Risk Assessment Methods for Urban Water Distribution Networks: A State-of-the-Art Review of Indicator, Statistical, and Machine Learning Approaches

Abstract

With the ongoing advancement of urbanization, urban water distribution networks (WDNs) are increasingly challenged by asset aging, corrosion, and pipe bursts, which collectively threaten the safe and reliable operation of urban systems. Consequently, rigorous risk assessment of urban WDNs has become essential. It enables the identification of high-risk segments and hotspots, and provides an evidence base for maintenance prioritization and network optimization. In this study, research progress on risk assessment and failure analysis of urban WDNs over the past 25 years was systematically reviewed. Mainstream approaches, including indicator-based scoring, statistical modeling, and machine learning (ML), were emphasized, and their fundamental principles, methodological characteristics, applicable contexts, and reported practical performance were comprehensively summarized. Indicator-based scoring methods are valued for their transparent structure and ease of implementation, and have been widely adopted in engineering applications. Statistical methods leverage historical records to develop failure models with explicit probabilistic interpretability. ML methods can capture complex nonlinear relationships and show strong predictive capability in data-rich settings. Nevertheless, prevailing approaches continue to face persistent limitations, including incomplete and heterogeneous data, constrained model transferability across systems, and substantial computational demands. Building on these findings, this study highlights future research priorities in enhancing multidimensional models, developing interoperable data-sharing platforms, and conducting life-cycle-oriented risk assessment, with the goal of supporting intelligent and sustainable management of urban WDNs.