演化演算法應用於結合正逆向物流之多目標區位網路設計問題之研究

A Study on the Application of Evolutionary Algorithm on Multiple Objective Location Network Design with Forward and Reverse Logistics

江鎮宇(Chen-Yu Chiang)

指導教授 ： 邱顯明

淡江大學/管理學院/運輸管理學系碩士班/碩士(2009年)

https://doi.org/10.6846/TKU.2009.00565

摘要

以企業經營實務而言，必需面對處理越來越多有關產品退貨、維修及回收等關係到顧客滿意度之服務；而且面對越來越嚴格的環保法律要求，廢棄產品再利用與再生資源之利用已經成為企業經營的關鍵議題之一。企業不管是面對顧客滿意度，以及法律與獲利等因素，都必須要正視產品回收的問題，這將是企業面臨這股綠色浪潮的機會與挑戰。
再者，現今企業已經花許多心力在採購生產和產品運輸等作業上，而運輸成本佔總成本的比例為最大宗，又因環境和法規的因素，企業不得不再投入成本在回收服務上，使得成本負擔更加吃力。根據Rogers et al.（2001）的研究指出，若將逆物流和正物流產品分別獨立處理，會花費許多成本在運送逆物流產品上。若有第三方物流公司（3PL）可以同時分擔正逆向物流服務的部份，提供倉儲設施執行儲存運送作業，廠商不需增設回收設施，不僅可以提供專家經驗意見和更廣的視野，也可以分散成本風險。
本研究在模式構建上主要是以第三方物流業者與廠商兩方面為出發點，考量系統總成本最小化、顧客涵蓋數最大化及設施平均利用率差異最小化等目標，來構建多目標區位指派模式。問題的求解則是以C 語言自行撰寫改良之演化演算法進行求解。
以自行設計的兩個小型範例配合窮舉法來驗證模式的正確性，並利用此結果來進行演算法的正確性測試；並以區位標竿例題來測試本研究設計之演算法，對於其它區位模式之精度與效度；與張立偉（2001）所設計之基因演算法比較測試，以範例二與大型問題（一）來比較兩者的求解品質，由測試結果可知本研究演算法之求解品質較佳。最後分別求解大型問題（一）、（二）之柏瑞圖最佳解，與情境敏感度分析。經由多次的測試，發現最小系統總成本與最大顧客涵蓋數之間呈反向關係；最大顧客涵蓋數與最小設施平均使用率差異之間呈反向關係；最小系統總成本與最小設施平均使用率差異之間呈些許正向關係，但不完全正向。由權衡結果可知合乎一般的邏輯性，更可證明本模式正確性，具有應用價值，可以作為實務單位營運決策之參考。

關鍵字

多目標規劃 ； 柏瑞圖最佳解 ； 演化演算法 ； 逆物流 ； 區位問題

並列摘要

In business practical operation, enterprises have to face problems related to services of customer satisfaction such as product returning, maintenance and recycling; with the ever-increasingly stricter laws and requests related to environmental protection, the recycling of discarded product and use of recycling resources have become one of the key issues of business management. From the perspectives of customer satisfaction, laws or profit, enterprises have to face the problem of product recycling which will be their opportunities and challenges under this green wave.
Moreover, nowadays many enterprises have devoted efforts in purchase, production and transportation, among them, transportation is the most cost, but due to environment and laws, they have to invest funds in recycling services which will further increase their cost. The study of Rogers et al. (2001) showed that if enterprises deal with products of reverse and forward logistics respectively, then the cost of transporting products of reverse logistics will be higher. A third-party logistics (3PL) company can deal with products of both reverse and forward logistics, provide warehousing facilities and implement warehousing and transportation operations, so that enterprises do not have to set up recycling facilities. They can not only offer advices of past experiences and professional opinions but also provide a broader view and lower the risk of cost.
In model construction, considering the minimization of total cost of system, the maximization of customer coverage number, and the minimization of difference of facility average utilization rate, this study constructed a multi-objective location-allocation model from the perspective of third-party logistics companies and enterprises. To solve the problem, a comprehensive solution procedure is developed based on Evolutionary Algorithm using C language.
This study used two self-designed examples coupled with enumeration to verify the validity of the model with solid result. In addition, the proposed procedure was applied to location benchmark examples to verify the validity of the algorithm. Compared with the genetic algorithm designed by Chang Li-Wei (2001), this study used example 2 and large question (1) to compare their solution quality; the test results indicated that the solution quality of the algorithm designed in this study was better. Finally, this study tried to find out Pareto optimal solution of large question (1) and (2) and conducted scenario sensitivity analysis. After many times of testing, it was found out that system total cost minimum and customer coverage number maximum was negatively correlated; customer coverage number maximum and minimum of facility average utilization rate difference was negatively correlated; system total cost minimum and minimum of facility average utilization rate difference was slightly positively correlated but not totally positively. The results were logical and the validity of this model was verified, and it could be used as decision support for operational decisions within enterprises.

並列關鍵字

Multi-Objective Programming ； Pareto Optimal Solution ； Evolutionary Algorithm ； Reverse Logistics ； Location Problem

參考文獻

18. 陳啟嘉（2006），「基因結構探勘於承接式子群體基因演算法求解多目標組合性問題」，元智大學工業工程與管理學系碩士論文，民國95年。

6. 林豐澤（2003），「演化式計算上篇：演化式演算法的三種理論模式」，智慧科技與應用統計學報，頁1-28，民國92年。

23. 楊文正（2005），「廢棄物物流系統規劃之研究」，國立成功大學工業管理科學系碩士論文，民國94年。

10. 洪子盛（2003），「以模糊多目標規劃法求解消防站配置問題之最佳化」，國立成功大學工業管理科學研究所碩士論文，民國92年。

15. 許義男（2008），「都市公共醫療設施區位規劃模型之設計與應用」，國立台北大學都市計畫研究所碩士論文，民國97年。