Advances in Water Pipe Failure Prediction: Methods, Challenges, and Future Directions – A Review

Maintaining the integrity of water distribution systems is crucial for public health, economic stability, and urban resilience, particularly in the context of sustainable development goals. As global infrastructure ages, pipe failures are increasing in frequency and severity, threatening water security and climate resilience. This review synthesizes the latest advancements in water pipe failure prediction from 2000 to 2025, tracing the transition from traditional statistical models to advanced machine learning, hybrid systems, and network-aware Graph Neural Networks that explicitly account for topological and temporal dependencies. We critically analyse the expansion of predictive factors—from basic physical attributes to complex environmental, operational, socio-economic, and climate-change influences—and evaluate model performance, interpretability, and applicability across diverse contexts, with particular attention to data-scarce regions in the Global South. Despite notable progress, persistent challenges remain in data quality, model transparency, uncertainty quantification, and practical implementation. The review concludes by outlining essential future directions, including explainable AI frameworks, physics-informed hybrid models, integrated decision-support systems linking predictions to life-cycle cost and renewal planning, and robust approaches for incorporating downscaled climate projections. By bridging predictive modelling with sustainable asset management, this work provides actionable insights for building climate-resilient and equitable water infrastructure worldwide.