Abstract
To avoid potential quality loss, inline defect inspection is used to monitor equipment health via sampling a processing lot every fixed period in semiconductor manufacturing. However, inspection needs cost and prolongs cycle time. Optimizing inspection sampling strategy is critical to enhance fab productivity and maintain competitive advantage of semiconductor companies. This study aims to construct a dynamic defect sampling decision framework for equipment health monitoring in semiconductor manufacturing. In particular, we focus on two sub-problems to enhance the effectiveness and efficiency of defect sampling. First, this study optimizes the sampling period allocation for each equipment using Bayesian decision analysis and mathematical programming model. Second, this study develops a scan lot evaluation hierarchy from the information value perspective to enhance cost-effectiveness. The empirical study was conducted in a leading semiconductor company in Taiwan. This study discussed the total expected quality loss in different scenarios and provided risk evaluation of scan lot reduction. In addition, this study offline simulated the selection mechanism based on historical data to evaluate performance, and implemented inline process. The result showed that the practical value for scan lot reduction and thus reduced workforce loading.