An Equivalent Radiation Source Based on Artificial Neural Network for EMI Prediction

Abstract

In this paper, an equivalent radiation source based on the artificial neural network (ANN) is proposed for the electromagnetic interference (EMI) prediction of an unknown noise source. Firstly, the unknown noise source is equivalent to a dipole array, and the magnetic field over the plane above the unknown noise source is scanned. From this information a set of linear equations is obtained for the solution of the dipole array. Next, in order to consider the multi-reflections between the unknown source and its nearby components on the same PCB, and also the possible nonlinearity interaction between the circuits and electromagnetic fields, the original dipole array equivalent source is extended to the equivalent source based on the ANN. A numerical example shows that the proposed ANN equivalent source can be better in predicting the shadowing effect of the components around the unknown EMI source. This study provides a novel possible solution for the EMI source reconstruction through the near-field scanning.