Research Of Wideband Metamaterial Absorber Based On Machine Learning And Optimization Algorithm

Metamaterial absorbers have a wide range of application scenarios in many national defense and civil fields such as stealth technology,electromagnetic shielding,and wireless communication.In practical engineering applications,expanding the working bandwidth of metamaterial absorbers has always been the difficulty and focus of research work.However,traditional design methods mainly rely on full-wave simulation for multiple iterations,which is time-consuming and labor-intensive.Therefore,it is of great significance to explore how to efficiently and quickly design high-performance metamaterial absorbers.This paper starts with exploring the reverse design method of broadband metamaterial absorbers,and conducts research from the following aspects:First,under the requirement of transparent and flexible absorbing,a method based on single-layer RFSS and optimization algorithm is proposed to expand the bandwidth for the problem of how to expand the absorbing bandwidth without increasing the thickness.Taking into account the two degrees of freedom of metamaterial unit structural design and spatial distribution.First,a series of units with continuous distribution of absorbing peaks are selected based on the equivalent circuit model of the SSL-type unit,and then the optimal distribution of the absorbing array is obtained according to the reflect array theory and optimization algorithm.The whole design process forms a closed loop in theory,avoiding the full wave Simulation saves time and effort.Finally,a broadband transparent flexible metamaterial absorber is designed based on the proposed method.The final experimental results are basically consistent with the simulation results and theoretical calculation results,which verifies the correctness and efficiency of the method.The absorber can achieve an absorption rate greater than 80% in the broadband range of 6.22GHz-19.22 GHz,and has large-angle stability under both TE polarization and TM polarization,and the absorption rate is greater than 45° at the incident angle.remains stable at more than 80%.In addition,the light transmittance of the absorber is good,more than 63% in the wavelength range of 500nm-800 nm.Second,a prediction model of MMA unit forward reflection spectrum based on artificial neural network is established.Firstly,the collection process of the training data set and the method of data preprocessing are introduced;secondly,the construction and training process of the ANN model for the SSL-type metamaterial unit is described;The forward prediction from the design parameters to the reflection spectrum is carried out for the type metasurface element,and the output results of the model are basically consistent with the full-wave simulation results.It is proved that the established ANN model has good prediction ability.Although it is a neural network model constructed for a metasurface unit of one structure,the method can also be extended to the forward spectrum prediction problem of other more complex structures.The subsequent reverse design optimization of metamaterial absorbers is expected to replace traditional full-wave simulation,thereby saving design time and improving efficiency.Third,a method based on the DGCNN model for array analysis of metamaterial absorbers composed of SSL cells of different sizes is proposed.In this method,the relationship between the units in the array is equivalent to the relationship between each node in the graph structure,and it is represented by an adjacency matrix,which is used in the dynamic graph convolutional neural network(DGCNN)to describe the relationship between each node.The connected link matrix can be updated adaptively during the training process of the entire model,so it can more accurately describe the mutual coupling effect between the units in the array,which helps to improve the discriminative ability of the entire network model.In the DGCNN model we established,the input is the size parameter of each unit in the absorbing array,and the output is the reflection coefficient of the entire array.Once the model is trained,it can give the results quickly and accurately,and the predicted results are consistent with the full-wave simulation results.Basically the same.The design efficiency is greatly improved.In addition,in this chapter,based on the trained DGCNN model combined with the optimization algorithm,a broadband metamaterial absorber with a scale of 10 × 10 is designed,which can achieve more than 90% absorption in the broadband range of 6.13GHz-18.98 GHz.wave effect.Since electromagnetic absorbers can effectively absorb the incident electromagnetic waves through some physical effects by converting electromagnetic wave energy into thermal energy,they play an important role in national defense and civilian fields.With the application of new radar detection technology to weapon systems,electromagnetic absorbers occupy a special strategic position in modern warfare.In addition,with the rapid development of information technology,electromagnetic stealth and interference have attracted more and more attention in the daily life.Researches on the application of high-performance absorbers for shielding or eliminating electronic interference have become a hot topic in recent years.Aimed at the strict requirments on microwave absorbers in practical engineering,such as wideband,low profile,light weight,strong strength etc,this work focuses on the high-performance absorbers,where low-profile broadband technology and dynamic tunable methods are proposed and employed for the design of the microwave absorbers.The main research includes the following aspects:Firstly,a set of segmented measurement covering the main microwave frequency band is proposed.If the test frequency is above 2GHz,the free space measurement system is proposed,which is simple and flexible.If the frequency to be tested is below 2GHz,a transverse electromagnetic wave(TEM)cell measurement is built,which has the advantages of high reliability and low cost,making up for the lack of low-frequency measurement as well.Secondly,a dynamically tunable wideband microwave absorber based on graphene and random metasurface is proposed.By stacking four layers of graphene,the tunable range of graphene sheet resistance is lowered to 380Ω/sq-80Ω/sq.In addition,by choosing twelve proper elements of metasurface and distributing them randomly,more resonant frequencies and phase responses can be achieved,thus improving the bandwidth of the microwave absorber and reducing its profile simultaneously.The equivalent circuit analysis is applied to explain the mechanism of absorption.By tuning the sheet resistance of graphene via bias voltage from 0V to 4V,the absorption of proposed absorber can be tuned from 80% to 50% in a wide bandwidth from 5 GHz to 31 GHz.The prototype of proposed absorber is fabricated and measured.The reasonable agreement between the simulated and measured results verifies the accuracy of our approach.Thirdly,two kinds of ultra-wideband microwave absorbers are proposed.Based on the principle of multilayer interference cancellation and ohmic loss,four layers of resistive films with different resistance values and different patterns are placed at certain intervals to achieve an absorption rate of more than 82% in the ultra-wideband from 0.5GHz to 18 GHz.On the basis of this absorber,the second one is improved by loading a high-permittivity dielectric plate and piercing metallic via to improve the absorption performance and incident angle stability simultaneously.Finally,the two samples are processed and measured to show the wideband absorption performance of the absorber.