A PVM based parallel sparse matrix equation solver to speed up computation of MEI method

摘要

The conventional method of moments is capable of solving wave scattering problems of circumferential dimension of 100 wavelengths. Using the MEI (measured equation of invariance) method the problem of a cylinder of 10,000 wavelengths is within the storage capacity of a personal computer (PC). The MEI method requires the generation of a large sparse matrix for the boundary equations. The MEI matrix coefficients are computed from numerical integrations of the metrons. Interpolation and extrapolation techniques can be employed to save integration time. This means that the solution of the sparse matrix equation of the MEI method is the bottleneck of computation. Since the MEI equation matrix is a sparse matrix with nonzero diagonal elements and very small number of nondiagonal elements at the upper right and lower left corners for 2 dimensional problems, a matrix decomposition algorithm derived by Chen (1973) can be employed for the purpose of decomposition and parallel computing (using parallel virtual machine-PVM).