本研究之目的為在客製化(CTO)環境下，針對每一個超級物料清單(super BOM)裡的半成品和原物料作生產策略的決策，當我們從超級物料清單選擇部分我們所需要的物料時，會有機率的產生，本文將在第三章討論機率的呈現，在本文我們假設物料的需求率是發生在最上層的階級(class)，而且是一個隨機過程(stochastic process)且產能與服務的時間是趨近M/G/1等候線，最後我們要決策每個物料是要庫存生產(MTS)或是訂單生產(MTO)，此外，庫存生產的物料會用存貨策略((Q,r) inventory policy)來控管物料，而且全部是由外購得到，反之，訂單生產的物料全部由廠內自製。 在客製化的環境下，我們建構一個非線性、整數的模型來決策每個物料是要庫存生產或是訂單生產，而且同時去決定再補貨數量(Q)和再訂購點(r)的值。不同於其他篇論文，我們考慮很多客製化環境下的特性，例如: 超級物料清單(super BOM)、相容性(dependency)、層級(class)以及特性(characteristic)等等，本文的目標是要最小化成本以及生產反應時間，最後會用啟發式解法以及AMPL/MINOS這套軟題找出比較好的解。 我們在最後也會作數值的分析去評估本文的啟發解與模型是否有效益，我們會發現用啟發解雖然不是找出最佳解，不過與最佳解的差異並不會太大，而且執行決策的時間縮短很多。 We study how to decide a suitable planning strategy for each assembly and raw material included in a Super Bill of Material in the Configurable-to-Order (CTO) environment. The condition probability was induced by selecting the BOM item. It is the fixed probability for each item. Demand rate is occurred at top-level class. The demand of finished item follows a stochastic process. We assume that the facility with limited capacity is approximated as an M/G/1 queue. We try to determine each BOM item whether it would be assigned to make-to-order (MTO) planning strategy or make-to-stock (MTS) planning strategy. Furthermore, MTS items follow an inventory (Q, r) inventory/production policy and are procured externally whereas MTO items are produced internally. On the basis of CTO manufacturing environment, we conduct a nonlinear, integer programming model to decide each BOM item whether it would be assigned to MTO or MTS and determine the replenishment batch Q and reorder point r for MTS item at the same time. Different from other papers about CTO, we take more features of CTO into account, e.g. super BOM, dependency, class, characteristic and so on. Our objective is to lower the production lead time and minimize the total cost such as setup cost, holding cost and backorder cost. We extend the heuristic developed by Chen (2003) and use the AMPL/MINOS mathematical software to solve the model. We also use a numerical test to evaluate the heuristic and we can find there are several factors having impact on the MTS/MTO decision and the (Q, r) policy. Finally we can find the number of iterations generated by the total enumeration approach is significantly larger than our heuristic so the average run time in our heuristic is shorter than the total enumeration approach. Besides, in our model application, the difference in objective value between total enumeration and our heuristic is lower than 16.65%.
