2018-04-23T22:28:12Z (GMT) by Kenneth Jansen
<p>Turbulence evolves through highly nonlinear interactions of a broad spectrum of spatial and temporal scales. Modeling these interactions involves a tradeoff between accuracy and computational cost. Uncertainty quantification and design analyses often require solutions for numerous parameter combinations, rendering Reynolds Averaged Navier-Stokes (RANS) the only feasible option. This work seeks to reduce the cost of accurately predicting separating turbulent boundary layers using data-centric analysis, multi-fidelity modeling, and machine learning.</p>