Automated X-Ray Inspection (AXI) on Surface Mount Technology Resistor (SMT-Res) Defects Detection Using GAN-YOLOv8n Model

Abstract

Product quality is a crucial factor in the electronic manufacturing sector, with inspection playing a significant role. However, human inspectors' accuracy fluctuates due to factors like fatigue, turnover, experience, and inconsistent fault categorization. These inconsistencies result in longer inspection times and operator-specific quality variations. To address this challenge, AI automation within Industry 4.0 framework is increasingly adopted to replace human involvement and reduce annual costs in quality control and testing. This paper proposes an Automated X-ray Inspection (AXI) system that develops a defects detector framework for detecting and classifying Surface Mount Technology-Resistor (SMT-Res) on PCBs, using a private dataset from Intel Technologies. The YOLOv7 and YOLOv8 models are examined as detectors, and experimental results highlight the exceptional performance of the GAN-YOLOv8n model. With just 3.01M parameters, it achieves 92.9% mean average precision (mAP) at Intersection over Union (IoU) 0.5 and 71.5% mAP at IoU 0.95, demonstrating excellent accuracy and efficiency. The proposed AXI system, leveraging AI automation, offers a promising solution for improving inspection processes, reducing costs, and enhancing product quality in electronic manufacturing