Research On Microwave Filter And Antenna Design Method Based On Extreme Learning Machine

With the rapid development of modern wireless communication technology,there is an increasing requirement of microwave passive components in communication system.In order to meet the different requirements of communication system,it is necessary to design and optimize the microwave passive components.It has become an inevitable trend to use algorithms to design and optimize the microwave passive components aiming at improving the design efficiency of microwave passive components and reducing the time costs.Machine learning is a simple and fast method to establish mapping relationship,but it needs a large number of training samples,which leads to high time complexity in practical application.In this paper,we study the design methods of microwave filters and antennas based on extreme learning machine(ELM)aiming at improving the design efficiency of microwave filters and antennas.The main works are summarized as follows:(1)The theories of extreme learning machine,space mapping and brain storm optimization algorithm(BSO)are studied.The principles and implementation processes of the three algorithms are described in detail,and the opposition learning strategy is introduced to improve the convergence speed and population quality of brain storm optimization algorithm.(2)A design method of microwave filters based on extreme learning machine and space mapping is proposed.The coarse model and the fine model of microwave filters are established by space mapping.The mapping relationship between the design parameters of the coarse model and the fine model is trained by extreme learning machine.In this way,the optimization is carried out in the coarse model,and the fine model only verifies the optimization results,which effectively reduces the time and labor costs of the design of microwave filters,and improves the design efficiency.The design examples of ridged substrate integrated waveguide(SIW)filter and low temperature co-fired ceramic(LTCC)filter show that the method is universal and feasible.(3)A design method of low-dimensional parameters microwave antennas based on improved extreme learning machine is proposed.The input weights and thresholds of extreme learning machine are optimized by brain storm optimization algorithm.The design parameters of the microwave antennas are predicted according to the target response after establishing the mapping relationship between the design parameters and the responses of the microwave antennas.In this way,the number of training samples can be reduced and the prediction accuracy of the extreme learning machine can be improved.The design examples of dielectric resonator antenna and low profile magnetoelectric dipole antenna show that the method has the advantages of reducing time costs and improving design efficiency.(4)A design method of multi-dimensional parameters microwave antennas based on latin hypercube sampling and improved extreme learning machine is proposed.Latin hypercube sampling(LHS)is used to obtain a number of antennas' design parameters,which solves the problem of obtaining training samples in the training process of extreme learning machine.At the same time,HFSS-Matlab-Api is used to realize the automatic operation of the design process,which improves the design efficiency of the microwave antennas and reduces the labor costs.The design example of stacked microstrip antenna shows that the method is effective.