應用啟發式演算法與家族式派工於多部傳輸整合步進機在小批量情境下之排程問題

Abstract

半導體工廠中，傳輸整合步進機為瓶頸機台，其在小批量的情況下，會因為機台埠區的限制造成產能損失之問題，而過去文獻中已對此問題作過研究，但和現實狀況比較，其假設過於簡化，因此本研究有三項假設：第一，多部傳輸整合步進機。第二，不同的電路佈局之工件，進入傳輸整合步進機加工，當更換工件時，會產生更換光罩的設置時間。第三，使用相同光罩之工件會將之排序在一起 ，即家族式派工法(Family-based)，以減少設置時間之發生，故在求解多機之工件排程問題上，工件會面臨三項重要之決策，分別為工件族需要分配給哪一個機台(Family Assignment)，以及工件族在各別機台內的排序問題(Among Family Sequence)，工件族內各工件的順序 (Within Family Sequencing)，而本研究提出七種演算法，經由多次的實驗後，顯示GA-Tabu方法優於其他的演算法，此外，比較家族式派工法與單獨派工法於此排程規劃之績效，更證明家族式派工法則之必要性。

An in-line stepper, a bottleneck machine in a semiconductor fab, may have capacity loss in a small-lot scenario. Prior studies have examined scheduling methods to reduce such a capacity loss. Yet, their contexts of study are relatively simplified, in contrast to that of real-world cases. This research aims to examine scheduling methods in a more general context—with three distinct features. First, there are multiple in-line steppers. Second, jobs that require the same setup are grouped into a family in scheduling. Thirdly, a setup is required while switching to process different families and not required while processing jobs within the same family. Such a context leads to three scheduling decisions: job-family assignment (how to assign jobs families to in-line steppers), among-family sequencing (how to determine the sequence among families), and within-family sequencing (how to determine the job sequence within each family). Seven meta-heuristic algorithms are developed. Numerical experiments indicate that the GA-Tabu algorithm outperforms the other six ones in most test cases. Moreover, experiments also indicate that such a family-based scheduling approach in most test cases outperforms the individual-based scheduling approach, in which jobs are not necessarily grouped into families.