Académique Documents
Professionnel Documents
Culture Documents
} (1.2)
where the instantaneous velocity is decomposed into mean and uctuating components,
u = U +u
(1.3)
and
U =
1
t
_
t+t
t
udt (1.4)
=Density, u=velocity vector, U =average velocity,u
eff
_
U +U
T
__
(1.5)
eff
is the eective viscosity accounting for turbulence, and is given by
eff
= +
t
(1.6)
p is the modied pressure and is given by dened by
p = p +
2
3
k +
2
3
t
U (1.7)
where
t
is the turbulent viscosity. For k model, turbulent viscosity is given by
t
= C
k
2
(1.8)
C
is a constant with value 0.09. k is the turbulent kinetic energy and is dened as the
variance of the uctuations in velocity. is the turbulence eddy dissipation (rate at which
turbulence kinetic energy is dissipated). The values of k and come directly from the dif-
ferential transport equations for the turbulence kinetic energy and turbulence dissipation
rate:
(k)
t
+ (Uk) = [
_
+
t
k
_
k] +P
k
(1.9)
()
t
+ (U) = [
_
+
t
_
] +
k
(C
1
P
k
C
2
) (1.10)
Kareem Akhtar Chapter 1 9
Where C
1
, C
2
,
k
and
2
3
U (3
t
U +k) +P
kb
(1.11)
P
kb
is buoyancy production term and depends on the buoyancy turbulence, which are given
in CFX manual. Based on k, SST model accounts for the transport of the turbulent shear
stress and gives highly accurate predictions of the onset and the amount of ow separation
under adverse pressure gradients. Although, the Baseline (BSL) k model combines the
advantages of theWilcox k and the k model, but still fails to properly predict the
onset and amount of ow separation from smooth surfaces. The main reason is that both
models do not account for the transport of the turbulent shear stress. This results in an over
prediction of the eddy-viscosity. The proper transport behavior can be obtained by a limiter
to the formulation of the eddy-viscosity given by
t
=
a
1
k
max(a
1
, SF
2
)
(1.12)
where S is an invariant measure of the strain rate. F
2
is a blending function, which restricts
the limiter to the wall boundary layer, as the underlying assumptions are not correct for free
shear ows. The blending function is given by
F
2
= tanh
_
2
2
_
(1.13)
2
= max
_
2
y
,
500v
y
2
_
(1.14)
Where y is the distance to the nearest wall, is the kinematic viscosity, is the turbulent
frequency and
t
(r
) + (r
) = 0 (2.1)
t
(r
) + (r
)) = r
+
_
r
_
U
+ (U
)
T
__
+r
g +M
(2.2)
where refers to the phase, r
=
M
(2.3)
and
M
= M
D
+M
L
+M
LUB
+M
TD
+.... (2.4)
Kareem Akhtar Chapter 2 37
Interphase Drag
The following general form is used to model interphase drag force acting on phase due
to phase :
M
= c
(d)
(U
) (2.5)
c
(d)
=
3
4
C
D
d
r
|U
beta
U
alpha
| (2.6)
For spherical particles the Schiller Naumann Drag Model the drag coecient C
D
is given
by
C
D
=
24
Re
_
1 + 0.15Re
0.687
_
(2.7)
The multiphase versions of turbulence models are equivalent to the single phase version,
with all ux and volumetric source terms multiplied by volume fractions. The eddy viscosity
hypothesis is assumed to hold for each turbulent phase. Diusion of Momentum in phase
is governed by an eective viscosity:
eff
=
+
t
(2.8)
For the k model, the turbulent viscosity is modeled as:
t
= c
_
k
_
(2.9)
Kareem Akhtar Chapter 2 38
The transport equations for k and in a turbulent phase are assumed to take a similar
form to the single-phase transport equations:
(r
)
t
+
_
r
_
+
t
k
_
k
__
= r
(P
) +T
(k)
(2.10)
()
t
+ (U) = [
_
+
t
_
] +
k
(C
1
P
k
C
2
) (2.11)
where C
1
, C
2
,
k
and
tc
=
ts
+
tp
(2.12)
2.3 Mixing Layer Geometric Details
The two-dimensional computational domain used in simulations is shown in gure 2.1(a).
