Thermophysics and Aeromechanics, 2015, Vol. 22, No.

Buoyancy effects in steeply inclined air-water bubbly

shear flow in a rectangular channel
K. Sanaullah1, M. Arshad2, A. Khan1, and I.R. Chughtai2
1

Faculty of Engineering, UNIMAS, Kota Samarahan, Malaysia

Department of Chemical Engineering, Nilore, Islamabad, Pakistan

E-mail: skhairuddin@feng.unimas.my
(Received August 16, 2014; in revised form October 28, 2014)
We report measurements of two-dimensional (B/D = 5) fully turbulent and developed duct flows (overall
4
length/depth, L/D = 60; D-based Reynolds number Re > 10 ) for inclinations to 30 from vertical at low voidages
(< 5 % sectional average) representative of disperse regime using tap water bubbles (46 mm) and smaller bubbles
(2 mm) stabilised in ionic solution. Pitot and static probe instrumentation, primitive but validated, provided adequate
(10 % local value) discrimination of main aspects of the mean velocity and voidage profiles at representative
streamwise station i.e L/D = 40. Our results can be divided into three categories of behaviour. For vertical flow (0)
the evidence is inconclusive as to whether bubbles are preferentially trapped within the wall-layer as found in some,
may be most earlier experimental works. Thus, the 4-mm bubbles showed indication of voidage retention but the 2-mm
bubbles did not. For nearly vertical flow (5) there was pronounced profiling of voidage especially with 4-mm bubbles
but the transverse transport was not suppressed sufficiently to induce any obvious layering. In this context, we also
refer to similarities with previous work on one-phase vertical and nearly vertical mixed convection flows displaying
buoyancy inhibited mean shear turbulence. However, with inclined flow (10+ degrees) a distinctively layered pattern
was invariably manifested in which voidage confinement increased with increasing inclination. In this paper we
address flow behavior at near vertical conditions. Eulerian, mixed and VOF models were used to compute voidage and
mean velocity profiles.
Key words: gas-liquid bubbly flow, void fraction, mean shear velocity.

Introduction
Two-phase gas-liquid bubbly flows in vertical and horizontal tubes [1] and inclined
channels [2] have gained huge attention from researchers and scientists and, thus, in comparison to the largely established knowledge both for vertical and horizontal air-water bubbly
flows, there have been far fewer systematic investigations on flow patterns and profile measurements on the effects of inclination. Such initial studies, e.g., [3, 4] largely focused extension
of existing regime maps presented in terms of broad parameters, normally implied in relation to
overall void fraction and pressure drop. For small inclination from horizontal, the authors of [5]
applied their criterion for the existence of horizontal bubbly flow, which implies that the mean
shear turbulence to be greater than the segregative buoyant force. Their simple force balance
seems inappropriate for steeply inclined bubbly flows. In the work [6], transformation to
the slug flow from bubbly flow due to the buoyant upward movement of bubbles and accumulation near the upper wall was detected. The transition criteria of that work are due to the earlier
relationship of [5], which provides for a maximum bubble speed owing to the maximum

K. Sanaullah, M. Arshad, A. Khan, and I.R. Chughtai, 2015

463

K. Sanaullah, M. Arshad, A. Khan, and I.R. Chughtai

bubble size, above which bubbly flow cannot be sustained in a base flow with prescribed mean
shear turbulence characterized in terms of wall friction, u0. The transformation in an inclined
pipe of 51 mm i.d at inclinations of 20 to 40 from vertical was shown experimentally in [2].
The inclined bubbly flow was also investigated in [7] in connection with the consequences
of phase segregation for instrumentation response in oil well production logging. The effect
of inclination on wall shear stress and heat transfer was investigated in [8, 9].
We call the flow regime observed in inclined bubbly flows ( 10 degrees from vertical),
segregated-disperse [10], it is characterized by two-layer flow (schematically shown in Fig. 1) in
which bubbles are buoyantly confined to the upper layer but dispersed within it, not only due to
the mean shear turbulence but also due to the turbulence associated with bubble fluctuating
motions.
For two-phase air-water bubbly duct flows, inclination of just 5 from vertical suffices
departure from nominal similarity, however, possessing distinctively skewed voidage profiles
though not considerably confined to describe regime as segregative pattern. For this nearlyvertical flow, the normal turbulence dispersion is manifestly reduced by buoyancy gradient but
is certainly not completely vanished as occurs in the layered regimes observed at higher inclinations; see [11]. Importance of segregative-disperse regime was discussed with particular
attention on interior shear layer in [10]. Here another potentially important connection, drawn
from our study, links to mixed convection in vertical (or nearly) single-phase flows where
symmetric normal velocity profiles of pumped flows are distorted by axial buoyancy forces
induced by sufficient wall temperature excess or deficit. Such mixed convection regime has
been studied in detail, for example, in [1215] in relation to phenomenon of buoyancy-induced
laminarisation and its capability to causing disastrous collapse for intensive heat exchangers.
The reduced turbulence due to laminarisation, can be assessed by thickening of viscous sublayer [16], however, was caused by impaired turbulence eddy shear stress in the vicinity
of the wall. Alternatively, cooled wall provides negative buoyancy which enhances the wall shear
stress and thus the heat transfer coefficient is increased in comparison with forced flow.
It is reasonable to explore application of above findings to the nearly vertical two-phase
bubbly flows and seek whether the underlying physical issues can be understood in terms
of buoyancy associated with the voidage profiles and whether they are sufficiently confined close
to the wall as to reduce the mean shear and hence restrict turbulence production. Another facet
for nearly vertical bubbly flows is to assess whether the turbulence associated with bubbles is
sufficient to compensate for the reduced mean shear. Both aspects ask for additional scaling of
bubble size characterized by ratio of rms shear dispersion as a measure of voidage confinement
and the rms fluctuations arising from the bubbles. Indeed these matters are important to boiling
water heat exchangers as an diagnostic for dry out due to collapse of shear turbulence caused
by bubbles, preferentially accommodate at the upper wall.

Fig. 1. A schematic view of two-layer flow, also showing buoyancy stabilized shear layer
in steeply inclined bubbly flows.
464

Thermophysics and Aeromechanics, 2015, Vol. 22, No. 4

We describe experimental results of nominally two-dimensional fully turbulent and developed duct flows (B/D = 5, Re > 104, L/D > 40) for inclination departures from vertical up to
30o at low voidages representative of disperse regime ( < 5 %) using both typical tap-water
bubbles (46 mm) and smaller bubbles (2 mm) stabilized in ionic solution. The experimental
results are compared with computations obtained by the application of an Eulerian multiphase flow model that uses a RNG k- model (renormalization group based k- model) for
the turbulence.
Equipment and instrumentation
Main elements of the experimental set-up (schematically shown in figure 2) and findings
were reported in [10] and [17]. Here we briefly describe how gas is injected into pumped duct
flow. The bubbles were injected via a manifold through 201-mm holes located on the upper
face at the entry section and turbulently dispersed transversely into the main body of flow.
However, for smaller bubbles orifices with hole diameter 0.2 mm for the air injector manifold
were adopted in line with the findings of [18] for rigid sparger holes delivering bubble sizes
of about 2 mm. Twenty holes of 9 mm were made through the front section of duct. Rubber
bungs, each with three capillary tubes (2.5 mm apart) with i.d of about 0.2 mm piercing them
were inserted into 9-mm holes so that the surface of the bung was flush with the inside surface
of the front wall.
Pitot and static probe instrumentation, primitive but validated, used for this study were
described in paper [10]. Incorporated in the mouth of our pitot tube (4.5-mm o.d) was a thin
steel mesh (perforations less than 0.3 mm) to inhibit the ingress of bubbles into the manometer

Fig. 2. Schematic facility for two-phase air-water inclined flow (up to 30 from vertical).
465

K. Sanaullah, M. Arshad, A. Khan, and I.R. Chughtai

lines whilst still retaining a reasonable time response. The voidage is estimated from the transverse (static) pressure gradient, where the local void fraction depends only on normal distance
from the upper wall and the resolved component of gravity across the flow. The pressure signals from these tubes were recorded using the differential pressure transmitter (Honeywell
model type 4101) generated a signal over the scale range 4-20 mA which was fed to
the analogue digital converter (DCP microdevelopments, Interbeeb).The ADC output ranges
from zero to 255 bits (0.48 to 2.4 volts) and the ADC output was processed using a BBC
(model B) computer.
Computational methods
A geometry consisting of rectangular duct is created in GAMBIT processor as shown
in Fig. 3, representing a cross section of 0.2502.97 m. A mesh of 0.5 mm size has been created
that provided a total of 362 044 nodes for calculation for more accurate and reliable results.
Simulation Parameters. Euler-Euler multiphase flow model has been chosen for CFD
simulation of the subject system. Standard k- turbulence model for each phase has been chosen to model the turbulence in the system.
Euler-Euler approach. In this approach the two phases are inter-penetrating continua and
the volume fractions are considered to be continuous functions of space and time and their sum
is equal to 1. Each phase has its own conservation equations and empirical relations so in our
case the two phases have been dealt on Euler-Euler approach basis for capturing the characteristics of flow turbulence through realizable k- model whereas drag calculations are performed
using the symmetric model. The purpose of usage and details of these models are given below.
There are three different Euler-Euler models, which are generally used, they are; the mixture model, the volume of fluid model (VOF) and the Eulerian model.
The Mixture Model. When two phases got mixed with each other to form a homogeneous mixture then mixture model is applied. In such cases the mass, momentum and energy
equations are solved for the mixtures.
The Volume of Fluid Model. VOF is a numerical technique for tracking and locating
the free surface (or fluid-fluid interface). This method is characterized by a mesh that is either
stationary or is moving in a certain prescribed manner to accommodate the evolving shape
of the interface. The NavierStokes equations describing the motion of the flow have to be
solved separately in this case.
The Eulerian Model. The Eulerian model is the most versatile and complex model that is
applicable to all types of problems involving two phase flows, it is the most expansive model
that needs a large computational resource to solve the problem. There is a separate set of equations involving mass, energy and momentum transfer for each phase in this model and coupling
between phases is achieved using the pressure and inter-phase coefficients.

