Turbulence Resolution Requirements for Delayed Detached Eddy Simulation in External Automotive Aerodynamics

Abstract

Computational Fluid Dynamics (CFD) is currently a widely used tool in the analysis of external automotive aerodynamics due to its flexibility and efficiency in providing detailed analyses of designs. Nevertheless, the accurate prediction of flow fields with complex geometry at high Reynolds numbers still presents a significant challenge. Low-fidelity methods such as Reynolds Averaged Navier-Stokes (RANS) methods are unable to consistently provide correct predictions of three dimensional flow fields due to their inability to accurately resolve turbulent fluctuations and account for non-local effects. On the contrary, while high-fidelity methods such as Large Eddy Simulation (LES) are able to resolve unsteady turbulent structures, these methods are too expensive for complex geometries and high Reynolds number scenarios. In the past two decades, many Hybrid RANS/LES Methods (HRLM) were proposed to reduce the cost of LES. Detached Eddy Simulation (DES) and its variants are amongst the most commonly employed in the industry due to their simplicity and widespread usage. However, their mesh resolution requirement for automotive external aerodynamic simulations has not been well studied, which makes meshing an experience-oriented activity and often reduces computational performance, either in the form of reduced accuracy or excessive mesh refinement. In this thesis, two simplified vehicle models are used to provide representative examples of flow topology around the geometries at 0 and 8 degree yaw angles. These cases are used as the basis for a detailed investigation into the mesh sensitivity of Delayed Detached Eddy Simulation (DDES) methods by assessing different sub-grid length scales and their mesh topologies, resolutions and strategies, in order to improve understanding of mesh sensitivities in key regions. We find that although the shear-layer adapted model, a flow based sub-grid length scale, can improve the level of resolved turbulence, a sufficient mesh resolution is still key to correctly predicting a flow field. The traditional octree-based refinement method used in the automotive industry is observed to be suboptimal, where large regions of the mesh can be removed without adverse impact influence on the flow prediction. A Taylor micro-scale based mesh quality index is developed to indicate the local mesh resolution requirement. By applying the Taylor micro-scale with a single-step adaptive mesh refinement method, we achieve a similar level of accuracy to octree-based refinement while saving up to 70% of the total mesh size.

Date of Award	1 Aug 2023
Original language	English
Awarding Institution	The University of Manchester
Supervisor	Alistair Revell (Supervisor), Antonino Filippone (Supervisor), Ben Parslew (Supervisor) & Adrian Harwood (Supervisor)