It has the same dimensions as the test section of the mixing layer facility used by Roig et
al. (1997). It is a rectangular channel of width of (2b) = 400mm and height of H = 2.3m.
A splitter is inserted in the middle dividing the channel into two sides, side 1 to the right
and side 2 to the left of the plate. The geometric details of splitter plate trailing edge are
shown in gure 2.1(b). The plate height is 300mm, and its thickness is 2mm over most of
its length, except for the last 10mm where its thickness drops to 0.5mm at the trailing edge.
Kareem Akhtar Chapter 2 39
This design closely matches the trailing edge, which is described as a cusp, in the Roig et al.
(1997) experimental facility. The simulations reported here show that no vortices are shed
from the trailing edge of the splitter plate.
Figure 2.1: Mixing layer geometry and boundary conditions
2.4 Boundary Conditions and Grid Description:
The commercial software ANSYS ICEM CFD 12.0 is used for grid generation. The grid is
block structured as shown in gures 2.2 and 2.3. Two grids, denoted as coarse and ne grids,
are used to check the eects of grid resolution on the simulations accuracy; the number of
elements is shown in table 2.1. A Cartesian coordinate system is introduced with origin at the
trailing edge of the splitter plate, and the x-axis is vertically upward (streamwise direction)
and the y-axis is horizontal (transverse to the mixing layer). The gravitational acceleration
Kareem Akhtar Chapter 2 40
is downward. The boundary conditions are labeled in gure 2.1. We have three types of
boundary conditions: solid walls, inlets and outlet. The two vertical side walls and the sur-
face of splitter plate are assumed to be no-slip walls for both phases. At the bottom we have
two velocity inlets conditions; one on each side of the splitter plate, and at the top we impose
a pressure outlet condition. Subscripts L and G are used to denote liquid and gas phases,
respectively. Subscripts 1 and 2 refer to the right (y > 0) and left (y < 0) sides of the splitter
plate, respectively.
Figure 2.2: (a)coarse grid (b)coarse grid (c) coarse grid
Kareem Akhtar Chapter 2 41
Figure 2.3: (a)ne grid (b) trailing edge of splitter
Domain Nodes Elements
Coarse grid 274,400 136,145
Fine grid 940,800 468,476
Table 2.1: Fine and coarse grid details
2.5 Numerical Simulations Parameters:
Dierent numerical runs were performed by changing the void fraction () and magnitude of
inlet velocities. ANSYS CFX 12.0 is used to solve the RANS two-phase ow equations. The
parameters of dierent numerical runs are given in table 2.2; they correspond to conditions of
Roig et al. (1999) experiments. Experiment 2-1 is purely single phase (liquid water) because
the void fraction in both sides is zero. The void fraction () in each experiment is less than
Kareem Akhtar Chapter 2 42
Experiment 2-1 2-2 2-3 2-4
10
% 0.0 1.9 1.9 0.0
U
L10
(m/s) 0.615 0.53 0.51 0.58
20
% 0.0 1.9 0.0 1.5
U
L20
(m/s) 0.255 0.23 0.18 0.19
U
L0
= U
L10
U
L20
0.36 0.3 0.33 0.34
U
Lm0
=
(U
L10
U
L20
)
2
0.435 0.38 0.345 0.385
0
=
U
L0
2U
Lm0
0.414 0.395 0.478 0.506
L
24 11 13 14
Table 2.2: Inlet conditions for dierent runs
2%. The liquid velocity in each experiment is less than 1 (m/s). The mean bubble diameter
is 2mm. The Reynolds number Re = U
Lo
x
L
varies from 1.910
4
to 4.410
5
indicating that
the ow is fully turbulent, where x is streamwise distance measured from the splitter plate
trailing edge, as shown in gure 2.1(a), and U
Lo
is velocity dierence across the mixing
layer.
