Wideband High Gain Lens Antenna Based on Deep Learning Assisted Near-Zero Refractive Index Metamaterial
Huanran Qiu,
Liang Fang,
Rui Xi,
Yajie Mu,
Dexiao Xia,
Yuanhao Zhang,
Shiyun Ma,
Jiaqi Han,
Qiang Feng,
Ying Li,
Hong Xu,
Bin Zheng and
Long Li
Deep learning neural network (DLNN) has enormous potential in solving electromagnetic inverse design problem, and thus meet the growing demand for rapid high gain antenna design in current industrial applications and other complex questions. Here, we propose a wideband near-zero refractive index high gain antenna based on dual band near zero refractive index frequency selective surface (DB-NZRI FSS) with the aid of Fourier transform neural network (FTNN). FTNN employs a Fourier transform-based data simplification algorithm to address the prevalent issue of long training time in neural network for antenna design. We verify the universal adaptive and effectiveness of FTNN by rapid designing near-zero refractive index metamaterial working in adjacent bands. The proposed DB-NZRI FSS unit has transmission zeros at the magnetic resonance point and electric resonance point, and a stopband with high reflectivity exists between the two points. By integrating the initial highly directional radiation effect of zero refractive index metamaterial with the planar parallel cavity principle, the proposed lens antenna obtains the maximum gain of 12.64 dBi at 8.2 GHz. The FTNN has high accuracy and low loss of 0.0407. The designed DB-NZRI FSS has relatively low profile of 3 mm (8% wavelength at the central frequency of 8.2 GHz). Besides, the designed antenna has the characteristics of dual polarizations and wideband with the relative 3 dB gain bandwidth of 19.35% (7.56-9.18 GHz).
Submit
