Page 203 - Pipeline Risk Management Manual Ideas, Techniques, and Resources

8/180 Data Management and Analyses
an event with no length. The distinction often has more to do can be drawn to determine population density around the
with how the data are collected. For instance, depth of cover is pipeline.
normally measured at specific points and then the depth is These type of data are generally converted into continuous
inferred between the measurements. So even though the depth bands by assuming that each reading extends one-half the
itself is often rather constant, the way in which it is collected distance to the next reading.
causes it to be treated as point data.
Examples ofpoint Event Data Eliminating unnecessary segments
A pipe-to-soil measurement Data that are collected at regular intervals along the pipeline are
Soil pH measurements at specific points often unchanging, or barely changing, for long stretches.
Depth of cover-actual measurements Examples of closely spaced measurements that often do not
Drain volume calculations at specific points change much from measurement to measurement include CIS
Elevation data. pipe-to-soil potential readings, depth of cover survey readings,
and soil pH readings. Unless this is taken into account, the
Examples of Continuous Datu process that breaks the pipeline into iso-risk segments will cre-
Pipe specifications ate many more segments than is necessary. A string ofrelatively
Depth of cover (when estimated) consistent measurements can be treated as a single band of
Flowrates information, rather than as many separate short bands. It is inef-
Procedures score ficient to create new risk segments based on very minor
Training score changes in readings since, realistically, the risk model should
Maintenance score not react to those minor differences. It is more efficient for a
Earth movement potential knowledgeable individual to first determine how much of a
Waterways crossings change from point to point is significant from a risk standpoint.
Wetlands crossings. For example, the corrosion specialist might see no practical dif-
ference in pipe-to-soil readings of 910 and 912 millivolts.
Some of these continuous data examples are evaluation Indeed, this is probably within the uncertainty of the survey
scores, such as “Procedures score,’’ which is described equipment and process. Therefore, the risk model should not
elsewhere. distinguish between the two readings. However, the corrosion
specialist is concerned with a reading of 9 10 mV versus a read-
Inferring continuous data ing of 840 mV, and the risk model should therefore react differ-
ently to the two readings. The use of normal operating pressures
Because the risk model requires variables to be characterized is another example. The pipeline pressure is continuously
continuously along the pipeline, all data must eventually be in changing along the pipeline, but smaller changes are normally
continuous format. Special software routines can be used to not of interest to the risk assessment.
convert point event data into continuous data, or it can be done
manually. Creating categories of measurements
Some data are generated as point events, even though they
would seem to be continuous by their nature. In effect, the con- To eliminate the unnecessary break points in the event bands, a
tinuous condition is sampled at regular intervals, producing routine can be used to create categories or “bins” into which
point event data. There are an infinite number of possible meas- readings will be placed. For instance, all pipe-to-soil readings
urement points along any stretch of pipeline. The measure- can be categorized into a value of 1 to IO. There will still be
ments taken are therefore spot readings or samples, which are sharp delineations at the break points between categories. If a
then used to characterize one or more conditions along the reading of -0.89 volts falls into category = 4 and -0.90 volts
length of the pipeline. This includes point measurements taken falls into category = 5, then some unnecessary segments will
at specific points, such as depth of cover, pipe-to-soil voltage, still be created (assuming the difference is not of interest).
or soil pH. In these cases, measurements are assumed to repre- However, the quantity of segments will be reduced perhaps
sent the condition for some length along the line. vastly, depending on the number of categories used. The user
Other point event data are not direct measurements but sets the level of resolution desired by choosing the number of
rather the result of calculations. An example is a drain vol- categories and the range of each. A statistical analysis of actual
ume calculated based on the pipeline’s elevation profile. readings, coupled with an understanding of the significance of
These can theoretically be calculated at every inch along the the measurements, can be used to establish representative cate-
pipeline. It is common practice to select some spacing, gories. A frequency distribution of all actual readings will assist
perhaps every 100 ft or 500 ft, to do a calculation. These cal- in this categorization process.
culated points are then med into continuous data by assuming
the calculated value extends half the distance to the next Assigning zones of influence
calculation point. Other examples include internal pressure
and population density. Internal pressure changes continuously A special case of converting point data into continuous data
as a function of flowrate and distance from the pressure involves assigning a zone of influence. Some data are very
source. Similarly, as one moves along the pipeline, the popula- location specific but provide some information about the sur-
tion density theoretically changes with every meter, since each rounding lengths of pipe. These data are different from the
meter represents a new point from which a circle or rectangle sample data previously discussed since the event of interest is