• 목록

 In this study, we propose a framework for detecting, classifying, and visualizing unknown patterns in the field of semiconductor wafer defect analysis. The rapid advancements in semiconductor processes and equipment have led to the emergence of unknown types of defects, most of which are analyzed and identified based on engineers' experience and judgment. Current approaches face difficulties due to limited labeling, emerging defects, and class imbalance. Although pattern recognition and deep learning techniques have been applied in research, they do not provide a complete solution.

Against this backdrop, our research presents a method that can quickly detect various unknown defect patterns and ensure high classification accuracy for known defect types. We utilize One Class SVM and a Transfer Learning-based ResNet50 network, which can be easily implemented on-site. This approach overcomes the limitations of traditional defect analysis, supports the identification of unknown types of defects, and enhances the work efficiency of engineers. In conclusion, our research provides a comprehensive framework that addresses the challenges of limited labeling and emerging unknown defects, and the effectiveness of the proposed model is evaluated through various experiments.