Fig. 3. Illustration of grid generated to simulate air-water bubbly flow in a rectangular duct.
466

Thermophysics and Aeromechanics, 2015, Vol. 22, No. 4

Transport equations. For a multiphase flow system with n phases, the mass balance
equation for phase q is given by

n
q q + . q q vq = p =1 m pq m qp + Sq ,
t

(1)

where q is the volume fraction, q is the physical density, v q is the flow velocity, S q is
the source term for the phase q, and m pq and m qp is the mass transfer from phase p to q and
from q to p, respectively.
The volume fraction q is also regarded as the unknown in the above equation due to
the fact that each computational grid of computational domain is surrounded by the different
phases in different fractions of their volume. Equation (1) is similar to the equation for mass
balance for single phase flow with just little difference that it has in terms for volume fraction
and inter-phase mass transfer terms m pq and m qp . Total volume of the phase q, Vq is given by
Vq = q dv and = q =1 q = 1.
n

(2)

The density which changes with prevailing conditions is effective. It can be calculated as

effective = qq.

(3)

Two governing equations, representing the two phases are used to simulate the momentum
balance. The momentum balance for the phase q is given by

q q vq + . q q vq vq = q P + . q + q q g +
t

+ p =1 R pq + m pq vq m pq vqp + Fq + Flift, q + Fvm, q ,

n

(4)

where P is the pressure of fluids, i.e. steam and water, q is the stress-strain tensor of phase q
and g is the force exerted by the one fluid throughout the second fluid, Fq is the external body
force, Flift,q is the lift force, Fvm,q is the virtual mass force, and Rpq is the interaction force between phases that in this case exerted on phase q.
Multiphase Flow Models. The mixture model is a simplified multiphase model that can
be used to model multiphase flows where the phases move at different velocities, but assume
local equilibrium over short spatial length scales. The basic concept of the mixture model is
to consider the mixture as a whole, rather than two phases separately.
In the Eulerian model formulation, the transfer processes of each phase are expressed by
their own balance equations, thus it is anticipated that the model can predict more detailed
changes and phase interactions than the drift-flux (or mixture) model. This model is formulated
by considering each phase separately. Mixture mean velocity and density has been calculated
by using the formulas
2
v

2
k =1 k k k
vm =
, m = k =1 k k .
(5)

Boundary Conditions. The CFD simulation has been carried out for an inclination of 0
(vertical) to 30 in an inclined duct. The liquid flow rates varied from 1.25 l/s to 5 l/s corre4
4
sponding to a Reynolds numbers (based on flow depth, D = 50 mm) range of 10 210 .
The liquid superficial velocities ranged from 0.2 to 0.4 m/s and the air flow rates were varied
from 10 l/min to 40 l/min corresponding to superficial velocities of 0.013 to 0.07 m/s.
The mean void fraction was estimated as ranging from 0.015 to 0.05 approximately.
The simulation results were plotted at 40D downstream of entry; here D = 50 mm.
467

