Deep Learning-designed Coding Pattern Units Enabling Ultrathin Chessboard Metasurfaces for Effective Multiband RCS Reduction

The effective reduction of radar cross-section (RCS) is of critical importance for enhancing military survivability and maintaining civilian electromagnetic security. In this work, a coding metasurface based on printed circuit board (PCB) fabrication is proposed. The metasurface comprises coding cells with different phase responses that are arranged in a chessboard configuration designed to suppress backward scattering. The research integrates a convolutional neural network (CNN) and a genetic algorithm (GA). The CNN rapidly predicts the coding pattern phase responses, while the GA optimizes the coding pattern composition for effective RCS reduction across C/X/Ku bands. The simulation results demonstrate that the optimally designed chessboard metasurface effectively suppresses backward scattering and redirects the scattered energy in other directions. With a thickness of 2 mm, the metasurface designed for the C, X and Ku bands achieves a RCS reduction of more than -10 dB in the 5.95-7.72 GHz, 8.78-11.75 GHz and 12.51-16.54 GHz frequency ranges, respectively. Experimental measurements verify the low reflection characteristics of the coding metasurface designed for X-band in real scenarios, and the results are in close agreement with simulation predictions. This work innovatively applies the coding metasurface design concept to reflective phase unit design, providing a new solution for designing low RCS scatterers in specific frequency bands.