Free‐surface turbulent flow induced by a Rushton turbine in an unbaffled dish‐bottom stirred tank reactor: LDV measurements and CFD simulations

Abstract

Laser Doppler velocimetry measurements and computational fluid dynamic (CFD) simulations of turbulent flows with free-surface vortex in an unbaffled dish-bottom stirred tank reactor agitated by a Rushton turbine are presented. Measurements of the three mean and fluctuating components of the velocity vector are made in order to characterise the flow field and to provide data for CFD model validation. An Eulerian-Eulerian multiphase flow model coupled with a volume-of-fluid method for capturing the gas-liquid interface is applied to determine the vortex shape and to compute the flow field. Turbulence is modelled using the standard k−ε, shear-stress transport and the differential Reynolds-stress model with two variants of the pressure-strain correlation. The predicted mean flow field obtained using all four turbulence models are on the whole similar and generally in good agreement with measurements. However, the Reynolds-stress models provide somewhat better predictions of the mean axial velocity. The turbulent kinetic energy is well predicted in the flow below the impeller, near the bottom of the tank; whereas it is underpredicted in the region close to the impeller and near the wall by all turbulence models.

Free Surface Turbulent Flow in an Unbaffled Stirred Tank: Detached Eddy Simulation and VOF Study

“…2, we can see that the water surface profiles are unsteady due to the transient nature of the free-surface hydrodynamics in the stirred tank. The results obtained by using k-ε model are consistent with the results of Haque et al (2011) 11 , which indicates that our simulation strategies are reliable. By comparison, the profiles predicted by the DES model are deeper, which are thought to be much closer to the experimental results.…”

“…Quantificationally, larger differences with the LDV results were found in the regions near the impeller shaft, with the velocity underestimated by about 50 % at maximum. Haque et al (2011) 11 pointed out that their measurements of the radial velocities are much larger than those reported by others (see for example, Alcamo et al, 2005; 1, 22 . This may be one reason why such a large error was detected.…”

“…In the absence of experimentally measured free surface profiles, the numerically predicted water surfaces were compared with the flow patterns reported by Haque et al (2011).…”

“…For comparison purposes, the results obtained by Haque et al (2011) 11 are also presented. In these plots, the radial position r is normalized by the radius R of the stirred tank, and the mean velocity is normalized by the blade tip velocity u tip .…”

“…In order to validate the numerical methods, the predictions were compared with data reported by Haque et al (2011) …”