Antenna Structure Optimization And Performance Prediction Based On Machine Learning

With the rapid development of the Internet of Things and Big data,the wireless communication field is facing great opportunities and challenges.As the core component of communication systems,antennas can realize the mutual conversion of radio waves and electrical signals,playing a crucial role in wireless communication,radar,remote sensing,and other fields.Traditional antenna design requires experienced antenna engineers to use simulation software for time-consuming simulation analysis,which requires a lot of effort and has a long development cycle.Machine learning technology can help antenna design engineers efficiently process and analyze complex signals through learning and analysis of a large amount of data,better understanding the characteristics and performance of signals,and achieving adaptive and performance optimization of antenna systems.This article focuses on the research of the design of 5G microstrip antennas and the algorithms of machine learning to predict antenna performance and structural optimization,as follows:Firstly,a 5G tri-band microstrip antenna suitable for multiple application scenarios was proposed,with the main radiation elements being circular patch,Y-shaped strip with sector patch,and a single-pole impedance converter inserted inside the circular patch to adjust the operating frequency band.Through experimental analysis,the simulation and measured results of the antenna are relatively consistent,and the antenna can operate in the frequency range of2.38～2.53 GHz,3.29～4.11 GHz,and 4.72～5.01 GHz,with a maximum gain of 4.09 d Bi,providing omnidirectional radiation,appropriate gain,and sufficient bandwidth.Secondly,the SSA-SVR algorithm was used to predict the performance of a multi-frequency microstrip antenna,modeling with support vector regression,and optimizing the model using the sparrow search algorithm.The antenna is a tri-band single-pole microstrip patch antenna with defects,and the main radiation unit consists of a rectangular ring and a single-pole rectangular patch connected inside,with a defective rectangular ground plane with a circular arc on the back of the dielectric substrate.The performance prediction analysis is divided into two parts:the first part models multiple performance parameters of the antenna at the second frequency band（including resonant frequency,bandwidth,radiation efficiency,and standing wave ratio）as labels and compares multiple regression machine learning algorithms.The second part models and analyzes S₁₁ and VSWR in the continuous frequency range of 1～7 GHz,and compares the performance of other regression machine learning algorithms.Through experimental verification and comparison,the effectiveness of the proposed algorithm in predicting the performance of microstrip antennas is proven.Finally,to solve the problem of antenna designers needing to invest a lot of effort in the antenna design process and solve the non-linear mapping problem between multiple structural parameters and performance of antennas.The WSO-MLS-SVR algorithm was used to optimize the structure of the microstrip antenna.Firstly,a multi-output least squares support vector machine was used to analyze the structural optimization of the microstrip antenna,and the White Shark Optimizer algorithm was used to optimize the hyperparameters of the multi-output least squares support vector machine.At the same time,using multiple validation indicators and comparing with other multi-output machine learning regression algorithms,the effectiveness of the algorithm is proven.To further demonstrate the practicality of the proposed algorithm,a rectangular multi-frequency microstrip antenna was optimized using Matlab-Python-HFSS joint simulation,adjusting the resonance arm to meet the needs of 5G application scenarios.Through comparative analysis,this study obtained the target performance of the antenna in relatively few simulation analysis times using the WSO-MLS-SVR algorithm,and finally analyzed the performance of the optimized antenna.