Modeling of Cone Absorbing Material Based on Neural Network and Optimization Algorithm

When calculating the absorption capacity of periodic cone-shaped absorbing materials, full-wave simulation is often difficult due to the large electrical size. And the theoretical calculation formula of the transmission line can only be applied to the case where each layer is a uniform medium. Therefore, this paper uses optimization algorithms to extract the equivalent material parameters of absorbing materials with different volume ratios in the air, and trains the BPNN (back propagation neural network) based on this data, which can effectively calculate the effective material parameters on specified frequency, actual material parameters and volume ratio. The equivalent medium model established in this paper not only breaks through the limitation of the Maxwell Garnett model on the frequency of use, but also uses the optimization algorithm to avoid the inapplicability of the S-parameter extraction method in the case of high-order modes.

I. Introduction

When calculating absorbing materials in microwave anechoic chamber, GTEM cell and other environments, due to the large electrical size, the direct use of full-wave simulation method calculations often consume a lot of time. And the method is difficult to apply to the optimal design of absorbing materials. According to the transmission line theory, the theoretical formula can be used to calculate the input impedance of a uniform absorbing material with a multilayer structure, and from this, the return loss of the multilayer structure as a whole can be calculated. But under normal circumstances, the structure of the absorbing material will be designed as a pyramidal structure arranged periodically. If each slice of the absorbing material can be equivalent to a layer of uniform dielectric material, the return loss can be calculated using transmission line theory (Figure 1), which can greatly reduce the time required for the calculation.