A Multi-Objective Optimization Framework for Capital Allocation in Public Infrastructure Systems

Introduction: Public infrastructure agencies must allocate limited capital across heterogeneous assets while balancing lifecycle cost, safety and reliability risk, service criticality, and governance constraints. Conventional composite scoring and single-objective ranking can obscure trade-offs and embed implicit value judgments. Methods: We formulate capital allocation as a constrained multi-objective optimization problem that maximizes portfolio-level expected risk reduction and service criticality subject to a budget cap and a minimum compliance requirement. Pareto-optimal portfolios are generated to expose trade-offs, and a transparent, Pareto-first balanced selection rule is used to recommend one portfolio without obscuring alternatives. A reproducible synthetic portfolio (N=16) is specified via distributions and random seed; a composite-score baseline is implemented for comparison; sensitivity and robustness analyses vary budget, compliance, and key input perturbations. A scalability demonstration (N=500) uses an NSGA-II-style heuristic with random-key encoding and constraint repair to generate an approximate frontier. Results: For N=16, Pareto analysis yields multiple defensible portfolios that differ materially by governance preference. A balanced portfolio is selected by a mathematically defined minimum utopia-point distance rule on normalized objectives, and an alternative knee-point rule is reported for sensitivity. Compared with composite scoring under the same constraints, Pareto-based reporting exposes dominated (inefficient) choices and improves transparency. For N=500, the heuristic generates an approximate frontier in seconds, illustrating feasibility at realistic portfolio scale. Conclusions: The framework shifts capital planning from static prioritization to explainable optimization aligned with asset management guidance. It supports auditable decision-making under real governance constraints and is implementable within enterprise asset management and decision-support platforms.