對於一個多重作業階層、多計劃期需求量的大型生產存貨系統，如何求解最佳生產日程使得整體作業流程的總採購成本、總備置成本、以及總儲存成本之和為最小，即為多階多期動態經濟批量問題（Multi-echelon multi-period dynamic lot sizing problem）。若將所有各種生產情況（依照生產日程所包含的採購/備置次數分類）之最低總成本逐一解出串連，即構成完整最佳總成本曲線。搜尋最佳總成本曲線扁平區域內所包含的一些近似最佳（near-optimal）生產日程，將可提供生管與採購經理人員在可接受的總成本變動範圍內，獲知最佳的可行替代生產日程，使得日後制定批量決策更具有彈性。目前文獻上所提出的各種求解方法僅能尋求最佳生產日程或近似最佳生產日程，無法解出所有各種生產情況之最低總成本，亦即無法建構完整的最佳總成本曲線。為使各種生產情況之最低總成本、以及最佳生產日程皆可以保證被求解，本計劃將發展出新的最佳解法，並且利用資料庫儲存求解過程的大量計算資料，可以減緩演算此複雜問題所需動用的電腦主記憶體。本計劃也將此一新解法發展成一個強大的演算系統，使得多階多期動態經濟批量問題的求解將不受限於作業階層數。 ; The decision objective of the multi-echelon multi-period dynamic lot sizing (MDLS) problem is to determine an optimal production plan for minimizing the sum of ordering costs, setup costs, and inventory holding costs during the manufacturing process in a large production-inventory system. A completed optimal total cost curve can be generated by connecting the individual optimal solution of minimal total cost from all production situations (which is classified by the ordering/setup times of the feasible production plans). Searching the near-optimal solutions within the flat zone of the completed optimal total cost curve for the MDLS problem may provide decision maker flexible alternatives of optimal production plan under an acceptable variation in total cost. The various optimal and heuristic algorithms found in literatures were used to offer an optimal or near-optimal production plan under specific production situation but failed to obtain the completed total cost curve which resulted from all production situations. In order to ensure the minimal total cost of each production situation and optimal production plan can be solved, in this study, a novel optimal algorithm is developed for solving the MDLS problem instead of a heuristic algorithm. Additionally, a database is established for loading the enormous data into hard disk, which is generated during the solving process, to abate the requirement of computer memory in solving a large scale problem. The proposed solution procedure will be further coded as a powerful solving system for optimizing MDLS problem without restriction on the number of echelons. ; 研究期間 9708 ~ 9807
