基因演算法於多目標TFT-LCD模組廠排程問題之研究

A Genetic Algorithm for the Multi-Objective TFT-LCD Module Production Scheduling Problem

陳建隆(Chien-Lung Chen)

指導教授 ： 徐旭昇

元智大學/工程學院/工業工程與管理學系/碩士(2008年)

https://doi.org/10.6838/YZU.2008.00036

摘要

現今的消費者對產品的功能、品質、價格與新潮流行要求的日益提升，使得消費市場競爭更加激烈，各家廠商為了能在市場保有優勢，相對對於生產工廠之交期、成本與品質之要求也越來越嚴苛，導致生產排程系統變得更加複雜。近年來全球化經濟快速發展，企業獲利大幅下降，為確保企業能永續經營，惟有具備良好的生產系統才能改善企業的競爭力，而生產排程規劃更是最為重要的一環。在現今少量多樣的消費市場及要求搶先上市的環境下，以往僅重視生產效率的單一目標生產排程已不符合現今生產系統之需求，管理者於進行生產決策時需同時考量多項生產績效指標。若使用僅以訂單交期為績效目標的EDD排程方式，雖可滿足訂單之交期，但會因產能利用率不佳造成生產成本之增加進而影響公司之獲利。本研究以薄膜液晶顯示器 (TFT-LCD)產業之後段模組生產排程為研究主題，模組製程會因顧客的訂單需求不同而影響交期與產能等問題，因此本研究針對模組廠生產排程問題進行研究，將針對最大完工時間(Makespan)和總延遲完工時間(Total Tardiness)的流程型工廠排程問題，發展出菁英式基因演算法的生產排程，以最大完工時間結合總延遲完工時間最小化來提高機台使用率與降低完工延遲之時間，以增加產能提升企業之獲利，並降低因完工延遲所產生之懲罰性成本損失並提升企業形象，針對最大完工時間與總延遲完工時間之權重以簡易多屬性決策分析之SMART-ROC進行權重之決策，基因演算法之編碼採用順序編碼法之偏好串列為基礎的編碼方式(Preference list-based representation)方式進行編碼，基因複製採競爭式選擇法，交配方式採Job-based Order Crossover (JOX)法，突變方式採移動突變法，並以實驗設計法求得較佳之基因運算參數－母體數、交配率、突變率、菁英比率與演化世代數，使基因演算法模型能求得近似最佳解之排程。並與現行之實際生產排程方法進行比較分析用以評估此基因演算法模型之效益。

關鍵字

流程型工廠；SMART-ROC；多目標基因演算法；實驗設計

並列摘要

Due to the changing market and the short life span of the TFT-LCD products, the production system requires efficiency as well as effectiveness. To make more profit, a business corporation needs an efficient production scheduling system to respond quickly to the market demand. In many situations, what the management concerns is not single but multiple objectives.
The purpose of this thesis is to develop an intelligent decision making method that aims to help the management handle their customer orders wisely and efficiently. The proposed method consists of two steps: (1) Determine the weights for the objectives that the management is most concerned with; (2) Employ an efficient and effective genetic algorithm to find the best schedule based on the weights in (1). The thesis studies the module phase of the TFT-LCD manufacture, which can be viewed as a hybrid flow shop scheduling problem, since some types of products may skip one or more certain manufacturing stages. An efficient genetic algorithm (GA) is developed to solve this scheduling problem with the weighted sum of two objectives: minimizing makespan and total tardiness.
A real life instance is provided to test the performance of the proposed GA. The parameters of the GA, such as population size, crossover rate, and mutation rate, are set according to the method in the theory of design of experiment (DOE). The performance of the GA is compared with the dispatching rule EDD, which has been used by this company. The numerical results demonstrate that the GA is superior to the currently used method.

並列關鍵字

Hybrid flow shop scheduling； TFT-LCD industry； Genetic algorithms； Multi-objective optimization

參考文獻

27. 徐英哲，「混合基因演算法於流程型工廠排程研究-

31. 林水耕，「應用混合式基因演算法求解流程型工廠之多

29. 謝日章，「柔性計算於生產管理之應用」，博士論文，

21. 韓俊逸，「基因演算法解算交期限制零工型排程問題之

1. Bagchi, T. P., Multi-objective Scheduling by