You are on page 1of 8

13th European Conference on Mixing

London, 14-17 April 2009

HYDRODYNAMICS OF IMMISCIBLE LIQUID-LIQUID DISPERSIONS


IN STIRRED VESSELS

F. Laurenzi, M. Coroneo, G. Montante, A. Paglianti, F. Magelli*

Department of Chemical, Mining and Environmental Engineering,


via Terracini 28, I-40131, Bologna, Italy; e-mail: franco.magelli@mail.ing.unibo.it

Abstract. This work is aimed at investigating the fluid dynamic features of a stirred tank of standard
geometry for the dispersion of organics in water. In particular, the dispersion conditions at different
impeller speeds and the flow features of both the continuous and the dispersed phases have been
considered. The two-phase turbulent flow fields of a dilute liquid-liquid system as measured by means
of Particle Image Velocimetry (PIV) technique are presented. The mean velocity and the turbulent
characteristics of water and oil droplets flow are compared with those of the corresponding single-
phase flow and discussed; the influence of the dispersed phase on the continuous one is assessed.
Starting from moderately agitated conditions, corresponding to completely separated phases, the
different dispersion regimes taking place in the tank at increasing impeller speeds have been
identified. The capability of a CFD approach based on the solution of Reynolds averaged Navier-
Stokes equations and on an Eulerian-Eulerian description of the two phases in catching the dispersion
features has been critically assessed.

Key words: Mixing; liquid-liquid; PIV; CFD; RANS;

1. INTRODUCTION
Agitated vessels where an immiscible liquid-liquid dispersion is produced are widely used in
the chemical, pharmaceutical, food, polymerisation and petroleum industries. The production
of a dispersion of given characteristics is rather complex due to the occurrence and interaction
of several phenomena, namely drop break-up and coalescence, mean flow pattern and
turbulence, drop suspension, interfacial area and drop size distribution (DSD), possible phase
inversion, the influence of system composition as well as of small amounts of impurities, etc.
[1], which make this operation one of the most difficult to deal with.
The fluid dynamic interaction between the two phases plays a significant role in
determining the features of the dispersion, but it is far from being fully understood. Average
properties over the whole vessel are usually considered for system description and for scale-
up. The following main aspects have been studied: minimum agitation speed for complete
liquid-liquid dispersion [2]; correlation of mean drop size and DSD to energy dissipation rate
and mixer geometric parameters [3] as well as to energy dissipation rate and flow in the vessel
[4,5]; the influence of various impellers on the dispersion features [4-7]; description of the
interaction between the liquid phases in terms of intermittent turbulence [8-10]. Modelling of
these systems has also been attempted in recent years by fully-predictive, two-phase CFD
techniques coupled with break-up and coalescence models [11-13], which parallels the
analogous efforts to study gas-liquid systems: in spite of the high complexity of the system
and the significant simplifications thereof, the first results are encouraging [14].

1
One key issue for broadening the understanding of these systems and allowing CFD code
validation is to get more detailed information about their flow structure, namely mean flow
and turbulence of both phases and their interaction. This scope is experimentally very difficult
as intrusive techniques may disturb the flow, while the dispersed phase limits the optical
accessibility of the modern, non-intrusive measuring techniques (LDV and PIV). Recently,
Svenson and Rasmuson [15] studied the effect of the dispersed phase (up to 10 vol. %) on the
flow structure of the continuous one by means of the PIV technique combined with refractive
index matching; though, no information has been provided about the behaviour of the
dispersed phase.
The purpose of this paper is to contribute to the study of the fluid dynamics of liquid-
liquid systems. In particular, average and r.m.s. velocities of both phases have been
determined simultaneously at low dispersed-phase hold-up by means of a two-phase PIV. In
addition, the onset of liquid dispersion, Nid, and the “just dispersed” condition, Njd, have been
investigated by examining the drawdown of portions of the upper liquid by the stirrer both
experimentally and numerically to ascertain whether a CFD simulation method is capable to
describe these threshold conditions.

2. EXPERIMENTAL AND COMPUTATIONAL TECHNIQUES


The immiscible liquid-liquid dispersion was investigated in a stirred tank of standard
geometry consisting of a fully baffled cylindrical vessel (T=23.2 cm diameter, H=T height)
stirred by a Rushton turbine (D=T/3 diameter, C=T/2 clearance). For avoiding optical
distortion, the vessel was contained in a square tank filled with demineralised water, that was
the continuous liquid phase adopted in all the experiments. As the dispersed phase, n-
tetradecane and diesel fuel were selected for the PIV and the Nid and Njd determinations,
respectively.
The former was found to be particularly suitable for PIV measurements due to its optical
properties: it is colourless and transparent and it is practically immiscible with water. Its
refractive index, as measured by a Maselli LR-01 refractometer at the temperature of 24°C,
was equal to 1.427, while the value of 1.333 was measured for water at the same temperature.
The latter was preferred for Nid and Njd determination, since it is opaque and clearly
distinguishable from water. The physical properties of the two oils are similar: they are both
lighter than water (760 and 826 kg m-3 density for n-tetradecane and diesel fuel, respectively).
The vessel was always entirely filled with the two liquids and it was closed on top for
avoiding air entrainment. For the flow field measurements, the impeller rotational speeds, N,
was set at 2.5 and 3.3 s-1 and the dispersed liquid fraction was equal to 1 vol.%. Nid and Njd
was determined for 19 vol. % of oil.

2.1 The two-phase PIV technique


The ensemble-averaged velocity fields of the two liquids were measured simultaneously and
separately in a vertical plane midway between two consecutive baffles using the PIV
technique already applied to gas-liquid systems by Montante et al. [16]. A single pulsed
Nd:YAG laser and two cameras provided with appropriate light filters were adopted. The
light scattered by fluorescent Rhodamine-B seeding particles was captured with one camera
(HiSense MK II, 1344×1024 pixels CCD) and that of the oil droplets with the other (PCO,
1280×1024 pixels CCD). The time interval between two laser pulses was equal to 1875 µs
and that from a pulse pair to another was equal to a few seconds. The interrogation area was
set at 64×64 pixel, corresponding to about 15×15 mm, and the cross-correlation of the image
pairs was performed on a rectangular grid with 50% overlap between adjacent cells. As with
the gas-liquid systems, preliminary tests on the dependency of the ensemble-averaged mean

2
and turbulent flow field on the number of images have confirmed that about 2000 image pairs
are required to obtain sample-independent velocity fluctuation measurements for the
dispersed phase. This high value is probably due to the small numerical droplet concentration
in some zones, that lead to a high number of rejected vectors. Instead, for the continuous
phase a lower number of image pairs was sufficient, since the tracer particle concentration can
be suitably selected for maximising the percentage of validation.
The application of this technique to liquid-liquid systems is possible, but it is limited to
dilute conditions and modest impeller speeds so as to ensure almost stable conditions of the
dispersion and optical access of the laser sheet across the measurement plane. The above-
mentioned values of oil volumetric fraction and agitation speeds resulted from an empirical
combination of the two variables.

2.2 The determination of the dispersion conditions


The dispersion regimes as a function of the impeller speed, ranging from 2 to 5 s-1 and
corresponding to a variation of the rotational Reynolds number of the water phase between
1.24×104 and 3.10×104, were determined from the analysis of the captured images. Before
starting the acquisition, for each agitation condition the two liquids were mixed for about one
hour. The appearance of droplets below the impeller was considered as the indicator of Nid
achievement. On the other hand, Njd has been identified by the visual observation of the lid of
the tank and it was considered achieved when the oil layer almost completely disappeared.

2.3 The computational approach


The CFD simulation approach was similar to that adopted for the simulation of solid-liquid
and gas-liquid systems [17,18] with the purpose of assessing its applicability to the prediction
of liquid-liquid dispersion conditions.
It was based on the solution of the Reynolds averaged Navier-Stokes (RANS) equations
for each phase in an Eulerian framework. The interphase momentum transfer was accounted
for by the drag force only and for determining the droplet drag coefficient, that was evaluated
assuming the droplets as rigid spheres rising in a stagnant liquid, a mean droplet size was
fixed. For each impeller speed, the simulations were repeated considering the diameter of 50,
200 and 1000 µm for assessing the effect of this parameter on the results. The mathematical
closure of the problem was obtained by the standard k-ε turbulence model. The general
purpose CFD code Fluent 6 was adopted for the calculations.
A liquid-liquid flat interface was assumed as the initial condition. The computational
domain was divided into a structured hexahedral grid of about 5·105 cells over π. The
Multiple Reference Frame method was selected for treating the interaction of the stationary
baffles with the rotating impeller in a stationary fashion. The usually adopted numerical
settings for RANS-based simulations of baffled stirred tanks were selected (e.g. [17,18]).
The reference experimental conditions were those of the water-diesel fuel system at 19
vol. % of oil. The simulations were run at different impeller speeds varying from 2 s-1 up to
4.67 s-1 with a step size of 0.67 s-1.

3. RESULTS AND DISCUSSION


3.1 The hydrodynamics of the dilute liquid-liquid dispersion
The mean velocity fields of water and oil droplets in the measurement plane provide the well-
known picture of the double loop flow structure that has been either measured or predicted by
CFD methods in standard geometry stirred tanks a number of times in the past years. The
differences of the two-phase flow structure with respect to the single-phase case, though very

3
limited, can be appreciated by comparing the experimental velocity magnitude maps of water
in single- phase (Figure 1a) and two-phase conditions (Figure 1b) and of oil (Figure 1c).

(a) (b) (c)


Fig.1 Experimental maps of velocity magnitude and velocity vector plot at N=3.33 s-1. (a) water, single-phase;
(b) water, two-phase; (c) oil droplets. Colour scale units: m·s-1.

0.4 The first two maps reveal that the


shape of both the upper and lower
recirculation loops is modified only
0.2 slightly by the droplets. The centre of the
V/Vtip

flow recirculation in Figure 1(a) moves to


a different position in Figure 1(b); also, the
0
symmetry of the two loops typical of the
single-phase case is somewhat lost in the
-0.2 two-phase conditions.
0.0 0.2 0.4 0.6 0.8 1.0 As expected, the oil droplets follow the
z/T water motion closely and their slip with
Fig. 2 Comparison of mean radial velocity profiles. respect to water (Figure 1c) is very limited
◊ single-phase N=4.16 s-1; two-phase N=2.5 s-1; due to the small density difference between
∇ two-phase N=3.33 s-1 the two liquids and the tiny droplet size.

Therefore, the main information provided by the PIV results is that even a very small
amount of dispersed liquid modify the two-phase mean velocities though to a limited extent.
However, the impeller pumping capacity is not affected by the small amount of oil in the
system, as suggested by the results shown in Figure 2: the dimensionless radial velocity
profiles along the vessel height at r/T=0.19, i.e. just out from the impeller tip, are almost
overlapped for single-phase and two-phase conditions, also for slightly different values of
impeller speeds (they scale with Vtip as usually happens in fully turbulent single-phase
systems).
The effect of the dispersed phase on the turbulence characteristics of the continuous liquid
phase may be important to know especially for its direct outcome on mass transfer and

4
reaction rates. The radial r.m.s. velocity maps of plain water, water with 1 vol. % oil and oil
droplets are shown in Figure 3(a), (b) and (c), respectively.

(a) (b) (c)


Fig. 3 Experimental maps of r.m.s. radial velocity components at N=3.33 s-1. (a) water, single-phase; (b) water,
two-phase; (c) oil droplets. Colour scale units: m·s-1.

(a) (b) (c)


Fig. 4 Experimental maps of r.m.s. axial velocity components at N=3.33 s-1. (a) water, single-phase; (b) water,
two-phase; (c) oil droplets. Colour scale units: ms-1.

In the single-phase case, as expected, the r.m.s. values are much higher in the impeller
region than in the rest of the tank. The value of r.m.s. radial velocity fluctuation close to the
impeller tip is equal to 0.25 Vtip that is reasonably close to the value suggested by Lee and
Yianneskis [19] considering the size of our interrogation area. An increase in these values is
obtained with the oil dispersion in wide zones of the measurement plane, thus suggesting that
the action of the droplets is generally to promote turbulence. In turn, the droplets themselves
are characterised by radial r.m.s. velocity values greater than those of the continuous phase
where they are dispersed almost everywhere. The r.m.s. axial velocity components shown in

5
Figure 4 confirm the same behaviour of the radial ones. No dramatic differences between the
maps of the two components are found. The features of the two-phase turbulent characteristics
provide a useful guide in the evaluation of the simplified assumptions on which the available
turbulence models for the closure of RANS equations are based. In particular, efforts for
taking into account the effect of the dispersed phase on the continuous one and the capability
to predict the turbulent characteristics of the dispersed phase separately might be beneficial,
while the assumption of equal fluctuating velocity components appears to be less critical.

3.2 Characterisation of the dispersion conditions at different impeller speeds


Two typical pictures of completely separated phases and initial liquid drawdown conditions in
a portion of the vessel are shown in Figure 5.

(a) (b)
Fig. 5 Picture of the experimental oil-water system close to the interface. (a) N=2.5 s-1; (b) N=3.17 s-1.

At the lowest of the investigated impeller speeds and up to N=3 s-1 some large oil droplets
are entrained from the interface, but due to the action of the buoyancy force they do not reach
the impeller and quickly come back to the liquid-liquid interface. Above N=3 s-1, that may be
defined as the Nid condition, they reach the impeller and are broken into a number of little
droplets that are dispersed either below and above the impeller.

(a) (b) (c)


Fig. 6 Maps of computed oil hold-up at N=3.33 s-1. (a) droplet diameter 50 µm; (b) droplet diameter 200 µm; (c)
droplet diameter 1000 µm.

6
At a further increase of N, the continuous oil layer thickness decreases until it entirely
disappears at N equal to about 4.67 s-1, that may defined as the Njd.
The oil hold-up maps in a vertical plane midway between two baffles at N=3.33 s-1
predicted at different droplet diameters, i.e. by assuming different droplet rising velocities, are
shown in Figure 6. As can be observed, the effect of the droplet size is significant: when db is
assumed to be equal to 1000 µm the two liquids remain completely separated, while oil
entrainment is predicted for the two lower size cases. For db=200 µm, the oil is dispersed only
in the upper half of the vessel, while with a further decrease down to 50 mm the droplets are
present also below the impeller, thus matching the experimental observation more closely.
The simulation results relevant to the lower
impeller speeds, which are not reported here for
brevity reasons, show the absence of droplets
below the impeller in all cases but for N=2.67 s-
1
and db=50 µm, at which the droplets start to
appear though with a very low concentration.
The predicted just dispersed impeller speed is
close to the value observed experimentally only
for db=50 µm while a thick oil layer is still
present if the db was assumed to be equal to 200
µm, as can be clearly observed in Figure 7.
Overall, the simplified computational approach
adopted in this work, which cannot account for
the small scale phenomena contributing to the

(a) (b) oil dispersion, can roughly predict the dispersion


Fig. 7 Maps of computed oil hold-up at conditions. Clearly, the results are strictly
N=4.67 s-1. (a) db=50 µm; (b) db= 200 µm. dependent on the assumed droplet rising
Colour scale as in Figure 6. velocity, since the oil is allowed to go down only
when its rising velocity is smaller than the local mean velocity in the upper part of the vessel.

4. CONCLUSIONS
The liquid-liquid dispersion in a stirred vessel of standard geometry has been investigated
under specific conditions. The mean flow fields of both phases as well as the r.m.s. value of
axial and radial velocity fluctuation have been determined by means of a two-phase PIV
technique at low dispersed-phase hold-up. Generally, an increase in turbulence intensity of the
continuous liquid phase has been detected due to the presence of the dispersed one.
The evolution of the oil dispersion has also been monitored. Starting from the condition of
separate layers, the impeller speed was increased stepwise and the conditions of just
drawdown and just dispersed speeds have been identified experimentally. A simplified RANS
approach leads to a fair picture of the dispersion maps only when a suitable mean value of the
droplet size that is required for estimating the droplet rise velocity, is selected. The present
results suggest that improvements in the prediction capability might be expected by the
modelling of the droplet size distribution and the possible effect of turbulence on the droplet
rise velocity.

ACKNOWLEDGEMENTS
This work was financially supported by the Italian Ministry of University and Research
(MIUR) and the University of Bologna under PRIN 2006 programme.

7
REFERENCES
1. Leng D.E., Calabrese R.V., 2004. “Immiscible liquid-liquid systems”, in: Handbook of
Industrial Mixing: Science and Practice (E.L.Paul, V.A. Atiemo-Obeng and S.M. Kresta,
eds), Chapter 12, Wiley-Interscience, Hoboken, NJ, pp. 639-753.
2. Armenante P.M., Huang Y.-T., 1992. “Experimental determination of the minimum
agitation speed for complete liquid-liquid dispersion in mechanically agitated vessels”,
Ind. Eng. Chem. Res., 31, 1398-1406.
3. Calabrese R.V., Wang C.Y., Bryner N.P., 1986. “Drop breakup in turbulent stirred-tank
contactors. Part III: Correlations for mean size distribution, AIChE J., 32, 677-680.
4. Zhou G., Kresta S.M., 1998a. “Evolution of drop size distribution in liquid-liquid
dispersions for various impellers”, Chem. Eng. Sci., 53, 2099-2113.
5. Zhou G., Kresta S.M., 1998b. “Correlation of mean drop size and minimum drop size with
the turbulence energy dissipation and the flow in an agitated tank”, Chem. Eng. Sci., 53,
2063-2079.
6. Pacek A.W., Chamsart S., Nienow A.W., Bakker A., 1999. “The influence of impeller
type on mean drop size and drop size distribution in an agitated vessel, Chem. Eng. Sci.,
54, 4211-4222.
7. Giapos A., Pachatouridis C., Stamatoudis M., 2005. “Effect of the number of impeller
blades on the drop sizes in agitated dispersions”, Chem. Eng. Res. Des., 83, 1425-1430.
8. Bałdyga J., Bourne J.R., 1993, “Drop breakup and intermittent turbulence”, J. Chem. Eng.
Japan, 26, 738-741.
9. Bałdyga J., Podgórska W., 1998. “Drop break-up in intermittent turbulence: maximum
stable and transient sizes of drops”, Can. J. Chem. Eng., 76, 456-470.
10. Bałdyga J., Bourne J.R., Pacek A.W., Amanullah A., Nienow A.W., 2001. “Effects of
agitation and scale-up on drop size in turbulent dispersions: allowance for intermittency”,
Chem. Eng. Sci., 56, 3377-3385.
11. Alopaeus V., Koskinen J., Keskinen K.I., Majander J., 2002. “Simulation of the
population balances for liquid-liquid systems in a nonideal stirred tank. Part 2–parameter
fitting and the use of the multiblock model for dense dispersions”, Chem. Eng. Sci., 57,
1815-1825.
12. Maaß S., Gäbler A., Zaccone A., Paschedag A.R., Kraume M., 2007. “Experimental
investigations and modelling of breakage phenomena in stirred liquid-liquid systems”,
Chem. Eng. Res. Des., 85, 703-709.
13. Derksen J.J., Van den Akker H.E.A., 2007. “Multi-scale simulations of stirred liquid-
liquid dispersions”, Chem. Eng. Res. Des., 85, 697-702.
14. Van den Akker H.E.A., 2006. “The details of turbulent mixing process and their
simulation”, in: Advances in Chemical Engineering, vol. 31 (G.B. Marin, ed.), Academic
Press, pp. 151-229.
15. Svensson F.J.E., Rasmuson A., 2006. “PIV measurements in a liquid-liquid system at
volume percentages up to 10% dispersed phase”, Exp. Fluids, 41, 917-931.
16. Montante G., Horn D., Paglianti A., 2008. “Gas-liquid flow and bubble size distribution in
stirred tanks”, Chem. Eng. Sci., 63, 2107-2118.
17. Montante G., Magelli F., 2005. “Modelling of solids distribution in stirred tanks: analysis
of simulation strategies and comparison with experimental data”, Int. J. Comp. Fluid
Dynamics, 19, 253-262.
18. Montante G., Paglianti A., Magelli F., 2007. “Experimental analysis and computational
modelling of gas-liquid stirred vessels”, Chem. Eng. Res. Des., 85, 647-653.
19. Lee K.C., Yianneskis M., 1998. “Turbulence properties of the impeller stream of a
Rushton turbine”, AIChE J., 44, 13-24.

You might also like