Artificial Intelligence-Driven Approaches for Automatic Wafer Map Failure Detection in Semiconductor Manufacturing

Engineers could learn a lot about what went wrong with the semiconductor production process by looking at wafer maps. Improving the semiconductor production process, product performance, and cost-effectiveness may be achieved by efficient pattern detection of wafer map failures. Hence, this study presents a DL-based algorithm for accurate automated failure type detection of wafer maps. This study set out to investigate DCNNs as a tool for defect identification in semiconductor wafer maps during production. The study takes use of the WM-811Kdataset, which includes 811,457 wafer maps with different kinds of defects, to train a model that can correctly detect and categorise these faults. Preliminary performance parameters indicate that the DCNN model performed well with an accuracy rate 93.75%, precision 93.81%, recall 93.79%, an F1-Score 93.76%, all of which suggest that it is superior to traditional ML models, viz., the CNN-WBM model and Logistic Regression. Moreover, the study shows the possibility of practical application in improving quality assurance and yield coefficients in the production of integrated circuits. As a good starting point for wafer diagnosis in the future, their work offers an effective end-to-end DL model that accurately captures the features of wafer labelling.