Four turbulence models were used to predict the flow over a 25°Ahmed body at a Reynolds number characteristic of suburban driving speeds in the United States: the k-ω shear stress transport (SST) model, large eddy simulation (LES) with the Smagorinsky-Lilly subgrid scale model, LES with the wall-adapting local eddy viscosity (WALE) subgrid scale model, and improved delayed detached eddy simulation (IDDES). The Reynolds number based on body height was ReH=2.65×105 - this corresponds to a freestream velocity of 13.44m/s, or approximately 30mph. The k-ω SST model only showed promise in predicting the drag coefficient; otherwise, it significantly underpredicted the lift coefficient and predicted an incorrect wake structure in terms of flow separation and the size of the separation bubble behind the body. All three scale-resolving models predicted a wake structure in good agreement with experimental data, with the WALE model and IDDES providing the best agreement in terms of separation bubble size, streamwise velocity magnitude, and turbulent kinetic energy in the wake. The best predictions of the aerodynamic forces were those of the WALE model, with overpredictions of only 5.45% for drag and 5.65% for lift. Overall, both the WALE model and IDDES demonstrated promise in predicting the wake structure and aerodynamic forces, but the best agreements were achieved with the WALE model. Therefore, the WALE model showed the lowest dependence on Reynolds number when comparing lower Reynolds number simulations to higher Reynolds number experiments for this situation.
Read full abstract