A simplex cut-cell adaptive method for high-order discretizations of the compressible Navier-Stokes equations

Abstract

While an indispensable tool in analysis and design applications, Computational Fluid Dynamics (CFD) is still plagued by insufficient automation and robustness in the geometry-to-solution process. This thesis presents two ideas for improving automation and robustness in CFD: output-based mesh adaptation for high-order discretizations and simplex, cut-cell mesh generation. First, output-based mesh adaptation consists of generating a sequence of meshes in an automated fashion with the goal of minimizing an estimate of the error in an engineering output. This technique is proposed as an alternative to current CFD practices in which error estimation and mesh generation are largely performed by experienced practitioners. Second, cut-cell mesh generation is a potentially more automated and robust technique compared to boundary-conforming mesh generation for complex, curved geometries. Cut-cell meshes are obtained by cutting a given geometry of interest out of a background mesh that need not conform to the geometry boundary. Specifically, this thesis develops the idea of simplex cut cells, in which the background mesh consists of triangles or tetrahedra that can be stretched in arbitrary directions to efficiently resolve boundary-layer and wake features.

(cont.) The compressible Navier-Stokes equations in both two and three dimensions are discretized using the discontinuous Galerkin (DG) finite element method. An anisotropic h-adaptation technique is presented for high-order (p > 1) discretizations, driven by an output-error estimate obtained from the solution of an adjoint problem. In two and three dimensions, algorithms are presented for intersecting the geometry with the background mesh and for constructing the resulting cut cells. In addition, a quadrature technique is proposed for accurately integrating high-order functions on arbitrarily-shaped cut cells and cut faces. Accuracy on cut-cell meshes is demonstrated by comparing solutions to those on standard, boundary-conforming meshes. In two dimensions, robustness of the cut-cell, adaptive technique is successfully tested for highly-anisotropic boundary-layer meshes representative of practical high-Re simulations. In three dimensions, robustness of cut cells is demonstrated for various representative curved geometries. Adaptation results show that for all test cases considered, p = 2 and p = 3 discretizations meet desired error tolerances using fewer degrees of freedom than p = 1.