Performance Analysis of Electromagnetic Frequency Response Prediction Based on LSTM

Abstract

In the field of electromagnetic compatibility (EMC), it is very difficult to obtain accurate high-frequency data, both for experiments and finite-element simulations. At this stage, frequency response prediction techniques based on deep learning have been applied in the field of EMC, such as long short-term memory (LSTM). However, the current research is in the state of “one thing at a time,” and there is no systematic performance analysis or research on LSTM. In this letter, the performance analysis idea based on the feature-selective validation (FSV) method is proposed to comprehensively analyze the prediction performance of LSTM with the help of eight sets of classical electromagnetic data examples. Finally, the analytical conclusions obtained are applied to the prediction of the shielding effectiveness of metal boxes to verify the accuracy of the proposed analytical ideas.