Scenario reduction-based simulation method for efficient serviceability assessment of earthquake-damaged water distribution systems ​

Abstract Water distribution systems are vulnerable to earthquake damage, highlighting the need to assess their seismic serviceability. While existing simulation-based methods have shown promise in providing accurate assessments, their applicability to large networks is limited by the prohibitive computational burden associated with sampling a sufficiently large space and the prolonger time of simulations. To bridge this gap, this paper introduces a scenario reduction-based simulation method that efficiently evaluates the serviceability of earthquake-damaged WDSs. Specifically, Monte Carlo simulation is first employed to generate a sufficient number of earthquake-damaged scenarios (e.g., 10,000), while a novel linear pressure estimation method (LPEM) is developed to approximate nodal pressures for each scenario. Subsequently, the proposed approach categorizes and selects representative scenarios based on their pressure similarities, enabling the assessment of system serviceability using a reduced set of representative scenarios. This approach significantly reduces the computational load without sacrificing estimation accuracy. The feasibility of the proposed method is evaluated using four benchmark networks, namely Grid, Modena, C-Town, and Exnet, under varying seismic intensities. The application of the LPEM yields promising results, with most errors in nodal pressure estimation below 3 m and all R 2 values exceeding 0.9. Furthermore, the proposed scenario reduction approach, utilizing only five representative scenarios, effectively estimates system serviceability with nearly all errors below 10%. Notably, increasing the number of representative scenarios to fifty further reduces most errors to below 5%, demonstrating the effectiveness of the proposed method.