Research On Efficient Optimization Algorithms For Antennas And Microwave Components

With the development of 5G communication technology entering a new era,the demand for high performance antennas and microwave components is greatly increased.The market is expected to exceed 100 billion RMB.The development of miniaturized,lightweight,and highly integrated antennas and microwave components is highly desirable.Efficient optimization algorithms for antennas and microwave components are the key to the fast design of new generation RF microwave device.Taking the design of new generation antenna and microwave devices as the application background,this thesis systematically studies direct optimization algorithm,surrogate-assisted optimization algorithm,surrogate-assisted yield optimization algorithm and test suite of optimization algorithm evaluation.The specific contents include the following aspects.The contents of this thesis include:(1)Two adaptive differential evolution algorithms are studied for antenna arrays.Firstly,a three-phase adaptive differential evolution(TADE)algorithm is proposed for antenna array synthesis problems.In the TADE algorithm,the evolution process is divided into three phases(namely,the initial phase,the stable phase,and the precise phase)according to the population fitness distribution and iterative information.Based on the statistical information of population fitness values,a suitable mutation strategy for each phase is designed to enhance the search ability.Secondly,we propose a novel differential evolution optimization method named K-means-based multi-group differential evolution(KMGDE)algorithm for antenna array design.The KMGDE algorithm divides the population into three groups using the K-means clustering method in each iteration.We develop a mutation strategy plan that assigns different mutation strategies for each group according to its average fitness value.The fitness function considering the characteristics of the antenna array is proposed to speed up optimization.The performance of the KMGDE and the TADE is verified and analyzed using test functions and antenna array synthesis and design examples.(2)A novel surrogate-assisted Quasi-Newton enhanced optimization algorithm is proposed for antenna design.In this proposed method,the heuristic hypersphere sampling method is used to obtain representative samples.The surrogate model is built based on the low-fidelity model.The Quasi-Newton enhanced differential evolution method is designed to optimize the surrogate model.Finally,the optimal design of a high-fidelity model is obtained through a space mapping procedure.The proposed algorithm is verified through two antenna design examples including a dipole antenna with balun and a SIW cavity-backed slot antenna.The results show that the proposed algorithm finds a more accurate minimum value with less computational time than direct optimization using differential evolution.(3)A novel global surrogate-assisted optimization is proposed for filter design.Because microwave filters use electromagnetic simulation for design optimization.There are some problems,such as high optimization cost,low design efficiency and long design cycle.In this thesis,a new global alternative model aided optimization algorithm is proposed.The greedy sampling method is used to obtain high-quality samples as the input of modeling data,and the vector fitting method is used to obtain the poles and residues of transfer function as the output of modeling data,The radial basis function neural network is used to establish the input and output surrogate model,and the efficient optimization algorithm is used to optimize the surrogate model to obtain the optimal design parameters.The results of two band-pass filter examples verify the performance of the algorithm.(4)An efficient yield-constrained optimization using a polynomial chaos surrogate considering multiple objectives is proposed for microwa ve filter design.Yield optimization aims at finding microwave filter designs with maximal yield satisfying a certain performance objective under fabrication tolerance.The EM-based yield optimization methods are very expensive because a large number of EM simulations are required.Moreover,microwave filter design usually requires several performance objectives to be met at the same time,which is not considered by the current method.Two contributions of the proposed method are(a)the cost and accuracy of the polynomial chaos model are improved by a sampling reduction strategy and an adaptive order determination strategy;(b)An efficient yield-constrained design framework is implemented to obtain the optimal design solutions under yield constraints.The results of three microwave filter design examples verify the effectiveness and efficiency of the proposed algorithm.(5)We propose a method for designing a test suite for antenna design problems.Performance evaluation of antenna optimization algorithms is essential.It is desirable to develop an efficient test suite with characteristics of typical antennas to replace the high-computational-cost EM-based evaluation.Four typical antennas with respect to four typical objectives based on EM-simulated S-parameters are systematically investigated.Based on the investigation results,we propose a versatile performance-evaluation test suite allowing efficient benchmarking for antenna S-parameter optimization algorithms.The test suite consists of eight analytical functions.Each function matches a typical landscape of the EM-simulated antenna problems.The performance of six widely used global optimization algorithms are evaluated using the test suite.The results using the proposed test suite for all six optimization algorithms are consistent with those of the real EM-based antenna design problems.The time cost for performance evaluation using the test suite is only a fraction of the cost using the real EM-based antenna design problems.