具等候時間限制之雙流線型工廠排程

Abstract

本研究針對雙流線型工廠排程，加入等候時間限制之考量。等候時間限制發生於加工過程，限制工件在工作站中等候加工之時間。當等候時間超出此限制，工件將產生良率問題。故具有等候時間限制之生產模式，於實際生產前，將根據客戶提供之預期訂單預排排程。為避免工件於工作站中等候過久，超出等候時間限制，控制工件投料與到站加工時間是不可避免的。然而此作法可能導致產能無充分利用，增加總完工時間並降低產出，延長可允諾給顧客之交期，降低競爭力。因此本研究提出一個基因演算法，針對雙流線型工廠之排程，同時決策加工途程、加工順序，考量等候時間限制，績效指標為兩廠總完工時間之最小化，使排程結果能滿足良率與維持產出。本研究提出之基因演算法，可解讀出跨廠與不跨廠兩種加工途程。在加工順序的解讀，採用三種單一派工法、組合派工法，以及基因演算法自然演化等不同方式進行解讀。在不同生產情境的實驗中，結果顯示，本研究之基因演算法可根據情境之不同，建議適合之加工途程決策；在加工順序的解讀上，沒有一種派工法在所有情境中，皆能有最佳之排程績效。然而，組合派工法結合多種各具優勢之單一派工法。在不同生產情境下，其排程績效相對於其它方法，能具有較佳之排程穩健性。

This research examines a dual flow-shop scheduling problem, which allows cross-shop production and is with queue time constraint (also called Q-time window). Q-time window denotes the longest waiting time that a job is allowed to have between its two consecutive operations. If the waiting time of a job is longer than its Q-time window, the yield would be adversely affected. The scheduling objective is to minimize the makespan, the completion time of the last completed job. The scheduling problem involves two decisions: route assignment (assigning jobs to shops) and job sequencing. A genetic algorithm (GA) is proposed for making the route assignment decision, which is further varied by including different dispatching algorithms. These dispatching algorithms include single heuristic rules, combined dispatching rules, and evolutionary approach. Numeric experiments for comparing these scheduling algorithms are carried out. Experiment results indicate that none of the GAs could outperform the others in all scenarios. Yet, the GA equipped with the combined dispatching rules is relatively more robust.