Acquisition method for effective electromagnetic parameters of honeycomb structures based on FNN–INN

The effective electromagnetic parameters of honeycomb composite structures are crucial for studying their electromagnetic properties. However, most existing homogenization methods are based on static or quasi-static equivalence, considering only the effects of the electromagnetic parameters and the volume fraction of each phase medium on the effective electromagnetic parameters, while neglecting the influence of honeycomb aperture parameters. To address this, this paper proposes a novel homogenization method systematically analyzing the influence of honeycomb structure parameters, particularly aperture parameters, on the effective electromagnetic parameters. This method enables the direct acquisition of the effective electromagnetic parameters of honeycomb structures from their geometric parameters and the material parameters of each phase medium. First, this paper uses the geometric parameters of the honeycomb structure and the material parameters of each phase medium as inputs to construct a forward neural network for predicting the reflection coefficient (S11 parameter). Subsequently, the S11 parameter matrix is rapidly constructed using the reflectance formula. Correlation analysis and global feature extraction in two-dimensional space yield the feature vector matrix, representing the transformed S11 parameters. These are used as inputs to construct an inverse neural network to predict the effective electromagnetic parameters of the honeycomb structure. Finally, this paper reconstructs reflectance curves for two-layer and three-layer honeycomb structures, with mean absolute error below 0.26, validating the proposed method's effectiveness and accuracy.