Prediction of Field-to-wire Coupling Problems Based on Deep Neural Network

In this paper, the current response of stochastic electromagnetic fields to a transmission line above a ground plane is accurately predicted with a proposed deep neural network (DNN) method. The network takes port number of the wire and incident direction of electromagnetic wave as input, with training and testing sample set obtained by the Agrawal's model and the BLT (Baum, Liu, Tesche) equation. Through the training, the proposed DNN is able to make accurate prediction of the induced current values on terminations of the wire rapidly. The full-wave simulation is also adopted to verify the predicted results. Preliminary numerical experiments show that the percentage of prediction error about the current value can reach below 1%. The proposed network demonstrates strong approximation ability and has a great potential for real-time current response in more complex problems of field-to-wire coupling.