本研究討論在彈性車間環境中考慮平行批次處理(parallel batching)及預防性維護(preventive maintenance)的問題。 在半導體製造環境中，材料的更換依據其服務的工件批次數量或使用壽命，而在產能滿載的情況下材料經常依據服務的工件批次數量達到上限進行更換。並以極小化“stage-out”的數目及總完成時間(Makespan)為目標， 以同時滿足短期及長期排程規劃。 極大化“stage-outs”為半導體製造的每日績效之一， 我們將其轉換為極小化延遲工作總件數來進行優化。 而極小化工作總完成時間則視為長期目 標。 依據此問題特性，我們透過在分離圖(conjunctive graph)上呈現工作中各層級結束時連接終止節點(sink node)以計算其完成時間，並在各層間以分離弧(conjunctive arc)連接以表示各層級間的順序中。 這種方法使我們能夠基於多個終止節點(sink node)的關鍵路徑(critical path)來定義鄰域結構(neighborhood structure)， 並引入重疊值(overlapping value)來優化搜索過程。 我們定義一種以兩個目標為基礎的節省法(saving method)，用於評估移動(move)操作對於兩個目標優化效果。通過結合於多條關鍵路徑與兩個目標中切換的搜索策略(search strategy)中，期望優化局部搜索過程。;This study discusses the consideration of incorporating parallel batching andpreventative maintenance in a flexible job shop scheduling problem. In the semiconductor manufacturing environment, materials are replaced according to the number of batches they serve or their service life, and at full capacity, materials are often replaced when the number of batches served reaches the upper limit. The goal is to minimize the number of stage-outs and the makespan to satisfy both short and long-term schedules. One of the daily performance metrics in semiconductor manufacturing is to maximize stage-outs, which we optimize by transforming it into minimizing the total number of tardy stage-outs. While minimizing makespan can be regarded as a long-term goal. Based on the characteristics of this problem, we propose a method for calculating completion times by representing the sink nodes at the end of each level in a conjunctive graph and connecting them with conjunctive arcs to indicate the order between layers. This approach allows us to define a neighborhood structure based on multiple critical paths to sink nodes and introduces overlapping values to optimize the search process. We define a saving method based on two objectives to evaluate the optimization ability of move. By combining this method with a search strategy that switches between multiple critical paths and two objectives, we aim to optimize the local search process.
