GPU-accelerated SVD-CBFM And Its Application In Electromagnetic Scattering Analysis

The analysis of target scattering characteristics plays a vital role in today's electromagnetic research.With the increasing demand for electromagnetic parameters in the military and daily life,more and more electromagnetic algorithms have begun to appear.The method of moments(MoM)is the most basic and representative algorithm of numerical algorithms.With the increase of electrical size,in order to solve the problem of excessive time and memory,the singular value decomposition-based characteristic basis function method(SVD-CBFM)was developed on the basis of the MoM.In recent years,with the rapid development of computers,the parallel computing capabilities of graphics processors(GPU)have also received widespread attention.This thesis takes SVD-CBFM as the research basis for this topic,and performs GPU-accelerated calculations on SVD-CBFM based on RWG basis functions and SVD-CBFM based on SWG basis functions,respectively.And their applications in electromagnetic scattering analysis were studied.The main work of this thesis is divided into three parts.The first part is the basic theory and method of this thesis.First,the electromagnetic integral equations are derived from the Maxwell equations,and the surface and volume integral equations are introduced,respectively.Then the basic principles of the MoM,basis functions and test functions,and radar cross section(RCS)are introduced in turn.Finally,the GPU and compute unified device architecture(CUDA)are illustrated,which include the development history,hardware structure of GPU,CUDA programming model,thread hierarchy and storage hierarchy.The second part proposes GPU acceleration of SVD-CBFM based on RWG basis functions,and studies its application in electromagnetic scattering analysis.Firstly,the RWG basis function is introduced,and the surface integral equation is discretized with the RWG basis function.Then combined with SVD-CBFM,it is applied to analyze the scattering from the conductor target.Next we use the GPU to accelerate it,not only the filling of the impedance matrix,but also the matrix multiplication in the reduced matrix filling.The third part proposes GPU calculation of SVD-CBFM based on SWG basis functions,and also studies its application in electromagnetic scattering analysis.First,the SWG basisfunctions and SVD-CBFM based on the SWG basis functions are introduced.Then for the SVD-CBFM of the SWG basis function,we implemented its parallel computing acceleration on the GPU.Through CUDA,GPU acceleration is performed on the matrix multiplication part of the impedance matrix filling part and the reduced matrix filling.And the GPU-based LU decomposition algorithm is proposed.In this thesis,electromagnetic scattering analysis is performed on the conductor target and the dielectric target,and the efficiency and accuracy of the proposed method are verified through several numerical examples.Compared with the traditional SVD-CBFM,the numerical results obtained by the numerical examples are consistent and have an extremely high acceleration ratio.