以共演化限制式遺傳演算法解決供應鏈產銷整合問題-以台灣織布業為例

APPLYING CO-EVOLUTIONARY CONSTRAINT SATISFACTION GENETIC ALGORITHMS TO SOLVE SUPPLY CHAIN INTEGRATION PROBLEMS-BASE ON THE TEXTILE INDUSTRY IN TAIWAN

余芳佩(Fang-Pei Yu)

指導教授 ： 張應華

大同大學/經營設計學院/資訊經營學系所/碩士(2008年)

摘要

由於市場快速變遷，企業為維持競爭力、永續經營的理念，及達成規模經濟、全域避險等因素考量，紛紛至海外設立新據點，全球佈局、放眼國際的跨國企業比比皆是，尤其是製造業，在面臨區域成本差異及全球化競爭下，更加積極至世界各地設廠，但隨著通訊科技的蓬勃發展及國際市場激烈競爭的影響，企業國際化雖然帶來龐大的市場商機，確也造成企業營運與管理的強烈衝擊，企業不再單打獨鬥，反而更積極拓展與全球各產業、企業合作，以尋求國際分工與產業整合，並強化各地域企業間之合作綜效；如何有效掌握全球供應鏈，以快速、有效率的回應顧客需求，提升顧客滿意度，並降低營運成本、提高訂單達交率及降低庫存，將成為企業是否具有國際競爭力的關鍵因素。
隨著全球化的快速發展，供應鏈整合技術已日趨複雜，規劃項目橫跨地域疆界，由於供應鏈上據點數增加、產品與原物料種類繁多、資源限制各異等影響，整合規劃機制須具備更強大的能力來因應目前變化。過去解決全球化供應鏈網路設計問題，較常用的方法為數學規劃法及啟發式演算法，若採用數學規劃法求解，一旦決策變數過多或限制條件過於繁複，將費時曠日導致求解效能不彰且易陷入局部最佳解；若使用一般的啟發式演算法求解，當求解變數過多、限制條件擴增而造成複雜度提高時，往往無法同時考量企業資源限制及最佳化解答。
有鑑於此，本研究將運用具最佳化搜尋特性的遺傳演算法，結合可依循準則變化而動態進化之共演化模式，與採用限制滿足模式之縮小搜尋空間能力以快速求解的特性(以下簡稱共演化限制式遺傳演算法)，來解決全球供應鏈整合網路設計問題，並以數學規劃法、一般遺傳演算法、共演化式遺傳演算法、限制滿足式遺傳演算法與共演化限制式遺傳演算法做結果與求解時間比較，以驗證本研究所提新方法之效益。
本研究以織布業為實驗對象，建構四階層多產品供應鏈整合模式，在考量配銷中心的配銷限制、織布廠之產能上限及紡紗供應商之原料供應限制下，以求取最小成本為目標。為確保本論文之方法可用於織布業各供應鏈規模，本論文依據訪談所獲得的織布業概況，區分為大、小規模共十個測試案例，並以比例性資訊做為模式參數之設定依據，以驗證本研究所提方法之適用性。

關鍵字

供應鏈網絡設計 ； 遺傳演算法 ； 共演化模式 ； 限制滿足模式 ； 織布業

並列摘要

Because the market condition changes fast, if enterprises want to keep its competence, sustainable management and achieve scale economy with avoiding risks in all areas, they set new sites overseas to have a general grasp of the world. There are many transnational enterprises in the world, especially in manufacture. In face of regional cost differences and globalized competition, manufacture tends to have more factories in every corner of the world. However, with the prosperous development of communication technology and the influence of international market’s fierce competition, the internationalization of enterprises bring out enormous business opportunities, It also causes great impacts on operation and management of enterprises. Enterprises are not alone any more, they search for cooperation with enterprises of other industries in the world actively in order to seek international division of labor and industries’ integration. Meanwhile, the integrated efficiency of cooperation among various enterprises of different regions has been enhanced. How to seize SCN efficiently at a world perspective, meet customers’ requirements efficiently, raise the content rate of customers, reduce the cost of operation and raise the success rate of purchase orders as well as reducing storage are pivotal factors that whether an enterprise has international competence.
With the rapid development of globalization, the integrated technology of SCN becomes more and more complicated. Projects are not regional any more. Because the data points of SCN increased, products and raw materials are various and the resources constraints are different, to integrate the planning mechanism should include much more powerful capacity to respond present changes. In the past, to solve globalized SCN network design problem, the usual way is Mathematical Programming and Heuristics Algorithm. If use Mathematical Programming to find the answer, once the decision variable are too many or constraint conditions are too complicated, it would cost much time which means low efficiency and it is easy to be trapped in partial optimum solution; if use general Heuristics Algorithm, when there are too many variables, constraint conditions are increased, the degree of intricacy will raise which usually makes people can not think about the resource constraints of enterprises and optimum solution simultaneously.
Therefore, this research will use GA with optimum search features. Together with Co-Evolutionary Mode that is in accordance with the variations of criteria and evolves dynamically and Constraint Satisfaction Mode’s capacity of narrowing down the space of search which can help us rapidly find the solution (in following the term will be Co-Evolutionary Constraint GA) to solve global SCN integration network design problem. In addition, with Mathematical Programming, general GA, Co-Evolutionary GA, Constraint Satisfaction GA and Co-Evolutionary Constraint GA to do comparison between result and processing time to verify the benefit of the new method provided in this research.
This research takes textile industry as the research object and establishes four echelons, multiple products SCN integration mode. Based on the consideration of distribution centers’ constraints, maximum productivity of textile factories and the raw material suppliers’ supply constraints, to achieve the lowest cost is the target. In order to make the method in this research can be applied in textile industries of various SCN scales, this research have 10 testing cases range from big scale to small scale based on the textile general condition information that acquired through interview and use proportionaliny information as the setting criterion of mode parameter to verify the applicability of the method in the research.

並列關鍵字

Supply Chain Network Design ； Genetic Algorithm ； Co-Evolutionary Constraint ； Textile Industry

參考文獻

[8] 李德威（2006），《建築廢棄物收容處理場所設置最佳區位評選之研究》，碩士論文，國立中央大學營建管理研究所。

[13] 陳朝文（2005），《以基因演算法求解跨期物流網路之配送規劃問題》，碩士論文，國立成功大學工業管理研究所。

[4] 王永宏（2003），《運用遺傳演算法求解供應鏈產銷整合問題》，碩士論文，國立成功大學工業管理研究所。

[7] 李月娥（2006），《多廠區多階製程緊急訂單分配與評估》，碩士論文，元智大學工業工程與管理研究所。

[20] David E. Goldberg(1989), “Genetic Algorithms in Search, Optimization and Machine Learning”, Addison-Wesley Longman Publishing Co., Inc., Boston, MA.