Visual-Learning-Based Convolutional Neural Network Application for Electromagnetic Susceptibility Event Classification

Abstract

The increase in the number of digital interfaces and monitoring devices has led to the need for more accurate and quicker decision-making solutions in electromagnetic-interference (EMI)/electromagnetic-compatibility (EMC) testing. Conventional monitoring and inspection methods, which have been carried out by human intervention, are not able to give the desired success criteria for accuracy in a reasonable confidence level due to the complexity of the electronic system’s susceptibility. With this study, an experimental convolutional neural network (CNN) architecture, which is to learn from the equipment under test (device(s) under test) operational behavior by training and detection of several events, including failure modes in a high-speed scan rate, is introduced. An automatic event detection and diagnosis model is successfully designed and then evaluated. The outcomes of this work demonstrate accurate and high-speed EMI monitoring, and event classification is achievable by the proposed method.