具終站批次機台之分散彈性零工式排程

Abstract

本研究探討製造系統中包含終站批次機台的分散彈性零工式排程(distributed and flexible job-shop, DFJS)問題，研究目標為最小化總延遲時間。在此情境中，每個工件都必須先經過DFJS系統加工，最後再進行批次機台加工，批次機台能夠同時加工多個工件，但加工時間相對來的比較長，而為了有效的批次決策與機台利用，到達批次機台的工件會產生即使機台可以利用也必須等待後續工件的狀態，因此批次決策會是非常重要的決策之一，本研究將此排程問題分成兩個階段，首先會先使用基因演算法(GA)求解DFJS問題，接著利用該資訊搭配本研究所提出的批次啟發式規則來求出最終指派結果。 本研究整和過去三位學者求解DFJS問題所提出的三種演算法，加入批次決策將該三種演算法稱為GA_OP_B、GA_JS_B與GA_IGA_B，實驗結果顯示，當終站批次機台的利用率低的時候，GA_OP_B演算法的解品質會優於另外兩種演算法；當終站批次機台的利用率越高的時候，三種演算法的解品質表現相近，彼此無顯著差異，而無論是高或低機台利用率，運算時間皆以GA_OP_B演算法最為快速。

This research investigates a scheduling problem in a manufacturing system which is a distributed and flexible job-shop (DFJS) with a terminal batch workstation problem. The scheduling objective is to minimize total tardiness. In the context, each job has to go through the DFJS system and finally proceed to the batch workstation. A batch machine can process multiple jobs simultaneously; arrival jobs even facing an available batch machine may need to wait for forming an effective batch for scheduling. Batching decision is therefore very important. The scheduling problems are solved in two stages. First, we use a genetic algorithm (GA) to solve the DFJS scheduling problem. Then, based on the obtained DFJS scheduling results, we develop a batching algorithm for scheduling the batch workstation. Integrating three prior GAs with the proposed batching algorithm, we develop three algorithms called (GA_OP_B, GA_JS_B, and GA_IGA_B). Experiments reveal that GA_OP_B outperforms the other two algorithms when the utilization of the batch workstation is low (less than 80%). Yet, in the case of high utilization, the three algorithms perform almost equally well.