Practical framework for automatic pipe burst detection and location: application to a real case study

This paper presents a framework for automatic detection and localisation of pipe bursts in real water distribution systems, structured in s four main stages: data collection, time series processing, burst detection and burst location. A real-world case study is used, with five physically recreated burst supported by a dedicated data collection programme. The network’s configuration is modified by closing isolation valves to increase friction losses and pressure sensitivities to burst events, and five additional pressure sensors are installed at optimal locations. The framework successfully identifies the burst location at the street level for three of the five recreated events. The unsuccessful location of two events is due to specific modelling challenges, namely difficulties in calibrating the hydraulic model and the effect of high non-revenue water levels. Overall, the obtained results are encouraging, demonstrating that the framework is robust and comprehensive enough for real-world applications.