- 學位論文
以基因演算法求解石英震盪器廠之平行機台排程問題
Applying Genetic Algorithm to Parallel Machines Scheduling problem for Crystal Manufacturer
摘要
中文摘要 由於電子產品的生命週期短。相對的，對於製造業者如何掌控及跟隨著整個大環境的變動。製造業的競爭力不只是降低製造成本；同時，對於縮短整個生產時間來達到如期的將貨品交給客戶及百分百的滿足客戶需求也是一個提昇競爭力的主要因素之一。所以，生產線的作業安排的重要性是可想而知。 在限制理論(TOC)中，對於排程方法也充分的指出，即以瓶頸工作站為排程規劃的對象。而本研究旨在探討在一定的資源下平行機台於生產現場之排程研究，其目標為求得工作安排之下達成總完成時間最小化。 本研究中將以石英振盪器製造業為研究對象，提出大家所熟知的派工法(EDD、SPT、LPT)進行相互的搭配，且實地的對平行機台進行派工。同時也運用基因演算法(GA)來解此排程上的問題，以相互比較不同的方法之間的排程績效。 本次研究是以總流程時間最小化為目標。由統計分析及改善率來看；當工單筆數在33筆時，其GA演算法與啟發法相較之下其效果有顯著的差異。從改善率所得出的結果為GA演算法在工單數為33筆時，改善率可以達到38.63%較其它的啟發法佳；但工單數增加(大於33筆)後，啟發法與GA演算法在ANOVA檢定後，對總流程時間並沒有顯著的差異。亦即，當工單筆數增加之後的總流程時間的是相同。 由於GA演算法在計算上往往所需要的時間較久，對於生產線在安排派工時較難以滿足時效性及緊迫性；相反的，啟發法在時效性確是優於GA演算法，而且對T公司之派工法也有不錯的效果。
並列摘要
ABSTRACT Owing to life cycle is shortage for 3C’s product. So that manufacturing will be how to control and monitor marketing. The manufacturing must reduction of cost and that reduction operation time is rise up competitiveness. Therefore operation in the production line arrangement will be very important. Bottleneck process will be focused schedule planning in the Theory Of Constrains (TOC). The target is minimization for total finished time by parallel machine. Scheduling method will be used to EDD, SPT and LPT which is clear and understood in this researched. At the same time, this subjection is focus on electrical component for XTAL manufacturing. At last, this thesis use to Genetic Algorithms it can compare to different for performance. Form statistic analysis and improved ratio, it has variance of the result to be compare the Genetic Algorithms and Heuristics when there are 33 orders. Therefore the improved ratio has 38% to use GA to compare others. But this order will be added great than 33 orders it is not different the average value by total finished time. Because GA calculates to spend time so long to compare with others, it is not satisfied for GA to be efficiency and efficacy. By the way, the thesis develops that will be the same with GA and it is short time than GA.