DocumentCode :
418616
Title :
Modeling large almost periodic structures using a non-overlapping domain decomposition method
Author :
Vouvakis, Marinos ; Zhao, Kezhong ; Lee, Jin-Fa
Author_Institution :
ECE Dept., Ohio State Univ., Columbus, OH, USA
Volume :
1
fYear :
2004
fDate :
20-25 June 2004
Firstpage :
343
Abstract :
Domain decomposition is a tool that is introduced artificially to ease large scale computation or that is natural in some situations. Domain decomposition methods are a very natural way to exploit the possibilities of multiprocessor computers, but such algorithms are very useful even when used on single process PC environment. The idea is to decompose the computational domain into smaller subdomains. The equations are solved on each subdomain. We first describe a domain decomposition method, a non-overlapping Schwarz algorithm, for solving large almost periodic electromagnetic problems. For large almost periodic structures, when using domain decomposition approach, the FEM matrix only needs to be evaluated once for all domains that are of the same building block. The solving process of domain decomposition is iterative and can become prohibitively slow. We describe a procedure that results in superior speed-up of the solving process.
Keywords :
computational complexity; computational electromagnetics; decomposition; finite element analysis; iterative methods; matrix algebra; periodic structures; FEM matrix; almost periodic structures; computational domain; electromagnetic problems; iterative process; large scale computation; matrix equation; multiprocessor computers; nonoverlapping Schwarz algorithm; nonoverlapping domain decomposition method; single process PC; solving process; Boundary conditions; Concurrent computing; Convergence; Iterative algorithms; Laplace equations; Large-scale systems; Mesh generation; Modems; Partitioning algorithms; Periodic structures;
fLanguage :
English
Publisher :
ieee
Conference_Titel :
Antennas and Propagation Society International Symposium, 2004. IEEE
Print_ISBN :
0-7803-8302-8
Type :
conf
DOI :
10.1109/APS.2004.1329643
Filename :
1329643
Link To Document :
بازگشت