A new perspective on efficient and high-fidelity prediction of unsteady flow in airfoil dynamic stall

High-accuracy simulation of unsteady flow remains one of the key challenges in computational fluid dynamics. Large Eddy Simulation (LES) and Detached Eddy Simulation (DES) can achieve high fidelity by resolving the spatiotemporal evolution and temporal characteristics of small-scale turbulence; however, their high computational cost limits their applicability in engineering practice. In contrast, conventional Unsteady Reynolds-Averaged Navier-Stokes (URANS) approaches often exhibit significant inaccuracies in unsteady flows involving large-scale separation, such as dynamic stall. This study presents an innovative perspective: for unsteady turbulent flows characterized by pronounced large-scale flow structures, accurate prediction of macroscopic unsteady aerodynamic coefficients (e.g., lift, drag, and pitching moment) does not necessarily rely on the explicit resolution of the unsteady temporal evolution of small-scale turbulence. Instead, it can be achieved by constructing a steady turbulence model with high accuracy in the time-averaged sense (such as improved eddy-viscosity or Reynolds-stress models) and embedding it into an unsteady numerical framework. By solving the unsteady governing equations that describe the evolution of large-scale flow structures, the dominant dynamic variations of aerodynamic coefficients can be effectively captured. The proposed approach is validated using dynamic stall simulations of three representative wind turbine airfoils, namely S809, S810, and S814. Large-scale flow separation, vortex shedding, and aerodynamic hysteresis phenomena are numerically investigated. The results demonstrate that, based on the proposed simplified strategy combining a high-fidelity steady data-driven turbulence model with an unsteady framework, the predicted aerodynamic coefficient response curves are in good agreement with experimental data. Compared with URANS simulations based on the conventional Spalart-Allmaras (S-A) model, the prediction error is reduced by more than 50%, enabling accurate prediction of key features of the dynamic stall process and confirming that large-scale flow evolution plays a decisive role in governing macroscopic unsteady aerodynamic loads.