Rapid Design of tunable metasurface using Deep Neural Networks for Field Localized Wireless power transfer

다국어 초록 (Multilingual Abstract)

Wireless power transfer (WPT) is a convenient method of delivering energy to multiple devices without connecting wires. To further enhance WPT efficiency, unique characteristic metamaterials, such as electromagnetic field focusing, have been successfully utilized. Normally, metamaterial characteristics depend on multiple parameters. Several metamaterial designs require a significant amount of time to complete numerical simulation. In this work, we propose a rapid design metamaterial method using a deep neural network (DNN). When DNN is used, the results show an accuracy of 98.1% and an accumulated mean square error (MSE) less than 0.3×10^-3. For synthesizing the design parameters, the MSE is less than 8×10^-3. Besides, the computation time of the 1000 samples can be reduced 85×10³ times compared to the HFSS simulation.