Electromagnetic induced transparent (EIT) metasurfaces have extremely high transmission coefficients

which have been widely used in terahertz modulators

sensors and other fields in recent years. However

the full wave simulation of large EIT metasurface array often consumes enormous computational resources and time. In order to accelerate the design process of EIT metasurface array

a design method based on combination of transfer function and neural networks model is proposed. The vector fitting technique based on transfer function is utilized to fit the electromagnetic response of the metasurface. In the proposed model

the poles and residues of the transfer function replace the discrete electromagnetic response as inputs to greatly reduce the input dimension of the model and simplify the structure of the network. An EIT metasurface based on high and low resonant rings is selected as an example to verify the effectiveness of the proposed model. In addition

compared to directly using discrete electromagnetic response as input to the neural network model

the proposed neural netwrok model based on transfer function has higher efficiency. The proposed model in this study can quickly generate corresponding structural parameters to match the expected electromagnetic response

thereby significantly simplifying the design process.