Improving Resilience of Water Distribution System through Burst Detection

A burst in a water distribution system (WDS) results from the pipe rupture resulting in loss of water and disruptions to customer service. Rapid burst detection increases resilience by reducing the time to identify that a break has occurred to isolate the failed pipes and to recover service. This study compares three univariate and three multivariate Statistical Process Control (SPC) methods with respect to their detectability. The univariate SPC methods are the Western Electric Company (WEC) rules, the Cumulative Sum method (CUSUM), and the Exponentially Weighted Moving Average (EWMA) method, while three multivariate methods are Hotelling T2 method and the multivariate versions of CUSUM and EWMA. A method's detectability is determined by detection efficiency and effectiveness. To test the methods, nodal pressures and pipe flow rates were generated from a real hydraulic model and detectability is computed for flow and pressure meters in alternative configurations.