晶圓廠機台妥善率組合決策之研究

Abstract

一座晶圓廠是由多種機台群所構成，生產績效與機台群表現息息相關，維持機台在最適合的妥善率相當重要。本研究探討晶圓廠機台群的妥善率方案組合決策問題。亦即在供應商主導的商業模式下，一個給定產品組合和機台組合的晶圓廠，如何決定各機台群的妥善率，使該晶圓廠的獲利最大。本研究發展一包含兩模組的演算法，模組一藉由等候理論模型來評估在一妥善率方案組合下的績效。模組二利用基因演算法與邊際配置法，尋找最佳的妥善率方案組合，使得在滿足生產流程時間上限的情況下，獲得最高的總系統利潤。案例結果指出，基因演算法在一可接受的計算時間內，其求解出之最佳解優於邊際配置法之最佳解。

This research proposes an approach to optimize the determination of tool availability for semiconductor fabs. A semiconductor fab is consisting of many tool groups. The performance of a semiconductor fab highly depends on the availability of tool groups in fabs. Therefore, maintaining the availability of tool groups is very important. This research aims to identify an optimal tool availability portfolio under the following scenario. The objective function is to maximize the revenue of a fab subject to a target cycle time constraint. The solution approach involves two main modules. First, we use a queueing network model to evaluate the performance of a tool portfolio. Secondly, we use a genetic algorithm (GA) as well as a marginal allocation (MA) algorithm to find a near-optimal tool portfolio. Experiment results indicate that the GA method outperforms the MA method at an acceptable computation time requirement.