Proposing a model for pipe failure estimation in water supply network

Abstract

    In urban area, most of water supply pipes are located under roads or pavements. Therefore when the pipe failures occur, a significant amount of water lost. At the same time, contaminants from outside are likely to enter the pipes, causing widespread water pollution in the water supply network, making possible risks of massive contamination for water consumers. In order to minimize water loss and to ensure water consumers' health, it is necessary to identify the probability of pipe failures in the water supply network, so that pipes maintenance and replacement planning is enabled to make in time. This paper analyzes quantifiably the factors causing pipe failures in the water supply network and proposes the use of decision tree model to estimate the probability of pipe failures. The proposed model is validated in an actual statistics and evaluated by AUC indicator.