晶圓製造廠設施規劃改善方法之研究

A Study of Improvement Type Algorithms for the Facility Planning of Semiconductor Fab

10.6358/JCYU.200503.0041

陳建良(James C. Chen)；楊榮鏜(Rong-Tang Yang)；彭國銘(Linus Kuo-Ming Peng)；王昭志(Chao-Chi Wang)

背脊式佈置 ； 陡降程序法 ； 門檻接受法 ； Semiconductor Fabrication ； Facility Planning ； Steepest Descent ； Threshold Accepting

中原學報

33卷1期（2005 / 03 / 01）

41 - 55

繁體中文

良好的晶圓製造廠設施規劃可有效的減少物料搬運距離，提昇生産效率，進而增加工廠的競爭力。晶圓製造廠的佈置通常採用功能式佈置的背脊式(Spine)佈置爲主，自動搬運系統(Automated Material Handling System, AMHS)置於工廠中央，平分工廠爲上下（南北）兩區，上下兩區再由擁有類似加工功能的功能式工作區所組成。 本研究提出一個由陡降程序法與門檻接受法二者整合的混合搜尋演算法則藉以改善晶圓製造廠佈置及減少物料搬運距離，此混合法先進行陡降程序搜尋區域解，而後應用門檻接受法跳脫區域解，反覆安排工作區的佈置組合以減少物料搬運距離。最後將此建議搜尋方法，與其他三種搜尋方法加以比較，以8至20個工作區之晶圓製造廠的資料加以測試。模擬結果顯示，混合法相較其他三種啓發式方法，雖然其搜尋時間較長，但其求解品質績效最佳，與最佳解的差距小於4%。.

A good facility layout of semiconductor fabrication plant can reduce material handling distance and increase production efficiency, and therefore increase productivity and competitiveness. Semiconductor fabrication plants generally use a spine configuration with an automated material handling system (AMHS) in the central area and functional bays surrounding AMHS. Each bay consists of equipment with similar functions. This study proposes a hybrid searching heuristic to improve the layout of wafer fabs and reduce the wafer transportation distance. It integrates Steepest Descent method (SD) and Threshold Accepting method (TA). This heuristic first uses SD to find a local optimal solution and then uses TA to search for a better solution. Based on this, the location of the bays in a wafer fab can be re-arranged to reduced wafers' transportation distance. The empirical data of real semiconductor fabrication plants consisting of 8 to 20 bays are used to evaluate the proposed hybrid approach and the other three approaches. Simulation results indicate that the hybrid searching algorithm leads to the best performance (less than 4% from the optimal solution) but requires longer searching time, compared to the other heuristic approaches.