Large-Eddy Simulation/Reynolds-Averaged Navier–Stokes Simulations of High-Speed Mixing Processes

Abstract

A large-eddy simulation/Reynolds-averaged Navier–Stokes model is applied to three high-speed mixing layers and three sonic injection flows to generate data suitable for evaluating two current Reynolds-averaged Navier–Stokes models for turbulent mass diffusivity. These models solve transport equations for concentration variance and dissipation rate and differ in the constitutive relation for the turbulent mass diffusivity and the form of the evolution equation for the dissipation rate. The predictive capability of the large-eddy simulation/Reynolds-averaged Navier–Stokes model is assessed through simulations of an air–air mixing-layer experiment and a sonic ethylene injection into a Mach 2.0 airstream. This model provides good predictions of the mean velocity, turbulence intensity, and rms temperature fluctuation throughout the shear layer flowfield but slightly over-predicts the spreading rate of the mixing layer. The simulation of the sonic injection is in generally good agreement with the experiment but underpredicts the level of jet entrainment. Four other large-eddy simulation/Reynolds-averaged Navier–Stokes data sets, involving the replacement of the lower airstream in the mixing-layer configuration by an argon stream and a helium stream, and a replacement of the injectant gas in the sonic injection experiment with argon and helium are also generated. All data sets are “mined” to extract turbulent mass diffusivities, concentration variances, and scalar dissipation rates associated with the resolved eddy motion. The results show that both models can be optimized to achieve a good match with the exact scalar variance production rate.