Pipe bursts are the most common failures in water distribution networks (WDNs). Numerous methods have been proposed to detect and locate failures based on the WDN measurements of pipe flow and pressure. The authors have developed a suite of tools using advanced metering infrastructure (AMI) data. The simplest is comparing network inflows and withdrawals in a mass balance approach to test for anomalies. If the AMI meters are supplemented with pressure gages, optimization or deep learning tools linked with hydraulic simulation can compare observed and computed pressures to identify and locate failures. The mass balance approach is better in detecting leaks during low demand periods when leakage rates are highest in magnitude and relative to the network demand. On the other hand, the pressure driven method is more effective when demands are high and head losses are greater. The latter also can guide burst location. To exploit the benefits of each approach, this study develops a comprehensive WDN burst detection tool by combining the mass balance and pressure driven approaches in a single algorithm. To evaluate the integrated model’s performance, the proposed model is tested for several WDNs that have different characteristics such as size and topology. Realistic sized bursts (e.g., 3 L/s for 150 mm pipes) are used to assess detection and localization performance using standard metrics (detection probability, rate of false alarms, and average shortest path distance estimation. Integrating the two approaches resulted in faster detection times compared to applying the methods independently.
