Development of a turbulent separated flow validation test case: Experimental and computational (RANS) studies

Abstract

A new validation test case for CFD of turbulent separated flows is investigated through a combination of experiments and simulations. This work is part of an ongoing collaboration between the University of Washington and Boeing, which aims to contribute to the development of a high-quality validation test case for turbulent separated flows and the improvement of RANS modeling for turbulent separated flows. A three-dimensional speed-bump-like geometry, which causes separation as a result of the surface curvature, was chosen for this study. Tests were conducted to determine the influence of Reynolds number and confinement on the flowfield and to test the ability of common RANS turbulence models to accurately predict the flow. Experimental data was used to formalize the required inflow length so that the bump inflow matched between the simulations and experiments. The parameter used to match the inflows was the Reynolds number based on momentum thickness, $Re_{\theta}$, of the incoming boundary layer. The experimental data collected from the incoming boundary layer was also used to prove that the flow upstream of the bump was fully turbulent for all levels of confinement. Five Reynolds numbers were tested in the experiments and the simulations, corresponding to a freestream velocity range of 60 m/s - 20 m/s and four vertical confinement levels were analyzed in the experiments. Six turbulence models were examined: $k-\omega$ SST, two versions of Spalart-Allmaras (SA and SARC) and three versions of $k-\epsilon$. The geometry was determined to be a challenge to RANS models, which was demonstrated by the pressure coefficient data because the simulations predicted an opposite Reynolds number trend to the experiments in the separated region. Furthermore, the simulations were not able to predict distinctive pressure coefficient profiles seen in the experimental results, such as an inflection point along the streamwise centerline in the separated region and a double-peak in the spanwise direction across the top of the bump. However, the simulations did predict that the flowfield is insensitive to Reynolds number at and above $Re_L = 2.46 \times 10^6$, which is in agreement with the Reynolds number insensitivity determined experimentally. Increasing confinement increased the magnitude of the pressure coefficients over the bump, and the pressure at the peak of the bump went from about -1.2 in the least confined case to about -1.5 in the most confined case. However, there was no change in the shape of the profile. The examination of the various RANS turbulence models concluded that, for this curved geometry, the turbulence models with the curvature correction, $k-\omega$ SST and SARC, corresponded more closely to the experimental flow than those without a curvature correction. None of the $k-\epsilon$ models predicted separation, but SARC, SA and $k-\omega$ SST all did. It was determined that SARC, SA and $k-\omega$ SST all displayed similar flow features to those observed in the experimental flow visualizations, such as the general shape of the separated region and counter-rotating surface vortices that were symmetric across the streamwise centerline. However, the extent, location and width of the separated region varied. Along with variation in the size of the separation bubble predicted by each turbulence model, they also all predicted different values and profiles for pressure and skin friction coefficient in the separated region. From this experimental and computational analysis, it is clear that this geometry poses a sufficient challenge to current RANS models, due to their inability to accurately predict the location, size and values of pressure coefficients within the separated region. Therefore, is a good choice for a turbulent separated flow validation test. Future work on this project will focus on detailed flowfield comparisons between experiments and simulations.