Author(s): Qiang Xu; Qiuwen Chen; Jinfeng Ma
Keywords: Pipe failure model; Non-homogeneous poisson process; Lognormal function; Exponential function
Abstract: Pipe failure often occurs in water distribution networks (WDN) and results in large water loss and social-economic damage. To effectively control the water loss from the WDN and ensure its safety, much interest has been attracted to study the pipe failure behavior. Physically based models, statistical models and data mining models have been developed to discover the rules that the pipe failure follows in the literature. Poisson process is often applied to do failure analysis because of its capability to depict the stochastic characteristic of system failures as well as its feasibility to do uncertainty analysis. In this study, the non-homogeneous Poisson process (NHPP) was selected as the framework to model the pipe failures where the intensity variable was related to the pipe age by superposing the lognormal function and the exponential function. This study took part of the WDN of Beijing as the study case and collected recent 4 years’ pipe failure data to develop pipe failure model. The results showed that the model based on the proposed approach could well describe the pipe failure behavior. Finally, uncertainty analysis was performed for the pipe failure model which was very useful for the utility managers to make cost-effective pipe failure detection and pipe maintenance plans.