Results
Results in this chapter are divided into seven sections: liquid (water) supercial velocity
proles, mixing layer growth rate, gas supercial velocity proles, void fraction, ne grid sim-
ulations, comparison between dierent turbulence models and comparison with experimental
data.
Kareem Akhtar Chapter 2 43
2.6 Liquid (Water) Supercial Velocity Proles:
The water volume fraction is greater than 98% in each case. So, the water ow gives us the
global structure of the ow. A single phase simulation was performed rst. It is expected
that RNAS model works well for single ow as reported in the literature. The water velocity
contours for single phase ow run 2-1 are shown in gure 2.4. The red color shows the high
velocity side. For two-phase ows, the supercial velocity counters of liquid in runs 2-2 and
2-4 are shown in gures 2.5 and 2.6, respectively. The supercial velocity is the product of
the local velocity and local volume fraction. Contours for run 2-4 are signicantly dierent
from those for run 2-2. This is due to the dierence in the inlet values of void fraction. At
the inlet, the void fraction is zero in the high speed side for run 2-4, whereas for run 2-2, the
void fractions are equal on both sides of the mixing layer. The mixing layer drifts towards
the side of higher void fraction. This result is supported by experimental data. Even though
the void fraction in run 2-4 is only 1.5 %, yet it has a strong eect on the overall behavior of
the mixing layer, hence indicating how the ow is sensitive to the void fraction.
Kareem Akhtar Chapter 2 44
Figure 2.4: Velocity contours for single phase ow run 2-1
Figure 2.5: Water supercial velocity for run 2-2
Kareem Akhtar Chapter 2 45
Figure 2.6: Water supercial velocity for run 2-4
The water velocity proles for single-phase ow run 2-1 at dierent streamwise stations
x=-1cm, 20cm, 30cm, and 40cm are shown in gure 2.7. The proles show the spreading
of the mixing layer with the streamwise distance. SST turbulence model is used here. The
supercial velocity proles for two-phase ow runs 2-2 and 2-4 at dierent stations are plotted
in gures 2.8 and 2.9, respectively. An important observation can be seen that the velocities
to the left and right of the mixing layer almost remain constant and equal to inlet conditions;
only the mixing layer thickness increases. The same observation can be seen in the single
phase ow.
Kareem Akhtar Chapter 2 46
Figure 2.7: Water velocity proles for single phase run 2-1
Figure 2.8: Water supercial velocity proles for run 2-2
Kareem Akhtar Chapter 2 47
Figure 2.9: Water supercial velocity proles for run 2-4
2.7 Mixing Layer Growth Rate:
In order to quantify mixing layer growth for single-phase and two-phase runs, we introduce
non-dimensional similarity variables U
+
L
and
L
dened by
U
+
L
=
U
L
U
L2
U
L1
U
L2
(2.13)
L
=
L
y y1
2
x
(2.14)
Where y
1/2
is dened by U
L
(y
1/2
) = U
Lm
and
L
is a measure of mixing layer growth rate,
which will be compared with the experimental lateral expansion. If the spreading rate of the
mixing layer is measured by the angle as shown in gure 2.10, then we dene
L
as
L
=
1
tan
L
=
1
L
(2.15)
Kareem Akhtar Chapter 2 48
Figure 2.10: Mixing layer thickness
So a small value of
L
means greater expansion rate of mixing layer. A good approximation
of the velocity prole of single phase mixing layer is the Gortlers prole (Schlichting, 1979
seveth edition), which is given by
u(
L
) = 0.5[1 +erf(
L
)] (2.16)
where erf(
L
)is the error function given by
erf() =
2
_
0
e
2
dx (2.17)
In gures 2.11, 2.12, 2.13 show a comparison between the computed velocity proles, ex-
pressed in terms of similarity variables U
+
L
and
L
, and Gortlers prole. For each run, the
value of (
L
) is determined to obtain the best match between the two proles. This was done
by varying
L
and then visually inspecting the plots of the two proles. The best value for
L
for single phase is 24. Extending the same concept for two phase ow, we present liquid
velocity proles for two phase runs 2-2 and 2-4 in gures 2.12 and 2.13, respectively. For run
2-2 the value of
L
was found to be 21 while for run 2-4 having bubbles only on one side the
value of
L
was found to be 19. This indicates that the spreading rate measured by 1/
L
is
Kareem Akhtar Chapter 2 49
dierent from single phase ow and depends on the void fraction. The dierence is due to
the presence of bubbles which generates buoyancy eects in the ow and density dierence
between the two sides. We note that the prole at the edges i.e. near |
L
| = 1 is not matching
well with Gortlers prole. The comparison is more favorable as we rene the grid as will be
shown in section 2.10.
