Computational and experimental studies of mixing for industrial multiphase processes

  • Richard Maltby

Student thesis: Doctoral ThesisPhD


The main aim of this PhD project is to develop computational and mathematical models to describe industrial-scale gas-liquid flows using the computational fluid dynamics (CFD) software ANSYS CFX. Two particular industrial applications are studied, namely industrial biotechnology and sugar refining. The models are based on a comprehensive literature review covering the use of CFD modelling in gas-liquid stirred tank systems. Where there is no consensus in literature for the preferred choice of model, different available options have been compared. The CFD model has been applied to a novel single-use-technology bioreactor, which has been designed and operated as part of an industrial collaboration. The 1,000 L cubic reactor incorporates a single floor-mounted magnetically-driven impeller and fourteen individual gas spargers, meaning that the mixing characteristics cannot be evaluated using traditional design correlations. The hydrodynamic model has been solved over a range of different stirrer rotational speeds and air flow rates. The flow patterns have been shown to change from buoyancy dominant to impeller dominant between 200 and 300 RPM, accompanied by a much greater distribution of the gas phase. The mass transfer performance, reported in terms of volume-averaged k_L a, is modelled using five different mass transfer models existing in literature and compared to experimental measurements. There is a large spread of k_L a values predicted by the different models, with the experimental value sitting within these predictions. However, due to the limited access to take experimental measurements within the full-scale system, laboratory-scale validation is also performed against multiple parameters in a 9.4 L square-bottomed glass tank. The liquid-phase velocity, measured using laser Doppler velocimetry, provides a reasonable fit to an equivalent model of the validation tank at stirrer speeds of 100 to 400 RPM, whereas a qualitative analysis of the gas dispersion shows a good match between the model and experiments. The bubble size distribution is approximated experimentally using a watershed function applied to multiple images. As with the full-scale system, the measured k_L a falls within the range of values predicted by the model. Two of the models are found to provide a good fit to the experimental k_L a measurement, with the slip velocity matching the measured k_L a most accurately across the full-scale and validation tanks. However, the model considerably over-predicted the k_L a in the validation tank at 400 RPM, which is proposed to predominantly result from the specification of the population balance parameters based on the full-size bioreactor, where the influence of the impeller action in bubble break-up mechanisms is reduced.

The CFD modelling work has shown that the bubble size is a much more significant factor in interphase mass transfer than the mass transfer coefficient, which remains relatively constant across different conditions, and therefore using sub-millimetre bubbles, i.e. microbubbles, may lead to vastly improved mass transfer. This has been investigated experimentally by using a commercially available microbubble pump to measure the mass transfer of oxygen from air to water in three different geometry tanks with varying volumes of water from 7.62 to 200 L. The results show that the k_L a is independent of the tank geometry as the microbubbles are observed to be dispersed evenly throughout each tank. Furthermore, introducing mechanical agitation is shown to provide no improvement in mass transfer, since microbubbles have a much greater stability than larger bubbles, meaning that stirring costs may be reduced for many applications. The pump is also shown to create a significant supersaturation of oxygen in the liquid due to the high pressure in the pumping circuit and within the individual microbubbles, which is also beneficial for interphase mass transfer. The mass transfer rate achieved per volume of gas is significantly improved by using the microbubble pump, however the large pumping capacity of the pump is likely to limit its applicability for many larger-scale mass transfer processes.

CFD modelling developed has also been applied to an industrial-scale continuous gas-liquid contactor used during the carbonatation process, which is a clarification step in the refining of cane sugar. The hydrodynamic model shows that the column is operating in the churn-turbulent bubbling regime under normal operating conditions. This means that the liquid-phase is well mixed and there are no significant concentration gradients between the top and the bottom of the column. The mass transfer is shown to improve with increasing gas flow rate up to 0.7 t hr-1, however the shear stress also increases significantly within the operating range, resulting in a predicted optimum gas flow rate in the region of 0.5 t hr-1.A model of the complex series of reactions occurring during carbonatation has been developed within the CFD software environment. The model has been used to predict the local and outlet concentration profiles under two distinct phases of column operation, with and without excess calcium hydroxide present from start-up, and for a range of different operating conditions. The reaction model has been validated against a laboratory-scale model system, consisting of a closed system of water and calcium hydroxide, which is continually bubbled with carbon dioxide gas. The model provides a good fit to experimental measurements of the pH and both solid and dissolved carbonate concentrations for three different gas flow rates, although the fit is less good at higher hydroxide concentration.
Date of Award19 Feb 2020
Original languageEnglish
Awarding Institution
  • University of Bath
SupervisorJohn Chew (Supervisor) & David Leak (Supervisor)

Cite this