Assessment of Turbulence Models in Predicting the Heat Transfer of Supercritical Carbon Dioxide

Abstract

Supercritical fluids are used as coolants in one of the Generation-IV reactors (i.e., supercritical water reactor) owing to their good diffusivity, low viscosity, and high specific heat. Additionally, these fluids exist at higher pressure and temperature which allows high thermal efficiency. Two heat transfer phenomena are related to supercritical fluids: heat transfer deterioration and enhancement. These phenomena made it difficult for Reynolds-averaged Navier–Stokes simulation (RANS)-based turbulence models to accurately predict the heat transfer. In this study, an assessment of RANS-based turbulence models is conducted for supercritical carbon dioxide under two different flow conditions (i.e., horizontal flow and natural circulation vertical flow). The two cases are simulated using current turbulence models (i.e., SST k-ω, k-ε, RNG k-ε) and a newly developed model based on the algebraic heat flux model (AHFM), hereafter called UniPi. It was found that for the horizontal flow case, the SST k-ω model captured the temperature difference induced by buoyancy between different regions of the wall, however, with poor accuracy in predicting wall temperatures. The RNG k-ε models captured the behavior of wall temperature across all regions with underestimated values. The enhanced wall treatment gives good predictions of wall temperatures compared to experimental data, but it underestimates the deterioration and recovery of heat transfer. In the natural circulation case, the recently developed model, which is based on AHFM, yielded better results compared to k-ε and the SST k-ω models. This is mainly because it explicitly considers the buoyancy production term and the turbulent heat flux.