Abstract
A number of inspection and measurement stations are set in the fabrication process to assure that the quality of wafer meets the specific requirements. Due to the limitation of inspection machine capacities, reducing sampling frequency to save capacity under the acceptable quality level becomes important. However, reaserchs about sampling plan usually focus on statistical aspect rather than economic aspect. In this study, we developed a cost-based heuristic for statistically determining the sampling frequency in wafer fab based on the Baye's decision analysis. The expected loss funciton is derived by integrating the loss of inspected and the loss of non-inspected lots. We aimed to determine the optimal sampling frequency that minimize the expected loss by trading-off between sampling costs and the quality loss. An empirical study is practiced with data from a wafer fab to validate the strucure and find out the crucial issue in reducing sampling frequency decision. As the resaults, we concluded that failure probability of population is the pivotal factor in sampling frequency decision. However,MTBF, the length between to assignable cause, is another keypoint. Thius, we suggested that further study should develop the distribution of MTBF from existing data in fabs or develop the method to monitor random cause and assignable cause separately.