Parameter Prediction of Microstrip Filter Based on Convolutional Neural Network
A parameter prediction method for microstrip bandpass filter based on a convolutional neural network is proposed,and the corresponding prediction is performed for a three-band bandpass filter.Firstly,the HFSS electromagnetic simulation software training da-ta is used to obtain the data set of S parameters.The S parameters of the microstrip filter are used as the input,and the physical struc-ture parameters are used as the output.The convolutional neural network is used for training.Finally,the target S parameters are used as the input for parameter prediction.Compared with the simple fully connected neural network,the convolutional neural network can not only greatly reduce the network parameters,but also effectively avoid the occurrence of over fitting,and solve the problem of long time consumption of the fully connected neural network.Moreover,because the prediction of structural parameters is direct,even for begin-ners,the convolutional neural network can save a lot of design time.The simulation results show that the fitting degree between the tar-get S parameters and the S parameters predicted by the convolutional neural network is very high,which proves that the method has a high accuracy in predicting the physical structure parameters of the microstrip filter.
国家科技期刊平台
NSTL
万方数据
