The Use of data-driven methodologies for prediction of water and wastewater asset failures

Water and wastewater asset failure analysis, according to the author, take two approaches. The first accumulates data from utilities as an industry segment and derives performance indicators (PIs). The second method extracts information from the particular utility and defines performance indicators. The combination of historical data on pipe attributes and their repair history supplies some of the data required for modeling water and wastewater performance, consistent with the second method. "The scope of such analyses is discovering patterns in asset data for describing pipe failures (e.g., sewer blockages or collapses and water distribution pipe bursts)" (p. 182).


Difficulties in applying the second method within water utilities result in the typical structure of their data.
The author cited the disparity in the quality and quality of asset data and the lack of integration between asset attribute and repair history data. Insufficient data prohibits a longitudinal analysis of pipe performance.
To overcome this deficiency, the author suggested forming pipe classes, "aggregating pipes into homogeneous groups" (p. 183). By utilizing the model, Evolutionary Polynomial Regression, with this data utility managers have a tool to identify variables to predict pipe failures.


A case study on a United Kingdom water utility, that expanded a fourteen year period from 1986-1999, demonstrated the application of the model. The data included the following: "pipe diameter, material, year laid, length, number of properties supplied and the total number of bursts recorded" (p. 184). The data lacked information on the selection criteria for the area, a map interface, and the specific timing and duration of each pipe burst. The study only had the total number of pipe bursts. For water distributions systems the model confirmed that age, length, and diameter contribute to pipe failure. Wastewater systems, based on a case study of two United Kingdom sewer systems, with sewers characteristically older, small in diameter, and in residential areas tended to experience higher incidences of blockage and collapse.


Savic, D. A. (2009). The use of data-driven methodologies for prediction of water and wastewater asset failures. In P. Hlavinek et al. (Eds), Risk management of water supply and sanitation systems (pp. ). Dordrecht: Springer.