Multicriteria Decision-Making Framework for Proactive Maintenance of Water Distribution Pipelines in South Africa

South African municipalities face significant challenges in implementing proactive infrastructure maintenance due to financial constraints, limited technical capacity, and aging water infrastructure, often leading to reactive maintenance approaches. To address this challenge, this study develops an integrated decision-making framework that combines machine learning survival analysis (MLSA) and multicriteria decision-making (MCDM) techniques to optimize pipeline maintenance prioritization. The MLSA component predicts pipeline failure probabilities using historical asset management data, while the MCDM approach ranks maintenance priorities based on pipeline condition, performance, and criticality. Specifically, this study integrates extreme gradient boosting survival embeddings (XGBSEKaplanTree) with the analytic hierarchy process (AHP) and the technique for order preferences by similarity to ideal solutions to improve predictive accuracy and decision transparency. A survey of 173 industry experts identified pipeline performance as the most critical maintenance criterion, with pipe pressure emerging as the most significant subcriterion. The framework's reliability was validated through multiple statistical methods, including the AHP consistency ratio, Cronbach's alpha, McDonald's omega, and integrated Brier score, while validity was confirmed using Spearman's rho correlation and expert review. By integrating data-driven predictive modeling with structured decision-making, this framework enables municipalities to transition from reactive, condition-, and cost-based maintenance to a proactive, data-driven approach. This shift can reduce water losses, improve auditability, optimize resource allocation, and enhance water security, contributing to advancing Sustainable Development Goal 6.