Investigation of a space-time adaptive method for bluff body flows

Abstract

The problem of large computational costs for achieving high accuracy uid dynamics simulations remains challenging. This thesis investigates an unstructured spacetime adaptive framework for unsteady ow simulations of the compressible Navier-Stokes equations. A fully-unstructured discretization of space and time is used: for d-dimensional spatial problems, (d + 1)-dimensional meshes are generated, where time is treated as an additional dimension. A high-order discontinuous Galerkin discretization is combined with an output-based anisotropic mesh adaptation framework for numerical approximations on simplex meshes with arbitrary orientation and anisotropy. This framework has been applied to study (2 + 1)d unsteady flows around a single circular cylinder and two cylinders in tandem arrangements at Re = 100. High-order solutions with sucient degrees of freedom are able to capture ow unsteadiness, shown as a von Karman vortex street formed behind the cylinders. The adapted results are compared with time-marching solutions to determine the eciency and reliability of this method.