Research On Optimization Of Antenna Structure Based On Data-driven And Electromagnetic Computation

With the development of wireless communication technology,the antenna with the function of transmitting and receiving electromagnetic wave is becoming more and more important.However,the traditional trial-and-error method which is driven by design experience gradually becomes difficult to meet the requirements of antenna design.In order to solve this problem,two kinds of antenna design schemes are proposed,namely,data-driven surrogate model and structure design based on electromagnetic computation.The surrogate model method can greatly reduce the operation cost of the original model by fitting the original model.For the antenna design,the surrogate model is constructed by parameterizing the important dimensions of the antenna structure,which can exactly fit the original model.In this paper,firstly,machine learning is chosen as the method of constructing surrogate model by comparing several methods.Then,a multi-layer surrogate model method based on ensemble learning is proposed.Finally,the effectiveness of the proposed multi-layer surrogate model in antenna design is demonstrated by comparing the actual antenna structure with the one designed by the surrogate model.The design of antenna by data-driven model needs a lot of data preparation and the choice of surrogate model construction method to achieve a good surrogate for the original model.In addition,in the design of the modified antenna structure,the parametric antenna structure can not be used for data-driven modeling.Therefore,an antenna structure design scheme based on electromagnetic computing is proposed to solve this antenna design problem.Firstly,the key matrices Z,I and V of the antenna structure are constructed by the method of moment(MoM).Then,the structure of the antenna is modified by combining the removal of the basis function and the genetic algorithm to realize the target antenna performances.Finally,the effectiveness of the design scheme is illustrated by the structural optimization of the half-wave dipole antenna and the butterfly antenna.