Dr. Poza-Diaz, Abner

The multiscale hybrid-mixed (MHM) method is extended to the Stokes and Brinkman equations with highly heterogeneous coefficients. The approach is constructive. We first propose an equivalent dual-hybrid formulation of the original problem using a coarse partition of the heterogeneous domain. Faces may be not aligned with jumps in the data. Then, the exact velocity and the pressure are characterized as the solution of a global face problem and the solutions of local independent Stokes (or Brinkman) problems at the continuous level. Owing to this decomposition, the one-level MHM method stems from the standard Galerkin approach for the Lagrange multiplier space. Basis functions are responsible for upscaling the unresolved scales of the medium into the global formulation. They are the exact solution of the local problems with prescribed Neumann boundary conditions on faces driven by the Lagrange multipliers. We make the MHM method effective by adopting the unusual stabilized finite element method to solve the local problems approximately. As such, equal-order interpolation turns out to be an option for the velocity, the pressure and the Lagrange multipliers. The numerical solutions share the important properties of the continuum, such as local equilibrium with respect to external forces and local mass conservation. Several academic and highly heterogeneous tests infer that the method achieves super-convergence for the velocity as well optimal convergence for the pressure and also for the stress tensor in their natural norms.

This work proposes and analyzes a new local projection stabilized (LPS for short) finite element method for the nonlinear incompressible Navier–Stokes equations. Stokes problems defined element-wisely drive the construction of the stabilized terms which make the present method stable for P1 × P1, for continuous pressure and P1 × P0 for discontinuous pressure, in two- and three-dimensions. Existence and uniqueness of a discrete solution and a nonsingular branch of solutions are proved under standard assumptions. Also, we establish that the LPS method achieves optimal error estimates in the natural norms. Numerics assess the theoretical results and validate the LPS method in the three-dimensional case.