Analysis of an aggregation‐based algebraic two‐grid method for a rotated anisotropic diffusion problem

Abstract

Summary A two‐grid convergence analysis based on the paper [Algebraic analysis of aggregation‐based multigrid, by A. Napov and Y. Notay, Numer. Lin. Alg. Appl. 18 (2011), pp. 539–564] is derived for various aggregation schemes applied to a finite element discretization of a rotated anisotropic diffusion equation. As expected, it is shown that the best aggregation scheme is one in which aggregates are aligned with the anisotropy. In practice, however, this is not what automatic aggregation procedures do. We suggest approaches for determining appropriate aggregates based on eigenvectors associated with small eigenvalues of a block splitting matrix or based on minimizing a quantity related to the spectral radius of the iteration matrix. Copyright © 2015 John Wiley & Sons, Ltd.

Convergence theory of exact interpolation scheme for computing several eigenvectors

“…We apply this algorithm recursively to obtain quadruples of nodes and groups of eight strongly connected nodes. Although this algorithm may not be optimal, as reported in , we show that it may yield better results than GC if the problem possesses a strong anisotropy, as in Example . The odd number of nodes along each coordinate direction is involved in every coarse basis vector when we use GC, while the even number of nodes is aggregated in AC.…”

Advanced Algebraic Multigrid Solvers for Subsurface Flow Simulation

“…Firstly one can improve the interpolation quality so that it represents the direction of the anisotropy at the coarse levels. For aggregation-based algebraic multigrid , the theoretical analysis in Chen et al (2015) shows that if the aggregates are aligned with the direction of anisotropy, the convergence rate can be improved. The second possible remedy is to use ILU type smoother, which has been a powerful smoother in multigrid framework (Stuben, 2000, Trottenberg et al, 2000.…”

The transport of nanoparticles in subsurface with fractured, anisotropic porous media: Numerical simulations and parallelization

“…[25] proposed an improved aggregation strategy to improve the convergence. In [25] a new automatic aggregation algorithm determines appropriate aggregates based on eigenvectors associated with small eigenvalues of a block splitting matrix. The generated aggregates by this approach mostly are lines and aligned with the direction of the rotated anisotropy.…”