A b-spline-based generative adversarial network model for fast interactive airfoil aerodynamic optimization

Abstract

Airfoil aerodynamic optimization is of great importance in aircraft design; however, it relies on high-fidelity physics-based models that are computationally expensive to evaluate. In this work, we provide a methodology to reduce the computational cost for airfoil aerodynamic optimization. Firstly, we develop a B-spline based generative adversarial networks (BSplineGAN) parameterization method to automatically infer design space with sufficient shape variability. Secondly, we construct multi-layer neural network (MNN) surrogates for fast predictions on aerodynamic drag, lift, and pitching moment coefficients. The BSplineGAN has a relative error lower than 1% when fitting to UIUC database. Verification of MNN surrogates shows the root means square errors (RMSE) of all aerodynamic coefficients are within the range of 20%–40% standard deviation of testing points. Both normalized RMSE and relative errors are controlled within 1%. The proposed methodology is then demonstrated on an airfoil aerodynamic optimization. We also verified the baseline and optimized designs using a high-fidelity computational fluid dynamic solver. The proposed framework has the potential to enable web-based fast interactive airfoil aerodynamic optimization.