Validation and Improvement of RANS models for Turbulent Flow over a Three-Dimensional Bump

Accurately predicting 3D non-equilibrium turbulent boundary layer behavior remains challenging for eddy viscosity-based (EVM) and Full Reynolds Stress (FRSM) Reynolds Averaged Navier-Stokes (RANS) models. A number of recent experimental efforts explore flow over 2D and 3D bumps, allowing validation and improvement of existing RANS models. In this context, we study the BeVERLI Hill flow configuration tested at Virginia Tech, at 45^o and 30^o orientations to evaluate the performance of five different turbulent models: the standard Spalart-Allmaras one-equation model, a data-augmented Spalart-Allmaras model, the Menter k-ω SST two-equation model, the Chien k-ε two-equation model, and the seven-equation SSG-LRR Full Reynolds Stress Model (FRSM). The simulations are performed on hill height-based Reynolds numbers of 250,000 and 650,000. The results show good agreement among the two SA models, the SST k-ω model, and the FRSM, while the Chien k-ε model returns less accurate results. We compare these model results with BeVERLI Hill experimental data. We investigate the ability of the RANS models to predict spanwise flow asymmetries observed in the experiments. Finally, we attempt to improve models using experimental data and low Reynolds number LES/DNS results.