A Lot Release Scheduling Problem for Semiconductor Cluster Tools with Parallel Chambers

This paper addressed a lot release scheduling problem for semiconductor cluster tools, aiming to minimize makespan and lot opening time by considering job splitting, dedicated machines, and resource constraints. In this scheduling problem, each chamber within the cluster tool can be regarded as a parallel machine, and a lot can be split and processed across its eligible chambers. In addition, the number of lot types that can be processed simultaneously within a cluster tool is limited by the number of loadports. This study proposes a Mixed-Integer Linear Programming (MILP) model to minimize makespan and lot opening time. To reflect the priority of the two objective functions in the proposed model, a hierarchical optimization approach was used to minimize the makespan first and then sequentially minimize the opening time of lots within the obtained solution space. The proposed MILP model can obtain optimal solutions for small-sized problems, but it can’t solve large-sized problems considered in actual industrial settings within a limited time. Therefore hybrid metaheuristic algorithms that integrate two metaheuristics with a greedy-based heuristic rule are proposed to solve large-sized problems. The proposed algorithms are compared with the MILP model for small-sized instances and with dispatching rules for large-sized instances to evaluate performance. The performance of the algorithms was compared using numerous randomly generated instances of various sizes. According to the experimental results, the Greedy-Simulated Annealing (SA)-SA algorithm showed an average gap of 5.25% in makespan and 5.00% in lot opening time from the MILP solutions in small-sized experiments, and achieved the best overall performance in large-sized experiments compared to the Greedy-SA-Tabu Search (TS) algorithm and other dispatching rules.