QUANTIFYING UNCERTAINTIES IN TURBULENT FLOW SIMULATIONS

The Reynolds averaged Navier-Stokes equations represent an attractive alternative to direct numerical simulation of turbulence due to their simplicity and reduced computational expense. In the literature it is well established that structure of Reynolds averaged turbulence models are fundamentally limited in their ability to represent the turbulent processes introducing epistemic model-form uncertainty into the predictions. Sensitivity analysis and probabilistic approaches have been used to address these uncertainties, however there is no well established framework within the turbulence modeling community to quantify this important source of error. This work introduces a new approach for addressing epistemic uncertainty which is then demonstrated for the ow over a 2D transonic bump configuration. The well known SST k-ω turbulence model is considered. The reported quantities are the wall pressure, separation location, and reattachment location along the bottom wall of the domain. The results show the new method is able to introduce bounding behavior on the numerical and experimental predictions for these quantities.