Multiobjective Risk-Based Renewal Prioritization of Water Pipeline Systems Using Genetic Algorithm

Abstract

Water utilities must choose a small set of mains to renew each year under tight budgets and delivery limits, yet practice relies on heuristic rankings and ad hoc adjustments, with little evidence on how to design multiobjective, constraint-aware portfolio models. We have developed a project-level formulation that treats likelihood of failure (LOF), consequence of failure (COF), water-loss propensity, and equity, legacy-material and other indicators as exogenous inputs, aggregates high-risk segments into constructible corridor projects, and uses a multiobjective genetic algorithm (GA) to search portfolios that maximize risk reduction, minimize budget deviation, and improve equity and disruption/packaging surrogates in three anonymized US drinking-water utilities. Model behavior is examined through an Evaluation–Verification–Validation program: four verification experiments against utility baselines plus a scenario-based validation where experts review algorithmic recommendations. Across the three systems, GA portfolios match or exceed incumbent and rank-based programs in risk-reduction-per-dollar while satisfying the same budget and delivery limits, shift spending toward high-equity and legacy-burdened areas with only modest loss of total risk capture, and remain stable across random seeds and objective-weight perturbations. Experts agree with scenario-level decisions in most cases, and remaining disagreements cluster around program boundaries and explicit policy rules, showing that the framework is a transparent, evidence-based decision-support tool for renewal planning.