結合歷史需求EWMA值與TOC補貨模式改善半導體存貨管理

Abstract

存貨管理是用來評判一間工廠績效好壞的重要指標，限制理論(Theory of Constraints, TOC)所提出的需求拉動補貨(Demand-Pull)結合緩衝管理(Buffer Management)為一套供應鏈存貨管理的技術，並有許多文獻證明此為一套有效的供應鏈存貨管理方法。然而，當傳統TOC補貨模式運用於產品較長的補貨前置時間以及需求變化大特性時，僅使用傳統的TOC補貨模式恐因為產品的補貨前置時間長的關係反應不及造成存貨過多或是發生缺貨。本研究提出一套以傳統TOC補貨模式為基礎結合歷史需求的EWMA值與自定存貨緩衝調整法則。有別於傳統TOC，本研究以補貨前置時間來決定EWMA權重及目標庫存調整的時機，透過補貨前置時間內EWMA值的變化判定需求是否出現趨勢，並提早進行調整。為驗證本研究所提出方法適用範圍及適用性，首先先透過不同需求情境進行模擬，最後再利用國內某晶圓製造廠所提供之產品歷史資料進行實例試驗。其結果證明本研究所提出的方法確實能夠應用於需求變異大且補貨前置時間長之特性產品。

Inventory management is an important indicator of the running performance of a factory. Theory of Constraints, or TOC, proposed using Demand-Pull and Buffer Management to manage supply chain inventory, and many researches have shown it to be effective. However, within the context of products with long replenishment lead time and high demand variance, TOC is likely to fail to respond quickly, resulting excessive stock or shortage. This thesis develops a new replenishment strategy, combining TOC with exponentially-weighted moving average (EWMA) of history demand and custom inventory adjustment rules. Unlike TOC, replenishment lead time will decide the weight of EWMA and the timing of target inventory level adjustment. Trends of demand can be spotted by examining the EWMA value, managers can then move accordingly. To test the generality of the method this research proposed, we simulate different scenarios, and apply it on the history data of a real product made by a local IC fabrication plant. The results show that our method works well on products with long replenishment lead time and high demand variance.