利用CUSUM整合需求資訊改善半導體產業需求拉動補貨機制

Abstract

在各種產業上存貨管理都是備受關注的議題，而限制理論(Theory of constrains, TOC)所提出之需求拉動補貨(Demand-Pull)結合緩衝管理(Buffer Management)應用在管理供應鏈上的存貨，與傳統的補貨模式(s, S)、(s, Q)、(R,S)、(R, s, S)等補貨策略相比，已多次被證明其為有效管理供應鏈存貨的方法。然而在半導體產業中，其產品特性多為生命週期時間短、前置生產時間長、需求變化大。若僅使用限制理論所提出之方法來管理存貨，會因為無法即時針對需求趨勢做出反應而造成存貨過多或是缺貨的現象發生。本研究提出一套方法，藉由產品需求趨勢結合累積和(Cumulative Sum, CUSUM)與市場對於需求的滾動式預測(Rolling Forecast)資訊制定補貨調整時機，並依照預計庫存及庫存水準進行補貨調整量的調整。為驗證本研究所提出補貨模式之適用範圍，首先利用國內某晶圓製造廠提供之實際產品資訊進行試驗，再將本研究所提出的方法應用於模擬資訊後發現，本研究提出的方法應用在補貨前置時間長、需求變化大且生命週期短的產品比傳統需求拉動補貨策略結合緩衝管理更能維持在滿足顧客需求情況下，維持在較低的平均庫存。

Inventory management is essential in all industries. The fact that the demand-pull in collaboration with buffer management within the theory of constraint (TOC) is superior to traditional replenishment policy (s, S), (s, Q), (R, S), (R, s, S) has been proven in many studies. However, many semiconductor products are with short product life cycle, long production lead time andfluctuating demand. The traditional demand-pull replenishment and buffer management proposed by TOC may not be able to respond to changes in demand in time thus tends to result in high inventory or out of stock situations. This research proposed a method by integrating the trend of product demand and rolling forecasts of market demand with cumulative sum (CUSUM) method to make stock replenishment and buffer adjustment decisions. To test the validity and feasibility of the method this study proposed, both real historical demand data of semiconductor products provided by a wafer foundry in Taiwan and simulated demand data were analyzed. Per the result of computational experiments, the proposed method resulted in lower average inventory and higher service levels in the majority experiments conducted than either the traditional method suggested by TOC or the method proposed in previous study.