Optimal Design Of Microwave Devices By PSO Algorithm And Gaussian Process

The traditional microwave device optimization method which is electromagnetic simulation software combining with optimization algorithm is very inefficient.Therefore,improving the design efficiency of microwave devices is a hot topic.In response to this problem,researchers have proposed many solutions,such as Artificial Neural Network(ANN),Support Vector Machines(SVM),and Gaussian Process(GP).This paper mainly introduces the fitness inheritance method,that is,the fitness of the child inherits the fitness of the parents in a certain way.These methods can not only avoid the time cost of sampling,but also maintain the excellent performance of the algorithm on the basis of greatly reducing the number of calculations of the true fitness.In this thesis,the particle swarm optimization(PSO)algorithm is combined with the GP modeling method to optimize the various antennas and filters.The main research work is as follows:(1)The thesis introduces the basic principles of PSO algorithm and Gaussian process,and describes the implementation of GP modeling and the method of Matlab calling HFSS.(2)The thesis studies the method of fitness estimation based on particle swarm optimization algorithm(FEPSO).The prediction model of the fitness of the particle is constructed according to the explicit evolution formula of PSO algorithm.Therefore,the fitness of the particle can be given by the prediction model and the prediction result replaces the time-consuming full-wave electromagnetic simulation.The effectiveness of the proposed method is proved by the optimal design of the Yagi microstrip antenna(MSA)and the SIR bandpass filter.(3)In order to make the prediction result of fitness estimation method more stable and the applicable more widely,the thesis introduces a Self-updating Fitness Estimation method based on PSO algorithm(SFEPSO).During the process of optimization iteration,the prediction model will be proofed every several generations.If the prediction model accuracy is lower than the threshold,the prediction model will be updated and continue iterating until the particle reach the constraint.The effectiveness and stability of the method are validated by the optimal design of E-shaped dual-band microstrip antenna and WLAN/WiMAX multi-band microstrip antenna.(4)The thesis introduces a method of GP model exploiting Fitness-Estimation-based(FEGP)PSO algorithm.When constructing GP model of microwave antennas,some of training data comes from precise simulation values of full-wave electromagnetic simulation software,and some of the training data comes from the fitness-estimation-based PSO algorithm.The modeling time is obviously decreased.The proposed method is verified by optimizing an inverted-F antenna and a GPS and Beidou dual-mode microstrip antenna,which proves its high efficiency and accuracy.