標題: 多晶圓廠短期產能交易模式
A short-term capacity trading method for multiple semiconductor fabs
作者: 王照瑋
Wang, Chao-Wei
巫木誠
Wu, Muh-Cherng
管理學院工業工程與管理學程
關鍵字: 多廠;半導體;短期產能規劃;產能交易;類神經網路;基因演算法;Capacity trading;Semiconductor manufacturing;Neural Network;Genetic Algorithm
公開日期: 2012
摘要: 晶圓製造廠的設備十分昂貴,因此機台的利用率越高,企業的競爭力越好。晶圓製造廠為了因應市場需求的變化,在不斷的產能擴充下,大多同時擁有多個工廠進行生產。為了達到最佳產能利用率,廠際之間的互相支援亦成為重要的課題。儘管有專門負責中長期產能規劃的IE部門,以及專門負責生產線現場排程與產能規劃的生管部門,仍有種種原因使得晶圓製造廠發生機台利用率不平衡的情況。過去有許多學者與前輩針對晶圓廠多廠間的規劃進行研究,其中張文珍 (2007)針對晶圓廠短期產能供需問題,發展了一套兩廠間的產能交易決策模式,找出近似最佳之交易組合提升公司整體的晶圓產出量。由於各晶圓廠經過多年的擴廠,均已有超過兩個廠以上的規模,因此本研究希望延伸張文珍 (2007)的產能交易決策模式至三個以上的晶圓廠,發展以最大化多廠的晶圓總產出為目標的產能交易決策模型。
In semiconductor industry, production tools are extremely expensive. A semiconductor company usually establishes multiple semiconductor factories (fabs) in a cluster. That is, these fabs are located in neighborhood in order to share the use of tool capacity. In a short-term period (e.g., one week), some types of tools in a fab might be highly-utilized and some other types might be lowly-utilized. Due to the production of different product mixes and some unexpected events, different fabs might have different tool utilization profiles. To increase the total throughput, tool capacity trading needed to be carried out among fabs. Wu & Chang (2007) have developed a method to find an optimal capacity trading portfolio between any two fabs. Their method involves the use of simulation, neural network, and genetic algorithm. As an extension of their method, this research develops a short-term capacity trading decision-making mechanism for multiple (more than two) fabs to maximize the total throughput of a semiconductor company. Experiment results reveal that the proposed method is effective in increasing total throughput by providing better tool trading portfolio, due to the enlargement of trading solution space.
URI: http://140.113.39.130/cdrfb3/record/nctu/#GT079863509
http://hdl.handle.net/11536/71736
顯示於類別:畢業論文