Reynolds number effect correction of multi-fidelity aerodynamic distributions from wind tunnel and simulation data

Abstract

Conventional wind tunnels cannot simulate actual flight Reynolds numbers because of testing condition constraints. The changes in aerodynamic characteristics caused by differences in the Reynolds number are especially sensitive on transonic supercritical airfoils and wings. Accordingly, this study proposes a correction method for the Reynolds number effect based on feature extraction of multi-fidelity aerodynamic distributions. Correcting wind tunnel test results at low Reynolds numbers with large amounts of numerical simulation data across various Reynolds numbers allows this method to achieve greater precision on airfoil pressure distribution predictions at high Reynolds numbers. The first step is to extract the primary features of the surface pressure distribution of the supercritical airfoil using the proper orthogonal decomposition technique. Next, a multi-fidelity neural network model is employed to relate the (low-fidelity) simulation data to the (high-fidelity) wind tunnel data acquired at low Reynolds numbers. Finally, the model is used to predict the airfoil pressure distributions at high Reynolds numbers, and the predictions are validated using wind tunnel test data collected from the RAE2822 airfoil. The results indicate that this approach can provide more accurate estimations than numerical simulations and single-fidelity models. Additionally, it can effectively reduce the impact of the Reynolds number difference on the transonic aerodynamic properties of supercritical airfoils.