A COST-EFFECTIVE METHODOLOGY FOR A RUN-BY-RUN EWMA CONTROLLER

• Main Authors: HUANG, LI-SHIA, 黃麗霞
• Other Authors: RUEY-SHAN GUO, ARGON CHEN
• Format: Others
• Language: zh-TW
• Published: 1997
• Online Access: http://ndltd.ncl.edu.tw/handle/46416386368512376478

碩士 === 國立臺灣大學 === 商學研究所 === 85 === As the competition in semiconductor market is getting more and more severe, quality improvement in manufacturing has become an important source of competitive advantage. During IC fabrication in a wafer fab, wafer outputs might deviate from the target values because of process abnormality such as a drift in equipment or material performance. The traditional Shewhart control chart can not detect this kind of small drift quickly, which usually results in a significant quality loss. In order to solve this problem, this study provides a more sensitive statistical process control method to detect the small drift, and at the same time provides a cost-effective control algorithm to maintain the wafer outputs close to the target values. In this study, EWMA (Exponentially Weighted Moving Average) control chart is used to detect the small drift in a manufacturing process. In order to select the optimal weight in the EWMA control chart, both sensitivity and robustness are characterized using ARL1 (as a measurement of robustness) and ARL2 (as a measurement of sensitivity). This is accomplished by deriving the run length distribution using Markov-chain formulation. We also propose a cost-effective methodology for a run-by-run controller. The controller combines the advantages of statistical process control and feedback control. It adjusts the equipment settings only when EWMA control chart detects an abnormal trend. We use simulations to characterize the mean square error, false alarm, and adjustment number under different drift conditions and select an optimal weight for a minimum cost. As the simulation results demonstrate, the cost-effective run-by-run controller is able to maintain the process outputs close to the target in a drifting process with only few runs of adjustment. When the cost of adjusting the equipment settingsis significant or the underlying abnormal patterns are not certain, the benefits of cost-effective run-by-run controller can be further enhanced.