Uncertainty Quantification of Structural Uncertainties in RANS Simulations of Complex Flows

A method is presented for building a statistical model for the structural uncertainties in RANS turbulence models. This approach compares the results of RANS calculations to direct numerical simulation for a collection of simple, randomly generated geometries. The adjoint method is used to eciently solve an inverse problem for each geometry to determine the RANS turbulent viscosity eld that most accurately reproduces the mean ow eld computed by direct numerical simulation. The discrepancy between the inferred turbulent viscosity and the turbulent viscosity predicted by RANS is modeled as a Gaussian random eld that is scaled according to the local ow eld properties. The uncertainty in the turbulent viscosity eld is then propagated to the engineering quantities of interest.