Detection of faults in electronic printed circuits with image processing methods

Detection of faults in electronic printed circuits, especially short circuits, is very important for industrial applications. Short circuits are faults that occur when the conductors in the circuit are contacted or isolated outside their functionality. This can cause the current in the circuit to deviate from its normal flow path and often lead to negative consequences such as overheating, damage to the circuit elements used, or complete dysfunction of the circuit. As the complexity of electronic circuits increases, manual detection and accurate localization of short circuits becomes more difficult. The causes of short circuits include factors such as improper assembly, faulty soldering processes, material quality deficiencies and physical damage. Understanding and properly managing these factors is critical for improving the reliability of electronic circuits and preventing or early detection of short circuits. In this context, computer-aided solutions to this problem can be produced. Image processing methods, especially using deep learning techniques, offer an effective approach for the detection of short circuits. In this study, the potential of deep learning models DenseNet, VGGNet, ResNet, AlexNet, GoogleNet, ResNext and YOLOV8 in short circuit detection is evaluated using image processing methods. Using the aforementioned deep learning models, fault detection in electronic printed circuits has been performed under various scenarios. This study supports the usability and reliability of deep learning models in industrial applications with their high accuracy and precision rates, fast detection times and low error rates. The models reduce the need for manual control when detecting short circuits, thus minimizing human error.