Page 287 - Pipeline Risk Management Manual Ideas, Techniques, and Resources

731264 Stations and Surface Facilities
Table 13.2 Variable risk contribution weighting 4. Volume of product stored, product hazards, prevention, and
mitigation systems all drive the magnitude of consequences.
Conditiondthreats
Process
5 Variable can easily, independently cause failwe-highest weight
4 Variable can possibly independently cause failure To outline a risk model based on the optimum number of vari-
3 Variable is significant contributor to failure scenarios ables from all of the possibilities shown in the database, the
2 Variable, in concert with others, could cause failure
1 Variable plays minor role in this failure mode-lowest weight following procedure can be used
?‘reventiom/mitigations 1. Conceptualize a level of data collection effort that is accept-
able-perhaps in terms of hours of data collection per sta-
5 Variable can easily, independently prevent failure-highest weight tion. This can be the criterion by which the final variable list
4 Variable can possibly independently prevent failure is determined.
3 Variable is significant obstacle to failure scenarios 2. Begin with an extensive list of possible risk variables, since
2 Variable, in concert with others, could prevent failure any variable could be critical in some scenario. See the sam-
1 Variable plays minor role in this failure mode-lowest weight
ple variable list at the end of this chapter.
3. Filter out variables that apply to excluded types of threats-
ones that will never be a consideration for facilities assessed
1. Results from older surveys and inspections (e.g., tank (e.g., if there is no volcano potential, then the volcano-
inspections, CP readings) will have less impact on risk related variables can be filtered out; similarly, threats from
assessments. The “deterioration” of information value meteors, hurricanes, freezes, etc., might not be appropriate).
depends on many factors and is specific to the survey/ 4. Examine the total variable count, estimated cost of data, and
inspectioniequipment type (see Chapter 2). distribution of variables across the failure modes-if
2. Estimated data will have less impact on risk scores than data acceptable, exit this procedure, determine how best to com-
with a known level of accuracy (e.g., depth of cover, coating bine the variables, and create data collection forms to popu-
condition) (see Chapter 8). late a database.
5. To minimize the level of detail (and associated costs) of
Uncertainty is further discussed in Chapters 1 and 2. the model, examine the lower weighted variables and filter
When deciding on a particular risk model structure, many cost out variables that have minimal application. In effect, the
and effectiveness factors should be considered, such as minimiz- model designer is beginning at the bottom of the list of
ing duplication of existing databases, efficiently extracting infor- critical variables and removing variables until the model
mation from multiple sources, capturing experts’ knowledge, becomes more manageable without sacrificing too much
and periodically collecting critical data. All risk model data are risk-distinguishing capability. This becomes increasingly
best gathered based on data collection protocols (e.g., restricted subjective and use-specific.
vocabulary, unknown defaults, underlying assumptions) as dis-
cussed in earlier chapters. A lower level risk model should be At any time in this process, variables can be edited and new
structured to allow “dnlling down” to assess individual equip- ones added. As implied in this procedure, care should be taken
ment, whereas a high-level risk model may be structured to allow that certain failure modes are not over- or underweighted. This
assessment at only the overall station level. procedure can be applied for each failure mode independently
The following are general risk beliefs that, if accepted by the to ensure that a fair balance occurs. Each failure mode could
model designer, can be used to help structure the model. also have a preassigned weighting. Such weighting might be
the result of company incident experience or industry experi-
1. A more complex facility will generally have a higher likeli- ence. This should be done carefully, however, since drawing
hood of failure. A facility with many tanks and piping will attention away from certain failure modes might eventually
have a greater area of opportunity for something to go negatively change the incident frequency.
wrong, compared to one with fewer such facilities (if all Having determined the optimum level of detail and a corre-
other factors are the same). A way to evaluate this is sponding list of critical variables, the model designer will now
described on pages 265-266. have to determine the way in which the variables relate to each
2. A manned facility with no site-specific operating proce- other and combine to represent the complete risk picture, The
dures and/or less training emphasis will have a greater following sections describe some overall model structures in
incomet operations-related likelihood of human error than order to give the designer ideas of how others have addressed
one with appropriate level of procedures and personnel the design issue. Most emphasis is placed on the first approach
training. since it parallels Chapters 3 through 7 ofthis text.
3. A facility handling a liquefied gas, which has the mechani-
cal energy of compression as well as chemical energy and
the ability to produce vapor cloud explosions, creates con- 111. Risk assessment model
siderably more potential health and safety-related cons-
quence than does a low vapor pressure liquid, which has no This approach suggests a methodology to generate risk assess-
mechanical energy and is much harder to ignite. On the ments that are very similar to those generated for the pipe-only
other hand, some nonvolatile liquids can create considerably portions of a pipeline system. It is based on the evaluation sys-
more environmentally related consequences. tem described in Chapters 3 through 7. For facilities that are for