Studying the aerodynamic signature of an airofoil structure beyond the experimental measuring limits of a wind tunnel using the ANN algorithm

This paper presents a comprehensive investigation into the utilization of an artificial neural network (ANN) to explore the aerodynamic characteristics of an airfoil. The ANN model, trained with an extensive experimental dataset, accurately predicts the lift and drag coefficients of different fabricated airfoils across a wide range of angles of attack (AoA) and wind speeds. The airfoil performance model is developed using practical measurements of lift and drag forces obtained from a diverse set of fabricated airfoils. The dataset covers wind speeds ranging from 4 to 10 m/s and AoA ranging from − 20 to 20°, providing a rich source of data that surpasses the limitations of traditional experimental setups. To enhance the accuracy and range of the dataset, a Levenberg–Marquardt algorithm (LMA) is employed for data regression, resulting in a significantly expanded dataset consisting of sub-million points. The accuracy of the developed model is rigorously tested and validated against experimental setups, demonstrating a maximum deviation error of 13% and an average accuracy exceeding 91%. The model's effectiveness is further assessed by applying it to real-world wind speed profiles from three selected wind farms in Egypt. The model accurately predicts lift and drag coefficients not only below 10 m/s but also at an AoA of 10°.