Detection of Faults in Electronic Printed Circuits with Image Processing Methods

The detection of short circuits in electronic printed circuits is critical for industrial applications. Short circuits occur when conductors come into contact or are not isolated, which can lead to problems such as current deviation, overheating, component damage or circuit dysfunction. As circuits increase in complexity, manual detection of short circuits becomes more difficult. Improper assembly, faulty soldering, material quality deficiencies and physical damage are among the causes of short circuits. Understanding and managing these factors is important to improve circuit reliability. The aim of this study is to evaluate short circuit detection in electronic circuits using deep learning models DenseNet and VGGNet. In this study, short circuit detection using deep learning models DenseNet and VGGNet with image processing methods was evaluated. Short circuit detection was performed with DenseNet and VGGNet models under various scenarios. In the first scenario, accuracy, F1 Score, precision, recall, MCC, specificity was 0.7566, 0.7363, 0.7836, 0.7441, 0.5193, 0.7845 for DenseNet without data augmentation, and 0.7562, 0.7488, 0.7853, 0.7607, 0.5439, 0.7680 for VGGNet. With data augmentation, 0.7570, 0.7485, 0.7061, 0.7661, 0.5591, 0.7432 were obtained for DenseNet and 0.7758, 0.7640, 0.7986, 0.7695, 0.5659, 0.7584 were obtained for VGGNet. Without data augmentation, the accuracy, F1 Score, precision, recall, MCC, specificity was 0.9804, 0.9802, 0.9810, 0.9819, 0.9629, 0.9638 for DenseNet, and 0.8641, 0.8482, 0.9029, 0.8617, 0.7556, 0.9287 for VGGNet. With data augmentation, 0.9412, 0.9405, 0.9419, 0.9419, 0.9414, 0.8834, 0.9371 for DenseNet and 0.7375, 0.7375, 0.7588, 0.7299, 0.4803, 0.7423 for VGGNet.The results show that both models work with high accuracy and precision. This study supports the usability of deep learning models in industrial applications. The models reduce the need for manual inspection and minimise human error when detecting short circuits.