- 學位論文
混合反覆式代理人賽局及遺傳演算的工廠地點設置及供應鏈成員選擇問題演化求解法
A Repeated Agent Gaming and Genetic Algorithm Hybrid Method for Factory Location Setting and Factory/Supplier Selection Problems
摘要
本研究在二階及三階層的供應鏈流程架構下提出一個「工廠地點設置及零售商選擇工廠問題」和一個「工廠地點設置及零售商選擇工廠與工廠選擇供料商問題」，據以探討運輸成本和產品生產成本對於供應鏈成員選擇的影響。前者「工廠地點設置及零售商選擇工廠問題」是一個包含零售商與工廠的兩階層的供應鏈設計問題。當中探討運輸成本與產品製造成本對於零售商選擇工廠以及工廠地點設置決策變數的影響。而後者「工廠地點設置及零售商選擇工廠與工廠選擇供料商問題」則是一個包含零售商、工廠、以及供料商的三階層的供應鏈設計問題。此問題同時探討上述兩類成本和原料購買成本對零售商選擇工廠、工廠選擇供料商、以及工廠地點設置決策變數的影響。本研究並針對這兩類問題發展一個「混合反覆式代理人賽局及遺傳演算的演化求解法」，並實作求解系統名為「混合代理人賽局和遺傳演算的優化系統」。此法使用交互凍結的演算架構結合遺傳演算和反覆式代理人賽局求解技術有效求解上述兩類問題。為比較本法的求解效能和品質，本研究同時建立單純遺傳演算法及交互凍結遺傳演算法進行比較。經由範例驗證與比較，本演化求解法對本研究提示的兩類問題的求解品質與收斂效果都比單純遺傳演算法以及交互凍結遺傳演算法好。此外也設計各種個案測試，探討不同情境下的供應鏈設計結果，提供供應鏈設計參考之用。
並列摘要
This paper presents two supply chain design problems: 1) a factory location setting and factory selection problem, and 2) a factory location setting and factory/supplier selection problem. The first problem involves a number of location fixed retailers choosing a factory to supply their demands and a number of factories whose locations are to be determined. The goal is to minimize the sum of transportation and manufacturing costs. In the second problem, the raw material procurement cost is additively added, since a number of material suppliers are involved and each factory needs to select one supplier to provide the raw material. Economic factors are considered in the cost evaluations. Therefore, the partner selections (factory selection for retailers and material supplier selection for factories) will influence the manufacturing and procurement costs significantly. To solve these problems, a repeated agent gaming and genetic algorithm hybrid method (AGGAHM) is proposed. The AGGAHM uses a mutual frozen mechanism to consecutively enable and disable the evolutions of agent gaming and genetic computation. Based on this proposed method, a prototype system, namely agent gaming/genetic algorithm hybrid optimization system (AGGAHOS) is developed to test several postulated numerical examples. Computation results of the AGGAHM are compared with those from methods of genetic algorithms and mutual frozen genetic algorithms. Results show that AGGAHM outperforms the methods solely using genetic algorithms. In addition, several supply chain design scenarios are established to verify the solutions against the practical applications.