Figure 2.11: Mean velocity prole for single phase run 2-1
Kareem Akhtar Chapter 2 50
Figure 2.12: Mean water supercial velocity for two-phase run 2-2
Figure 2.13: Mean water supercial velocity for two-phase ow run 2-4
Kareem Akhtar Chapter 2 51
2.8 Gas Supercial Velocity Proles:
The mean supercial air velocity proles for run 2-2 and run 2-4 at dierent streamwise
locations x = 6cm, 20cm, 30cm, and 40cm are shown in gures 2.14 and 2.15, respectively.
As show in these gures the air velocity is high compared to liquid velocities. The slip
velocity (U
G
U
L
) for run 2-4 (having air bubble only in one side) is 0.21 m/s. There is
a peak in the velocity prole for run 2-4, which corresponds to a peak in the void fraction.
The experimental work by Roig et al.(1999) also show some peaks of void fraction for some
sections but not in the entire ow.
Figure 2.14: Mean supercial air velocity proles for run 2-4
Kareem Akhtar Chapter 2 52
Figure 2.15: Mean supercial air velocity proles for run 2-2
2.9 Void Fraction:
Void fraction proles are plotted in the gure 2.16. In the experimental work performed by
Roig et al. ( 1999 and 2006) peaks in void fraction that depends on the inlet conditions are
also observed. These peaks are due to the boundary layer developing on the splitter plate,
which depend on inlet conditions.
Figure 2.16: Void fraction run 2-4
Kareem Akhtar Chapter 2 53
2.10 Fine Grid Simulations:
Fine grid simulations are run to make sure that our calculations are independent of grids.
The computational domain and all boundary conditions are identical to that of the coarse
grid. The value of
L
is calculated for ne grid. We nd
L
= 18 for run 2-4 (for coarse grid it
is 19) and
L
= 20 for run 2-2 (for coarse grid the value is 21). The water supercial velocity
in similarity variables are plotted in gures 2.17 and 2.18 for run 2-4 and 2-2, respectively.
Better agreement between simulations and Geortlers prole is evident.
Figure 2.17: Mean velocity in the liquid phase run 2-4
Kareem Akhtar Chapter 2 54
Figure 2.18: Mean velocity in the liquid phase run 2-2
2.11 Comparison Between Dierent Turbulence Models:
The eects of turbulence models on RANS predictions of two-phase mixing layers are reported
here. Three models known as SST model, k model and SSGR model are tested. Results
show that SST model gives better prediction in comparison to other models. The k model
prediction are close to SST model, but SSGR model predictions show large deviation in the
spreading rate of bubbly mixing layer. Figure 2.19 shows plots of water velocity proles
for three turbulence models compared with Geartlers law. The plots indicate that the SST
model gives better agreement with Geortlers law for
L
equal to 19. The k model deviates
slightly from SST but the SSGR model deviates a lot for the same value of
L
.
Kareem Akhtar Chapter 2 55
Figure 2.19: Liquid supercial velocity proles for dierent turbulence models
2.12 Comparison with Experimental Data:
Comparison is being made with the experimental work performed by the Roig et al. (1999).
