Application of uncertainty quantification machine learning tools to airfoil structural behavior and health monitoring

The accurate prediction of aerodynamic and structural responses in airfoil systems is essential for their design, control, health monitoring, and reliability in modern wind and/or aerospace applications. However, uncertainties arising from manufacturing imperfections, material variability, and fluctuating flow conditions can significantly affect airfoil performance and safety. This study presents a framework integrating machine learning with uncertainty quantification techniques to enhance the robustness of airfoil performance predictions. Data-driven surrogate models are trained to approximate complex aerodynamic and structural relationships while providing probabilistic confidence bounds. The framework is validated using benchmark data of real-world applications which have been previously researched, but not from the uncertainty quantification point of view. Sensitivity analyses are conducted to evaluate the influence of input uncertainties on lift, drag, and structural deformation responses. Results demonstrate that uncertainty quantification-enhanced machine learning models can capture nonlinear dependencies, identify critical parameters contributing to uncertainty, and improve prediction reliability compared to deterministic approaches, supporting safer and more efficient design optimization.