Adding heterogeneous inputs to transition networks to improve unsteady airfoil lift prediction.

The prediction of lift during in-flight conditions poses significant challenges, and the development of effective tools for predicting aerodynamic loads is of great importance. In the current investigation, a data-driven approach is used to predict the lift generated by a NACA0015 airfoil under highly separated flow conditions at a Reynolds number of 5.5x10⁵ using a discrete set of surface and flow field sensors as input. The airfoil was pitched around its static stall angle, α_ss = 20°, with a pitching amplitude of 8°. The experimental data included six cases with reduced frequencies k from 0.025 to 0.15. A weighted-average based transition network was applied to measurements from 36 surface pressure taps, with 20 installed on the suction side and 16 on the pressure side of the airfoil. Predictive results were compared against those also incorporating planar flow field data, such as vorticity levels in specific zones near the airfoil, to improve the accuracy of lift predictions. The degree and nature of improvement gained by expanding the inputs to the transition network will be discussed in terms of phase-averaged performance of the prediction, and performance in capturing cycle-to-cycle variations in the experimental data.