The invention discloses a
metamaterial wave absorber construction method and device based on
deep learning and proxy optimization. The method comprises the following steps: constructing a plurality of
metamaterial structures; calling a CST-
MATLAB-API interface, and controlling CST
software to perform
electromagnetic field simulation on the plurality of
metamaterial structures to obtain a three-dimensional
structure diagram of each metamaterial and S parameters under different frequencies; performing inversion on the S parameters under different frequencies to obtain the absorptivity of each metamaterial under different frequencies, converting the design layer structure of the three-dimensional
structure diagram into a two-dimensional grey-
scale map, taking the absorptivity and the two-dimensional grey-
scale map as a group of data, and repeating the steps to obtain N groups of data; constructing a
deep learning model by using the N groups of data, taking the two-dimensional grey-
scale map as input and taking the corresponding group of absorptivity as output; and based on a
deep learning model, taking the longest interval length continuously greater than a threshold value in a group of absorptivity as a
fitness function, and utilizing a proxy optimization method to optimize the two-dimensional grey-scale map to obtain an optimal structure of the metamaterial.