Assessment of a novel Large Eddy Simulation model in separating flows around a wing section

Simulating turbulent flows in engineering applications is challenging due to the high computational cost of resolving a wide range of time and length scales. While Direct Numerical Simulation (DNS) offers ideal accuracy, its computational demands make it impractical for industrial use. Consequently, Reynolds-Averaged Navier-Stokes (RANS) models remain the industry standard, offering rapid solutions by predicting only the mean flow. However, Large Eddy Simulation (LES) is becoming increasingly viable. By resolving large-scale fluctuations and modelling the rest, LES provides superior accuracy and is rapidly gaining industrial relevance thanks to advancements in high-performance computing and subgrid-scale modelling. In the present work, a novel LES approach, the non-linear Tensorial Viscosity Model (TVM), and its Shear-Improved variant (SITVM) are tested and benchmarked against the established eddy-viscosity model, the Dynamic Smagorinsky Model (DSM). Relying on a new decomposition of the subgrid stress tensor, the TVM incorporates velocity gradients and the inertia properties of the grid element, while the SITVM further subtracts the mean velocity gradients to correctly yield zero subgrid stresses at the walls and in laminar regions. The models are evaluated on a NACA 4412 airfoil at a 15° Angle of Attack (AoA) and a chord Reynolds number of Re_c = 66 667. Freestream turbulence is introduced to trigger the transition of the separated laminar boundary layer and promote flow reattachment. All simulations are performed using the Finite Volume Method (FVM) code T-Flows, and the results are compared against DNS and Implicit LES (ILES) data. The findings demonstrate that, despite suffering from numerical stability issues under specific conditions, both tensorial models provide enhanced predictive capabilities compared to the traditional DSM and ILES, showing a closer agreement with high-fidelity DNS data in capturing laminar separation bubble and turbulent reattachment. However, zones characterised by permanent separation continue to pose significant challenges for all evaluated approaches, highlighting the inherent limitations of current subgrid-scale modelling and underscoring the need for further advancements in aerodynamic applications.