Akademik Veri Yönetim Sistemi
IEEE Access, 2026 (SCI-Expanded, Scopus)
Printed circuit boards (PCBs) are widely used in the production of many electronic devices today, making the accurate detection and classification of defects that may occur during manufacturing critically important. In recent years, deep learning-based approaches have gained prominence in this field. In this study, a novel deep learning approach is proposed to accurately detect and classify defects in electronic components on PCBs. An open-source dataset containing defect samples was selected for the research. This dataset includes 12 different classes across 4 types of PCBs.To address the class imbalance in the dataset, data augmentation and preprocessing techniques were applied. For feature extraction, the models CSPResNet50, EfficientViT, RepGhostNet, and ResNet50 were employed. While the cross-entropy function was used as the loss function in the training of traditional models, a cosine-based loss function was specially designed to align with the structure and purpose of the proposed model.This study involves both PCB classification and defect detection processes. When examining the performance on the defect detection validation dataset (which is the key task) the proposed method achieved the highest performance with a precision of 0.90, an F1-score of 0.8961, a Matthews Correlation Coefficient (MCC) of 0.87453, and a Cohen’s Kappa score of 0.8736.