半導體廠訂單允交策略

Abstract

半導體廠訂單允交策略 研 究 生：許建鴻 指導教授：巫木誠 博士 國立交通大學工業工程與管理研究所 摘要 本研究主要針對半導體廠允諾週期時間訂定與訂單服務水準間取捨的決策問題提出一解決方法。在半導體生產的環境下，產出量、生產週期時間、允諾週期時間及達交率間的關係相當複雜，無法以一個簡單的模式來加以分析。本研究利用模擬建構半導體廠的生產情境，蒐集訂單允交策略相關數據，並利用此數據建立兩種類神經網路，以訂單的允諾週期時間、達交率來預測半導體廠的產出量。再藉由類神經網路預測所得到的數據、透過不同遲交成本函數的計算，找出最佳的允諾週期時間、達交率組合，使半導體廠的利潤最大。 關鍵字：倒傳遞類神經網路、允諾週期時間、達交率

Delivery Commitment and Fulfillment Strategy for Semiconductor Manufacturing Student：Chien-Hung Hsu Advisor：Dr. Muh-Cherng Wu Institute of Industrial Engineering National Chiao Tung University ABSTRACT This research formulates and solves a decision problem for the trade-off of cycle time commitment (CCT) and service level for semiconductor foundry fabs. In semiconductor manufacturing, the relationships among throughput, manufacturing cycle time (MCT), committed cycle time (CCT) and on-time delivery rate are quite complicated and cannot be modeled by an analytical function. This research uses simulation to obtain a sampling data set. The data set is subsequently used to establish two back-propagation neural network for predicting the throughput, mean delay time for given CCT and mean delay rate (MDR). The neural network models are used to find the optimal combination of (CCT, MDR) to maximize the profit in scenarios with different penalty cost functions. key word：back-propagation neural network、committed cycle time、on-time delivery rate