A PCBA Solder Joint Defects Inspection System Based on Deep Learning Technology

This paper proposes a deep learning-based Printed Circuit Board Assembly (PCBA) solder joint defects inspection system. The proposed system is comprised of a camera module, an AI edge computing module, a conveyor, and a PCBA quality management platform. Therefore, the proposed system can use the camera module to capture the PCBA snapshots of the inspection area on the conveyor. Then, the PCBA snapshots are forwarded to the AI edge computing module to inspect whether the solder joint defect occurs via deep learning object detection technology. The experimental results show that the proposed system can inspect three common solder joint defects with average precision (AP), recall, and precision scores ranging from 72.3% to 100.0%, 0.98 to 1.00, and 0.98 to 0.99, respectively, and provide an adequate solder joint defects detection within 1.2 seconds, improving production yield.