A macroelement stabilization for mixed finite element/finite volume discretizations of multiphase poromechanics

Julia T. Camargo, Joshua A. White, and Ronaldo I. Borja

Computational Geosciences

Abstract

Strong coupling between geomechanical deformation and multiphase fluid flow appears in a variety of geoscience applications. A common discretization strategy for these problems is a continuous Galerkin finite element scheme for the momentum balance equation and a finite volume scheme for the mass balance equations. When applied within a fully implicit solution strategy, however, this discretization is not intrinsically stable. In the limit of small time steps or low permeabilities, spurious oscillations in the piecewise-constant pressure field, i.e., checkerboarding, may be observed. Further, eigenvalues associated with the spurious modes will control the conditioning of the matrices and can dramatically degrade the convergence rate of iterative linear solvers. Here, we propose a stabilization technique in which the mass balance equations are supplemented with stabilizing flux terms on a macroelement basis. The additional stabilization terms are dependent on a stabilization parameter. We identify an optimal value for this parameter using an analysis of the eigenvalue distribution of the macroelement Schur complement matrix. The resulting method is simple to implement and preserves the underlying sparsity pattern of the original discretization. Another appealing feature of the method is that mass is exactly conserved on macroelements, despite the addition of artificial fluxes. The efficacy of the proposed technique is demonstrated with several numerical examples.

Link to Full Paper

https://doi.org/10.1007/s10596-020-09964-3

Highlight

highlight image
The standard mixed finite-element/finite volume scheme is intrinsically unstable, leading to near-singular matrices for either small timesteps or low permeabilities. This can have a substantial impact on linear solver convergence. Here, we see the proposed stabilization dramatically reduces the number of linear iterations needed to converge.

Updated: