Unrelated Parallel-Machine Scheduling Problem with Time-Changing Effects and Dynamic Job Arrivals

With machine deterioration, time-changing effects are observed in the ion implantation work center of wafer fabrication. Furthermore, jobs arrive in a dynamic pattern. A mixed integer stochastic programming model is formulated to address the unrelated parallel-machine scheduling problem at the work center. The objective is to minimize the average flow time of wafers. Time-changing effects and dynamic job arrivals are considered simultaneously. A genetic algorithm with a reinforcement learning procedure (GA-RL) is developed to solve real-size problems. The reinforcement learning procedure aims to determine the optimal confidence levels during the search for the best solution. Computational analyses demonstrate the efficiency of the GA-RL. Sensitivity analyses provide valuable managerial insights into actual production processes.