Modeling of Flow in Impeller Stirred Tanks using Computational Fluids Dynanlics

ABSTRACT

The efficiency of mixing processes in impeller agitated tanks depends highly on the

hydrodynamics. Computational fluids dynamics (CFD) provides a method of predicting the

complex tlow structures in stirred tanks. As with any approximate numerical method, CFD

methods are subject to errors due to assumptions in the underlying mathematical models, as well

as errors due to the numerical solution procedures. The aim of this thesis was to present a CFD

method that accurately models the hydrodynamic properties of the 110w in stirred tanks.

The general purpose eFD software Fluent 6. f was used to develop the model of a laboratoryscale

stirred tank. Numerical experiments were conducted to investigate the effects of the

computational grid density, discretization schemes, turbulence models and impeller modeling

method on the accuracy of the simulated tlow. The results were validated with Laser Doppler

Velocimetry data from the literature.

It was found that the density of the numerical grid had more influence on the predicted turbulent

quantities than on the mean velocity components. For the mean velocity components, reasonable

agreement with the experimental data was observed even on relatively coarse grids. The choice of

discretization scheme was found to have significant effect on the predicted turbulent kinetic

energy and Power numbers. Very good agreement with experimental data was achieved for both

these flow variables when higher order discretization schemes were used on fine grids. This is an

important finding as it suggests that the generally reported underestimation of turbulence in

literature is caused by numerical errors in the CFD simulation as opposed to inadequacies in the

turbulence models as suggested by most researchers.

Steady-state and time-dependent impeller models were compared and found to have little etlect

on the mean velocity and turbulent kinetic energy. Ilowever impeller Power numbers calculated

from the time-dependent simulations were found to be in better agreement with the experimental

values. A comparison was also made between the standard k-s and RNG models. It was found

that the standard k-s turbulence model gave better predictions of the flow than the RNG- k-s

turbulence model.