科研之友
微信
新浪微博
Facebook
分享链接
摘要
The increase of the time step size significantly deteriorates the property, of the coefficient matrix generated from the Crank-Nicolsonfinite-difference time-domain (CN-FDTD) method. As a result, the convergence of classical iterative methods, such as generalized minimal residual method (GMRES) would be substantially slowed down. To address this issue, this article mainly concerns efficient computation of this large sparse linear equations using preconditioned generalized minimal residual (PGMRES) method. Some typical preconditioning techniques, such as the Jacobi preconditioner, the sparse approximate inverse (SAI) prcconditioner, and the symmetric successive over- relaxation (SSOR) preconditioner, are introduced to accelerate the convergence of the GMRES iterative method. Numerical, simulation shows that the SSOR preconditioned GMRES method can reach convergence five times faster than GMRES for some typical structures.
