Leak detection in water distribution networks based on graph signal processing of pressure data

Abstract

Leakages in water distribution networks (WDNs) affect the hydraulic state of the entire or a large part of the network. Statistical correlation computed among pressure sensors monitoring network nodes aids the detection and localization of such leaks. This opens the possibility to work with water network databases, where graph signal processing (GSP) tools aid in understanding changes in pressure signals due to leakages in the hydraulic system. This paper presents a methodology for time-varying pressure signals on graph structures. The core of this methodology is based on changing of pressure, due to leaks, that modifies the graph structure. Computing for each time step a new topology of the graph and applying centrality analysis based on PageRank, it is possible to identify the presence of new leaks at the water system. A confusion matrix evaluates the precision of the proposed methodology on defining where and when such leakages start and end. Seven leaks are used to validate the process, which presented 86% in accuracy terms. The results show the benefits of the method in terms of speed, computational efficiency, and precision in detecting leakages. HIGHLIGHTS A novel data analysis methodology for leak detection in water distribution networks.; Correlation of pressure data for the creation of temporal graphs.; Analysis of water networks as graphs and vertices ranking by the PageRank metric.;

Keywords

• graph signal processing
• graph theory
• leak detection
• pagerank metric
• water distribution networks