Fast Prediction of Electromagnetic Scattering Characteristics of Targets Based on Deep Learning

In order to solve the problem that traditional numerical method can not meet the requirement of high dynamic multi-target RCS prediction in a short time, this paper introduces the deep learning method, and proposes a novel idea of “generate ahead of time, call on site”. By constructing the deep learning model, the RCS simulation results of the target under different influence factors are taken as the training set of the deep learning model. Finally, the target RCS is predicted by the test set. Through analysis, the mean absolute error (MAE) of the proposed method for single and two triangular pyramid models is less than 1dB and 6dB respectively, and the prediction time is within 0.6s. The results show that the proposed method is effective for fast prediction and analysis of multi-target electromagnetic scattering characteristics.