Collaborative Optimization of Semiconductor Manufacturing Production and Opportunistic Maintenance Scheduling based on QMIX

Minimizing production delays due to equipment maintenance is crucial in the semiconductor industry, driven by high costs and competitive pressures. This paper introduces opportunistic maintenance (OM) to optimize equipment utilization and reliability in semiconductor manufacturing. OM strategically schedules maintenance actions based on buffer utilization and predicted failure times to minimize operational disruptions and maintenance costs. Recognizing the limitations of traditional mathematical programming in handling the complexities of real-world production systems, this paper presents an integrated decision-making framework tailored for intelligent semiconductor workshops. Central to this approach is the QMIX algorithm operating under the CTDE (Centralized Training with Decentralized Execution) paradigm. This enables joint training of policies while supporting independent decision-making during execution. The framework leverages independent policies assigned to each machine group, coordinated through a decision collaboration module to synchronize outputs with real-time workshop states, thereby optimizing production and maintenance scheduling. This methodology enhances performance in dynamic environments with limited local observations, thereby advancing operational efficiency in semiconductor manufacturing.