- 學位論文
半導體覆晶封裝廠多階不相關平行機台排程問題之研究
The Study of Unrelated Parallel Machine with Multiple Processors Scheduling Problem in Semi-conductor Flip Chip Assembly House
摘要
在半導體覆晶封裝生產系統中，每一階段的製程是由一台以上各自獨立並可執行相同工作的新舊機器設備所組成，如何完成各階段製程中工件之指派及各不相關機台上工件加工順序之安排，即所謂的多階不相關平行機台排程問題。此外，覆晶封裝製程是不同產品有不同製造途程的流程型生產型態，屬於容許部份製程加工時間為零的一般流程型工廠。由於一般流程型多階不相關平行機台之排程問題是屬於複雜的NP-hard問題，故本研究將採用精英政策之基因演算法（Genetic Algorithm），針對覆晶封裝製程之特性、加工順序相依的設置時間及不同製造途程的產品有不同交期等限制條件，以總延遲時間最小化為目標，進行此一問題之排程研究，並與傳統的EDD、SPT、LPT派工方式進行分析比較。研究結果顯示，本研究採用之基因演算法進行一般流程型多階不相關平行機台之排程，比案例公司現行傳統的EDD排程方式，約有12％的改善率，比其他傳統的SPT、LPT派工方式更有高達80%以上的改善率，具有良好的實際應用價值。
並列摘要
In a semiconductor flip chip production system each process is typically handled by more than one machine, which could be either new or old. This problem is referred to as an "Unrelated Parallel Machine with Multiple Processors Scheduling Problem". In addition, the flip chip assembly process has different flow-shop based on different products and allows zero process time for some processes. Therefore, the flip chip production system belongs to a “General Flow-Shop Factory”. Due to unrelated parallel machines with multiple processors, the scheduling problem in the above situations is a NP-hard problem. To solve the problem, this study applied a genetic algorithm for flip chip assembly scheduling problem that has different sequence-dependent setup times and different cycle time requirements. The result of this study indicates that the proposed GA method has 12% improvement to a real case company when compared with EDD scheduling method. Besides, the proposed method has over 80% improvement when compared with SPT and LPT methods.