Similar sample selection approach based on sequence alignment; application to semiconductor industry

In semiconductor manufacturing system, diagnosis method consists in measuring and comparing similar samples that are taken at the same process level, processed by the same equipment and the same instructions in order to identify the defect root cause. However, the number of similar samples available for diagnosis analysis in semiconductor industries is often very limited. In this case, select more similar samples for the analysis is a great challenge. First, the different factors (same process level, measurable sample, comparable sample) need to be considered. Second, the various data (process historic data, product models data, measurement data) need to be analyzed. This paper proposes a Similar Samples Selection (3S) approach based on the activity sequences of products in the process. Local sequence alignment algorithm is used to align the activities in order to determine the similar regions and to calculate the similarity index. A prototype has been developed in order to test the proposed approach on real data in semiconductor manufacturing system. Result shows a large number of similar samples that could be selected for the measurement as well as a significant reduction of analysis time.