Multifractal Analysis for Leak Detection in Urban Water Systems with Continuous Acoustic Monitoring

Water leaks constitute a growing concern in modern cities, resulting in financial losses, public health risks, and environmental impacts. Recently, above-ground, continuously monitoring acoustic sensors, such as hydrophones, have shown promising results in detecting the distinct sound generated by leaks. The present study aims to provide insight on the acoustic trace of pipe leaks by employing multifractal analysis on acoustic signals collected in an existing water distribution system. Multifractal analysis is a technique for measuring self-similarity in signals, and thus, can detect the dynamic characteristics of a signal. Dynamic characteristics, including variability and roughness of signals based on time scale, are quantitively evaluated by multifractal parameters, and the values of these parameters are expected to change at the occurrence of a pipe leak. In this study, various field tests simulating pipe leaks were performed in a real water network, and acoustic data was collected during both leak and no-leak conditions using hydrophones attached to fire hydrants. The multifractal parameters extracted from leak signals appear to deviate from the distribution of the corresponding parameters of no-leak signals, indicating that analyzing signal self-similarity can help discern between leak and no-leak conditions. The proposed approach was tested on data collected at different locations and demonstrates promising results in the presence of variable conditions (e.g., water usage, pipe layout, material, and size). Overall, this work contributes to assessing the feasibility and limitations of continuous acoustic monitoring for leak detection in urban water systems.