Title :
ACA method and SPAI preconditioner for acceleration BICGSTAB solver in MoM based solver TriD
Author :
Gabriadze, G. ; Bogdanov, F. ; Tsereteli, P. ; Jobava, R. ; Yavolovskaya, E.
Author_Institution :
EMCoS Ltd., Tbilisi, Georgia
Abstract :
In this paper effective acceleration technique for solving of electrically large EMC problems is described. It is known that for solving of large linear systems iterative solvers are very practical. The main problems for solution of such problems are related to large memory requirement for matrix and poor convergence of iterative solver. For solution of these problems the following techniques are applied: ACA (Adaptive Cross Approximation) technique for matrix compression and SPAI (Sparse Approximate Inverse) preconditioner to speed up convergence of iterative solver. For solving linear system BICGSTAB (Biconjugate Gradient Stabilised) solver is used. The advantages of considered solution are demonstrated using several numerical experiments.
Keywords :
approximation theory; conjugate gradient methods; electromagnetic compatibility; iterative methods; method of moments; ACA technique; MOM; SPAI preconditioner; acceleration BICGSTAB solver technique; acceleration biconjugate gradient stabilised solver technique; adaptive cross approximation technique; electrically large EMC problem; large linear systems iterative solver; large memory requirement; matrix compression; solver TRID; sparse approximate inverse preconditioner; Approximation methods; Computers; Convergence; Linear systems; Matrix decomposition; Moment methods; Sparse matrices;
Conference_Titel :
Direct and Inverse Problems of Electromagnetic and Acoustic Wave Theory (DIPED), 2011 XVth International Seminar/Workshop on
Conference_Location :
Lviv
Print_ISBN :
978-1-4577-0897-8
