Multi-objective <i>Q</i> -learning approach of operation mode control and predictive maintenance for phased mission manufacturing systems

Phased Mission Manufacturing Systems (PMMSs) experience multiple operational phases throughout their lifecycle, each characterized by distinct mission profiles and production demands. In each phase, manufacturers typically prioritize different objectives when making Operation and Maintenance (O&M) decisions. However, existing O&M optimization methods generally focus on a single objective, such as minimizing makespan or maximizing remaining useful life, which does not fully address the varying demands of different phases. To bridge this gap, we propose a multi-objective optimization method for equipment Operation Mode Control (OMC) and Predictive Maintenance (PdM) in PMMSs that adapts to shifts in mission phases. Our main contributions include: 1) the development of an operation status index system for equipment within PMMSs that integrates optimization objectives of productivity, reliability, workpiece quality, and energy consumption, 2) the use of dynamic weight coefficients to identify and prioritize key objective functions in different phases and the application of the TOPSIS method to formulate the optimization objective based on these weighted indices, and 3) the implementation of a Q-Learning algorithm for mode control and PdM optimization. The effectiveness of the proposed method has been validated through an industrial case study involving a nuclear fuel rod shielding component manufacturing system.