Grey-Area Mitigation for the Ahmed Car Body Using Embedded DDES

Abstract

The Ahmed car body represents a generic car geometry which exhibits many of the flow features found in real-life cars despite its simplified geometry. It is a challenging test case for the turbulence modelling community as it combines both 3D separation and the formation of counter-rotating vortices, which interact together to produce a recirculation region behind the car body. It is shown that none of the RANS models tested are able to correctly predict the size of the recirculation region, regardless of modelling level, mesh resolution or the choice of the length scale (i.e. $\omega$ or $\varepsilon$). All of these models under-predict the turbulence levels over the slanted back and as a consequence over-predict the separation region. The DDES simulations (regardless of the underlying URANS model) offer an improved predictive capability compared to the RANS models when the mesh resolution is sufficient. When the mesh resolution is insufficient the DDES models produces worse results than either of the URANS models. In both cases, the grey area problem is demonstrated, wherein a lack of both modelled and resolved turbulence in the initial separated shear layer results in an over-prediction of the separation region. A one-way embedded DDES approach is shown to give the best compromise between accuracy and simulation cost. It accurately predicts the level of resolved turbulence in the initial separated shear layer and thus compared to non-embedded DDES and URANS, the injection of synthetic turbulence upstream of the separation point allows for the correct level of turbulence at the onset of separation. The resulting separation zone is correctly predicted and the grey-area problem is reduced.