An Assessment of CFD for Prediction of 2-D and 3-D High-Speed Flows

Two-dimensional hypersonic entropy layer and shock/boundary layer interactions with high-temperature effects and three-dimensional supersonic base flows are employed to examine the ability of a spectrum of CFD methods to predict surface thermo-mechanical quantities of interest, as well as flowfield structure. The 2-D cases correspond to a subset of those adopted by the RTO-AVT Task Group 43 and include the flow past a cylinder with axis normal to the flow and that past a 25 o 55 o axisymmetric double-cone configuration. A separate 3-D situation considered is the flow in the base region of a cylinder with axis parallel to the flow, but with an afterbody bent at 10 o to simulate angle of attack. Different MUSCL-based high-order upwind methods are employed for the inviscid fluxes, while standard central second order differences discretize the viscous terms. The range of parameters yield laminar and turbulent situations, as well as frozen and thermo-chemical non-equilibrium effects. For distinctly laminar situations, reasonable prediction accuracy can be attained, including in thermo-chemical non-equilibrium regimes for which commonly employed chemical reaction sets such as the Park models are appropriate. However, the solutions depend on the specific inviscid flux algorithm chosen (including limiter), particularly when the carbuncle is observed. One double-cone test case yields an unsteady asymptote in which the separation region extends to the leading tip. It is postulated that for this case, the laminar approximation may be inappropriate and a three-dimensional simulation with Direct or Large-Eddy Simulation (DNS/LES) may be required. For base-flows at angle of attack, Detached-Eddy Simulations (DES) are compared with experimental data and previous Reynolds Averaged Navier-Stokes (RANS) results. Although significant unsteadiness is observed in the wake region, base surface pressures with DES fluctuate only by roughly 3% about a mean that is slightly higher than RANS which in turn overpredicts experimental measurements. Averaged DES and RANS results however reproduce experimental visualizations of 2-D projections of a complex non-intuitive 3-D structure which is substantially different from the axisymmetric situation. Instead of circumferential structures, the 3-D base flow yields a streamwise-oriented horse-shoe structure, whose two legs point forward and form the clearly observed vortical core pair.