Research On Port And Parallel Computing Technology In Discontinuous Galerkin Time Domain Method

In recent decades,frequency-domain methods such as finite element methods have been intensively studied by researchers in computational electromagnetics,which can be quickly calculated to obtain accurate electromagnetic results.However,the frequency-domain approach requires a lot of recalculation of frequency points when target wide band characteristics are required,and the time domain approach can solve this problem well.However,the traditional time-domain finite element needs to solve a large coefficient matrix equation system at each time step,so the time-domain discontinuous Galerkin method with discontinuous numerical flux is a good research direction.Time-domain discontinuous Galerkin method is a combination of discontinuous Galerkin method and time-domain finite element method.This method can obtain mass matrices with small dimensions,each small matrix represents a spatial unit,so it has natural parallel characteristics.In this thesis,based on the Maxwell's curl equations,a flexible tetrahedral grid element is used in space,a vector basis function is used in basis function selection,and a leapfrog method is used in time stepping.According to the different numerical flux introduced at the unit interface,the time-domain discontinuity Galerkin method can be divided into the central flux time-domain discontinuity Galerkin method and the upwind flux time-domain discontinuity Galerkin method,and compare the advantages and disadvantages of upwind flux and central flux.Secondly,this thesis proposes to use the quick sorting method to solve the preprocessing mapping relationship of the tetrahedral mesh,which greatly reduces the preprocessing preparation time.For the calculation of microwave waveguides,the total field / scattering field boundary condition is introduced to mode excitation.The uniaxial perfect matching layer is used as the absorption boundary condition,and the first-order absorption boundary condition is added to the outermost layer of the matching layer to better absorb incident waves.In addition,this thesis discusses the field distribution of the waveguide mode at the excitation surface,introduces the coaxial port and the rectangular port,and studies and verifies the characteristics of the uniaxial perfectly matched layer.Secondly,it verifies and discusses the port scattering parameter extraction method.As the scale of computing continues to increase,the demand for memory is increasing,and the requirements for computing time are also increasing.Parallel technology can be used to solve this problem.Therefore,the time-domain discontinuous Galerkin method is designed in parallel to solve large-scale electromagnetic problems.In this thesis,MPI technology is used to realize the parallelism of the time-domain discontinuous Galerkin method,and the parallel design process of the method is introduced in detail.Finally,an example verifies the correctness and effectiveness of the parallel program.