Comparison of Mixing Length, Standard k-epsilon, and Renormalized Group k-epsilon Turbulence Models in GOTHIC 7.2

Abstract

Scores of CFD codes are available for computing flow properties in complex geometries. Many of these codes require significant effort for producing the models and large amounts of computational power for even simple simulations. Within the nuclear industry, large-scale transient simulations are required that would take considerable time to compute with a standard CFD code. Instead of utilizing a CFD code, the GOTHIC thermal-hydraulic code can be used to compute various thermal and flow properties for these large-scale simulations. For calculating flow properties, GOTHIC has several turbulence models that can be used for computing the Reynolds stresses. The purpose of this work is to evaluate the effectiveness of the mixing length, k-ε, and RNG k-ε turbulence models in predicting flow properties using coarse meshes in GOTHIC. Experiments given in literature for an axi-symmetric jet, mixing layer flow, and channel flow are modeled with GOTHIC and compared to the experimental results.

Various flow properties such as the turbulent kinetic energy, spreading rates, velocity profiles, mixing region growth rates, and centerline velocities generated by GOTHIC are compared with experimental data. The results show that the mixing length model does poorly at predicting any of the flow properties, especially the turbulent kinetic energy. The RNG k-ε model does significantly better at predicting the flows, but takes much longer to run. Overall, however, the standard k-ε model provides the best replication of the experimental results for the coarse meshes utilized. In conclusion, the standard k-ε model should continue to be used for the default turbulence model in future validation of GOTHIC.