In Computational Electromagnetics, iterative techniques for solving algebraic linear system of equations are of fundamental importance, since actual problems give rise to linear systems too large to be practically solved by direct methods. In this work we investigate as performances of the major Krylov subspace iterative solver (i.e., GMRES), is affected by different choice of these set of functions. Specifically, we consider the algebraic linear system of equations obtained by reducing the electrical field integral equation (EFIE) from the TMz scattering of a plane wave by a metallic strip. It can be observed that exists a critical threshold ?0 such that, whenever either the basis or the weight pulses are given with an amplitude greater than ?0, then the total number of internal loops necessary for taking the relative residual under a definite tolerance ε>0 increases all of a sudden, in such a dramatic way that it can even prevent the process at all from convergence. We try to explain this numerical behavior by inquiring the relationship between the MoM matrix condition number and the number of overall iterations necessary to numerical convergence

Convergence of Krylov Solvers and Choice of Basis and Weighting Set of Functions in the Moment Method Solution of Electrical Field Integral Equation

ANGIULLI, Giovanni
;
2008-01-01

Abstract

In Computational Electromagnetics, iterative techniques for solving algebraic linear system of equations are of fundamental importance, since actual problems give rise to linear systems too large to be practically solved by direct methods. In this work we investigate as performances of the major Krylov subspace iterative solver (i.e., GMRES), is affected by different choice of these set of functions. Specifically, we consider the algebraic linear system of equations obtained by reducing the electrical field integral equation (EFIE) from the TMz scattering of a plane wave by a metallic strip. It can be observed that exists a critical threshold ?0 such that, whenever either the basis or the weight pulses are given with an amplitude greater than ?0, then the total number of internal loops necessary for taking the relative residual under a definite tolerance ε>0 increases all of a sudden, in such a dramatic way that it can even prevent the process at all from convergence. We try to explain this numerical behavior by inquiring the relationship between the MoM matrix condition number and the number of overall iterations necessary to numerical convergence
2008
Computational Electromagnetics; Krylov Solvers; Method of Moments
File in questo prodotto:
Non ci sono file associati a questo prodotto.

I documenti in IRIS sono protetti da copyright e tutti i diritti sono riservati, salvo diversa indicazione.

Utilizza questo identificativo per citare o creare un link a questo documento: https://hdl.handle.net/20.500.12318/8075
Citazioni
  • ???jsp.display-item.citation.pmc??? ND
  • Scopus 0
  • ???jsp.display-item.citation.isi??? ND
social impact