Printed Circuit Board Defect Detection Using CNN Based Deep Learning Technique

Being in the competitive area of electronics manufacturing, PCB defects have to be detected for reliability of products and maintaining the quality standards. This paper deals with the new advanced methodology for PCB automatic inspection by using CNN-based methodology for the detection of defect on PCBs. The models were highly trained and tested with a huge dataset that was designed especially for the identification of defects in PCBs. For effective identification and classification of various defects like the irregularities, missing components, gaps and electrical shorts, this paper uses the CNN architecture. The achieved results of experiments with the model's accuracy up to 95.57%, precision up to 95.44% and recall up to 93.76% underscore the importance of achieving high performance. Ensuring such stable and accurate representation of the CNN - based method makes this approach valuable in improving the production workflow. This automatic defect detection system enables man inspection to be minimized. Hence the possibility of human error becomes low and thus forms a reliable solution for production efficiency improvement and strict quality control throughout the electronics manufacturing process by the manufacturers.