Calculate The Parallel Method Of Moments In Electromagnetics And The Realization Of The Cluster Environment

Because of the requirement of engineering, the analysis of electrical large objects becomes increasingly interesting. However, as the electrical dimension of the target increases, the memory requirement and CPU time are barriers to the solution by a single PC. To overcome this problem, parallel computing was employed in the computation of electromagnetics. Recently, clusters with distributed memory are popular for parallel computing systems, and MPI is the actual standard of message passing programming. Therefore, the application of clusters based on MPI to computational electromagnetics is studied in this dissertation. The methods are MoM based on the integral equation method in conjunction with the preconditioning technique to speed up the iterative solution and a matrix order reducing method, namely the Characteristic Basis Function Method (CBFM).To solve an electrically large problem with clusters, the parallel solution procedure is very complex and the demands of the users'ability of development and application are very critical, such as the complex system and software environment, parallel programming and debugging. The parallel MoM is developed in a cluster called lenovo Deepcom 1800 in this research, and the result shows a good speed-up performance. The Electrical Field Integral Equation (EFIE) and the RWG basis function are used. The impedance matrix is decomposed by the row, stored in distributed memory of the nodes. Besides, the processors exchange messages through the MPI communication library. The numerical results indicate the accuracy and the high performance of the method. Because the memory requirement and the computation complexity of the iterative solver are both O(N2), where N is the number of unknowns, as the unknowns increase, the memory requirement of MoM solution turns so huge that it becomes the key barrier to the overall solution. To accelerate the iterative solution and reduce the iterations, a physically based near-field preconditioner is employed and parallelized. The numerical results show a good preconditioning performance.Notifying that the selection standard of the near field preconditioner is physically based, we make some modifications and design a novel data-based preconditioner,...