Condition-based maintenance for multi-component systems: A scalable optimization model with two thresholds

Many original equipment manufacturers (OEMs) provide customized after-sales service contracts tailored to their customers' specific requirements. While these customized offerings may increase customer satisfaction and loyalty, they are likely more expensive, and that in times that OEMs are striving to reduce maintenance costs and ease the workload of their service engineers. To address this challenge, we introduce a quantitative methodology for shaping maintenance policies that minimize overall maintenance costs for systems comprising multiple heterogeneous components over a finite lifespan. Our proposed two-threshold condition-based maintenance policy incorporates scheduled visits and semi-urgent interventions, using component-level control thresholds to preventively trigger component replacements. Scheduled visits provide opportunities for grouping component replacements, capitalizing on positive economic dependencies. Our optimization model and solution methodology are tailored to systems consisting of numerous components. We assume a fixed time interval between consecutive scheduled visits, allowing us to employ a decomposition approach that ensures model scalability. The optimal policy for each component is determined by formulating a single-unit replacement model as a finite horizon Markov Decision Problem solved via dynamic programming. Simultaneously, at system level, we optimize the maintenance interval for the entire system using an iterative approach. The results of our case study highlight that complementing scheduled visits with semi-urgent interventions- both triggering preventive replacements- leads to a reduction in the volume of corrective maintenance. From a practical standpoint, our findings offer valuable insights for OEMs seeking to enhance their after-sales service contracts while concurrently reducing maintenance-related costs. Specifically, our model is particularly valuable for OEMs that service systems with high shutdown costs, rely heavily on customer satisfaction, and favor more attractive policies for their service engineers (i.e., reducing the number of emergency corrective interventions).