LES and URANS simulations of hydrodynamics in mixing tank: Comparison to PIV experiments

Dissipation rate of turbulent kinetic energy is a key parameter in a number of chemical and biochemical processes. In stirred tank, its overall value is readily estimated; however its distribution is far from homogeneous. The experimental determination of dissipation rate is difficult and often limited to small regions of a reactor. Therefore, CFD appears to be a useful tool to estimate the distribution of the turbulence dissipation rate. In this work, two kinds of simulation have been performed: a simulation based on the URANS equations and a Large Eddy Simulation with the Smagorinski subgrid scale model. The aims of the present paper are (1) to compare URANS and LES simulations to PIV data in terms of velocity, kinetic energy and turbulence dissipation rate and (2) to study the influence of the constant in the subgrid scale model used in LES. In term of mean velocity, URANS simulation and especially LES are in good agreement with PIV experiments. In term of total kinetic energy, both simulations also give good results; however URANS simulation fails to predict the repartition of total kinetic energy between its periodic and turbulent components in the vicinity of the impeller. The mean velocity and the kinetic energy are shown to be independent of the value of the constant CS in the Smagorinski model. The overall dissipation rate is well predicted by URANS simulation but not its distribution in the impeller stream. For LES, the amplitude of local and overall dissipation rate is shown to be strongly dependent of the value of the constant CS. However, its distribution in the stream is similar regardless of the value of CS and best reproduced than by the URANS simulation.