Modeling aerodynamic characteristics of a wing airfoil using artificial neural networks

Abstract

The paper considers the application of artificial neural networks to solve the direct and inverse aerodynamic modeling problems using the example of a two-dimensional NACA2415 wing profile. Based on the numerical solution of the steady-state Navier–Stokes equations, a training dataset is formed, which includes aerodynamic lift and drag coefficients for various values of the geometric
parameters and angle of attack. A neural network with two hidden layers of 10 neurons and a sigmoid activation function is built and trained on datasets with regular and random distribution of parameters. The feasibility of solving the inverse problem of recovering the geometric parameters of an airfoil and angle of attack from given aerodynamic coefficients with an error of no more than 5% is demonstrated. The results confirm the effectiveness of neural networks for modeling and inverse design of aerodynamic profiles.