基于全连接神经网络的天线阵列参数优化方法OA

Aiming at the problem that traditional antenna array parameter optimization relies on empirical parameter tuning and is computationally inefficient,an antenna array parameter optimization method based on Fully Connected Neural Network(FCNN)is pro-posed.First,a dataset for training the network is constructed.Secondly,an FCNN network with multiple hidden layers is designed to learn the nonlinear mapping relationship between the radiation characteristics and the 16-dimensional amplitude and 16-dimensional phase parameters,to achieve the optimal antenna parameter output under the condition of maximizing the main lobe gain and enhancing the side lobe suppression ratio.The experimental results show that the antenna excitation parameters obtained by the constructed FCNN network achieve a main lobe gain of 8.97dBi in the 1.6GHz band,and the side lobe suppression ratio is improved by 11.71dB and 7.41dB com-pared with the artificial optimization method and support vector regression(SVR)algorithm,respectively.In the 2.5GHz band,the anten-na main lobe gain of this method reaches 13.45dBi,and the sidelobe suppression ratio is still significantly better than the comparison method,which is improved by 7.85 dB and 6.38 dB,respectively.In terms of comprehensive performance index(weighted score of main lobe gain and side lobe suppression ratio),the proposed method is 57.9%and 29.3%higher than manual optimization and SVR algorithm in the 1.6GHz band,respectively.In the 2.5GHz band,it is increased by 30%and 22.2%,respectively.This method breaks through the technical bottleneck of low efficiency and long simulation time of traditional manual parameter adjustment and provides an efficient solu-tion for large-scale array design.