Automatic Defect Classification (ADC) solution using Data-Centric Artificial Intelligence (AI) for outgoing quality inspections in the semiconductor industry

Quinn Killough, Onder Anilturk - Landing AI and NXP

Abstract:

Machine vision significantly improves the efficiency, quality and reliability of defect detection & classification and has been applied successfully in many domains.In the semiconductor industry, one way to assure the quality of the processed wafers is to inspect the wafers via inline inspections, and then determine if the defectivity observed is abnormal or accepted via non-yield or quality impacting defects. Reliable classification of the defects when human operators are involved, requires excessive time, expertise and training of the individuals; which increases the cost and time to detect abnormalities on the wafers; along with large variation of correct classification. In this study, we present an automatic defect classification (ADC) application for outgoing quality inspections. In outgoing inspections, all of the defects were manually classified to reject or accept the inspected die with the defect classification. Earlier adoption of ADC systems usually emphasizes both accuracy (recall) and purity (precision) as output metric to deploy the system to classify the defects. In our implementation, purity is targeted as the main output metric for classification of clearly defined defects in the training set. This allowed us to deploy automatic defect classification of defects with high purity and benefit from its automatic classification earlier in the adoption process with immediate impact on workload reduction, while working on less pure defect classes.