Page 286 - Pipeline Risk Management Manual Ideas, Techniques, and Resources

Station risk assessment 131263
Model design Table 13.1 Typical database fields for risk variables
For those desiring to develop a custom station risk model, a Database, field Example entries
database-structured approach to model development could be
used. Here, a database of all possible variables is first created. Type of data (used to Engineering: data that are directly
Then, depending on the modeling needs, a specific risk model estimate the cost of counted or measured with common
is created from a selection of appropriate variables. modeling the variable) measuring tools
The comprehensive station risk variable database will Frcquenq: measurable events that
occur often enough to have
identify the contribution of any and all possible risk variables. predictive power
The user will then be able to quantify the relative risk benefit Semiquantitative: combination of
or penalty of an action, device, design specification, etc. frequency data and forecasting
However, more than 400 variables can be readily identified (see (where frequency data are rare, but
page 288) as possible contributors to station risk. Some add to potential exists) and/or ajudgment
the risk, others reduce it, and they do not impact the risk of quality
equally. One of the initial objectives of a model design should Type of failure mode Third-party damage
Corrosion
be to determine the critical variables to be considered, which is Design
a function ofthe level of detail desired. Incorrect operations
A costhenefit balance will often need to be struck between a Type of impact Health
low- and high-level risk assessment. A comprehensive, high- Environmental
resolution station facilities risk model will include all possible Business
variables, rigorously defined to allow for consistent quantita- Type offacility Aboveground storage tanks
tive data gathering. A more manageable low-resolution (high- Underground storage tanks
level-screening only) station model will include only Collection sumps
Transfer racks
variables making a larger impact to risk. The large volume of Additive systems
detailed data necessary to support a detailed risk model often Pumps
has initial and maintenance data gathering costs that are many Compressors
times the costs of gathering a moderate volume of general data Engines
that can be filtered from existing sources. Piping
The risk variables database should be structured to allow Level of detail High-se only for very detailed
sorting, filtering, and selection of variables based on any of the models
database fields to provide optimum flexibility. The evaluator Medium-use for models of moderate
complexity
can easily create multiple custom risk models, or continuously Low-use for all models
change a model, depending on requirements for level of
detail, cost of evaluation, or changes in the perceived impor-
tance of specific variables. Within the context of overall risk
assessment, making adjustments to the list of variables will not
diminish the model’s effectiveness. On the contrary, customiz- 2. The ability to compare modeling results is better preserved,
ing for desired resolution and company-specific issues should even if the choice of variables changes from user to user or
improve the model’s effectiveness. the model structure changes. For example, the relative risk
To support this approach to model design, each potential of failure due to internal corrosion can be compared to
model variable should be classified using several database assessments from other models or can be judged by an
fields to allow for sorting and filtering. The fields shown in alternate selection of variables.
Table 13.1 are examples, selected from many possible database
fields, that can define each variable.
For example, a variable such as pump motor type would be Weightings
classified as a high-level-of-detail variable, applying to pumps,
when consequences of business interruption are considered in Each variable in the database should be assigned a weight based
the model; while a variable such aspopulation density would be on its relative contribution to the risk. Whether the variable rep-
a low-level-of-detail variable that would probably be included resents a potential conditiodthreat (risk-increasing factor) or a
in even the simplist risk model. Screening of the database for preventiodmitigation (risk reduction factor), it can be first
appropriate variables to include in the model is done using the assessed based on a scale such as that shown inTable 13.2.
fields shown in Table 13.1, perhaps beginning with the “Level The number of variables included in the model will deter-
of detail” field. This initial screening can assist the evaluator in mine each variable’s influence within the model since the total
identifying the appropriate number of variables to include in risk is distributed among all the variables. This raises a model
high-, medium-, or low-resolution models. resolution issue: The more variables included in the model, the
The grouping of variables by failure modes is done for two smaller the role of each variable because of a dilution effect if
reasons: all weightings sum to 100%.
Overall company risk management philosophy guidelines
I, Data handling, analysis, and reactions are enhanced because should be established to govern model building decisions.
specific failure modes can be singled out for comparisons, Example guidelines on how risk uncertainty can be addressed
deeper study, and detailed improvement projects. include these: