Pergamon Chemical Engineering Science, Vol. 51, No. 2, pp. 233-249, 1996 Copyright © 1995 Elsevier Science Ltd Printed in Great Britain. All rights reserved 0009-2509/96 $9.50 + 0.00 0009-2509(95)00270-7 LDA MEASUREMENTS AND CFD MODELLING OF GAS-LIQUID FLOW IN A STIRRED VESSEL K. E. MORUD and B. H. HJERTAGER* Department of Process Technology, Telemark Institute of Technology and Telemark Technological R&D Centre, Kjolnes Ring, N-3914 Porsgrunn, Norway (First received 30 May 1994; revised manuscript received 21 July 1995; accepted 4 August 1995) Al~trae~Turbulent two-phase flow in a stirred vessel has been investigated experimentally and numer- ically. Mean and turbulent gas velocities are measured using a laser/phase Doppler anemometer (LDA/PDA). The effects of varying gas flow rates and impeller rotational speeds on axial, radial and tangential mean and turbulent velocities at three levels of the vessel are investigated. Furthermore, total gas fractions are measured by observing the level of the liquid surface. A two-dimensional computational fluid dynamics (CFD) two-fluid model, with a standard k-e turbulence model, is used to predict the gas-liquid flow. Impellers and baffles are modelled by introducing source and sink terms in the appropriate momentum equations. The numerical results are verified against the experimental data. l. INTRODUCTION 1.1. Motivation Gas-liquid reactors are often used in industrial processes based on biotechnology. The processes that occur in these bioreactors often have contradictory demands. The bubbles should be well dispersed to ensure sufficient oxygen mass transfer. On the other hand, care should be taken not to damage the shear sensitive microorganisms. Up till recently, empirical correlations have been used as tools for design, scale- up and operation of bioreactors. In view of the com- plex turbulent gas-liquid flow and microbial kinetics, such limited empirical relations are clearly inad- equate. To fully understand the processes occurring in bioreactors, there is a need for mathematical models describing the local turbulent gas-liquid flow and bioconversion characteristics. 1.2. Previous work The basis for general prediction models of bio- processes in bioreactors is the modelling of turbulent flow in these reactors. Issa and Gosman (1981) simulated the two-phase flow in a baffled, agitated vessel. They used a three-dimensional model, with a simplified equation for the gas momentum, and the k-e turbulence model. Pericleous and Patel (1987) modelled the flow in stirred vessels using an algebraic slip model. Momentum equations were solved using mixture properties. The spatial distribution of bubbles was calculated by the concentration equations, as- suming constant bubble size. Experimental results and three-dimensional calculations of the flow pro- cesses occurring in a stirred reactor were presented by Bakker and van den Akker (1991). Single phase laser *Corresponding author. Fax:+ 47-35-55-75-47; email: bhh@tmih.no Doppler measurements were performed, and for gas-liquid flow local gas fraction, bubble size and local oxygen mass transfer rate were measured, From a single-phase calculation and assumptions of the phase slip, they calculated the distribution of gas in the vessel. Patterson (1991) used the LDA technique to measure liquid velocities with and without gas sparged. A three-dimensional calculation of the flow with the FLUENT code was applied, where the indi- vidual bubbles were tracked by a Lagrangian ap- proach. Br6ring et al. (1991) and Fisher et al. (1992) used an improved ultrasound-Doppler technique for extensive measurements of bubble velocity compo- nents in stirred vessels equipped with three and one Rushton impeller respectively. Gosman et al. (1992) calculated the gas-liquid flow in a stirred vessel with a three-dimensional calculation domain. The simula- tion results on local gas fraction were compared against experimental data. They used a two-fluid model and a k-E turbulence model, extended for two- phase flow. Most modelling of stirred vessels is on single-phase flow, and here the impeller region has been modelled in different ways. Harvey and Greaves (1982) cal- culated turbulent single-phase flow in an agitated vessel with baffles. They used a two-dimensional de- scription and the standard k-e turbulence model and specified k, ~ and velocity at the impeller surface. The swirling motion of fluid on the impeller boundary was assumed to be solid body rotation. Some predictions of mean flow and turbulence were compared with experimental data. Placek and Tavlarides (1985, 1986) have predicted and measured the flow patterns and turbulence properties in a two-dimensional baffled vessel. They applied a three-equation turbulence model and assumed anisotropic turbulence in the impeller region. The model requires proper assump- tions concerning the impeller boundary, thus they got 233