Detecting Bursts in Water Distribution System via Penalized Functional Decomposition

Detecting bursts in water distribution systems, as anomalies from normal daily usage, is of critical importance for urban infrastructure maintenance. Existing methods based on conventional statistical process control fall short of accurate estimations of burst magnitude and starting time. This research combines a functional basis expansion of water flow data stream and a penalized decomposition to parameterize and estimate the normal water usage profile and detect spars burst with a comparatively small magnitude. The effectiveness of the proposed method is demonstrated by a high-fidelity simulation case study.