Real-time monitoring of water states in large-diameter aqueducts - learning from distributed acoustic sensing signals.

Large-diameter gravity aqueducts are essential for water supply systems but face performance and safety risks from complex flow conditions. Effective flow-state monitoring is critical for hydraulic performance and infrastructure safety. However, conventional monitoring techniques like closed-circuit television (CCTV) inspection and ultrasonic sensing have limited real-time accuracy in distinguishing flow states. Here we show a real-time, distributed flow monitoring framework based on distributed acoustic sensing (DAS). A hierarchical clustering model, called DAS-Hydro HierarchyNet, was developed to analyze low-frequency acoustic signals and classify water flow states using a multi-level approach. The framework enables continuous flow monitoring along large aqueducts, overcoming point-based measurement limits. A 6 km case study in the Pearl River Delta demonstrates this approach's feasibility and effectiveness. The results confirm that DAS combined with advanced AI classification enables accurate flow-state monitoring, water location detection, and flow velocity estimation, offering a scalable, intelligent solution for large-scale transmission monitoring.