多目標遺傳演算法求解供應鏈整合性庫存控制與設施定址問題

A Multi-Objective Evolutionary Approach for an Integrated Location-Inventory Distribution Network System

謝佳琳(Chia-Lin Hsieh)

指導教授 ： 廖述賢

淡江大學/管理學院/管理科學研究所博士班/博士(2011年)

https://doi.org/10.6846/TKU.2011.00547

摘要

供應鏈配送網路系統在提供一個最佳化的平台來追求供應鏈需求者的時間效率與供應者的成本效益，以期追求在成本上有效率以及在時間上可快速回應的供應鏈管理。有效率的供應鏈管理主要目的在於減少並降低作業時的成本：例如設施定址成本，庫存作業成本與運輸配送成本等。快速回應的供應鏈管理主要目的，則是為了能快速回應市場上需求的急速變化，以滿足大多數的顧客。然而，成本與顧客滿意度這兩個目的之間常常是互相抵觸的。
我們的研究主要是將一個包含設施定址、庫存控制與網路配送等三種供應鏈決策規劃的議題，並以兩種相互衝突目標：需求者時間效率與供應者成本效益為追求最佳化的標竿，設計了一個整合性的多目標規劃模式稱為多目標定址庫存問題此數學模式，簡稱為MOLIP。該數學模式同時包含了三個目標函數：分別為供應鏈總成本，顧客服務水準（或訂單達交率）與供應鏈彈性（或顧客回應水準），因此，我們所建立的模式乃是一個包含非線性混合整數規劃的最佳化的問題。此問題乃是在求解最佳的分派中心設址地點並將所有不同地區的顧客與需求，指派到最適當的分派中心，並找出最佳的柏拉圖最適解。
本研究探索以多目標遺傳演算法中稱為「菁英式非支配排序遺傳演算法」(NSGA-II) 來求解MOLIP模式的可行性。為了有效求解此問題，我們以接近實際供應鏈分銷網路問題設計了模擬的問題，包含了15間分銷中心位址與50位潛在顧客，並進行相關的數值分析以驗證求解方法的成效，結果發現，該方法所獲得的答案是令人滿意的。另外，本研究亦進行了相關的敏感性分析與情境分析，用來評估該模式所呈現的不同結果，並提出相關的管理意涵給決策者作為決策參考之用。

關鍵字

供應鏈管理 ； 整合性多目標庫存與定址網路分派問題模式 ； 多目標基因遺傳演算法 ； 敏感性分析 ； 情境分析

並列摘要

Supply chain distribution network system provides an optimal platform for efficient and effective supply chain management. There are trade-offs between demand time efficiency and supply cost effectiveness. In this dissertation, an integrated two-echelon distribution network system consisting of one supplier, multiple distribution centers, and multiple customer zones is formulated under a vendor managed inventory (VMI) setup which simply assumes the vendor (supplier) manages the inventory of the customers and stores them at different distribution centers. The system also integrates the effects of facility location, distribution, and inventory issues and includes conflicting objectives such as cost (for effectiveness), volume fill rate and responsiveness level (for efficiency). With these considerations, we present a Multi-Objective Location-Inventory Problem (MOLIP) which results in a Mixed-Integer Non-Linear Programming (MINLP) formulation.
The MOLIP model consists of two steps. The first step makes the strategic decisions to determine the optimal number, sites and capacity of opening distribution centers (DCs) to be used, as well as the establishment of distribution channels and the amount of products to distribute from the supplier to assigned buyers via DCs. In the second step, the model in turn determines the inventory levels and safety stocks, economic order quantities of different facilities in the tactical level. However, the model is difficult to solve with existing optimization algorithms due to the considerable number of decision variables and constraints resulting from the integration. To obtain feasible and satisfactory solutions to the integrated MOLIP model, a hybrid multi-objective evolutionary approach is presented which is preliminarily based on a well-known NSGA-II evolutionary algorithm with a non-dominated sorting mechanism and an elitism strategy. To facilitate the genetic search and improve the search results, a heuristic method is designed to generate a well-adapted initial population.
To investigate the possibility of the proposed evolutionary approach for MOLIP model, we implemented on three experiments. First, an experimental study using practical data was then illustrated for the efficacy of the proposed approach. The hybrid approach has been successfully applied for providing promising solutions on a base-case problem with 50 buyers and 15 potential DCs. Computational analyses has presented a promise solution in solving such a practical-size problem.
Second, we implemented several scenario analyses to understand the model performance and to illustrate how parameter changes influences its output. The scenario analysis illustrates that excess capacity in the supply chain network design is beneficial for volume fill rate and responsiveness level and has only little expense of total costs. In additions, the results of the scenario analyses implied that the distribution network flexibility and competitiveness level sought by the supply chain managers is warranted. The model proposed in this research is helpful in adjusting the distribution network to these changes.
Finally, we tested and compared our NSGAII-based algorithm with the one based on the improved Strength Pareto Evolutionary Algorithm (SPEA2) by developing a test set of random problem instances of the MOLIP model to understand the efficiency between two approaches. In these test instances, two algorithms obtained similar approximations of their Pareto frontiers but NSGAII algorithm outperformed in terms of the diversity quality of the approximation to the Pareto frontier. However, the SPEA2-based algorithm was more efficient in terms of execution time in small or tight capacity instances. This suggested that the propose hybrid algorithm can be an efficient approach for providing feasible and satisfactory solutions to large-scale difficult-to-solve problems.

並列關鍵字

Supply Chain Management ； Integrated Location-Inventory Distribution Network Problem ； Multiobjective Evolutionary Algorithm ； Trade-off Analysis ； Scenario Analysis

參考文獻

[1] Akinc U. (1993). Selecting a set of vendors in a manufacturing environment, Journal of Operations Management, 11, pp. 107-122.

[2] Altiparmak F., Gen M., Lin L. and Paksoy T. (2006). A genetic algorithm approach for multi-objective optimization of supply chain networks, Computers and Industrial Engineering, 51(1), pp. 196-215.

[3] Ambrosino D. and Scutella M. G. (2005). Distribution network design: new problems and related models, European Journal of Operational Research, 165(3), pp. 610–624.

[4] Averbakh I. and Berman O. (2002). Minmax p-traveling salesmen location problems on a tree, Annals of Operations Research, 110(1-4), pp. 55–68.

[6] Aytug H., Khouja M. and Vergara F. E. (2003). Use of genetic algorithms to solve production and operations management: a review, International Journal of Production Researches, 41(17), pp. 3955-4009.