Research on Aircraft Aerodynamic Modeling Method Based on Point Cloud Deep Learning

In the pre-research and selection stage of aircraft that requires continuous modification of the shape, constantly changing the geometric parameters of the shape will significantly increase the computational and labor costs of numerical simulation. The application of artificial intelligence technology in predicting the aerodynamic performance of aircraft is of great significance. An aerodynamic modeling method for aircraft based on point cloud deep learning is proposed in this article. Taking a certain wing layout unmanned aerial vehicle (UAV) as an example, a panel method is used to construct aerodynamic datasets for different wing profiles with the maximum thickness, the span length, the root chord length, and the tip chord length. A point cloud deep learning model is established based on the PointNet++ network to predict the surface pressure coefficients of the UAVs. The results show that the point cloud deep learning method can accurately predict the surface pressure coefficient of the UAVs with different shape parameters. The mean values of the six tests for MSE and R2 are 0.002590 and 0.9692, respectively.