In-Situ Adaptive Mesh Refinement for High Fidelity Simulations of Compressible Turbulence

Abstract

A validated, curvilinear, direct numerical simulation (DNS) and large-eddy simulation (LES) flow solver is augmented with a novel curvilinear adaptive mesh refinement (AMR) for resolving turbulent flows with minimal grid density for hypersonic applications. The AMR code implementation, CRoCCo-AMR, is validated against the DNS and LES data of a Mach 3 shock turbulent-boundary layer interaction (STBLI). Such data was previously generated using the same code without AMR on the CRoCCo code. The CRoCCo-AMR implementation enables in-situ AMR; namely, the mesh adapts dynamically based on the local flow solution during the simulation, and there is no need to stop the simulation, re-mesh and re-partition the load balance amongst parallel processors. For the STBLI, we demonstrate an overall reduction in grid points of 40% and 27% for DNS and LES, respectively. Mean flow variables, turbulent intensities and spectral quantities are presented and compared with good agreement to the original, experimentally validated datasets. Specific considerations for enabling curvilinear AMR for DNS and LES are discussed.