A Surrogate ANN–BEM Framework for Aerodynamic Modeling of Smart Wind Turbine Blades

Abstract

Accurate prediction of aerodynamic coefficients is essential for the design and control of smart blades featuring morphing airfoils. This study presents a data-driven metamodel based on Artificial Neural Networks (ANNs) developed to predict the aerodynamic behavior of airfoils within the NACA series. The model accepts geometric descriptors of airfoils along with the angle of attack (AoA) as input and outputs corresponding lift (Cl) and drag (Cd) coefficients. A high-fidelity aerodynamic database was generated through systematic simulations across a wide range of AoAs and NACA profiles to train and validate the ANN. The model is trained on NACA 4-digit series profiles covering a wide range of AoA and geometric parameters. The ANN model achieved a mean squared error of 2.10805 e-3 and an R² above 0.997 on test data. The trained metamodel demonstrates excellent generalisation accuracy while drastically reducing computational requirements compared to conventional CFD or BEM-based methods. The model is particularly suited for integration into larger simulation frameworks, such as Blade Element Momentum (BEM) codes or adaptive control systems, enabling real-time performance estimation for morphing smart blades. This work contributes a scalable and efficient surrogate modelling approach for aerodynamic prediction across diverse airfoil geometries.