Antenna Inverse Problem Analysis Based On Neural Network

In recent years,communication technology has been developing with each passing day,and these new technologies have had a huge impact on human beings.While people enjoy the convenience brought by high-performance microwave/RF components,they also have more stringent performance requirements.EM inverse algorithm(EM inverse algorithm)has been widely used in microwave/RF fields because it can invert geometric parameters according to the electrical performance of microwave/RF devices.However,EM inverse problems are mostly nonlinear and ill-posed,which makes the solving cost of inverse algorithm extremely high.In recent years,neural network has achieved great success in the fields of image and voice,etc.How to use neural network NN to solve electromagnetic inverse problem and obtain the algorithm of electromagnetic inverse problem with less resource consumption,high precision and short time is one of the important research directions at present.The research results of this paper are as follows:(1)Firstly,this paper describes the antenna near-field power synthesis model in the inverse electromagnetic problem,introduces a variety of electromagnetic inverse algorithms to solve the antenna near-field power synthesis,and chooses the PSO algorithm for performance comparison.Then,this paper presents the antenna near-field power synthesis algorithm based on neural network,and designs a feasible algorithm to solve the phase at the source point of the array antenna according to the near-field power of the array antenna.Experimental results show that the model of average error is 2.25×10-2,then the numerical experiments prove that the average prediction accuracy of the model is 92.19%,and using the PSO algorithm performance comparison and contrast,according to the results of the algorithm under the premise of convergence of the algorithm training time is 17.5% of the PSO,9.19% higher prediction accuracy than PSO algorithm,has high precision and low cost advantages,resources in the field of electromagnetic inverse problems such as microwave thermal therapy has a broad application prospect.(2)In view of the disadvantages of the traditional antenna design process,this paper proposes a microstrip antenna design algorithm based on the neural network,and utilizes the powerful nonlinear fitting ability of the neural network to improve the design process of the traditional antenna.The experimental results show that the algorithm design of microstrip antenna frequency 1.9 GHz and 2.44 GHz,target resonance frequency of 1.9GHz and 2.45 GHz,result is obtained by NN training model,the first target frequency matching accuracy,second only to deviate from the 10 MHZ,deviating from the small scope,design time only 58% of the traditional methods,this method can greatly improve the simulation efficiency and save computational resources,and has unique advantages in the field of complex antenna optimization and design(3)Based on neural network design of microstrip circularly polarized antenna,using neural network modeling technology,electrical parameters and geometric parameters of the antenna model,the experimental results show that: the antenna design time only 49%of the traditional method,the simulation efficiency,axial ratio and S11 characteristics consistent with the original antenna,the average prediction error is 0.13 mm,this method is beneficial to quickly locate initial parameters using smaller computation cost,a shorter cycle of antenna design and simulation optimization.