Inverse Airfoil Design Using Attention-Enhanced Deep Neural Networks

This paper presents an attention-enhanced deep learning-based framework for airfoil inverse design that leverages the PARSEC parameterization methodology. A model is developed to predict the PARSEC parameters that define a given airfoil shape using various aerodynamic coefficients as inputs. The model is obtained by training a neural network utilizing an aerodynamic database generated using XFOIL. The proposed model successfully predicts airfoil shapes, demonstrating its ability to capture the relationship between aerodynamic coefficients and airfoil geometries. By leveraging the attention mechanism, the model identified and prioritized critical input features, enhancing its robustness. Results showed that the model could generalize well across a variety of airfoil shapes.