Professional Documents
Culture Documents
www.elsevier.com/locate/ces
Abstract
Ejectors are gasliquid contactors that are reported to provide higher mass transfer rates than conventional contactors. Detailed experiments
were performed and computational uid dynamics (CFD) modeling studies were undertaken to understand the hydrodynamic characteristics
of the ejector geometry. The CFD model provides a basis for quantifying the effects of operating conditions on the ejector performance. CFD
studies shows that there is an optimum ratio of nozzle area to throat area (area ratio), at which the liquid entrainment rate is the highest. This
can lead to substantial economic benet in the industrial practice. The liquid entrainment rate correlates with pressure difference between the
water surface in the suction chamber and the throat exit for a wide variety of ejector geometries and operating conditions.
2007 Elsevier Ltd. All rights reserved.
Keywords: Hydrodynamics; Multiphase ow; Ejector; Entrainment; Pressure drop; CFD
1. Introduction
An ejector is a device in which a uid is pumped through a
nozzle at a high velocity (called the motive uid). The high velocity jet transports momentum to the outside uid and causes a
suction of the surrounding uid (called the entrained uid). The
mixing of the motive uid jet emerging from the nozzle and the
entrained uid in the mixing tube leads to the dispersion of one
phase into another in the throat of ejector. The diffuser section
after the mixing tube/throat helps in pressure recovery. Based
on the uids involved, three types of ejectors have been developed and been used over the years. The gasgas type ejectors are
used for the generation of vacuum. The liquidliquid ejectors
are used as mixing units in the mixersettler setup (Mukherjee
et al., 1988). Gasliquid ejectors use the jet of either gas or
liquid from the nozzle to entrain the other uid surrounding
the nozzle. The gasliquid ejectors are used in chemical industries for absorption and stripping (Ben Brahim et al., 1984) and
in biochemical industry. Ejectors produce high mass transfer
Corresponding author. Tel.: +91 22 2414 5616; fax: +91 22 2414 5614.
722
2. Experimental setup
A schematic diagram of the experimental setup is shown in
Fig. 1 with air as the motive uid and water as the entrained
Air outlet
Separator
Baffle
POUT
Pressure outlet boundary
condition (B.C.) (1 atm)
Straight tube/Column
Water
Pressure point P2
Two-phase
mixture
DC
Water
Diffuser
Liquid level
(LH)
HT
Throat
DT
Suction Tank
Converging section (inlet diameter = DEC)
Air mass flow rate B.C. at nozzle tip
Nozzle
DN
D0
PIN
Rotameter
Compressor
Air inlet
Valve
Fig. 1. Gasliquid ejector experimental apparatus for measuring liquid entrainment and boundary conditions used in CFD simulations.
Values
0.0254
0.004, 0.006, 0.008, 0.01, 0.012
0.02, 0.0254, 0.04
0.05, 0.1
0.06
1
0.25, 0.3, 0.35, 0.4, 0.45, 0.5
1013114 000
0.00260.026
the nozzle tip). The local pressure values were acquired using
an AD card in a personal computer for every second for about
10 min. These values were time averaged to get local average
pressure and hence the pressure drop across the two points.
The differential pressure measurements were used to calculate the gas hold-up using Eq. (1)(4):
P = mixture gh.
(1)
(2)
(3)
(4)
723
0.005 m. Peger and Becker (2001) have used a bubble diameter of 0.004 m. This value was considered to be constant
in all simulations. Similarly in stirred tank simulations, Deen
et al. (2002) have used 0.002 and 0.004 m bubble diameter and
Khopkar et al. (2005) have used 0.004 m constant bubble diameters. The droplet diameters in ejectors are generated by the
shear created near the nozzle region. Thus, the droplet diameter
will be different for different gas velocities.
The CFD simulations were carried out for two phase ow
in pipes. In the vertical pipe simulations, the droplet diameter
was varied so that the predicted pressure drop and the hold-up
match with the experimental data reported by Anderson and
Mantzouranis (1960) for a range of gas velocities. This enabled
us to establish a relationship between the droplet diameter and
the hydrodynamic conditions within the pipe and the gas velocity. This relationship (given later in the manuscript) has been
used in all further simulations of ejectors studied in this work.
