Intelligent Detection of Circuit Board Faults based on Convolutional Neural Networks

With the increasing functionality and complexity of modern aviation electronic equipment, the requirements for diagnosing circuit board faults have become more stringent. Traditional testing and diagnostic methods typically require specialized personnel and electronic measuring instruments. These methods not only incur high time and personnel costs but also cannot fully meet the production and maintenance requirements of modern products. To address these issues, this study developed a circuit board fault detection and localization method based on deep convolutional neural networks. In this study, experimental circuit boards were designed to simulate 48 typical faults and 1 normal state. The normal voltage and fault voltage data of these circuit boards were collected and underwent outlier screening and normalization preprocessing operations to enable better learning of these data by the neural network. Finally, the processed data was input into the designed deep convolutional neural network, which can classify the data and convert the fault localization problem into a classification problem.