Equivalent Radiation Source Reconstruction based on Artificial Neural Network for Electromagnetic Interference Prediction

In this paper, a source reconstruction method is proposed based on artificial neural network (ANN). Equivalent dipole array is used for the source reconstruction, which includes z direction electric dipole, x direction and y direction magnetic dipoles. The parameters of the dipole array are extracted by ANN, which input is the Green's function of the dipoles and the expected output is the electromagnetic field data. The conventional source reconstruction approaches use linear equation to fit the nonlinear relationship between dipoles and fields, resulting in errors. In contrast, the proposed method adopts ANN to model complex and nonlinear circuit characteristics with its powerful self-learning ability. A patch antenna is used as an example to validate the accuracy of the proposed ANN-based method, which shows that the error of the method is about 4%.