Predicting the behavior of a pipe network using the "critical Z-score" as its performance indicator

Abstract

The paper deals with pipe reliability assessment in two networks using the Discriminant Analysis and classification (DAC) method. The pipes of each network are divided in two groups based on whether they failed at least once (failures group) or not (successes group). Several scenarios resulting from combining pipe characteristics (such as length, diameter, wall thickness, operating pressure, grade, and product (fluid in the pipe), lifetime) are being analyzed. A sensitivity analysis of the data available takes place to check the stability of the results. The criterion of the "critical Z-score" is finally used as an indicator predicting the pipe's future state (fail or not). The goals for each network are to develop a model that can correctly classify network pipes to successes or failures; define the pipe characteristics to be "blamed" for the pipes' behavior; and predict whether a pipe will fail or not. Studying the results of the DAC method application at the case study networks, a SWOT analysis is attempted in order to find out whether and under which presuppositions DAC can be successfully applied to water pipe networks. (C) 2009 Elsevier B.V. All rights reserved.