Multiparameter Modeling Technique Based on TF-ANN for FSS Design
Yuhong Ma, Weiwei Wu, Wentao Yuan, Jingjian Huang, Ximeng Zhang, Zhaoyu Huang, Naichang Yuan
Abstract
In this communication, a novel transfer function-artificial neural network (TF-ANN) model is proposed for parametric modeling of the frequency-selective structure (FSS). In the proposed model, the pole–residue-based TF is used to fit the frequency responses of the training samples. The introduction of the modified response improves the robustness of the TF-ANN model for FSS and solves the order-changing of the TF between different samples. Then, the ANN is used to learn the nonlinear relationship between the physical parameters and the coefficients of the TF. Finally, two wideband FSSs are used as examples to validate the effectiveness of the proposed model. The modeling results show that the proposed model is a good candidate to accelerate the design and optimization processes of the FSS without using any exact empirical model.