Rotor Blade Dynamic Stall: From Large Eddy Simulations To Reduced Order Models

Abstract

We investigate deep dynamic stall over a pitching NACA0012 airfoil using wall-resolved Large Eddy Simulations at Reynolds number 135 000 and reduced frequency 0.1. The influence of the span-to-chord ratio on unsteady aerodynamic behavior is assessed through simulations with varying span lengths from 0.2c to 1.2c. Results demonstrate that smaller spans introduce significant two-dimensional effects, affecting lift, drag, and vortex dynamics. The simulation data are compared against experimental measurements and state-of-the-art RANS and hybrid RANS/LES methods. Furthermore, a reduced-order model based on deep learning is developed using a flow-matching diffusion approach with a DiT architecture. This model successfully predicts flow fields across a wide range of operating conditions, with high fidelity and a fraction of the computational cost. The integration of high-fidelity simulations and efficient ROMs offers a robust pathway toward real-time aeroelastic simulations and advanced control strategies in rotorcraft design.