Predicting the Performance of Camber Morphing Airfoils Using Machine Learning Techniques

Conventional aircraft have various control surfaces, such as rudders, flaps, ailerons, and elevators, to maneuver by altering aerodynamic performance. While these are straightforward and traditional methods for controlling flight envelopes, they also present significant challenges. Most of the aircraft's noise and vibration originates from the control surfaces when they are deflected. Meanwhile, a camber-morphing airfoil may solve this issue by smoothly and substantially adjusting its shape in a conformal manner. However, it is quite challenging to analyze the aerodynamic performance of such morphing wings due to the airfoils' unconventional shapes, which complicate the analysis. Most well-known airfoils, such as those defined by the NACA, are well-established in databases, enabling users to efficiently select and utilize them for analysis and aircraft design. Among the various approaches to understanding the aerodynamic performance of camber-morphing airfoils, traditional computational fluid dynamics (CFD) tools can analyze them through multiple simulations of different morphing shapes, which can be time-consuming and computationally expensive. This study uses machine learning (ML) tools to predict the shapes of conventional and unconventional airfoils by training on the University of Illinois at Urbana-Champaign (UIUC) airfoil database. The database contains a table of 1,600 airfoils and their corresponding 2D coordinates. The study demonstrates that a selected ML model can accurately predict the aerodynamic performance of camber-morphing wings, including the lift coefficient (CL), drag coefficient (CD), and moment coefficient (CM), through training and learning. Among the ML models, including Random Forest, Support Vector Regression, and Multi-layer Perceptron, Gradient Boosting is the most effective at accurately predicting the aerodynamic coefficients (CL, CD, CM). The superior performance is primarily due to Gradient Boosting’s ability to capture complex, nonlinear relationships by sequentially minimizing residual errors, making it particularly well-suited to the highly nonlinear, shape-sensitive aerodynamic behavior of camber morphing. By enabling rapid, accurate predictions without iterative CFD simulations, this approach offers a scalable alternative that significantly reduces computational costs and design iteration time. Doing so opens a new path for real-time aerodynamics performance estimation, on-the-fly optimization of morphing strategies, and accelerated design cycles for next-generation adaptive aircraft technologies. CFD comparisons confirm that the ML model reliably predicts aerodynamic trends for moderate morphing, with minor deviation at higher camber.