Page 325 - Pipeline Risk Management Manual Ideas, Techniques, and Resources

14/302Absolute Risk Estimates
Table 14.13 Example of predicting leaklbreak probabilities overtime based on relative risk scores

Time Day 1 Year1 Year IO Year20 Year30 Year40 Years0 Year60 Year 100 Year150 Year 199
Third-party 100 60 60 60 60 60 60 60 60 60 60
damage index
Corrosion index 100 90 85.5 81 76.5 72 67.5 63 45 22.5 0.45
Design index 100 90 85.5 81 76.5 72 67.5 63 45 22.5 0.45
Incorrect operations 99 70 70 70 70 70 70 70 70 70 70
index
Index sum 399 310 294.5 265.05 225.29 180.23 135.18 94.62 47.31 11.83 0.06
Relative failure 1.00% 65.98% 69.30% 72.44% 75.42% 78.23% 80.86% 83.33% 91.50% 97.87% 100.00%
probability score

safety level, indicating almost no chance of failure. In other VII. Ignition probabilities
words, a brand new system, successfully pressure tested and
placed into operation will probably not immediately fail. At The possibility of ignition of a flammable pipeline product is a
year 1, the chances for something to go wrong have increased part of most hazard scenarios for hydrocarbon pipelines.
and failure probability is dominated by third-party damage Ignition is usually thought to increase consequences, but can
potential and human error, neither ofwhich are sensitive to age also theoretically reduce them. A scenario where immediate
of the system. Gradually, the time-dependent mechanisms ignition causes no damage to receptors but eliminates a con-
become dominant, until at year 200, there is nearly a 100% tamination potential (preventing groundwater contamination
chance that the segment has had at least one failure since instal- or shoreline damage from an offshore spill, for example) is
lation. This assumes there is no integrity verification, inspec- such a case.
tion, or other means of “resetting” the clock to demonstrate that Ignition probability is, of course, very situation specific.
deterioration is not actually occurring. So, in year 100, there is Countless scenarios are possible for most pipelines. Ignition of
still a 70% chance of survival of third-party related failures a flammable gas release can occur at either the source or a loca-
according to the third-par@ index, but due to assumed deterio- tion some distance away. A buoyant gas such as hydrogen or
ration impacts over time, the survival probability from corro- natural gas will rise rapidly on release and limits the formation
sion effects is only 40%. The relative failure probability scores of a flammable gas cloud in open space. With the assumption
are based on the statistical combination of the four failure prob- that most ignition sources are at or near ground level, this
abilities (this calculation is discussed on page 300). They can be reduces the probability of remote ignition for these lighter
converted into leakibreak rates with the proper establishment of gases.
space and time parameters-determining that the relative prob- In some cases, the source of ignition is related to the loss of
abilities relate to the chances of one or more failures perXfeet containment itself, such as sparks generated by impact from
per Yyears. machinery or heat generated by the release process itself
Table 14.13 shows a very rudimentary and perhaps overly (including static electricity or sparks from flying debris colli-
simplistic example with some very generalized assumptions sions). Other sources of ignition include
such as assuming that each index score is actually a survival
probability-an assumption that is potentially disputed given Vehicles or equipment operating nearby
the complex relationships between risk variables and the use Grinding and welding
of a simplifying scoring regime. This table is provided only Residential pilot lights or other open flames
to illustrate the one possible approach for leaWbreak predic- External lighting or decorative fixtures (gas or electric).
tion using risk scores. An actual forecasting tool should be
calibrated by historical failure rate data and careful scrutiny It is not uncommon during gaseous product release events for
of the model to ensure that key survivability factors such the ignition source to be created by the release of energy,
as wall thickness are appropriately considered in the index including static electricity arcing (created from high dry gas
scores. velocities), contact sparking (e.g., metal to metal, rock to rock,
The leakibreak rate should be continuously adjusted by risk rock to metal), or electric shorts (e.g., movement of overhead
factors andor inspection results to arrive at an assessment for power lines).
all lengths of pipe. Historical breaks and leaks would normally Estimates of ignition probabilities can be generated from
be considered to be evidence that overrides previous estimates company experience, pipeline failure databases, or obtained
of failure probability. via literature searches. The following empirical formula is rec-
Converting corrosion scores to corrosion rates might be ommended for use in quantitative risk assessments for gas
especially important when making repair versus replace deci- pipelines in Australia [67]:
sions. Such decisions are typical in distribution systems where
leak indications are often the main source of integrity informa- Ignition probability = 0.01 56(release rate in kgis)o.”2
tion (see Chapter l l). A discussion ofusing corrosion scores to
estimate deterioration rates (corrosion rates) can also be found Some other examples of ignition probability estimates are
in Chapter 1 1. showninTables 14.14 through 14.18.