An evolutionary algorithm-based event-adaptive rolling-horizon framework for dynamic wafer probing rescheduling with flexible preventive maintenance

Wafer probing is a critical stage of semiconductor manufacturing, where effective production scheduling is essential for improving workshop throughput. In practical wafer probing workshops, dynamic disruptions with different impacts can significantly reduce schedule stability. Production scheduling and flexible preventive maintenance (FPM) decisions are also strongly coupled, further complicating decision-making. Moreover, generic evolutionary algorithms may not adequately exploit problem-specific characteristics or capture the interactions among coupled subproblems. To address these challenges, this paper investigates a dynamic wafer probing rescheduling problem with flexible preventive maintenance (DWPR-FPM) in a heterogeneous-resource environment. An enhanced evolutionary algorithm-based event-adaptive rolling-horizon framework, termed EEA-EARH, is proposed to coordinate disturbance-specific rescheduling, FPM-state evolution, and efficient optimization. Within this framework, an event-adaptive rolling-horizon (EARH) method maps disruptions to event-specific triggering times and response strategies according to their urgency and operational impact. A failure-aware FPM mechanism is further developed to model heterogeneous maintenance policies and dynamically update FPM windows. A problem-specific enhanced evolutionary algorithm (EEA) is also developed by incorporating an integrated three-vector representation, FPM-aware decoding, tailored initialization and neighborhood search into a conventional evolutionary framework. Experimental results demonstrate that EEA-EARH produces high-quality schedules under dynamic disruptions. Compared with competing methods, it reduces the makespan, normalized machine idle rate, and schedule instability by 5.30%, 10.30%, and 3.89% on average, respectively.