需求不確定性兩階段下單模式之探討-以IC設計M公司為例

Abstract

因產品生產前置時間長與市場波動大造成預測困難，如何進行投料規劃控管產品交期與庫存為IC設計公司生產規劃的主要議題。因預測誤差一般會隨著越接近實際需求點而縮小，若為縮短生產週期，則須付出趕工成本，將使公司成本大量增加，競爭力將隨之流失。本論文探討在需求不確定性兩階段投料模式，研究情境為假設客戶需求為不確定，服從均勻分配，且該分配的平均值不確定，也服從均勻分配。第一階段投料之後，進行需求資訊的收集，修改預測進行第二階的投料，而第二階段投料的前置時間較短需付出趕工成本。在上述研究情境，在收集觀察值之後利用貝氏方法修正事後機率，並以利潤最大化為目標建構數學模式，利用動態規劃手法求得兩階段各自最佳投料量，最後利用案例說明求解方法，由敏感度分析說明管理意涵。

Due to long producing cycle time and the uncertainty of customer demand in design industry, the manager is difficulty to decide the releasing plan under the consideration of the shortage cost and the inventory cost. In general the accuracy of the forecast of customer demand depends on the lead time. If the producing cycle time is shortened, then the accuracy of forecasting can be improved. Therefore in this study we propose a two-stage releasing model. At stage 1, based on the initial forecast of customer demand, a releasing quantity with the normal producing cycle time is decided. Then at stage 2, based on new information to modify the initial forecast of the customer demand, an additional releasing quantity with the rush producing cycle time is decided. The modification of the forecast of customer demand is based on the Bayesian method. In this study, first we construct a mathematical model of the two-stage releasing model. Then we propose a dynamic programming method to solve the problem. Finally some numerical examples are given to illustrate the proposed method and the sensitivity analysis.