A Parallel Finite Element Discretization Algorithm Based on Grad-Div Stabilization for the Navier–Stokes Equations

Abstract

We present and study a parallel grad-div stabilized finite element discretization algorithm based on entire-overlapping domain decomposition for the numerical simulation of Navier–Stokes equations. The algorithm is easy to implement on top of existing sequential software, in which each subproblem used to calculate a local solution in its designated subregion is actually a global problem with vast of degrees of freedom coming from its own subregion, and hence, can be solved independently with other subproblems. We derive error bounds of the approximate solution by employing the technical tool of local a priori estimate, and investigate the effect of grad-div stabilization term on the approximation solutions. Numerical comparisons, with both inf-sup stable and unstable mixed finite elements pairs for the velocity and pressure, show that our present algorithm has an amazing superiority to its counterpart without stabilization in the sense that accuracy of the approximate velocities could be improved by two orders of magnitude when the viscosity \(\nu \) is small. While compared with the usual standard serial grad-div stabilized finite element method, our algorithm saves lots of CPU time in computing a solution with comparable accuracy.