Automated Detection of Pipe Bursts and other Events in Water Distribution Systems

Abstract

This paper presents a new methodology for the automated near real-time detection of pipe bursts and other events which induce similar abnormal pressure/flow variations (e.g., unauthorised consumptions) at the District Metered Area (DMA) level. The new methodology makes synergistic use of several self-learning Artificial Intelligence (AI) techniques and statistical data analysis tools including wavelets for de-noising of the recorded pressure/flow signals, Artificial Neural Networks (ANNs) for the short-term forecasting of pressure/flow signal values, Statistical Process Control (SPC) techniques for short and long term analysis of the pipe burst/other event-induced pressure/flow variations, and Bayesian Inference Systems (BISs) for inferring the probability of a pipe burst/other event occurrence and raising corresponding detection alarms. The methodology presented here is tested and verified on a case study involving several DMAs in the United Kingdom (UK) with both real-life pipe burst/other events and engineered (i.e., simulated by opening fire hydrants) pipe burst events. The results obtained illustrate that it can successfully identify these events in a fast and reliable manner with a low false alarm rate.