- 學位論文
接單後生產系統中考量零組件共同性之最佳零組件備料模式
Optimal Material Stocking Policy Model in Assemble-To-Order Systems by Considering Material Commonality
摘要
共同性零組件備料問題是供應鏈重要的問題之一。共同性零組件由於具備替代多種專屬性零組件的弁遄A而存在有降低零組件需求變異程度之風險共攤價值與減少向供應商訂購頻率之訂購成本共攤價值。然而目前針對共同性零組件備料模式的研究，多忽略零組件之訂購成本，並基於每期補充至相同數量零組件之假設，僅從零組件購買成本、存貨成本、與最終產品缺貨成本等因素進行模型建構與共同性零組件價值之分析。但在實務上，每期均訂購零組件、尤其是均同時訂購專屬性零組件以及共同性零組件之備料模式實則會造成高昂的訂購成本，進而影響製造系統之整體成本。為此，本研究納入零組件訂購成本於模式中，並建構為一整數線性規劃問題，同時考量專屬性零組件與共同性零組件之訂購成本、購買成本、存貨成本與最終產品的缺貨成本，為製造商求解出最佳之備料決策。 但是當產品結構趨於複雜、模型期數延伸時，由於零、壹變數的數量增加，造成此問題在求最佳解的過程中所需的時間會大幅上昇，甚至超越既有軟體之解題能力。因此本研究運用拉氏釋限法(Lagrangian relaxation)，發展一啟發式演算法，以迅速地求出本研究備料模式之近似解。透過情境設計之結果，可證明演算法在求解時間與誤差比率上在各情境中均有相當好的表現。最後比較本研究之共同性零組件備料模式與之前學者所提出之共同性零組件備料模式之表現，並分析訂購成本對共同性零組件備料模式的影響。
並列摘要
Recently, stocking policy by considering component commonality has become an important issue in supply chain management. Common component can reduce component demand variation and lower the order frequency through the effect of risk polling and order pooling, respectively. However, current researches on stocking policy related to commonality are most based on myopic assumption and base stock policy. Their models only consider component purchasing and inventory cost as well as product shortage cost but ignore component ordering cost. In practice, ordering components in every period, especially for both unique and common components, may result in huge ordering cost and, thus, increase the overall manufacturing cost. Therefore, in this research we construct a mixed integer linear programming model to find an optimal stocking policy for common components including ordering cost in objective function. As our model is extended to longer period or accommodated with more complex product structure, this mixed integer linear programming model cannot be solved optimally due to large number of 0-1 variables. Hence, in this research, we develop a Lagrangian relaxation-based heuristics to solve this model efficiently. Under the tests of different scenario design, our heuristics perform well in computational time and duality gap in all scenarios. Finally, we compare the results of our model with previous models based on myopic assumption. This comparison shows that our model produces much lower cost if the component ordering cost is significant.