Equivalent Circuit-Guided GAN Sample Generation of Metasurface for Low-RCS Scanning Array

In this article, a conditional deep convolutional generative adversarial network (cDCGAN) for the metasurface (MTS) to reduce the radar cross section (RCS) of the phased array is proposed. To reduce the calculation burden of collecting training data for cDCGAN, an equivalent circuit model is presented based on the mechanism of MTS. This method suppresses the sample space and largely reduces the number of collected samples from full-wave simulations. The designed MTS is loaded with resistors, and it absorbs the incident wave and transports the radiating wave. The resistor is represented as a binary image during the training procedure. The trained cDCGAN is utilized to generate satisfying meta-atoms. Two low-RCS wideband scanning arrays are designed, and the validity of the proposed method is verified with simulated and measured results. The monostatic RCSs of the arrays are reduced by 10 dB from 2.5 to 7.6 GHz and from 3.1 to 7.5 GHz, and the insert losses of MTSs are below 1.4 and 1.1 dB.