The Design And Optimization Of Dielectric Waveguide Filters

With the development of communication systems,the design requirements for microwave devices are increasing.This is not only reflected in more stringent device indicators,but also poses a huge challenge to designers.How to design microwave devices quickly and accurately is extremely important.The optimization theory is widely used in the optimization design of electromagnetic devices.Traditional optimization algorithms have many defects.The swarm intelligence optimization algorithm proposed in recent years has a good optimization effect,so it has been widely used in the optimization problem of microwave devices and has achieved good results.This thesis mainly focuses on the application of different Swarm Intelligence(SI)algorithms in the field of microwave filter designs.Basically,different SI algorithms is used and proposed to improve the efficiency of microwave filters.Additionally,SI algorithm is combined with the machine learning to improve the efficiency of the optimization process.The specific work is as follows:1.A novel updating strategy,random current optimal boost(RCBB)strategy for SI algorithms,is proposed and combined with the standard flower pollen algorithm(FPA),which yields a novel improved FPA,i FPA.From the view of optimization,the synthesis of(N)-type and(N+2)-type coupling matrix is solved by the proposed algorithm.The experimental results show that compared with conventional methods,i FPA solves the synthesis problem shows its versatility and simplicity.In addition,compared with other existing state-of-art algorithms,i FPA exhibits the advantages of fast convergence and higher optimization accuracy in the considered problems.2.Based on a novel swarm intelligence algorithm,Harris Hawks algorithm(HHO),the optimization design of dielectric waveguide filters is investigated.First,the algebraic method is used to analyze the various basic update strategies of the HHO algorithm in detail to investigate the convergence behavior of the algorithm.Secondly,the HHO algorithm is used to solve the optimization problem of dielectric waveguide filter design to improve the convergence speed,efficiency and accuracy of the filter optimization design.Two dielectric waveguide filters are optimized and designed based on the HHO algorithm.The experimental results show that the convergence speed,efficiency and accuracy of the HHO algorithm are significantly better than the adaptive differential evolution algorithm,flower pollination optimization algorithm and gray wolf optimizer that are currently mainstream applications in the industry.3.Based on Extreme Learning Machine(ELM),a novel algorithm is proposed and applied to the establishment of the parameterized model of the microwave filter.In order to reduce the performance fluctuation of the standard ELM,multiple isomorphic sub-networks are generated in combination with the concept of ensemble learning,and the sub-networks are trained and evaluated through the bootstrapping method.The stability and predictive ability of the integrated network is improved by the combination of all these sub-networks.The differential evolution algorithm is used as a search engine to automatically obtain network training data.After the training network is obtained,the adaptive differential evolution algorithm is used to interact with the network to complete the optimization process.In order to verify the performance of the algorithm,the proposed algorithm is used to model and optimize a microstrip filter.For comparison,ELM,E-ELM and direct electromagnetic simulation optimization are considered to finish the optimization of filters.The experimental results show that the ensemble ELM network is more effective.Compared with the standard ELM,its performance is better,and the mapping of the physical structure can be extracted accurately.In addition,in these methods,the integrated network obtains the best optimization results,and the measurement results also prove the effectiveness of the proposed algorithm.