應用蟻群演算法於半導體晶圓廠之設施規劃問題

Applying the Ant Colony Optimization Algorithm on the Facilities Layout Problem for Semiconductor Fabrication Plant

陳譽升(Chen Yu-Sheng)

指導教授 ： 胡黃德

元智大學/工程學院/工業工程與管理學系/碩士(2005年)

摘要

半導體製造產業屬於高成本的投資，興建一座8吋晶圓廠需要新台幣200多億，而12吋晶圓廠則需超過600億元。晶圓由開始生產到製造完成，需經數十種的機器，數百道的加工步驟，因此如果能有一個規劃良好的設施方案，必可以提升生產的效率與降低製造的成本。現今晶圓製造廠的設施佈置以脊椎式 (Spine) 佈置為主，其可以有較高的彈性應付多樣化的產品，獲得較高之利用率與減少機台的投資成本。

本研究是應用蟻群演算法於半導體晶圓廠之設施佈置問題，針對其四種物料搬運模式 (單向、雙向、單向加一捷徑、雙向加一捷徑) 來加以探討。藉由田口實驗設計的方式，求取蟻群演算法的最佳參數組合，對12個設施佈置的案例 (8至20工作站) 進行測試與分析，並與文獻中的其他演算法 (如模擬退火、基因演算法) 所得之結果，加以比較，探討其物料搬運模式與物料搬運成本的差異，以期能夠在可接受時間下，求取一近似最佳解。其求解品質均超過94.14%，十分穩定，且以20 Bays為最複雜之測試例題，其求解品質也有98.14%的表現。

關鍵字

蟻群演算法 ； 脊椎式佈置 ； 田口實驗設計

並列摘要

The cost of fabrication plant in the semiconductor industry is a very expensive investment. It needs more than NT$ 20 billions dollars to build the fabrication plants of 8 inches wafer, and the 12 inches of wafer plant exceeds NT$ 60 billions. Wafer fabrication needs to pass more than ten kinds of machines and hundreds of processing procedures. Hence, to design a good facilities layout, it can increase production efficiency and reduce manufacturing cost. Semiconductor fabrication plants generally use a spine layout, which has more flexibility to adapt the various products, obtain higher capacity utilization, and reduce the machine cost.

This study applied the ant colony optimization (ACO) algorithm on the facilities layout problem with four different material flow directions (one-direction, bi-direction, one-direction with a short cut, and bi-direction with a short cut). We also employed the Taquchi method to find out the best combination of the design parameters and analyzed the total material handling cost to compare with other procedures (simulated annealing, and genetic algorithm). The layout improvement heuristics was evaluated on 12 study cases of facilities layout problem with 8 to 20 bays. The results showed that the quality of solution is more than 94.14% compared with the best solution, and even in the complicated case of 20 bays, it also has more than 98.14%.

Keywords: Ant Colony Optimization (ACO), Spine Layout, Taquchi Method

並列關鍵字

Ant Colony Optimization (ACO) ； Spine Layout ； Taquchi Method