Simulation of bubbly flow in a flat bubble column

Abstract: Numerical, transient simulations of a meandering bubble plume in a rectangular flat bubble column was carried out, using the commercial CFD package ANSYS Fluent 15 within the framework of a Eulerian-Eulerian description. A Sensitivity analysis of turbulence closure was performed using four turbulence models: standard, RNG, Realizable k − epsilon and the Reynolds stress model. The effect of dispersed phase turbulence on the continuous phase was also investigate. The interaction force term included drag, virtual mass and turbulent dispersion. Some simulations were also made using lift and wall lubrication forces. The available drag and dispersion models available in Fluent were evaluated. Lastly, the impact of numerical methods was investigated where the model was tested for discretization scheme, gradient limiter, time step and grid size. The standard and Realizable k − epsilon models resultsed in a meandering bubble plume that oscillated with a period of 32.5 seconds. The Reynolds stress model did not reach a quasi-steady state, as the period kept increasing during the flow-time. The RNG k − epsilon model produced an oscillating plume, but the recorded time series show that the velocity fluctuated in a more chaotic way. The analysis of the dispersed phase turbulence modelling showed that the chaotic oscillations predicted by the RNG k − ² was a results of under predicting the turbulent viscosity. Comparing drag models showed that the drag coefficient influenced the amplitude of the velocity measurements, and that higher drag coefficient yield a higher amplitude. It was also found that including the turbulent dispersion force is imperative to capture the dispersion of the bubble plume. The lift force did not influence the oscillation frequency of the plume, but only the dispersion. Including the lift force caused the non-physical effect of pushing the bubbles towards the back and forth wall. It was also shown that this effect could be counteracted by including wall lubrication, and keeping the dispersive effect of the lift force. The choice of time step had the largest impact on the oscillation frequency. Decreasing the time step below 4 ms caused the oscillation frequency to decrease rapidly. If a time step of 2.5 ms was used, the time between peaks doubled.