Comparative Assessment of Turbulence Models for Predicting Square Cyclone Separator Performance

Abstract

As with conventional cyclones, the gas-particle flow within a square cyclone is characterized by the consistent presence of turbulence. Therefore, selecting an appropriate turbulence model to accurately predict cyclone performance is crucial. This paper presents the findings of a computational fluid dynamics (CFD) study that assesses the impact of various turbulence models on the flow field, collection efficiency, and pressure drop predictions within square cyclone separators. In the simulation of the flow field, an investigation was conducted into five turbulence models falling under the Reynolds Averaged Navier-Stokes category, namely the Spalart-Allmaras, standard k–ε, RNG k–ε, standard k–ω, and Reynolds Stress models. Each turbulence model was coupled with a discrete phase model to represent solid particle flow within a square cyclone. The solution of the flow and particle transport equations was conducted using the software Ansys Fluent (version 21.1). Except for the flow field, for which experimental data were unavailable, the calculations were validated by comparing the predicted results for the square cyclone overall and grade efficiencies, and the pressure drop, with values from the literature. The findings of this simulation investigation demonstrate that all tested turbulence models exhibit comparable qualitative trends in predicting the tangential and axial velocity profiles. However, when evaluated in quantitative terms, the RNG k-ε model tends to overpredict both the tangential and axial velocities compared to the other models. Conversely, in the majority of cases, the k–ε standard turbulence model tends to underpredict these velocities in comparison to the other models. Regarding the predictions of the square cyclone performance, the Reynolds Stress Model is particularly noteworthy among the five turbulence models examined. It provides qualitative and quantitative closer agreement with the experimental data regarding the overall collection, grade efficiencies, and pressure drop. In general, while all models produced qualitatively similar trends in square cyclone performance predictions, the Reynolds Stress Model yielded the closest agreement with the experimental results, both qualitatively and quantitatively.