Equivalent Motor Radiation of an Electric Vehicle Based on Neural Network Approach

The feedforward neural network method is utilized in this paper for predicting the desired electromagnetic radiation properties of the motor of an electric vehicle. The neural network is first fully trained by simulation data and its predicted field values are then compared with those obtained through "traditional" FEKO simulations. The comparison has demonstrated a similar level of accuracy for both approaches, but the simulation time and cost can be dramatically reduced by employing the neural network method.