兩層級供應鏈中排程與配送整合模式建構

Abstract

在排程與配送整合型問題中，配銷中心必須處理零售商多樣化的產品需求且數量各異，決策者需決定訂單揀貨優先順序並同時考量車輛派遣組合與配送路線安排，待配銷中心備齊商品需求量並裝載至特定貨車後後再配送至各零售商，配銷中心為提升在同業中的競爭力，除降低生產成本外，也期望在零售商所需的時窗(time windows)限制內準時達交。 此研究考量的兩種情境分別為標準型問題與一般型問題，標準型問題求解排程結合傳統車輛途程問題(Vehicle Routing Problem, VRP)，一般型問題探討排程結合多車種之車輛途程問題(Heterogeneous Fleet Vehicle Routing Problem, HVRP)，兩情境中分別建立非線性規劃模型以求得此問題最佳解，然而，VRP與HVRP均屬於非線性時間中可求解的問題(NP-hard)，在整合排程問題後至少亦屬於一個NP-hard問題，因此，本研究以基因演算法(Genetic Algorithms, GAs)為基礎架構，演算邏輯中再分別考量順推模式與逆推模式進行求解，期望在有限的時間內求得問題的近似最佳解。 自適應的概念將使用在本研究的基因演算法中，自適應基因演算法(Adaptive Genetic Algorithms, AGAs)的求解效果將透過實驗與傳統基因演算法進行比較，實驗結果說明AGA在此問題環境的適用性，此外，為瞭解本研究之求解邏輯的搜尋能力，我們在相似的問題環境中與其他學者之方法進行比較，其結果說明本研究之邏輯有較佳的求解能力。透過此研究之演算法，我們針對地圖點位特性與多車種的使用進行敏感度分析，並在不同的環境組合提出相對應的管理建議，提供配銷中心之管理者進行參考。

In a production scheduling with delivery problem, there are different types of products processed by a distribution centre and then delivered to retailers. Each retailer order might be consisted of different products. The resolution of this problem is to determine the sequence of order operation, vehicle configuration and the visiting sequence of each vehicle for delivery goods within time windows. In this thesis, the standard problem and the generalized problem are proposed. The standard problem integrated with the production scheduling and conventional vehicle routing problem (VRP); generalized problem represented by production scheduling with heterogeneous fleet vehicle routing problem (HVRP). In this article, the nonlinear mathematical model is proposed. Due to the NP-hard complexity of VRP and HVRP, we have developed a genetic algorithm based framework for forward/backward correlations in a different mechanism. During this research, two adaptive genetic algorithms (AGAs) are designed and tested in variety of production and delivery scenarios. Illustrative examples, comparisons, and experimental analysis demonstrate the effectiveness of the propose framework. An experimental comparison of benchmark heuristic and our approach would be compared in similar scenarios. Moreover, the sensitivity analysis of changing in the geographical characteristic and introducing of new vehicle fleet group is performed. Finally, we present some management recommendations in different map features and fleet combined.