摘要
The cell process in a thin film transistor-liquid crystal display (TFT-LCD) manufacturing is characterized as a complicated multi-stage and parallel-machine production system. This system has multiple products, sequence-dependent setup times, multi-step production, multi-objective, and matching constraints between color filter lines and array lines. This complex and capital-intensive production environment needs an efficient capacity planning decision that determines the order release schedule, estimated machine start and finished time of each order, and machine allocations to improve the performance of the entire production process. This research first proposes a finite capacity planning system (FCPS) to cope with the capacity planning problem faced by the TFT-LCD cell process based on the pull philosophy and assumption of finite capacity. A modified multi-objective genetic algorithm (MOGA) embedded in FCPS is also developed to efficiently generate the order release decision, which simultaneously minimizes the machine workload balance, makespan, and lateness. The preliminary experiment identifies the fine MOGA tuning parameters using a two-level factorial experimental design. Finally, the full factorial experiment evaluates the significance and robustness of the proposed modified MOGA algorithm compared with other competitive algorithms under various circumstances.