Abstract
Machine-process grouping refers to what group of process should be processed in what group of machines. In highly reentrant high-product-variety production systems such as semiconductor manufacturing, optimization for machine-process grouping is very complex. This research proposed a grouping algorithm considering group overlapping, machine set-up and waiting times minimization, and machine availability. Mathematical models to estimate overall set-up time, waiting time under various machine availabilities and possible overlapping of machines grouping were established. Particle Swarm Optimization (PSO) technique was used to solve the minimization of a given period of production requirements. Flexsim simulation models were used to evaluate the performance of the grouping solutions generated by the PSO. The four performances indices observed are Throughput, WIP, Cycle Time and delivery rate. Real-world production data from a fab of a major Taiwanese Foundry were used to test the results. The algorithm was able to identify better grouping than non-overlapping and the company’s original grouping arrangement with significant improvements. The results indicated that regardless of the arrival rates and availability rates, allowing grouping overlap showed better performances. Contributions of this research include: 1. Establishing a dynamically adjustable algorithm to solve a complex machine-process grouping problem in highly re-entrant and high-product-variety fabrications allowing overlapping and availability variations. 2. Identifying a much improved grouping arrangement for the fab.