Pooled maintenance resource investment for multi-site production under uncertainty

This presentation examines machine degradation and maintenance optimization using Remaining Useful Life (RUL) prognostics to guide preventive maintenance (PM). The study focuses on a multi-site production system supported by a centrally managed shared pool of backup machines and technicians that can be flexibly allocated across sites. The goal is to determine the optimal investment in these shared resources to ensure reliable production under uncertain demand. In addition, demand uncertainty is addressed through a distributionally robust optimization model that incorporates a Mean-CVaR measure, allowing adjustable risk aversion with limited distributional information. The model is reformulated as a mixed-integer linear program for tractability. Computational results show that the shared pool reduces redundant investment, produces distinct resource configurations under different maintenance policies, and maintains high reliability under uncertainty while controlling costs.