Deteksi cacat solder pada PCB Through-hole menggunakan YOLOv11x untuk Inspeksi Otomatis

Abstract

Process soldering on Printed Circuit Through-hole PCB (PCB) is a crucial stage in electronic assembly, as defects such as under-soldering, over-soldering, and solder bridges can lead to device failure. This study proposes a YOLOv11x-based automatic detection system to identify solder defects in PCB through-hole in real time. Dataset consists of 1,300 the augmented images were split in a 70:20:10 ratio for training, validation, and testing. The model was trained using the YOLOv11x architecture with a CNN backbone and implemented in interface GUI Python based. The evaluation results show that the system achieved an accuracy of 97,5%, a precision of 99,3%, a recall of 98,4%, and an F1-score of 0,984. The model demonstrated optimal performance in Classification of three main types of defects and reducing reliance on manual inspection. This system has the potential to be implemented in quality control processes in the electronics manufacturing industry.