Prediction of Aerodynamic Coefficients for Multi-Swept Delta Wings via a Hybrid Neuronal Network

In this study, the prediction capabilities of a hybrid neural network with regards to aerodynamic coefficients of multiple swept delta wings are investigated. The quick evaluation of aerodynamic coefficients based on a few geometrical and flow parameters instead of cost-consuming computational fluid dynamics simulations or wind tunnel experiments could save time and costs during early aircraft design phases. The training data is based on the results of wind tunnel measurements for a number of multiple swept delta wing configurations with varying leading-edge sweeps. These datasets contain angle of attack slopes for the basic configurations, as well as measurements with deflected control surfaces and an applied sideslip angle. The results show, that different aerodynamic coefficients can be predicted accurately by machine learning models. The neural network shows great abilities in forecasting the aerodynamic characteristics and trends in coefficient slopes. Producing highly accurate predictions with respect to lift coefficients and its derivative, the prediction accuracy can vary for the pitching moment coefficient, mispredicting absolute values while still matching slope trends very well.     «

In this study, the prediction capabilities of a hybrid neural network with regards to aerodynamic coefficients of multiple swept delta wings are investigated. The quick evaluation of aerodynamic coefficients based on a few geometrical and flow parameters instead of cost-consuming computational fluid dynamics simulations or wind tunnel experiments could save time and costs during early aircraft design phases. The training data is based on the results of wind tunnel measurements for a number o...     »