The experimental data of Anderson and Mantzouranis (1960)
were used because the values of the gas hold-up reported in
their work ranged from 0.75 to 0.95, which are similar to those
observed in the present work. Thus, for all further ejector simulations, the droplet diameter is not a tted parameter making
the CFD simulations predictive in nature.
A 3D geometry of the pipe was created and meshed in
Gambit software with 0.2 million hexahedral structured cells.
More than the total number of cells, what is important is the
size of the grid in the region of high velocity gradients such
as the nozzle and the throat. In our previous work (Kandakure
et al., 2005, 2007), we have studied this aspect in detail. The
mesh size in the region close to the nozzle and the ejector wall
were small enough to capture the complex ow phenomenon.
During the initial stages of the simulations, course grids (0.08
million, 0.1 million) were used for simulation. In these simulations, the momentum and k. quantities did not converge to
the required convergence criteria. Simulations with 0.2 million
grid size gave good convergence. The entrainment rate and
liquid hold-up predicted using the CFD simulations with
0.2 million matched quite well with the experimental results.
All the CFD simulations were performed using Fluent 6.2 software. Since the ow is vertically upward in the pipe, the gravity
was taken in the negative Z direction and the operating pressure was taken as 1 atm. The standard k. model was used for
the modeling of turbulent behavior of the ow. FLUENT uses
its in-built slip velocity formulation given by Manninen et al.
(1996). The mass ow rates of air and water were given as the
inlet boundary condition. Since the pipe outlet is open to atmosphere, the outlet boundary was taken as 0 gauge pressure.
The no-slip boundary condition was enforced at the walls of
the pipe. The second order upwind discretization scheme was
used for the momentum, volume fraction, turbulent kinetic energy and turbulent energy dissipation rate and SIMPLE scheme
was used for the pressurevelocity coupling. For all variables
under-relaxation factor of 0.2 was used.
The CFD models used for the pipe simulations were used
to simulate the ejector. Fig. 1 also shows the boundary condition used in the ejector simulation. Air and water at room
temperature were considered as motive and entrained uids,
respectively. The air was assumed to obey an ideal gas law. The
mass ow rate of air, measured from experiments, was provided
at the nozzle tip as the boundary condition. The top liquid surface in the suction chamber was considered as the input with a
value of 0 (zero) gauge pressure. The ejector outlet is open to
atmosphere and, therefore, the pressure outlet boundary condition was used with 0 (zero) gauge pressure. For both the pipe
and ejector simulations, the solution was iterated until convergence was achieved, such that the residue for each equation fell
below 102 . In general, it was observed that the residue for
the momentum equations was below 103 ; that for the turbulent kinetic energy and turbulent energy dissipation rate were
well below 103 . The continuity equation residue was below
102 , while the residue for the liquid volume fraction was well
below 103 .
12000
724
8000
4000
0
0.000
0.002
0.004
Mass flow rate air, (kg/s)
0.006
0.002
0.004
Mass flow rate of air (kg/s)
0.006
dP = 1.778 10
VGT .
(5)
In this case, there are two major effects. As the gas velocity
increases, the shear rates in the throat increase. However, unlike other contactors, the liquid entrainment rate also increases
dramatically. This means that a larger quantity of liquid needs
to be dispersed in the gas. As a combined effect of these two
contrary effects, the liquid droplet diameter increases with the
gas velocity. In fact, the increase in liquid entrainment rate is
very large. There are no similar equations in the literature because his aspect has not been studied in the past literature. One
more point to be noted is that, CFD simulations were carried
out with mixture model in Fluent. In the mixture model, the interaction between the two phases is captured through the drag
forces acting on the droplets of the dispersed phase. The drag
0.95
For the vertical pipe, at particular ow rates of air and water, the simulation was performed to predict pressure drop and
gas hold-up. As discussed earlier, the droplet diameter was varied until the predicted values of pressure drop and gas hold-up
matched with the experimentally reported values. Similar procedure was followed for other ow rates of air and water.
Figs. 2A and B show the comparison of the predicted values
of pressure drop and gas hold-up with the experimental measurements of Anderson and Mantzouranis (1960). The gures
show that when the airow rate is increased, the pressure drop
and gas hold-up in the pipe also increases. Fig. 2B shows that
the predicted values of the gas hold-up compare well with the
experimental values. However, the experimental pressure drops
are slightly on the higher side (Fig. 2A). This could be because
Anderson and Mantzouranis (1960) measured the pressure drop
including the pressure drop across the air rotameter. Hence, the
pressure drop indicated in the experiments is actually a sum of
the pipe pressure drop and the rotameter pressure drop (not the
region of interest).
The droplet diameter values tted from the above simulations
were related to the supercial gas velocity in the following
manner:
0.85
0.75
0.65
0.000
725
0.5
0.4
Liquid hold-up (-)
1200
800
0.3
0.2
400
0.1
0
0.00
0.02
0.04
0.06
Mass flow rate air (kg/s)
0.08
0.10
0.96
120
140
160
Nozzle velocity (m/s)
180
200
Fig. 4. Comparison of liquid hold-up predicted by CFD with experimental results obtained with pressure measurements for DN =0.008 m, DT =0.0254 m,
HT = 0.1 m and LH = 0.5 m. : Experimental and : predicted.
0.92
0.88
0.84
0.00
0
100
0.02
0.04
0.06
Mass flowrate air (kg/s)
0.08
0.10
rate of 0.025 kg/s and the airow rate range from 0.0125 to
0.0875 kg/s were used. Figs. 3A and B compare the pressure
drop and gas hold-up values predicted by CFD with experimental values reported by Gill et al. (1965). The gures show that
the CFD predictions match quite well with those of the experiments. Hence, the relation developed to estimate the droplet
diameter was able to predict the pressure drops and the hold-up
values with good accuracy.
4.2. Ejector simulations
Eq. (5) developed using the vertical pipe simulations was
used for the ejector simulation. The air jet emerging from the
nozzle enters the ejector through the converging section and
it carries the entrained liquid along with it. The dispersion
726
0.7
QL (m3/hr)
0.6
0.5
0.4
0.3
75
125
Gas velocity (m/s)
175
5000
-5000
VG = 48 m/s
VG = 55 m/s
-10000
-15000
VG = 86 m/s
VG = 132 m/s
-20000
0.1
0.3
0.5
0.7
Axial distance (m)
0.9
1.1
Fig. 6A shows the comparison between the liquid entrainment rate predicted from CFD simulations with the experimental measurements for the nozzle diameter of 0.008 m,
throat diameter of 0.0254 m, throat height of 0.1 m and mass
ow rate of 0.0054 kg/s for different liquid levels. The liquid
entrainment rate increases with the increase in liquid level.
Fig. 6A shows that the predicted values of entrainment rate
match well with the experimental values as points. Fig. 6B
shows the variation of pressure at throat exit for different liquid levels for the same mass ow rate of 0.0131 kg/s. When
the liquid level was increased from 0.3 to 0.5 m, the pressure
at the throat exit decreased. The momentum generated by the
gas jet and the liquid head provided by the liquid level are the
driving forces for the entrainment rate. Hence, when the liquid level is increased, the pressure at the throat exit decreases
increasing the liquid entrainment rate.
4.5. Effect of area ratio
The ratio of nozzle area (AN ) to throat area (AT ) (area ratio) is one of the important parameters for the ejector design.
Fig. 7A shows both the predicted and measured liquid entrainment rates at a constant gas mass ow rate of 0.0031 kg/s for
different area ratios ranging from 0.025 to 0.893 (nozzle diameters from 0.004 to 0.024 mm for a xed throat diameter of
0.60
QL (m3/hr)
0.40
0.20
0.00
0.3
0.4
Liquid level (m)
0.5
-14000
-16000
-18000
-20000
-22000
727
gas jet. This causes a reduction in the annular area available for
water to ow (Fig. 7E). When the nozzle diameter approaches
the diameter of the throat, the area available for the liquid to
ow decreases. Water starts to climb on the outer walls of the
ejector due to reduction in the space available for entrained
liquid to ow into the ejector. Kandakure et al. (2005) have
reported similar optimum area ratio and shown a recirculation
pattern inside the ejector as the reason for reduction in the entrainment rate at higher nozzle diameters. It has been reported
that at high nozzle diameters (high value of AT /AN ), a substantial amount of recirculation occurs within the converging
section of the ejector. This is primarily due to the reduction in
available area for the ow of the entrained uid.
