Page 206 - Pipeline Risk Management Manual Ideas, Techniques, and Resources

Computer environments 8/183
Several opportunities for QNQC also arise after each 1. Problems with dates. The algorithms are generally set up to
section has been scored. The following checks can be made by accommodate either a day-month-year format or a month-
using queries against the database of scores: year format or a year-only format, but not more than one of
these at a time. The algorithm can be made more accommo-
Find places where scores are not being calculated. This dating (perhaps at the expense of more processing time) or
will usually be the result of an information gap in some the input data can be standardized.
event required by the algorithm. After the default assign- 2. Missing or incorrect codes. Non-scoring values (nulls) or
ment, there should not be any more gaps unless it is an errors are often generated when input data are missing or
event for which a default cannot logically be assigned incorrect. These create gaps in tte final scores.
(such as “diameter” or “product type”). Common causes of 3. Data gaps. As noted in item 2, these generally represent
non-scoring segments include misnamed events or condi- non-scoring values. Errors are easily traced to the initiating
tions, incorrect condition values, and missing default assign- problem by following the calculation path backward. For
ments. example, using the algorithms detailed in chapters 3 through
Find places where score limits are being exceeded. This is 6, an error in IndexSum means there is a error somewhere in
usually a problem with the algorithm not functioning as one of the four underlying index calculations (thdpty. corr,
intended, especially when more complex “if. . . then” condi- design, or incops). That error, in turn, can be traced to an
tional equations are used. Other common causes include date error in some subvariable within that index.
formats not working as intended and changes made to either 4. Maximum or minimum values exceeded. Maximum and
an algorithm or condition without corresponding changes minimum queries or filters can be used to identify variables
made to the other. that are not calculating correctly.
Ensure that scores are calculating properly. This is
often best done by setting up queries to show variables,
intermediate calculations, and final scores for the more
complex scores especially. Scanning the results of these VIII. Computer environments
queries provides a good opportunity to find errors such as
incorrect data formats (dates seem to cause issues in many The computer is obviously an indispensable tool in a data-
calculations) or point assignments that are not working as intensive process such as pipeline risk management. Because a
intended. great deal of information can be gathered for each pipeline sec-
tion evaluated, it does not take many evaluations before the total
These QNQC opportunities and others are summarized amount of data become significant. The computer is the most
below. logical way to store and, more importantly, organize and
Common input data errors include retrieve the data. The potential for errors in number handling is
reduced if the computer performs repetitive actions such as the
1. Use of codes that are not exactly correct, Le., “high” when calculations to arrive at riskvalues.
“H is required, or misspelled codes
2. Wrong station numbers, Le., a digit left off, such as entering Options
21997 when219997 iscorrect
3. Conflicting information, Le., assigning different conditions Many different software environments could be used to handle
to the same stretch ofpipeline, sometimes caused by overlap the initial data input and calculations. As the database grows,
of the beginning and ending stations of two entries. the need for programs or routines that can quickly and easily
(from the user standpoint) search a database and display the
Some QNQC checks that are useful to perform include the results of the search becomes more important. More sophisti-
following: cate risk assessment models will require more robust soft-
ware applications. A model that requires spatial analyses of
I. Ensure that all pipeline segment identifiers are included in information, perhaps to determine spill migration or hazard
the assessment zone perimeters, requires special software capabilities. Addi-
2. Ensure that only listed IDS are included. tional desired capabilities might include automatic segmenta-
3. Find data sets whose cumulative lengths are too long or too tion, assignment of zones of influence, or calculation of
short, compared to the true length of an ID. intermediate pressures based on source strength, location, and
4. Find individual records within a data set whose beginning flowrate.
station andor ending station are outside the true beginning
and ending points of the ID. Use computers wisely
5. Ensure that all codes or conditions used in the data set are
included in the codes or conditions list. An interesting nuance to computer usage is that too much
6. Ensure that the end station of each record is exactly equal to reliance on computers is potentially more dangerous than too
the beginning station of the next record when data are little. Too much reliance can degrade knowledge and cause
intended to be continuous. insight to be obscured and even convoluted-the acceptance of
7. Ensure that correctkonsistent ID formats are being used. ‘black box’ results with little application of engineering judg-
ment. Underutilization of computers might result in inefficien-
Common errors associated with risk score calculations cies-an undesirable, but not critical event. Regardless of
include: potential misuse, however, computers can obviously greatly