Resilient facility location optimization under failure scenarios using NSGA-III: A multi-objective approach for enhanced system robustness

The Uncapacitated Facility Location Problem (UFLP) is a classical optimization problem that seeks to determine the optimal placement of facilities to minimize the total distance between facilities and demand points. In the face of increasing disruptions, such as natural disasters and other operational failures, the need for robust solutions that can maintain high levels of performance before and after disruptions has become more pressing. This paper presents an improved approach to the UFLP using the Non-dominated Sorting Genetic Algorithm III (NSGA-III). Compared to its predecessor, NSGA-II, NSGA-III exhibits significant computational improvements, including a reduction in computation time and an enhanced capability to explore the Pareto front efficiently. The proposed method aims to provide decision makers with robust solutions that minimize the total weighted distance under normal operating conditions and maintain resilience in the event of facility failures. NSGAIII demonstrates superior computational performance by requiring less time to obtain high-quality solutions and achieving a more diverse set of Pareto-optimal solutions. Computational experiments demonstrate a 25% improvement in Pareto front diversity and a 15% reduction in computational time compared to NSGA-II. Unlike previous methods requiring probabilistic failure data or additional resource allocation for protection, the proposed approach directly incorporates resilience considerations into the optimization process. The method allows decision-makers to identify robust facility placements that minimize the total weighted distance during normal operations and maintain performance under interruptions. Practical insights for resilient infrastructure planning are provided, supported by benchmark dataset analyses. The results underscore the efficacy of NSGAIII in achieving computational efficiency and solution diversity, advancing the state-of-the-art in multi-objective optimization for complex FLPs.