Time Advancement of the Navier-Stokes Equations: <i>p</i>-Adaptive Exponential Methods

Abstract

An adaptive exponential time advancement framework is developed for solving the multidimensional Navier-Stokes equations with a variable-order discontinuous Galerkin (DG) discretization on hybrid unstructured curved grids. The adaptive framework is realized with cell-wise, variable-order DG refinements and a dynamic assembly of elemental Jacobian matrices. The accuracy and performance gain are investigated for several benchmark cases up to a realistic, three-dimensional rotor flow. Numerical results are shown to be more efficient than the use of uniform-order exponential DG for simulating viscous flows.