Prediction of Airfoil Lift Coefficient Based on Multilayer Perceptron

Airfoil lift coefficient prediction is pivotal to the study of aerodynamics, traditionally relying on time-consuming computational fluid dynamics (CFD) simulations or expensive wind tunnel tests. This study proposes a novel Multilayer Perceptron (MLP) model for predicting the airfoil lift coefficient, offering a significant computational cost reduction by eliminating the data preprocessing steps, thereby paving the way for practical application of deep learning algorithms in lift coefficient analysis. Our methods involve dimensionality reduction of the input airfoil coordinates, selection of activation functions, and incorporation of regularization terms in the loss function, all aiming to enhance prediction accuracy. Experimental findings reveal that dimensionality reduction of input coordinates effectively improves prediction precision. Remarkably, this model trims training time to only 2% and prediction time to 54% compared to previous Convolutional Neural Network model. Among different activation functions, Rectified Linear Unit (ReLU) converges fastest and provides the highest accuracy. Furthermore, introducing regularization terms into the loss function proves beneficial for prediction accuracy enhancement. In conclusion, this study offers a robust and computationally efficient approach for airfoil lift coefficient prediction, contributing to the application of deep learning techniques in aerodynamics.