PCB Classification using Convolutional Neural Network

For circuit-based applications to run successfully, the Printed Circuit Board (PCB) must be properly functioning, as, a defect in it can be dangerous for electronic circuits. Identification of simple images like fruits, scenery, etc. may be a convenient task for humans, but for images like that of a PCB, minute observation is required. When there is a huge dataset containing a variety of PCB images, it is very difficult for humans to classify the defective and non-defective PCBs. PCB has joint soldering, component placing, connections, etc. which can be the factors for classification. PCB defects can be then identified via model training through Machine Learning (ML). Convolutional Neural Network (CNN) model is proposed for classifying the PCB to be defective or not. The model is trained for two classes: defective and non-defective PCB images and then the classification is done as per the PCB faults and parameters. So, if a user wants to separate out defective PCBs from a mixed lot, such an ML-based model can be helpful. The proposed model achieved a detection accuracy of 92.8% for PCB.