A Novel Model for Electromagnetic Properties of Complex Microstructure Composites Based on Support Vector Regression

We propose a support vector regression (SVR) model to obtain dielectric properties of composites with complex microstructure. First, the composites are regarded as a homogenizing medium in a certain polarization direction, and a training database is calculated by the transmission line theory(TLT). The database consists of the effective permittivity of composites and corresponding transmission coefficients at different frequency points. Then, SVR is utilized to establish composite material electromagnetic characteristic model. The cross-validation method is employed to select model optimization parameters during the modeling. Finally, seventy percent of the database is selected for model training, and the remaining data which is not involved in training is for verifying model accuracy. The reversion results reveal that prediction electromagnetic characteristics of composites by SVR model is in excellent agreement with original data. The equivalent permittivity can be applied to the electromagnetic characteristics of fuselage structures in aerospace aircraft modeling, and can be used to analyze its electromagnetic scattering and shielding.