Square Tangential Metasurface Automatic Design Based on Deep Learning

To accelerate the design of electromagnetic metasurfaces, a machine learning-based approach has been proposed, which has resulted in a automatic design of square tangential metasurface. When the scattering parameter (S-parameter) curve of the metasurface, centered between 3.5GHz and 3.8GHz, is provided, the network immediately returns its dimensional parameters. A dataset of the electromagnetic response curves of the metasurface is extracted using a joint Matlab-CST simulator, upon which a neural network comprised of convolutional and fully connected layers is constructed. The network's output dimension is 2, representing two actual dimensional parameters of the square tangential metasurface. After feeding the validation set into the network, it was observed that the network exhibited a high level of accuracy.