A graph based analysis of leak localization in urban water networks

A graph based analysis is proposed to improve leakage management in water distribution networks. Starting from the model of the network, leakage scenarios, created through hydraulic simulation (EPANET), are considered as nodes in a graph, whose edges are weighted by the similarity between each pair of nodes (scenarios), in terms of pressure and flow variation due to the leak. The graph is then analyzed in the eigenspace of its Normalized Laplacian matrix and specifically into the eigensubspace spanned by the most relevant eigenvectors, allowing Spectral Clustering, which is more effective than traditional techniques but with much higher computational requirements, to be applied also to large scale problems. The results obtained in the eigenspace are eventually mapped back into the physical space where the capability of leakage localization may be further improved through the fusion with leak severity estimation. © 2013 The Authors.