A two-phase approach for leak detection and localization in water distribution systems using wavelet decomposition and machine learning

Water is a crucial resource for all forms of life, yet it is becoming increasingly scarce. A significant portion of water loss in urban and industrial areas is attributed to leaks. Addressing this issue is critical for enhancing efficiency, sustainability, and resource conservation. This paper presents a novel two-phase approach for leak detection and localization in water distribution systems using wavelet decomposition and machine learning for depth analysis of pressure signals. The first phase, Leak Detection, utilizes wavelet analysis to extract significant features from the daily pressure signal data. These features are then inputted into a Random Forest classifier, achieving a classification accuracy of 99% for distinguishing between "Leak"and "No Leak"scenarios. Following the detection, the Leak Localization phase aims to pinpoint the leak's location using strategically placed sensors within the system. To facilitate understanding and application of our methodology, we have developed a user-friendly, web-based application designed for the detection and localization of water leaks on any given day. Extensive testing in a WDS named "L-Town"has validated our system's ability to accurately identify leaks. The combination of wavelet-based signal analysis and the Random Forest algorithm forms an effective framework for advanced leak detection in water distribution systems. This approach holds great promise for future research and practical implementations in water management.