Airfoil aerodynamic/stealth design based on conditional generative adversarial networks.

The article focuses on the development of an airfoil inverse design model integrating aerodynamic and stealth characteristics using a conditional generative adversarial network (CGAN). The CGAN model, conditioned on parameters such as lift coefficient, drag coefficient, pitching moment coefficient, and radar cross section for vertical and horizontal polarizations, generates airfoil shapes with high accuracy, achieving an average prediction accuracy of 99.22% on the test set. The model employs multilayer perceptron neural networks for both generator and discriminator, demonstrating robustness, stability, and improved generalization compared to traditional neural networks. This approach simplifies and accelerates the multidisciplinary airfoil design process by directly mapping aerodynamic and stealth parameters to airfoil profiles without relying on computationally expensive simulations.