CFD modelling of stirred tanks: Numerical considerations

The aim of this paper is to demonstrate that the Multiple Reference Frames (MRF) impeller rotation model and the standard k–ε turbulence model, as commonly used in engineering CFD simulations of stirred tanks, can accurately model turbulent fluid flow provided very fine grids coupled with higher-order discretization schemes are used. The MRF model has been found to give adequate results for the steady-state simulation of stirred tanks but the k–ε turbulence model generally under or over-predicts turbulence. In this study the CFD software Fluent 6 is used to simulate flow in a small baffled tank of standard geometry agitated by a Rushton turbine impeller. Simulations are conducted on four grids of significantly different resolution using the upwind, central and QUICK discretization schemes. CFD model results are evaluated in terms of the predicted flow field, power number, mean velocity components and turbulent kinetic energy using published experimental data. The general flow field and mean fluid velocity predictions are not strongly influenced by either the grid resolution or discretization scheme. However, turbulent kinetic energy predictions are strongly influenced by both the grid resolution and discretization scheme. In this study a grid consisting of nearly 2 million control volumes in one half of a 15 cm diameter stirred tank, combined with a high-order discretization scheme, was required to accurately predict the turbulent kinetic energy. These represent conditions which are considerably more numerically intensive than used in most similar studies and suggests that the poor predictions of turbulence obtained using the k–ε turbulence model, often noted in the literature, may be due to numerical errors rather than inadequacies in the turbulence model.