فهرست نشریات علمی دسترسی آزاد ایران

The growing complexity of printed circuit boards (PCBs) in modern electronic devices has made quality control an essential requirement for ensuring manufacturing precision, product reliability, and overall system performance. However, conventional inspection methods are often limited in terms of accuracy, speed, and adaptability to diverse defect types. This paper presents an AI-driven framework for automatic PCB quality control, built upon a customized version of You Only Look Once (YOLO)-v8 object detection model, renowned for its high accuracy and speed. To enhance fault detection capability, several architectural modifications were implemented, including the reordering of Conv and C2f blocks, sequential stacking of C2f layers, integration of MaxPooling layers, and redesign of output connections. These improvements enhanced the model’s ability to accurately identify a variety of PCB defects. The proposed model was trained and evaluated using the publicly available PCB_DATASET comprising 692 images across five common defect types. To increase robustness and generalization, additional real-world defective samples were incorporated. Experimental results demonstrate that the customized YOLOv8 achieved an accuracy of 97.7%, recall of 96.2%, and F1-score of 96.9%, reflecting improvements of 1.6%, 0.3%, and 0.9% respectively, over the baseline. These findings confirm the effectiveness of the proposed method for fast and accurate PCB defect detection, suggesting its strong potential for deployment in industrial quality control pipelines.