Optimization of preventive maintenance plans in G/G/m queueing networks and numerical study with models based on semiconductor wafer fabs

Preventive maintenance planning is important to avoid unanticipated equipment breakdowns and ensure smooth production. Previous work seeking to consider mean cycle time in preventive maintenance planning focused on individual toolsets. We develop a method to jointly determine preventive maintenance plans for a network of G/G/m queues with the objective of minimizing the system's mean cycle time. Our contribution is the integration of two methods that have been developed previously: approximations for the mean cycle time in networks of G/G/m queues and preventive maintenance plan optimization for a G/G/m toolset in isolation. Preventive maintenance events are considered as non-preemptive high priority customers in the network model. We conduct numerical experiments via simulation to assess the performance of the optimization approach. We use network models inspired by publically available semiconductor industry datasets. The study suggests that cycle time improvements may be possible via fabricator-level preventive maintenance plan optimization.