Large-eddy simulation with modeled wall stress for complex aerodynamics and stall prediction

The aerodynamics of aircraft high-lift devices at near-stall conditions is particularly difficult to predict numerically. The computational requirements for accurate wall-resolved large-eddy simulations are currently prohibitive, whereas Reynolds-averaged Navier–Stokes (RANS) models are generally reliable only for low angles of attack with fully attached boundary layers. Methods such as detached-eddy simulation resolve unsteadiness of the outer boundary layer and can predict separation, but they rely upon a thick RANS layer and highly stretched cells that damp the resolved turbulent fluctuations near the wall. An alternative approach, adopted here, is to extend the LES down to the wall, employing a relatively large near-wall normal grid spacing and avoiding grid stretching and high aspect ratios near the wall. A boundary condition then applies the correct wall shear stress as provided by a semiempirical wall model. An adaptive formulation of this wall-modeled large-eddy simulation is presented here and validated using realistic test cases. Validation using a channel flow case at a range of Reynolds numbers demonstrates accurate results with a seamless transition between fully resolved (y⁺≈2) and wall resolved (y⁺≈50). Predictions of the MD-30P/30N airfoil using a modest grid with y⁺≈100 give excellent agreement with experiments and correctly predict C_Lmax. Finally, the method is demonstrated for the NASA High-Lift Common Research Model providing surface pressure coefficients and velocity profiles. The predictions using a 50-million-cell mesh (for a full aircraft half-model) are in good agreement with considerably finer-grid RANS solutions. The presented method has considerable potential because it can produce accurate solutions to challenging engineering problems involving separation with modest grid and computational requirements while being robust to variations in near-wall grid spacing.