Leak Localization in Operational Water Distribution Networks Using a Cross Correlation–Based Approach

Water distribution networks (WDNs) buried underground constitute a significant part of the urban infrastructure. Leaks in WDNs cause substantial water loss; hence it is critical to quickly detect, localize, and fix leaks. Acoustic sensors combined with cross-correlation signal processing methods are widely used to detect and locate leaks in WDNs; however, most studies have focused on simulated leaks in lab-scale test beds or in straight pipe segments where a line of sight exists between the leak and the measurement locations. This research presents leak localization results based on acoustic (hydrophone) data from a full-scale WDN in California, USA, that encompasses multiple bends and T-junctions. This system presents practical challenges in implementing conventional correlation techniques due to non-line-of-sight conditions. The problem of localizing a leak in this situation is solved using the cross-correlation-based maximum likelihood estimation method, which addresses non-line-of-sight and multipath challenges in this application. Additionally, an optimization-based framework is proposed to correct errors that occur in the time delay of arrivals resulting from desynchronized sensor clocks. The proposed method shows promising results with predicting a localization estimate with an approximate error in the range of 0.6%–3.06% of the total pipeline length surveyed.