The optimum nozzle diameter (0.010 m) from the CFD predictions is in good agreement with the experimental results.
Several authors also have reported in the past that there is an
optimum value of the area ratio, however, no explanations were
provided by them. The optimum DN /DT ratio in this work corresponds to 0.393 and is in good agreement with the reported
values in the literature. Biswas and Mitra (1981) have reported
optimum DN /DT ratio to be in the range of 0.2230.258. Rylek
and Zahradnik (1984) have reported the optimum DN /DT ratio to be 0.33. Bando et al. (1990) have reported the optimum DN /DT ratio to be 0.50.l6, for a HT /DT ratio of 20.
Zahardnik et al. (1997) have reported that, as the nozzle diameter approaches the throat diameter, the entrainment rate decreases because the throat gets entirely lled with liquid for a
gasliquid system.
4.6. Effect of throat height
-24000
-26000
0.3
0.35
0.4
Liquid level (m)
0.45
0.5
The effect of throat height of the ejector on the liquid entrainment rate was predicted with the help of the current model. It
was observed that the entrainment rate increases with decrease
in HT . Fig. 8 shows the effect of HT on the pressure prole
along the center line of the ejector. An ejector with no throat
was observed to have the maximum entrainment rate. This is
because the pressure value at the throat exit was smallest in this
case. This means the driving force for the suction that is the
pressure difference between the pressure at throat exit and atmosphere decreases with increase in throat height. Hence, the
entrainment rate decreases with increase in throat height. Similar results were reported by Kandakure et al. (2005) from the
CFD simulations of liquid jet ejectors. Henzler (1983), from
his review work on single phase horizontal ejectors, has given
the optimum ejector dimensions. He has observed that the optimum HT /DT ratio is 7.5 for (AT /AN ) of less than 2.5. Similar observations were made by Rylek and Zahradnik (1984).
The (HT /DT ) ratio used in Rylek and Zahradnik (1984) experiments ranged from 1.25 to 3.9. Bando et al. (1990) reported
that the optimum range of HT /DT ratio is 2030. Similar to the
present work results, Dirix and van der Wiele (1990) have reported that when throat height was increased, a decrease in the
efciency of the ejector was observed due to less entrainment
rate of gas. Havelka et al. (1997) have reported that the entrainment rate of gas increased with an increase in throat height.
However, they have also observed that the further increase in
728
0.40
0.30
0.20
0.10
0.00
0.0
0.2
0.4
0.6
0.8
1.0
AN/AT
4000
0
-4000
AR = 0.893 (DN = 0.024 m)
-8000
-12000
-16000
-20000
0.1
0.3
0.5
0.7
Axial distance (m)
0.9
1.1
Fig. 7. (A) Effect of area ratio (AN /AT ) on entrainment rate and throat pressure drop for LH = 0.5 m, HT = 0.1 m, DT = 0.0254 m and gas
mass ow rate = 0.0054 kg/s. : Experimental and : predicted. (B) Effect of area ratio (AN /AT ) on pressure at throat exit for LH = 0.5 m,
:
HT = 0.1 m, DT = 0.0254 m and gas mass ow rate = 0.0054 kg/s. : AR = 0.025 (DN = 0.004 m);
AR = 0.155 (DN = 0.010 m); - - - - - - - - - - - - -: AR = 0.223 (DN = 0.012 m) and : AR = 0.893 (DN = 0.024 m). (C) Predicted liquid volume fraction prole for LH = 0.5 m, HT = 0.1 m, DT = 0.0254 m and gas mass ow rate = 0.0054 kg/s for DN = 0.004 m.
(D) Predicted liquid volume fraction prole for LH = 0.5 m, HT = 0.1 m,
0
0.5
1.0
DT = 0.0254 m and gas mass ow rate = 0.0054 kg/s for DN = 0.010 m.
(E)
0
0.5
1.0
Predicted liquid volume fraction prole for LH = 0.5 m, HT = 0.1 m, DT = 0.0254 m and gas mass ow rate = 0.0054 kg/s for DN = 0.024 m.
