利用EWMA改善限制理論需求拉動補貨策略之緩衝調整時機─於半導體製造之應用

Abstract

存貨管理一直是各產業所關注的議題，而限制理論(Theory of Constraints, TOC)的需求拉動(Demand-Pull)補貨與緩衝管理(Buffer Management)相較於傳統的(s, S)、(s, Q)、(R, S)、(R, s, S)等補貨策略在供應鏈存貨管理上已被多次驗證能有較好的表現。緩衝管理利用存貨量所占緩衝的比例來決定是否需要調整緩衝，在半導體產業中，大部分的產品具有生命週期短、生產前置時間長、需求變化大等特性。在這樣的狀況下，若僅採用限制理論所建議的方法進行補貨與調整存貨緩衝，可能無法及時針對需求的變化做出反應，而造成存貨過多或缺貨過多。本研究提出一套方法，藉由歷史需求的指數加權移動平均(Exponentially Weighted Moving Average, EWMA)值與存貨水準決定存貨緩衝的調整時機，以改善單純使用需求拉動補貨策略與緩衝管理於存貨管理的效果。需求的EWMA值能結合現在與過去的需求資料，以移動平均的方式消除需求中的變動因子，較易觀察出趨勢。本研究設立數種偵測需求的EWMA值趨勢的候選法則，並將每種法則互相搭配，再利用國內某晶圓製造廠所提供之產品歷史資料篩選出表現最好的一組調整法則。為評估本研究所提出之方法的可行性與有效性，本研究將此方法用於分析實際案例資料與模擬資料。分析結果顯示本研究所提出的方法確實能夠加強應用需求拉動補貨策略與緩衝管理於管理存貨的效果，在需求變異大時，此種效果更為明顯。

Inventory management has been the topic of concern to all industry. The demand-pull replenishment policy and buffer management of TOC has been verified to have better performance than traditional replenishment policy such as (s, S), (s, Q), (R, S), (R, s, S) in supply chain inventory management. Buffer management use the ratio of inventory in buffer size to decide whether to adjust the buffer size, in the semiconductor industry, most of the products have short life cycle, long production lead time, and protean demand characteristics. In this situation, if only use the replenishment policy and buffer adjustment method proposed by TOC, may be unable to timely respond to changes in demand, cause too much inventory or out of stock. This study proposes a method by exponentially weighted moving average (EWMA) of historical demand and inventory level to decide timing of buffer adjustment to improve only use demand-pull replenishment policy and buffer management in inventory management. The EWMA value of demand can combine present and past data, eliminate variation factor in demand by moving average, it can easier to observe demand trends. This study set up a number of candidate rules to observe demand trends, and collocate with each different rule to each other, and then use historical data in the wafer manufacture plant to compare rule combinations to find the best set of adjustment rules. When the EWMA value of demand has the trends of adjustment rules, adjust the buffer size. By real date and simulation data can find the method proposed in this study have better performance compare with demand-pull replenishment policy and buffer management in inventory management when demand has large variance.