions previously
larger-sized spills to create a meaningful profile. noted. Therefore, for the purposes of the overall impact estimations,
the national data (DOT) is assumed to be representative of LPP’s
Embedded in this approach is the assumption that the national spill future risks for this impact.
size distribution (DOT data) is representative of the LPP’s future spill
An example of fatalities and injuries, is Case 1 shown in Table 3. It
size distribution. This implies that the following variables are also can be described in general terms as follows:
representative:
Statistically, one fatality is expected to occur for every 217
reportable leaks and an injury is expected to occur for every 48
Topography; reportable leaks.
Failure mechanisms that determine hole size; The industry average leak rate applied to this pipeline results in an
Leak detection capabilities; and estimate of 35 leaks over 50 years and, hence predicted fatalities and
Leak reaction capabilities. injuries of 0.16 and 0.72, respectively.