Machine learning model and strategy for fast and accurate detection of leaks in water supply network

Abstract The water supply network (WSN) is subjected to leaks that compromise its service the communities, which, however, challenging identify with conventional approaches before consequences surface. This study developed Machine Learning (ML) models detect in WDN. Water pressure data under leaking versus non-leaking conditions were generated holistic WSN simulation code EPANET considering factors such as fluctuating user demands, noise, and extent of leaks, etc. results indicate Artificial Neural Network (ANN), a supervised ML model, can accurately classify conditions; it, requires balanced dataset both conditions, which difficult for real mostly operate normal condition. Autoencoder neural (AE), an unsupervised further leak unbalanced data. show AE model achieved high accuracy when occur pipes inside sensor monitoring area, while compromised otherwise. observation will provide guidelines deploy sensors cover desired area. A novel strategy proposed based on multiple independent detection attempts increase reliability by found significantly reduce probability false alarm. trained tested testbed promising results. be readily deployed in-service WSNs using obtained internet-of-things (IoTs) technologies smart meters.