Scalable modeling of grounded coplanar waveguide for MMICs design using neural network with an effective sampling strategy

Abstract

In this paper, a scalable model of millimeter-wave grounded coplanar waveguide (GCPW) based on back propagation (BP) neural network is proposed. The relationship between sampling strategy and generalization accuracy is investigated, and the modeling idea of segmental sampling is proposed. The established multi-size GCPW model is first simulated using the electromagnetic simulation software High Frequency Structure Simulator to obtain the corresponding electrical characteristic parameters. Then different sampling strategies are used to obtain training data respectively, and the training data are used to train the neural network. The MSE of the GCPW neural network model obtained by using 30 sets of data was 0.0424, while the MSE of the neural network model obtained by using only 8 sets of training data were 0.43165 and 0.28671 respectively. The results show that the model established in this paper has high model accuracy and modeling efficiency, as well as good dimensional scalability and high generalization ability. The trained neural network model can be scaled to include a wide range of dimensions, and only the dimensions and frequency of the GCPW need to be used as input, and the neural network model can accurately output the corresponding electrical parameters. In addition, segmented modeling not only reduces the amount of training data and improves modeling efficiency, but also maintains accuracy. The neural network model proposed in this paper has high accuracy and can be used for accurate and rapid design and analysis of microwave circuits. The method can also be applied to the modeling of other transmission lines and can provide insight into the modeling of high frequency passive devices.