0.5
1.0
order to nd the predictive ability of the model to other geometries, it was decided to predict the performance of ejectors
reported in the past literature. For this purpose, the ejector geometry reported by Davies et al. (1967) was used for the CFD
simulations. The liquid entrainment rates reported by Davies
et al. (1967) were compared with the CFD predictions. Davies
et al. (1967) have preformed experiments with a gas jet ejector
729
Pressure outlet
HT = 0.2
HT = 0.1
HT = 0
0
0.1
0.3
0.5
0.7
0.9
1.1
DT = 1.56
-20000
-40000
DC = 0.0381
-60000
Axial distance (m)
Fig. 8. Effect of throat height on pressure prole from the center line of
ejector with the axial locations from the nozzle inlet for DN = 0.008 m,
:
DT = 0.0254 m, LH = 0.50 m. - - - - - - - - -: HT = 0;
HT = 0.1 and : HT = 0.2.
0.2667 m
DT = 0.0127
DN = 0.00095, 0.00133
Mass flow inlet
Water inlet,
Diameter = 0.01905 m
0.14
Liquid Entrinament rate (kg/s)
0.159 m
0.12
AR = 0.01
0.10
AR = 0.0056
0.08
0.06
0.04
0.02
0.00
0
50
100
150
Gas velocity (m/s)
200
250
730
0.7
0.6
0.5
0.4
0.3
0.2
0.1
0
0
5000
10000
15000
Pressure at Throat Exit (Pa)
20000
the total pressure drop across the ejector is the same as the
pressure before the nozzle inlet (PIN ). This was measured experimentally (the measurement location has been indicated in
Fig. 1). It can be seen from the gure that, for the same mass
ow rate, when the nozzle diameter was increased from 6 to
12 mm, there was a substantial reduction in the pressure drop.
For example, for a mass ow rate of 0.0075 kg/s, an increase
in the nozzle diameter from 6 to 12 mm leads to a reduction in
the pressure drop from 1.5 105 to about 1.05 105 N/m2 .
This reduction is close to 30%. Combination of Figs. 11A and
B shows that for a given mass ow rate of air, the extent of
contacting can be increased (higher L/G ratio) and simultaneously operating cost can be reduced (lower pressure drop) by
using an optimized value of nozzle diameter.
6. Conclusions
Fig. 10. Liquid entrainment rate correlation with pressure at the throat exit.
40
10 mm
L/G
30
20
4mm
10
6 mm
8 mm
12 mm
0
0.000
0.006
0.012
Gas mass flow rate (kg/s)
0.018
2.0E+05
6 mm
4 mm
1.5E+05
8 mm
Notation
10 mm
12 mm
1.0E+05
0.000
0.006
0.012
Gas mass flow rate (kg/s)
0.018
Fig. 11. (A) Effect of nozzle diameter on L/G ratio for constant DT =0.0254 m,
HT = 0.1 m, LH = 0.5 m. : DN = 0.004 m; :
: DN = 0.010 m
DN = 0.006 m; : DN =0.008 m;
and - - - - - - - - -: DN = 0.012 m. (B) Effect of nozzle diameter on total pressure drop across the ejector for DT = 0.0254 m, HT = 0.01 m and
LH = 0.5 m. : DN = 0.004 m; : DN = 0.006 m; : DN = 0.008 m; :
DN = 0.010 m; and : DN = 0.012 m.
AR
dP
DC
DN
DT
g
h
HT
LH
P
PIN
POUT
VG
VGT
Greek letters
G
L
G
L
mixture
Acknowledgment
The authors would like to acknowledge the nancial support
in the form of research grant & fellowship from the Department
of Atomic Energy (DAE), India (Project no. 47.01).
References
Anderson, G.H., Mantzouranis, B.G., 1960. Two phase (gasliquid) ow
phenomenaI, pressure drop and hold-up prole for two-phase ow in
vertical tubes. Chemical Engineering Science 12, 109126.
Balamurugan, S., Gaikar, V.G., Patwardhan, A.W., 2006. Hydrodynamic
characteristics of gasliquid ejectors. Chemical Engineering Research and
Design 84, 11661179.
