Predictive simulations of many complex flows of engineering interest require higher fidelity representations of the effects of turbulence than those available in current industry-standard…

This is a $616,604.65 cooperative agreement awarded by the National Aeronautics and Space Administration (NASA) Shared Services Center to the University of Texas at Austin. The goal is to develop and validate a hybrid RANS/LES (Reynolds-Averaged Navier-Stokes/Large Eddy Simulation) turbulence model to improve the accuracy of predictive simulations for complex aerodynamic flows. The proposed approach aims to address limitations of current hybrid methods, including the effects of grid anisotropy and inaccurate boundary layer separation predictions. The resulting model is intended to enable more accurate simulations of aerospace flows that are beyond the capabilities of existing RANS models yet computationally intractable for wall-resolved LES. This contract has no set-aside designation.

Predictive simulations of many complex flows of engineering interest require higher fidelity representations of the effects of turbulence than those available in current industry-standard reynolds-averaged navier-stokes (rans) turbulence models. By resolving more of the turbulent scales as in large eddy simulation (LES) but modeling some computationally expensive regions where rans models can perform well, like attached boundary layers, hybrid rans/LES methods have the potential to provide the necessary higher fidelity modeling paradigm at affordable cost. However, these techniques have yet to deliver on their promise of predictive simulations for a wide range of flow scenarios because of multiple difficulties. These difficulties include corruption of the resolved turbulence by artifacts due to inhomogeneous and anisotropic grids as well as inaccurate predictions of smooth wall boundary layer separation due to premature switching from rans to LES mode. The goal of the proposed work is to develop and validate a hybrid rans/LES model that overcomes these deficiencies without relying on ad hoc correction functions that must be tuned for each new flow. The proposed approach, which has been under development in the PI's group in response to problems observed in detached eddy simulation (DES) predictions of a flow separation control concept, is based upon four modeling improvements: 1) a novel subgrid stress model that uses the filtered two-point, second-order correlation and a measure of the grid anisotropy to define an anisotropic eddy-viscosity model; 2) an anisotropic diffusivity-based model of the effects of resolution inhomogeneity in LES; 3) a new energy conserving rans/LES switching approach accounting for both resolution anisotropy and the state of the resolved turbulence and designed to transfer energy between the resolved and modeled scales at a physically realistic rate; and 4) reliance on a rans model, specifically durbin's V2-f model, that provides reynolds stress anisotropy information and has no explicit dependence on distance to the wall, eliminating the need for ad hoc wall treatments in the resulting hybrid model. The proposed work will continue to pursue these fundamental modeling developments, extend these models to compressible flows, and apply the resulting models to a wide range of validation test cases. In pursuing these developments and validation studies, we aim to produce a broadly applicable, predictive, and computationally tractable hybrid rans/LES model. The potential benefits of such a model to NASA and the aeronautics community in general are quite dramatic. Specifically, the modeling approach proposed here could lead to accurate simulations of flows that are beyond the capabilities of current hybrid models, including many of standard test cases identified by the revolutionary computational aerosciences technical challenge. Further, the methods developed here provide a plausible path toward practical and predictive simulations of many complex engineering flows, like the full aircraft configuration identified in the CFD vision 2030 study, where rans models are not predictive and wall-resolved LES is computationally intractable.