Feed-forward artificial neural network as surrogate model to predict lift and drag coefficient of NACA airfoil and searching of maximum lift-to-drag ratio

The problem of computation time in numerical calculations of aerodynamics has been studied by many research centres. In this work, a feed forward artificial neural network (FF-ANN) was used to determine the dependence of lift and drag coefficients on the angle of attack for NACA four-digit families. A panel method was used to generate the data needed to train the FF-ANNs. Optimisation using a genetic algorithm and a neural metamodel resulted in a non-standard NACA aerofoil for which the optimal angle of attack was determined with a maximum L/D ratio. The optimisation results were validated using the finite volume method.

---

Opracowanie rekordu ze środków MNiSW, umowa nr POPUL/SP/0154/2024/02 w ramach programu "Społeczna odpowiedzialność nauki II" - moduł: Popularyzacja nauki (2025).