Page 249 - Pipeline Risk Management Manual Ideas, Techniques, and Resources

11/226 Distribution Systems
installation periods and of a dramatically increasing failure more leaks and can generate more extensive (and, hence,
rate observed in many locations. more statistically certain) information on leaks. This can
0 Many investigators report that an exponential relationship be useful for failure prediction, wherefailure is defined as
between the passage of time and future leaks is the most “excessive leakage.” Given the leak tolerances, the risk
appropriate forecasting model. That is, break rates increase assessments for lower pressure systems often make a distinc-
exponentially with the passage of time. Other investigators tion between leaks and breaks, where only the latter are
report constant or decreasing break rates for specific group- considered to be failures.
ings ofpipes in certain cities [41].
One reference characterizes current statistical break predic- Sectioning
tion models into deterministic, probabilistic multivariate,
and probabilistic single-variate models applied to grouped It may not be practical to examine each piece ofpipe in a distri-
data. Reference [40] reports that a three-parameter Weibull bution system, at least not for an initial risk assessment. It may
curve is generally accepted as the best predictor of time to be more important to examine the general portions of the sys-
failure, given adequate failure history. tem that are of relatively higher risk than other sections. In
Investigators use a variety ofvariables to characterize break- many cases, the higher risk areas are intuitively obvious. Areas
age patterns. These variables tend to divide the population of with a history of leaks, materials more prone to leaks, and areas
all breaks into groups that experience similar break rates with higher population densities often already have more
over time. The most widely reported variables influencing resources directed toward them. The more detailed risk assess-
break rate seem to be ment becomes useful when the risk picture is not so obvious.
Pipe material The subtle interactions between many risk variables will often
Pipe diameter point to areas that would not have otherwise been noticed as
Soil temperature being high risk.
Soil moisture content A geographical segmentation scheme might be appropriate
Previous break countirate in some applications. A segment could represent a page in a
Age of system. map book, a grid, a pressure zone, or some other convenient
Additional variables that appear in some break forecasting grouping.
models include To optimize the sectioning of a distribution grid (see also
Soil resistivities general Sectioning discussion in Chapter2) each section should
0 Joint type exhibit similar characteristics within its boundaries, but have at
Pressure least one differing characteristic compared to neighboring
Tree locations sections. This difference is the reason for the section boundary.
Traffic. A hierarchical list of sectioning characteristics can be created
In some models, variables are identified but not fully popu- as explained on page 26. For example, if the distribution system
lated for the analysis. They therefore serve as input locations to be examined is composed of more than one material of con-
(placeholders) for information that may be gathered in the struction, then “material type” could be the first characteristic
future. to distinguish sections. As the second attribute, perhaps the
Some investigators note that for cast iron, only a fraction of pressure reduction points or pipe diameter changes provide a
through-wall corrosion holes reveal themselves by becom- suitable break point. For instance, section 1A of Acme
ing breaks [41]. The holes cause leakage below detection Distribution System might be all polyethylene (PE) pipe oper-
thresholds or within leak tolerance. ated above 50 psig in the northeast quadrant of the city of
Many references report “as-new’’ conditions observed on Metropolis. Because steel distribution systems are often
pipelines, even those with more problematic materials such divided into electrically isolated sections for cathodic protec-
as cast iron that have been in service for many decades. tion purposes, this corrosion-control sectioning might be fol-
Reference [40] uses a median of 220+ years for cast iron lowed for risk assessment purposes also.
pipe failures and states that this is collaborated by inspection In certain cases, it might be advantageous to create noncon-
of some 75+-year-old cast iron pipe “that looks to be in fac- tiguous sections. In the preceding example, a section could
tory-new condition.” include all steel pipe operated at less than 50 psig. Such a sec-
Metal porosity and excessively large graphite flakes are tion would contain unconnected pieces of the distribution net-
sources ofweaknesses observed in gray cast iron pipe, espe- work. In this scheme, pipes of similar characteristics and
cially in larger diameters [42]. environment are grouped, even if they are geographically
Similar efforts (deterioration modeling and break forecast- separate.
ing) have been undertaken for sewer pipes.

Data IV. Assigning risk scores
Differences in leak tolerance and uses of inspection result in As previously noted, a risk model similar to that described for
differences in information availability for many distribution transmission pipelines in Chapters 3 through 7 can be used to
systems. As noted elsewhere, leakage information in the dis- assess distribution systems. The following sections discuss
tribution industry replaces inspection data in the hydrocar- similarities and differences and suggest changes to the assign-
bon industry. More leak-tolerant systems generally have ment ofpoints in the risk model.