摘要：AbstractWhile computational intelligence and its applications in machine learning have expanded among the different fields of industry, many semiconductor and electronics manufacturing companies are increasingly utilizing machine learning techniques in their manufacturing process in order to improve the yield of production. In a high volume electronics manufacturing facility it is important to be able to find out quickly the reasons for bad quality. The analysis of systematic defect patterns on wafer maps is an interesting approach for aiming at this. Because of the fast accumulating data that are typical for sensor manufacturing, it is obvious that the method for recognizing systematic patterns in wafer maps should be robust, adaptive and relatively fast. The purpose of this study is to find out if it is possible to form distinctive groups consisting of wafer maps having similar patterns and associated with cell-specific data to enable further actions such as the search for the root causes of bad quality. In this paper, the SOM method is combined with k-means clustering to extract systematic data patterns from spatially oriented wafer maps. Artificial data consisting of 700 individual wafers is used for validating this modelling approach.
