Observational and experimental insights into machine learning-based defect classification in wafers

This survey paper offers a comprehensive review of the methodologies utilizing machine learning (ML) classification techniques for identifying wafer defects in semiconductor manufacturing. Despite the growing body of research demonstrating the effectiveness of ML in wafer defect identification, there is a noticeable absence of comprehensive reviews on this subject. This survey attempts to fill this void by amalgamating the available literature and providing an in-depth analysis of the advantages, limitations, and potential applications of various ML classification algorithms in the realm of wafer defect detection. The survey provides a structured taxonomy that classifies these techniques into distinct categories and sub-categories, facilitating an in-depth understanding of their strengths and limitations. Based on an extensive review and experimental evaluation, the paper highlights that Residual Neural Networks (ResNet) demonstrate the highest accuracy (99%) and F1-score (98.88%) across all methods, with convolutional neural networks (CNN) also showing strong performance in handling complex defect patterns. XGBoost is noted for its balance between accuracy (94.8%) and computational efficiency, offering significant reductions in time compared to deep learning models. The findings underscore the efficacy of ML models in identifying wafer defects and provide insights into their computational trade-offs, positioning these techniques as vital tools in automating quality control processes in semiconductor manufacturing. Additionally, the paper identifies avenues for future research, such as improving the generalization capabilities of deep learning models and optimizing their computational efficiency for use in real-time applications.