The value of
L
is for single phase ow in the numerical simulations was found to be 24 which
agrees with the experimental value reported by Roig et al. The velocity proles in similarity
variables are compared in gure 2.20. For two-phase ow RANS simulations considerably
underestimate the growth rate of the mixing layer. The values of
L
in the experiments
for run 2-4 and 2-2 are 14 and 11, respectively, as compared to 20 and 18. If we use the
experimental values for scaling the RANS model predictions, we obtain the proles shown in
gures 2.21 and 2.22. Only if use the higher values of
L
, we can bring the RANS models
Kareem Akhtar Chapter 2 56
Run Experimental Value Numerical Value ( ne Grid)
2-2 11 20
2-4 14 18
Table 2.3: Numerical and experimental vlaues for
L
simulations to match with the Gortler prole or experimental data prole.
Figure 2.20: Comparison with the experiments for single phase ow
To investigate the turbulence kinetic energy, we plot the longitudinal velocity uctuations
of the liquid velocity in a non-dimensional form. u
/
1
2 =
u
/
L
2
U
2
L
The prole u
/
1
2 agianst
L
is
shown in the gure 2.23. The plot indicates that there is large dierence in the turbulence
kinetic energy between the experimental results and numerical results. This increase in
velocity or turbulence kinetic energy is due to turbulence induced by the bubbles, which is
Kareem Akhtar Chapter 2 57
Figure 2.21: Comparison with the experiments two-phase run 2-4
Figure 2.22: Comparison with the experiments two-phase run 2-2
Kareem Akhtar Chapter 2 58
not captured by the RANS models.
Figure 2.23: Turbulent energy comparison with experimental data for run 2-4
2.13 Conclusions:
We tested RANS models for a multiphase ow given by a bubbly mixing layer. The simula-
tions were performed with two grid resolutions. RANS models including SST model, k
model and SSGR models were used in simulations for single phase (water only) and multi-
phase (air and water) ow. The mixing layer thickness was calculated, and the results are
compared with experiments. For single phase ow the numerical predictions agree with the
experimental data for mean velocity and spreading rate. For multiphase the RANS models
predictions do not agree with experimental work. Numerical results show that the SST model
gives slightly better predictions for multiphase ow than the k and SSGR models. It is
Kareem Akhtar Chapter 2 59
found that RANS model do not predict the turbulence kinetic energy correctly. There is a
signicant dierence in turbulence kinetic energy of numerical predictions and experimental
data. There are also dierences in the liquid mean velocity proles. The failure of local
turbulence models are also reported by Lakehal et al (2002). The possible reason may be
that RANS work on constant bubble diameter, but in reality the bubble diameter varies and
has a distribution. The constant diameter of bubble ignores the coalescence and break up of
bubbles that induce local turbulence in the ow and changes the kinetic energy of the carrier
phase. The argument is presented by Min et al (2008) that a model with constant single av-
erage bubble diameter (SABD) does not predict correctly void fraction. The RANS models
need further investigation to include turbulence generation due to breakup and coalescence.
Kareem Akhtar Chapter 2 60
References:
Ayed, H., Chahed, J., and Roig, V., Hydrodynamics and mass transfer in a turbulent
buoyant bubbly shear layer AIChE Journal Volume 53 Issue 11, Pages 2742 - 2753.
Colin, C., Legendre, D., and Fabre, J., Bubble distribution in a turbulent pipe ow
Microgravity Research and Applications in Physical Sciences and Biotechnology, Proceedings
of the rst international symposium held 10-15 September, 2000 in Sorrento, Italy. edited by
Minster, O., and Schrmann, B., European Space Agency, ESASP 454, 2001., p.91
Dhotre, M.T., Niceno, B., and Smith, B.L., Large eddy simulation of a bubble column
using dynamic sub-grid scale model Chemical Engineering Journal Volume 136, Issues 2-3,
1 March 2008,Pages 337-348.
Dr. Schlichting, H., Boundary-Layer Theory translated by Dr. J. Ketin seventh Edition
[pages 737 and 738]
Hu,G., and Celik, I., Eulerian-Lagrangian based large-eddy simulation of a partially
aerated at bubble column Chemical Engineering Science Volume 63, Issue 1, January 2008,
Pages 253-271.
Lakehal, D., Smith, B.L., and Milelli, M., Large-eddy simulation of bubbly turbulent
shear ows Journal of Turbulence, Volume 3, N 25 May 2002.
