关键词: 库存控制, 滑动平均法, 指数平滑法, 预测模型, 半导体生产

Abstract: To satisfy changing market demands and complicated manufacturing process in semiconductor manufacturing enterprises, aiming at the uncertainties of the semiconductor manufacturing process and order arrival, Moving Average method (MA) and Exponential Smoothing (ES) were used to construct customer order prediction model respectively. Based on these two models, a production inventory control system model under discrete feedback status was put forward. Furthermore, the stability of the two different prediction models were analyzed, and the general stability criterion of the order policy parameters for a system, which was Bound Input, Bound Output (BIBO), with any lead time was given. Finally, based on the practical data taken from a semiconductor manufacturing factory, the effectiveness of these two prediction models were verified. Results showed that the MA prediction model could achieve less than 3% bullwhip effect in this factory, which was better than ES prediction model.

Key words: inventory control, moving average method, exponential smoothing, prediction model, semiconductor manufacturing