啟發法與遺傳演算法在半導體廠排程之研究

Research of the Heuristic Dispatching Rule and Genetic Algorithm in Semiconductor Manufacturing Factory

黃子晟(Tzu-Chen Huang)

指導教授 ： 張百棧 博士

元智大學/工程學院/工業工程與管理學系/碩士(1998年)

摘要

半導體製造業係一高投資、高風險產業，且具有異於傳統流程型或零工型生產方式的一些特性，晶圓廠的排程派工問題因此變的困難，更由於實際作業中派工的執行者為現場操作員，多數學者所發展的複雜派工法則，並不易為現場操作者所瞭解，且容易造成混淆[Johri, 1994]，而本研究運用遺傳演算法強大的平行運算能力並結合派工法則建立一個有效的的排程派工模式，並經模擬驗證可以在晶圓製造系統中產生良好效益，並提供簡單、充分而考量整體效益的結論供作管理決策者參考。此外，本研究並利用實驗設計來分析遺傳演算法中各參數設定對排程模式在平均流程時間和平均延遲時間的影響，分析結果顯示執行世代與突變率對本研究的求解品質有顯著的關係。

關鍵字

半導體 ； 派工法則 ； 遺傳演算法

並列摘要

Semiconductor manufacturing is a industry with high investment and risks. Some of its various characteristics are different from traditional flow shop or job shop producing arrangement and the dispatching problem become difficult, furthermore, just because the one who did the dispatch jobs practical in the shop floor was the operator. Most of scholar develop the complicated dispatching rules that were not easily understood by the operator, even more feel confused [Johri, 1994]. In this study, we combined the Genetic Algorithm (GA) with powerful、parallel computational ability and the heuristic dispatching rules to develop a scheduling / dispatching model ,and simulation results show that the model created great benefit in the semiconductor manufacturing fab and also provide the information simple、ample、thought about overall the fab for the management.An experimental desing was made to analyze some design parameters of the genetic algorithmg for the mean flow time and mean delay time problem. This results show that the factors of the generations and mutation rate are significant in solution in this study.

並列關鍵字

Semiconductor ； Dispatching Rule ； Genetic Algorithm

參考文獻

N