The Advantages and Costs of Higher-Fidelity Turbulence Modeling

This paper compares the results and the computational efficiency of a research code with that of a commercial code on the same problem. The research code (REACT-MB) is tested using its unsteady hybrid large-eddy simulation/Reynolds-averaged Navier-Stokes (LES/RANS) method as well as a more common steady-state Menter RANS method. CFD++ of Metacomp is tested on the same problem using its realizable k-ε turbulence model. Normal sonic ethylene injection through a circular injector into a Mach 2 cross-flow was simulated by each code. Time-averaged statistics of the hybrid LES/RANS computations and converged solutions from the RANS computations are compared with experimental contours of time-averaged mixture fraction. Scalability of the codes is also compared.