Integrating machine learning and multi-criteria decision analysis for health risk management in water distribution networks.

Leakages and breaks in water distribution networks (WDNs) cause significant water losses and pose health risks due to pathogen intrusion. The Water Safety Plan (WSP), developed by the World Health Organization (WHO), provides a comprehensive framework for identifying, assessing, and controlling risks within water supply systems. This study demonstrates the application of the WSP framework through a case study of a WDN in Sweden. Pipe break probabilities were estimated using three classification models: Logistic regression, random forest, and extreme gradient boosting (XGBoost), while hydraulic and health consequences were evaluated using hydraulic modelling and Quantitative Microbial Risk Assessment (QMRA) to quantify the overall health risk. A Multi-Criteria Decision Analysis (MCDA) approach, specifically the Technique for Order of Preference by Similarity to Ideal Solution (TOPSIS) was utilized to prioritize risk mitigation strategies through breakage and leakage control measures. The proposed approach integrates predictive modelling, consequence evaluation, and decision analysis, offering a structured method for water utilities in prioritizing interventions and improving the overall safety and reliability of WDNs.