Designing Of Electromagnetic Absorber For Metasurfaces Based On Deep Learning Method

As an electromagnetic material that cannot be directly obtained in nature,metasurfaces can achieve unique electromagnetic responses and flexibly modulate electromagnetic waves by becomingly designing and choosing materials.Since it was proposed,metasurfaces get extensive attentions from industrial and scientific communities.However,the previous design is raised elapsed time and computational consumption.Meanwhile,with the increase of materials characterization skills,the structures of metasurfaces have also become more and more complex.It is difficult to dig the deep relation between data manually.To improve the efficiency of design and optimization,many researchers use genetic and other algorithms to boost the metasurfaces design.Nevertheless,these algorithms are possibly limited by random search.Therefore,how to fleetly and exactly design and optimize metasurfaces functional devices has made a conundrum in this research field.With the fast growth of AI,some scholars design and analyze the unit structure,structure parameters of metasurfaces via deep learning(DL).This offers a fire-new way for the pursuit of metasurfaces functional devices.Although DL has been widely studied in this field,few research workers study metasurfaces electromagnetic absorber via DL.Metasurfaces electromagnetic absorber plays an important role in many applications(such as electromagnetic stealth,electromagnetic shielding).The conventional method has two main disadvantages: difficult to generate complex on-demand absorption spectra and longer design cycle.Because its practical value and to avoid the shortcomings of conventional method,DL is used to study metasurface electromagnetic absorber.The head tasks of this essay are expressed below:1.Design of the broadband metasurfaces electromagnetic absorber based on deep learning.To boost the design of the broadband metasurfaces electromagnetic absorber,the DLM is rendered.It can find the metasurfaces structures that researchers want with high predictional accuracy within ten seconds.To prove the DLM,we judge the metasurfaces absorber that consists of PMI material and alternate multi-layer metal rings.And these predicted curves are in correspondence with the desired curves.2.Design of the multiband metasurfaces electromagnetic absorber based on deep learning.In real usages,the single-band metasurfaces electromagnetic absorber can hardly meet the diverse requirements.Hence,it is necessary to design a metasurfaces electromagnetic absorber with multiple bands.Moreover,it takes a lot of time to collect datasets via the parameter scanning function of CST only,major time-consuming mismatch design aim.Therefore,we gain datasets by Equivalent Circuit Theory.At the same time,a new deep learning model is designed based on DLM.We delete the redundant network structure and propose an Improved DLM(IDLM).Through the meatures mentioned above,the design efficiency of the multiband metasurfaces electromagnetic absorber is greatly improved.Finally,the IDLM is used to carry out the inverse design of the multiband metasurfaces electromagnetic absorber and these results gratify design goal.