Integrated Predictive‑Maintenance and MPC Scheduling: Achieving High Availability in Smart Manufacturing

Open-shop scheduling in smart manufacturing requires both rapid rescheduling and high equipment availability. However, most model predictive control (MPC) schedulers assume perfect machine reliability. This leaves predictive maintenance (PdM) as an external problem, separate from the dispatcher’s control. This paper introduces a degradation-aware MPC(D-MPC) for unrestricted open shops. Our approach makes Remaining Useful Life (RUL)-triggered maintenance an internal endogenous part of the scheduling decision.We use a two-level architecture. The upper level converts RUL forecasts into feasible maintenance windows. The lower level then solves a mixed-integer linear program (MILP) to co-optimize job scheduling and maintenance timing. The controller is event-driven and optimizes over a fixed-size window. This design bounds the problem size at each step, enabling real-time deployment on standard hardware. We performed Monte-Carlo experiments on open shops up to 50 × 50 with stochastic disturbances. Compared to reactive and periodic baselines, our D-MPC framework consistently improves both productivity and availability. It reduces makespan by 7–9% and downtime by 54–60%, and lowers tail risk ( CVaR 95 ). Our results show that RUL-aware, opportunistic maintenance allows the scheduler to eliminate breakdowns with minimal planned stops. This offers a practical path to unify production scheduling and asset health in smart factories.