Download as docx, pdf, or txt
Download as docx, pdf, or txt
You are on page 1of 1

14th World Congress in Computational Mechanics (WCCM)

ECCOMAS Congress 2020


19 – 24 July 2020, Paris, France

USING THE MULTISCALE HYBRID MIXED METHOD (MHM) AS A


PRECONDITIONER

Jorge Ordonez¹, Philippe R. B. Devloo2, José B. Villegas S.1 and Omar Y. Durán3
1
Mechanical Engineering Department (FEM) - University of Campinas (Unicamp), 13083-839,
Campinas-SP, Brazil, www.labmec.org.br
2
Civil Engineering Department (FEC) - University of Campinas (Unicamp), 13083-839, Campinas-
SP, Brazil, www.labmec.org.br
3
Université Paris-Est, CERMICS (ENPC), 77455 Marne-la-Vallée cedex 2, and INRIA-Paris, 75589
Paris - France
Key Words: Elliptic boundary value problem, Iterative method, Preconditioning, Multigrid
technique, Multiscale Method.

Approximations of the Laplacian by the mixed finite element method on a very fine meshes
leads to a very large computational burden due to the indefinite symmetric system of
algebraic equations. This work is intended to be used as an improvement when using the finite
elements method. Moreover, the proposed method would reduce significatively the amount of
memory that the system of equations required by preconditioning the matrix that needs to be
solved. The mathematic implementation is made using a residual transfer from meshes (full
mesh to MHM mesh), a correction transfer (MHM mesh to full mesh) and a block iterative
process. The operations can be performed at the algebraic level or based on integrations of
function spaces. The advantage of using integration is that those operations fit in a framework
of approximating nonlinear differential equations. Furthermore, implementing the method
from the algebraic point of view is that the structure of the operations is easier to understand.
The C++ computational implementation is made on the context of the NeoPZ finite elements
object-oriented library.

REFERENCES

[1] Pearson, J. W., & Wathen, A. J. (2012). A new approximation of the Schur
complement in preconditioners for PDE‐constrained optimization. Numerical
Linear Algebra with Applications, 19(5), 816-829.

[2] Simon, R. and Zulehner, W. Multigrid Solvers for Saddle Point Problems in
PDE-Constrained Optimization. Johannes Kepler Universitat Linz, DVR 0093696,
2008.

[3] John R. Hauser, editor. Partial Differential Equations: The Finite Element
Method, pages 883–987. Springer Netherlands, Dordrecht, 2009.

[4] Devloo, P. R. B. (1997): PZ: An object-oriented environment for scientific


programming. Computer Methods in Applied Mechanics and Engineering, vol.
150, no. 1, pp. 133-153.

[5] Mario Paz and William Leigh. Static Condensation and Substructuring, chapter
8, pages 239–260. Springer US, Boston, MA, 2001.

You might also like