Active and passive acoustics to detect and localize leaks in water distribution networks

Problem Statement

Urban water distribution networks lose an estimated 20–30% of treated water before it reaches consumers—representing not only a major loss of a critical resource, but also the wasted energy and cost associated with its treatment and delivery. A significant portion of this loss occurs through undetected, underground leaks that can persist for weeks or even months without surfacing, especially within aging infrastructure. There is an urgent need for reliable, scalable, and real-time technologies that can detect and precisely localize leaks across complex pipe networks.

Approach

Our approach integrates acoustic sensing, advanced signal processing, and machine learning to detect and localize leaks across complex water distribution networks. We deploy both passive hydrophone-based sensors that capture naturally generated leak noise and active acoustic excitation for scenarios where passive signals are weak or masked. The recorded acoustic data is processed using Generalized Cross-Correlation and a Maximum Likelihood Estimation framework that leverages Time-Difference-of-Arrival to accurately infer leak location. To address the complexity of looped and branched networks, we introduced an Interior Points–based graph search method that efficiently pinpoints the most probable leak-bearing locations. These physics-guided methods are complemented by data-driven techniques, including time–frequency analysis, statistical detection theory, and machine learning models to improve leak detectability and robustness under field conditions.

Ongoing Work:

We are currently developing optimal sensor placement (OSP) strategies that combine acoustic attenuation modeling, graph theory, and optimization to identify the most effective locations for hydrant-based sensors, reducing deployment cost while maximizing leak detection and localization performance. Future efforts aim to integrate these capabilities with digital twins and utility SCADA systems for continuous, system-level monitoring.

 

Key Results and Visuals:

Figure: (a) A water distribution network (WDN) near Los Angeles instrumented with a leak detection device consisting of a hydrophone and a pressure sensor, (b) An active transmitter mounted to a plate for installation in a WDN, (c) Flow testing at a sampling station for leak simulation, (d) Leak location estimate results obtained using the developed localization framework (Agrawal and Narasimhan, 2005)

 

Relevant Publications

  • Kafle, M. D., Fong, S., & Narasimhan, S. (2022). Active acoustic leak detection and localization in a plastic pipe using time-delay estimation, Applied Acoustics, 187, 108482.
  • Cody, R. & Narasimhan, S. (2020). A field implementation of linear prediction for leakmonitoring in water distribution networks, Advanced Engineering Informatics, 45, 101103.
  • Agrawal, P. et al. (2023). Maximum Likelihood Estimation to Localize Leaks in Water Distribution Networks, ASCE Journal of Pipeline Systems Engineering and Practice, 14(4), 04023038.
  • Agrawal, P., & Narasimhan, S. (2025). Leak Localization in Operational Water Distribution Networks Using a Cross Correlation–Based Approach. Journal of Pipeline Systems Engineering and Practice, 16(4), 04025050.

 

 

Students

Pranav Agrawal