The electromagnetic radiated emission prediction of PCB with the improved artificial neural network model

This paper proposes that different electromagnetic sources have unique radiation characteristics. Thus, the radiation source type can be chosen as an important feature to train an artificial neural network (ANN) to predict the radiation of the sources. A new ANN model, including radiation source type and frequency as inputs were presented. The model took several seconds to train and calculate, and demonstrated very high accuracy. When the radiation source type was deleted from the features of the inputs, however, the accuracy of the ANN model degraded significantly. Different important parameters of the ANN model, such as validation checks, gradient and normalization, hidden layers and neurons, and learning rate are analyzed for the usage of the ANNs in the EMC field. The proposed method can be applied in the radiated emission prediction and practical design of PCB.