Modeling turbulent flow in stirred tanks with CFD: the influence of the modeling approach, turbulence model and numerical scheme

Single phase turbulent flow in a tank stirred by a down- and an up-pumping pitched blade turbine has been simulated using computational fluid dynamics. The effect of the modeling approach, discretization scheme and turbulence model off mean velocities, turbulent kinetic energy and global quantities, such as the power and circulation numbers, has been investigated. The results have been validated by laser doppler velocimetry (LDV) data. The stationary and time-dependent modeling approaches were found to have little effect on the turbulent flow, however the choice of the numerical scheme was found to be important, especially for the predicted turbulent kinetic energy. A first-order method was found to highly underestimate LDV data compared with higher order methods. The type of the turbulence model was limited to the k-epsilon and RNG models due to convergence difficulties encountered with a Reynolds stress model and there was found to be little effect of these models on the mean flow and turbulent kinetic energy. This latter quantity was found to be largely under-predicted in the discharge region of the down-pumping impeller in comparison with LDV data. Better agreement was found for the up-pumping pitched blade turbine. Estimated power numbers were found generally to be in good agreement for the down- and up-pumping data. However, the circulation number tended to be over-predicted by about 30%, and 40%, for the down- and up-pumping agitators, respectively. (C) 2003 Elsevier Inc. All rights reserved.