Martnez-Bazn, C., Montas, J.L. and Lasheras,J.C., Statistical description of the bubble
cloud resulting from the injection of air into a turbulent water jet International Journal of
Multiphase Flow Volume 28, Issue 4, April 2002, Pages 597-615.
Min, J. Bao, Y., Chen, L., Gao, Z., and Smith, J.M., Numerical simulation of gas
dispersion in an aerated stirred reactor with multiple impellers Ind. Eng. Chem. Res.,
Kareem Akhtar Chapter 2 61
2008, 47 (18), pp 7112-7117.
Neto, I.E.L, Zhu, D.Z., and Rajaratnam, N., Bubbly jets in stagnant water International
Journal of Multiphase Flow Volume 34, Issue 12, December 2008, Pages 1130-1141.
Niceno, B., Boucker, M., and Smith, B.L., Euler-Euler large eddy simulation of a square
cross-sectional bubble column using the Neptune CFD code Hindawi Publishing Corporation
Science and Technology of Nuclear Installations Volume 2009, Article ID 410272.
Rightley, P.M., and Lasheras, J.C. Bubble dispersion and interphase coupling in a free-
shear ow J. Fluid Mech. (2000), vol. 412, pp. 21-59.
Roig, V., Suzanne, C., and Masbernat, L., Experimental investigation of a turbulent
bubbly mixing layer International Journal of Multiphase Flow Volume 24, Issue 1, February
1998, Pages 35-54.
Qi, S., and Shuli, W. Measurement of turbulent of gas-liquid two-phase ow in a bubble
column with a laser velocitymenter Advances in Natural Science ISSN 1715-7862 Canadian
Research Development Center of Sciences and Cultures.
Tournemine, A.L.D. and Roig, V. Self-excited oscillations in buoyant conned bubbly
mixing layers Phys. Fluids 22, 023301 (2010).
Chapter 3
A Numerical Study of Gas Hold Up
in a Water Tank Supplied by a
Dual Jet of Air and Water
This chapter deals with the numerical simulations of mixing of two parallel jets of air and
water injected into a tank initially lled with water. The purpose is to explore this ow
conguration as a new conceptual design of a mineral separation machine. In mechanically
agitated otation machines, air is injected within an impeller placed in a tank of water and
minerals. Air break up into bubbles by a rotor installed in the tank. The injection point of
the air is important. It can aect the generation of bubbles and hence the separation process.
In this chapter, we simulate the mixing of an air jet with a parallel water jet. The eects
of the relative position of the two jets on gas hold up in the tank is an important design
parameter. In this study we determined the gas hold up in a tank for two cases by changing
62
Kareem Akhtar Chapter 3 63
the air entrance, in one case air jet is above water jet and in the other case air jet is below
the water jet.
It is found that when air jet is below the water jet, the gas hold up is more than the case
when the air jet is above the water jet. When air jet is below the water jet, air has more time
to travel and interact with the shear induced by the water jet, hence, gives more mixture.
3.1 Geometric Details:
The two-dimensional geometry is shown in gure 3.1. The domain width and height are
selected so as to make sure that the water and air jets do not deect and attach to the
bottom boundary. The jet deection and attachment are discussed in detail in chapter 1.
The total height of the domain is 1400mm and width of the domain is 600mm. The two jets
are supplied by two separate but parallel channels. Each jet width is 10mm. The main tank
is divided into two parts; and according to their initializations, we refer to them as a water
tank and an air tank, as shown in gure 3.2. The purpose of air tank is to provide a space
for the mixture generated due to the air entering into the water tank. So, we have four parts:
air tank, water tank, water jet and air jet.
Kareem Akhtar Chapter 3 64
Figure 3.1: Geometry details
3.2 Grid Generation and Geometric Details:
ANSYS ICEM CFD is used for grid generation. Dierent sections of the grid are shown in
gure 3.3. The dierent colors show the dierent parts of the domain discussed in section 3.1.
