遠距雙廠之短期產能交易模式

Abstract

本研究主要針對遠距雙廠之短期產能交易模式進行探討。在實務上，各廠負責生產不同的工單；然而市場需求變化頻繁，各廠所生產的產品組合也經常變動。因此工廠間必須進行短期產能交易，以達到總產出最大化的目標；換言之，原廠工件的某些製程會被轉移至其它工廠進行加工。在過去針對跨廠產能交易的研究中，張文珍 (2007)提出產能交易機制求解最佳交易組合，其研究假設兩廠間運輸時間為零；然而在現實情況中，兩廠之間所需要的運輸時間，將會影響到產能的交易。因此，本研究延伸張文珍 (2007)所提出的方法，並假設兩廠間運輸時間不為零，以找出遠距雙廠的最佳產能交易組合。本研究方法主要是透過抽樣模擬產生出的結果建立類神經網路，以快速評估產能交易組合的績效，並以基因演算法求解出近似最佳交易組合。實驗結果顯示，運輸時間確實對於最佳產能交易組合的選擇有顯著影響。

This research investigates the short-term capacity trading method for two plants located with far distance. In practice, each plant is responsible for the production of a particular set of manufacture orders. The product mix for each plant is frequently different due to market demand change. As a result, the two plants must weekly “trade capacity” in order to maximize total throughput. That is, some operations of a job originally assigned a plant are transported to the other plant for processing. Such a capacity trading mechanism has been investigated by Wu & Chang (2007); and they assume that the transportation time between the two plants is zero. However, in many real world cases, the transportation time between the plants may be substantial and cannot be ignored. This research therefore extends Wu & Chang’s proposed method to find an optimal capacity-trading portfolio for two plants with far distance. In the method, we use sampled discrete-event simulation results to establish a neural network for quickly evaluating the performance of a capacity-trading portfolio, and develop a genetic algorithm to obtain a near-optimal trading-portfolio. Experiment results indicate that the transportation factor indeed has a significant effect on the selection of capacity-trading portfolio.