A New Hybrid LES/RANSModel with Eddy Viscosity Transport (EVT) Based Outer-layer Length Scale

摘要

A new hybrid large-eddy simulation / Reynolds-averaged Navier-Stokes simulation (LES/RANS) turbulence model is presented in this work. The new model is then compared to earlier models developed at NCSU and the IDDES model as a 'state-of-art' competitor for various flow configurations. In common with other hybrid turbulence models developed at NCSU, this new model utilizes a flowdependent blending function that shifts the turbulence closure fromMenter's baseline(BSL)/shearstress transport (SST) RANS model to a sub-grid model of choice in the logarithm-layer as the distance from wall increases. This blending function is based on estimates of outer- and inner turbulence length scales. Unlike earlier models that used a problem-specific calibration of a model constant or ensemble-averaged the modeled turbulence kinetic energy (TKE) and specific turbulence dissipation rate, the estimated outer-length scale of the new model is a function of an eddy viscosity that is specifically used for this purpose. A transport equation for this eddy viscosity is solved in addition to other transport equations, with excessive growth of eddy viscosity due to fluctuating strain rates controlled by a destruction termbased on the von Karman length scale. As a result, this new model is completely local. The new model is tested through simulation of zero pressure-gradient flat-plate boundary layers at low and moderate Reynolds numbers, a supersonic flat-plate boundary layer, an airfoil near stall, flow over a wall-mounted hump, and flow over a simplified car body (Ahmed body).;For the low-Re incompressible flat-plate boundary layer (FPBL) case, the results given by the new model are comparable to wall-resolved LES as well as experimental data. For the moderate-Re FPBL case, the effects of model constant values and numerical schemes are discussed, and the model is further tuned. The new model is proven capable of predicting incompressible and compressible FPBLs reasonably well. For the airfoil case, the new model gives good surface pressure and skin friction predictions, as well as separation-bubble predictions near the trailing edge. It is not able to capture a laminar separation bubble near the leading edge. Extensive study on the wall-mounted hump case shows that the tuned new model can do a better simulation than Gieseking's model and the IDDES model on this case, and its results are close to those of Choi's model which required pre-calculation calibration. This study also shows that spanwise resolution and dimension and choice of numerical scheme can affect the boundary layer structure and separation predictions produced by the hybrid models. The Ahmed body is a full 3-D test case for the new model, and results are sub-satisfactory due to insufficient surface mesh resolution. Over all, the EVT based new model works reasonably well for a wide range of problems, producing results comparable to other NCSU LES/RANS models while retaining a completely local formulation.Meanwhile, further investigations relating to the sensitivity and mechanism of impact of numerical methods, 3-D implementations with adequate resolution, and turbulent flows that may not possess a statistically-steady state are needed.