The article discusses the significance of aircraft geometry in determining the aerodynamic coefficients and the limitations of traditional polynomial-based methods in accurately representing the 3D shape of a wing. It highlights the use of deep learning-based methods for extracting latent neural representations of the shape of 2D airfoils or 2D slices of wings. However, recent studies have shown that directly incorporating geometric features into the neural networks can improve the accuracy of predicted aerodynamic coefficients.
In line with this, the article proposes a method that incorporates Riemannian geometric features for learning Coefficient of Pressure (CP) distributions on wing surfaces. This approach involves calculating geometric features such as Riemannian metric, connection, and curvature, and combining them with the coordinates and flight conditions as inputs to a deep learning model. By doing so, the method aims to predict the CP distribution more accurately.
The article emphasizes the experimental results, which demonstrate the effectiveness of the proposed method compared to the state-of-the-art Deep Attention Network (DAN). The method achieves an average reduction of 8.41% in the predicted mean square error (MSE) of CP for the DLR-F11 aircraft test set.
This research is significant in the field of aerodynamics as it addresses the limitations of traditional methods in representing the complex geometry of wings in 3D space. By incorporating Riemannian geometric features, the proposed method provides a more accurate prediction of CP distributions on wing surfaces. This knowledge can be crucial in the design and optimization of aircraft for better performance and efficiency.
Moving forward, it would be interesting to see further exploration and refinement of incorporating geometric features in deep learning models for other aspects of aerodynamics. Additionally, the applicability of this approach to different types of aircraft and varying flight conditions should be investigated to assess its generalizability. Overall, this research opens up new possibilities for improving the understanding and prediction of aerodynamic coefficients, thereby enhancing the design and performance of aircraft.