Identification of leakages by calibration of WDS models

Abstract

Leakage detection is critical for the proper management of water distribution systems (WDS). This paper proposes a leak detection approach based on a Bayesian calibration method. The methodology uses a newly formulated index, μ, which takes into account the variation of roughness in pipes between the calibrated models with and without leaks. Case studies, which use literature networks, are presented to demonstrate how the approach can be used in identifying pipes with losses. The approach starts with a calibration method followed by the analysis of sensitivity matrices. The approach proved to be effective in finding leaks, but the results depend crucially on the number and quality of the observed data.