Adaptive mesh refinement for predicting the near wake of the Windsor body at yaw

In this paper, the influence of turbulence modelling on flow prediction around a simplified square-back model, the Windsor body, is investigated. Three turbulence models are used: the k − ω SST, and the hybrid RANS/LES models DDES-SST (Delayed Detached Eddy Simulation) and IDDES-SST (Improved Delayed Detached Eddy Simulation), which are based on the k − ω model. Numerical simulations are performed using the ISIS-CFD code to predict the flow around the Windsor body, a test case from the Automotive CFD Prediction Workshop. Two types of mesh are used: those generated by the AutoCFD Workshop organisers and those obtained with automatic mesh refinement, which are then adapted to the flow. The numerical simulations show that mesh convergence for results obtained with adapted meshes is much faster than that for results obtained with meshes generated by the workshop organisers. The results obtained with the hybrid RANS/LES models agree better with experimental results than those obtained with the k − ω SST model. The IDDES-SST model yields slightly better results than the DDES-SST model, regardless of the mesh type. Furthermore, results obtained with an adapted mesh and a turbulence model agree better with experimental measurements than those obtained with the Workshop organiser’s meshes. Consequently, mesh refinement and the hybrid RANS/LES IDDES-SST model are the solution for this type of flow.