Aerodynamics, as an ancient discipline, has always played a significant role in fields such as aerospace, shipbuilding, and wind power generation. Rapid and accurate solutions to aerodynamic problems have long been a goal pursued by researchers. In light of this, this paper distills the form of bird wing flapping motion and the mechanisms behind high lift generation, exploring the underlying principles of how wing flapping affects aerodynamic forces. Subsequently, by integrating numerical simulation, wind tunnel testing, and flight test data, the paper establishes an empirical formula for the correlation between shock-boundary layer interaction forces/thermal loads in ground-to-air conditions, corrects the pressure-thermal flux relationship, and confirms the objective existence of low-frequency oscillations in separation bubbles under real flight conditions. Finally, the paper introduces convolutional neural networks (CNNs) to conduct experiments on wing profile aerodynamic performance prediction based on CNNs. The study shows that by reasonably arranging the network structure and optimizing hyperparameters, the convolutional neural network can achieve high accuracy in predicting the lift-to-drag ratio of airfoils. The relative error distributions of the validation set and test set are almost consistent, with approximately 90% of the samples having a relative error below 1%. Thus, the model in this paper has high accuracy and can rapidly and accurately predict the lift-to-drag ratio of unknown airfoils.
