Analysis of a Polynomial Chaos-Kriging Metamodel for Uncertainty Quantification in Aerodynamics

New aerospace aerodynamic bodies increasingly require robust design methods that demand data of the key problem variables in the presence of uncertainty. When these bodies are subject to complex physics, such as turbulence, separation, or secondary flows, the uncertainty data become more difficult to produce economically. Metamodels (surrogate models) can be used to produce data in the presence of uncertainty more efficiently by propagating the uncertainty from the model parameters to the outputs. However, the chief difficulty of metamodels is in consistently producing statistical data of the full system from a sparse number of evaluations. Recently, the polynomial chaos and kriging metamodeling approaches have been combined to take advantage of both their benefits. This research explores the combined method's effectiveness on airfoil and aircraft engine nacelle examples. It demonstrates that, although the combined method can produce more accurate results than either method alone, there is always a compromise between accuracy and efficiency.