Printed Circuit Board Defect Detection Methods Based on Machine Learning and Deep Learning: A Survey

Abstract

This study offers a careful assessment and examination of many deep learning-based defect detection models for enhancing quality control in printed circuit board (PCB) manufacturing. Leveraging advancements We survey state of the art object ID models in picture handling and deep learning draws near, like Quicker R-CNN, RetinaNet, SSD, and YOLO variants, such as YOLOv3-tiny, YOLOv5s, YOLOv5x6, and YOLOv8. By analyzing over a hundred related articles spanning from 1990 to 2022, we aim to provide manufacturers with insights into the effectiveness of these models in accurately identifying diverse defects in PCBs. Metrics such as precision, recall, mean average precision (mAP), and computational efficiency are employed to assess model performance. The study reveals that YOLOv5x6 achieves a superior mAP of 99.9%, indicating its potential for significantly enhancing defect detection accuracy. As an extension, the paper proposes building a user-friendly front end using the Flask framework, facilitating user testing with authentication. This study progresses the field of computerized blemish recognizable proof. Systems in PCB production, offering guidance for manufacturers to improve quality control processes.