RUS  ENG
Full version
JOURNALS // Zhurnal Vychislitel'noi Matematiki i Matematicheskoi Fiziki // Archive

Zh. Vychisl. Mat. Mat. Fiz., 2025 Volume 65, Number 7, Pages 1118–1142 (Mi zvmmf12008)

General numerical methods

Adaptive block algebraic multigrid method for multiphysics problems

I. N. Kon'shinabc, K. M. Terekhovad

a Marchuk Institute of Numerical Mathematics of the Russian Academy of Sciences, Moscow
b Federal Research Center "Computer Science and Control" of Russian Academy of Sciences, Moscow
c I. M. Sechenov First Moscow State Medical University
d University of Science and Technology "Sirius", Sochi

Abstract: We propose the adaptive block algebraic method to solve the multiphysics problems arising from the collocated finite volume discretization methods. The method is specifically designed to solve multiphysics problems featuring various physics in various parts of the domain, resulting in block-structured saddle-point linear algebraic systems with variable block size. The adaptive algebraic multigrid method uses available information on the eigenvectors of the problem to construct prolongation and restriction operators. The information on the distribution of degrees of freedom within the blocks to form an initial set of vectors is used. It was shown that the arising linear systems are amenable to the solution with the proposed method. Various approaches to strong point selection, coarse space refinement, and bootstrapping the test vectors are discussed and analysed. In this work, we address the systems arising from coupled problems of free-flow and poroelasticity, frictional rigid body contact mechanics, and poroplasticity with fractures. All of the problems are of saddle-point nature.

Key words: sparse linear system, block multigrid method, saddle-point problem.

UDC: 519.615

Received: 02.04.2025
Accepted: 23.04.2025

DOI: 10.31857/S0044466925070055


 English version:
Computational Mathematics and Mathematical Physics, 2025, 65:7, 1495–1519

Bibliographic databases:


© Steklov Math. Inst. of RAS, 2026