A ne grid is generated at the center of the grid, where the two jets interact, and a coarser
grid is generated away from the center. Each part of the domain is initialized separately. The
water tank is initialized with water with zero velocity; air tank is initialized with air with
zero velocity. The air jet is initialized with air and water jet is initialized with water. The
boundary conditions are labeled in gure 3.2. The boundary conditions are inlet, opening
and no-slip wall. At inlets, mass ow rate is specied instead of velocity. The upper edge of
Kareem Akhtar Chapter 3 65
the domain is an opening for air only with zero gauge pressure. The bottom edge is an outlet
for water only. Continuity is imposed on both air opening and water outlet. The boundary
between the air and water tanks is an internal boundary which is treated as interface. An
interface is a ow through boundary where mass, momentum and turbulence parameters are
conserved.
Figure 3.2: Boundary conditions
Kareem Akhtar Chapter 3 66
Figure 3.3: Two-dimensional grid
3.3 Numerical Simulations:
Numerical simulations are performed using ANSYS CFX. The mathematical modeling for
two-phase ow is discussed in chapter 2 section 2.2. Numerical simulations are performed for
two cases: Case 1: air jet is injected below the water jet, Case 2: air jet is injected above the
water jet. The boundary conditions for each case are the same. The air jet velocity in both
Kareem Akhtar Chapter 3 67
cases is 3.19(m/s), and water jet velocity in each case is 1.7(m/s).; the only dierence in the
cases is the vertical stratication of two jets. The air bubble diameter is selected as 1mm.
3.4 Results:
Simulations are performed by solving the unsteady RANS equations. The air volume fraction
contours after dierent time steps for case 1 and case 2 are shown in gure 3.4 and gure
3.5, respectively. The air volume fraction contours provide visual evidence of better air
dispersion when air jet is injected below the water jet. Gas hold up is monitored in the water
tank (initially lled with water) as function of time for each case, and the results are shown
by gures 3.6 and 3.7. The gas holdup increases with time. The water volume fraction in
the air tank (initially lled with air) is also calculated after dierent time steps. Figures 3.8
and 3.9 show the water volume fraction for air tank for case 1 and case 2. As expected, the
water volume fraction in air tank also increases with time because the air volume fraction in
water tank increases. The total quantity of water and air in the domain (water tank + air
tank) remains constant.
Kareem Akhtar Chapter 3 68
Figure 3.4: Air volume fraction contours for case 1: air jet below water jet
Kareem Akhtar Chapter 3 69
Figure 3.5: Air volume fraction contours for case 2: air jet above water jet
Kareem Akhtar Chapter 3 70
Figure 3.6: Air volume (%) in water tank for air jet below water jet
Figure 3.7: Air volume (%) in water tank for air jet above water jet
Kareem Akhtar Chapter 3 71
Figure 3.8: Water volume (%) in air tank for air jet below water jet
Figure 3.9: Water volume (%) in air tank for air jet above water jet
Kareem Akhtar Chapter 3 72
3.5 Comparison and Conclusion:
Comparison is made between the gas hold up in the two cases. Figure 3.10 shows the air
volume fraction in water tank for cases 1 and 2. The gure shows that if the air jet is injected
below the water jet, enhanced air dispersion is achieved in a shorter time compared to the
case when the air jet is injected above the water jet. Furthermore, the enhanced gas hold up
for case 1 persists over time. Similarly, water volume fraction in air tank is compared in both
cases as shown in gure 3.11. Like the air volume fraction in water tank, the water volume
fraction in air tank is also greater in case 1 than case 2. Injection of air jet below the water
jet gives air more time to interact and be inuenced by the shear of the water jet. Whereas
when the air jet is injected above the water jet, buoyancy accelerates the upward motion of
the air jet with less time to be inuenced by the water jet. This oers a new concept for the
design of ecient otation cells. Stratication of layers of air jets and water (or slurry) jets
can be explored to enhance particle-bubble collision for less power consumption. This will
make the otation process more ecient and mechanically simpler than the case of impeller
agitated cells.
Kareem Akhtar Chapter 3 73
Figure 3.10: Air volume fraction comparison between case1 and case2
Figure 3.11: Water volume fraction comparison between case1 and case2