Premium
A Note on the Global Extraction Element‐by‐Element Method
Author(s) -
Kickinger F.,
Langer U.
Publication year - 1998
Publication title -
zamm ‐ journal of applied mathematics and mechanics / zeitschrift für angewandte mathematik und mechanik
Language(s) - English
Resource type - Journals
SCImago Journal Rank - 0.449
H-Index - 51
eISSN - 1521-4001
pISSN - 0044-2267
DOI - 10.1002/zamm.19980781553
Subject(s) - preconditioner , mathematics , finite element method , linear subspace , matrix (chemical analysis) , smoothing , stiffness matrix , eigenvalues and eigenvectors , combinatorics , mathematical optimization , iterative method , pure mathematics , physics , statistics , materials science , quantum mechanics , composite material , thermodynamics
In the engineering literature, the so‐called Global Extraction Element‐by‐Element (GE‐EBE) Methods have been successfully used for solving real‐life mechanical problems, especially, in 3D [2], although the EBE preconditioners C haven't any asymptotically preconditioning effects on the stiffness matrix K, i. e. the spectral condition number k(C ‐1 K) behaves like O(h ‐2 ), say, for elasticity problems. However, the method has quite good scaling properties. From a theoretical point of view, the GE‐EBE preconditioner is nothing else than an Additive Schwarz Method (ASM) preconditioner generated by an overlapping splitting of the finite element space into subspaces spanned by the element ansatz functions. The authors propose to use the GE‐EBE methods and their patch counterpart (GE‐PBP) either as smoothers, especially, in their multiplicative version, or as preconditioners in their multilevel additive version. In the multilevel version, such preconditioners are asymptotically optimal. The GE‐PBP smoother is especially suited for the use in Algebraic Multi‐Grid (AMG) Methods based on matrix graph coarsening strategies. The numerical results presented confirm the good smoothing properties of GE‐PBP smoothers as well as the preconditioning properties of multilevel additive GE‐PBP preconditioners. Finally, the authors discuss parallelization aspects.