Cascading-aware sensor placement for leakage detection in water distribution networks: A graph signal processing framework with risk propagation

Optimal pressure sensor placement is essential for leakage detection in water distribution networks (WDNs). However, existing studies often fail to capture the coupled effects of network structure, hydraulic operation, and leakage-related factors in real systems, leading to limited engineering applicability. To address this issue, this study proposes a cascading-aware sensor placement method (CASP). From a graph signal processing (GSP) perspective, CASP establishes an optimization framework that links structural attributes, hydraulic states, and leakage-related factors by integrating cascading effects with risk propagation. Leakage-triggered failures are conceptualized as cascading effects and represented through a risk propagation model that quantifies a composite risk for each pipe across the WDN, resulting in a risk-weighted adjacency matrix. A graph Laplacian–based spectral low-pass filtering approach is then employed to simulate multi-level risk transmission, with the optimal propagation depth determined through clustering stability analysis. Finally, the d -optimality criterion is adopted to optimize sensor placement, ensuring maximal information representativeness and leakage detection performance in the risk propagation space. The proposed method is validated using a real-world WDN. The results demonstrate that CASP achieves high leakage identification accuracy and low localization error under both random and realistic leakage scenarios, showing strong robustness under complex operating conditions.