Two-Stage Transient-Based Leak Detection in Water Supply Pipe Networks

Transient-based leak detection methods have been progressively developed and tested in pilot water supply pipe systems. In this context, so-called time-domain inverse transient analysis (ITA) is one of the most popular methods for detecting pipe leaks by minimizing the discrepancies between the measured and modeled transient signals. However, it suffers from a drawback when applied to large pipe networks as the inverse analysis requires optimization wherein both leak sizes and locations are estimated simultaneously, rendering the computational cost high and parameter space multidimensional. To address this problem, we propose a two-stage localization algorithm that tackles the estimation of leak locations and leak sizes separately using a fast and robust scheme tailored for pipe networks. The rationality of the proposed method has been theoretically evidenced by coupling the transfer matrix and maximum likelihood estimation, which is then verified through experimental and numerical applications to different pipe networks. The results have demonstrated remarkable efficiency of the proposed technique in comparison to the conventional ITA method. Additionally, the proposed method exhibited favorable accuracy when applied to different case studies with differing noise levels and leak sizes.