Boeing Contributions to the Second AIAA Uncertainty Quantification Challenge Problem for Aerodynamics

In mid 2025, an uncertainty quantification challenge problem was proposed by the AIAA Fluid Dynamics Technical Committee Uncertainty Quantification Discussion Group. The challenge problem is based on XFOIL panel method predictions of subsonic airfoil lift, drag, and pitching moment and it involves several parametric sources of uncertainty including angle of attack, Reynolds number, and a simple flap deflection angle. The Challenge Problem provided a set of artificial experimental data at a set of angle of attack and flap deflection combinations with a nominal Reynolds number and requested an uncertainty quantification analysis of XFOIL at a different set of angle of attack and flap deflection combinations. This work presents an approach to the problem that considers parametric uncertainty, discretization error, model form uncertainty, surrogate modeling error, and regression error. Gaussian Process regression and Artificial Neural Network regression are investigated for prediction of parametric uncertainty. The results show that the approach results in consistent uncertainty prediction across the experimental design space. Furthermore, the Artificial Neural Network regression outperformed the Gaussian Process regression in parametric uncertainty prediction accuracy.