Bando, Y., Kuraishi, M., Nishimura, M., Takeshita, I., 1990. The
characteristics of a bubble column with a gas-suction type, simultaneous
gasliquid injection-nozzle. International Chemical Engineering 30,
729737.
Ben Brahim, A., Prevost, M., Bugarel, R., 1984. Momentum transfer in
a vertical down ow liquid jet ejector: case of self gas aspiration and
emulsion ow. International multiphase ow 10 (1), 7994.
Biswas, M.N., Mitra, A.K., 1981. Momentum transfer in horizontal multijet liquidgas ejector. Canadian Journal of Chemical Engineering 59,
634637.
Cramers, P.H.M.R., Beenackers, A.A.C.M., 2001. Inuence of the ejector
conguration, scale and the gas density on the mass transfer characteristics
of gasliquid ejectors. Chemical Engineering Journal 82, 131141.
Davies, G., Mitra, A.K., Roy, A.N., 1967. Momentum transfer studies
in ejectors. Industrial and Engineering Chemistry Process Design and
Development 6 (3), 293299.
731
Deen, N.G., Solberg, T., Hjertager, B.H., 2002. Flow generated by an aerated
Rushton impeller: two-phase PIV experiments and numerical simulations.
Canadian Journal of Chemical Engineering 80 (4), 638652.
Dirix, C.A.M.C., van der Wiele, K., 1990. Mass transfer in jet loop reactors.
Chemical Engineering Science 45, 23332340.
Gill, L.E., Hewitt, G.F., Lacey, P.M.C., 1965. Data on the upwards annular
ow of airwater mixtures. Chemical Engineering Science 20, 7188.
Havelka, P., Linek, V., Sinkule, J., Zahradnik, J., Fialova, M., 1997. Effect
of the ejector conguration on the gas suction rate and gas hold-up in
ejector loop reactors. Chemical Engineering Science 52, 17011713.
Havelka, P., Linek, V., Sinkule, J., Zahradnik, J., Fialova, M., 2000.
Hydrodynamics and mass transfer characteristics of ejector loop reactors.
Chemical Engineering Science 55, 535549.
Henzler, H., 1983. Design of ejectors for single phase material systems.
German Chemical Engineering 6, 292300.
Kandakure, M.T., Gaikar, V.G., Patwardhan, A.W., 2005. Hydrodynamic
aspects of ejectors. Chemical Engineering Science 60, 63916402.
Kandakure, M.T., Patkar, V.C., Patwardhan, A.W., 2007. Characteristics
of turbulent conned jets. Chemical Engineering Processing, in press,
corrected proof available online 6 April 2007.
Khopkar, A.R., Rammohan, A.R., Ranade, V.V., Dudukovic, M.P., 2005.
Gasliquid generated by a Rushton turbine in stirred vessel: CARPT/CT
measurements and CFD simulations. Chemical Engineering Science 60,
22152229.
Manninen, M., Taivassalo, V., Kallio, S., 1996. On the mixture model for
multiphase ow. VTT Publications 288, Technical Research Centre of
Finland.
Mukherjee, D., Biswas, M.N., Mitra, A.K., 1988. Hydrodynamics of
liquidliquid dispersion in ejectors and vertical two phase ow. The
Canadian Journal of Chemical Engineering 66, 896907.
Pan, V., Dudukovic, M.P., Chang, M., 1999. Dynamic simulation of bubbly
ow in bubble columns. Chemical Engineering Science 54, 24812489.
Peger, D., Becker, S., 2001. Modelling and simulation of the dynamic
ow behaviour in a bubble column. Chemical Engineering Science 56,
17371747.
Rylek, M., Zahradnik, J., 1984. Design of Venturi-tube gas distributors
for bubble type reactors. Collection of Czechoslovak Chemical
Communications 49, 19391948.
Sanyal, J., Vasquez, S., Roy, S., Dudukovic, M.P., 1999. Numerical simulation
of gasliquid dynamics in cylindrical bubble column reactors. Chemical
Engineering Science 54 (21), 50715083.
Zahradnik, J., Fialova, M., Linek, V., Sinkule, J., Renickoca, J., Kastanek,
F., 1997. Dispersion efciency of ejector-type gas distributors in different
operating modes. Chemical Engineering Science 52 (24), 44994510.