Study On Inverse Problems In Electromagnetics Based On Genetic Algorithm And Parallel Computing

Inverse problems in electromagnetics as know as optimization of electromagne-tic equipments is always the hot issue since it's advanced in electromagnetics application fileds. By virtue of developments in computer science and electromagneti-cs methods of numerical analysis, inverse problems in electromagnetic has a wonderful prospect.This paper discusses the inverse problems in electromagnetic particularly,then takes team workshop problem 22 as typical issue.We use the finite element method to calculate the fuction fitness of problem 22. As a population-based algorithm, genetic algorithm, on the one hand, has intrinsic parallel characteristics, which is perfect matched for parallel computation; on the other hand, infeasible local search ability caused by premature make solutions easily trapped in the local optimum. Considering above advantages and disadvantages, an improved genetic algorithm with adaptive adjoint population is proposed to circumvent premature and improve search ability.Te-sting proposed algorithm with typical mathematical problems,it achieves better solution and less running-time steadily.Then use the improved algorithm to optimize SMES system and achieve a better optimization result.Generally, inverse problems in electromagnetics are known to have expensive evaluations and time-consuming iteration process.In order to High-performance calculation of Problem 22,we introduce parallel genetic algorithm. The parallelization of the proposed algorithm is implemented by one-level master-slave parallelization and two-level one, respectively. Programming base on C and MPI in Linux operation system, simulation results illustrate parallel genetic algorithm get more faster optimiz-ation speed,which explaines the method of this paper has not only great theoretical meaning but also engineering application value.