K. Sanaullah, M. Arshad, A. Khan, and I.R. Chughtai

Results and discussion

Experiments were conducted at 40D downstream of entry. The mean liquid velocity profiles were obtained for inclination angles ranging from zero (vertical) to 30 from the vertical.
The voidage and velocity profiles were also obtained for both ionic (sodium sulphate, NaSO4
solution, 5 g/l [19]) and non-ionic two-phase solutions. The liquid flow rates varied from
1.25 l/s to 5 l/s corresponding to a Reynolds numbers (based on flow depth, D = 50 mm) range
4
4
of 10 210 . The liquid superficial velocities range from 0.2 to 0.4 m/s and the air flow rates
were varied from 10 l/min. to 40 l/min., corresponding to superficial velocities of 0.013 m/s
to 0.07 m/s. The mean void fraction was estimated as ranging from 0.015 to 0.05 approximately.
Vertical flow. The present study recovered only velocity profiles for vertical flow (Fig. 4).
The results with 46 mm bubbles were certainly consistent with voidage excess retained
in the wall zone insofar as the velocities were higher than expectation for neutral buoyancy.
On the other hand, in our set-up the bubbles were introduced at the wall so this finding might
equally be explained in terms of insufficient test-section length to properly accommodate evolution to fully developed conditions. Establishment of equilibrium voidage profiles is known to
take considerably longer than for mean velocity and shear stress in uni-phase turbulent flows
because of the weak lateral forces associated with bubble transverse fluxes. Indeed, although
the mean velocity profiles obtained with 2 mm bubbles showed no obvious indication of walllayer confinement we equally cannot take this contrary result as satisfactory evidence that attraction is absent from the asymptotic approach to far downstream fully developed flow. More
studies are required to resolve these uncertainties, including systematic attention to bubble size
as independent parameter determining the bubble induced turbulence (burbulence) scale and
intensity for prescribed voidage independent of mean shear turbulence (shurbulence) scale and
intensity.
Nearly vertical flow. Inclination of just 5 from vertical is sufficient to encourage flow
evolution to a nominally fully developed state possessing distinctively skewed profiles of voidage (Fig. 5), though not sufficiently confined to warrant description as a segregative pattern.
Figures 5 and 6 show the local void fraction profiles for ionic and non-ionic solutions at angles
of inclination of 5 and 10. As expected for the smaller bubbles the bubbly layer deepens
somewhat and for near vertical conditions (5) the bubbly layer for 2-mm bubbles deepens
more as compared to results for 46-mm bubbles. These findings are further consolidated when
we examine the effect of gasrate on voidage profiles (Fig. 5). As expected, with increased
gasrate, the bubbly layer (Fig. 6) spreads further into the flow at both 5 and 10, until it covers
the whole width at gasrates over 10 l/min at 5. The photographic evidence (Fig. 7) also
indicates the considerable difference of spread of bubbles for 5. We see here the bubbly layer

Fig. 4. Velocity profiles for vertical bubbly duct flows obtained at Re=14000 and gasrate 15 l/min.
One-phase flow (1), 2-mm bubbles (2) and 46-mm bubbles (3).

468

Thermophysics and Aeromechanics, 2015, Vol. 22, No. 4

Fig. 6. Bubbly layer thickness for 5 and 10.

There are same conditions as for Fig. 5.
1 5, 2 10.

Fig. 5. Void fraction profiles for 5 and 10

inclination obtained for Re = 14000 and 15 l/min.

covers more than 90 % of the flow width

for ionic flow, whereas this was about 80 %
of the flow width for non-ionic flow.
The apparent reason for this is increased voidage due to the smaller slip speed associated with
smaller bubbles, although of course, the increased voidage also represents an increase
in the buoyancy flux, and hence also more turbulent energy was consumed working against
gravity.
The profiles with smaller bubbles more nearly represent the influence of buoyancy alone:
i.e, with bubble slip Reynolds number only 100 or so, as compared with perhaps 1000 or so for
the larger bubbles ( > 2 mm), there is a greatly reduced bubble dispersive turbulence. Peaking
near the upper wall here specially for smaller inclinations (10o or lower) is similar to that found
in mixed convection heat transfer of one-phase flows (see, e.g., [20, 21] and Fig. 8). In effect,
we are saying that buoyancy introduced by upper layer excess voidage must be effectively confined within this wall layer if laminarisation of the mean shear turbulence is to occur in bubbly
flow. Our results (Fig. 8) suggest that the upper bubbly layer may be sufficiently confined as to
introduce such inhibition; i.e., the buoyancy profiles are comparable to measurements in thermal mixed convection flows done in the works [21] and [22] among others.
Figure 8 compares velocity profiles in mixed convection flow with our results for bubbly
flows. Whilst for the work [22] the velocity peak was very close to the wall, it was found
in [20] that the velocity peak lies further
away from the wall, corresponding to
reduced confinement of buoyancy. Our
present results, although less confined
than in the work [22] certainly suffice
46-mm bubbles (1) and 2-mm bubbles (3) at 5 inclination,
46-mm bubbles (2) and 2-mm bubbles (4) at 10.

Fig. 7. Comparison of voidage, measured at

Re=14000 and gasrate 15 l/min for 2-mm
bubbles, with temperature gradient profiles
in one-phase mixed convection flows.
Results of the present work at inclinations of
5 (1), 10 (2) and the results of the works
[21] (3), [22] (4).

469

K. Sanaullah, M. Arshad, A. Khan, and I.R. Chughtai

Fig. 8. Comparison of velocity profiles at Re =
=14000 and gasrate 15 l/min for 2-mm bubbles
with those in mixed convection flows.
Results of the present work at the inclinations of 5 (1),
10 (2), and the results of works [20] (run No. 13) (3),
[22] (at Re = 4300) (4).

to indicate potential impairment of heat

transfer by buoyant inhibition of turbulence.
Computational results
The computation of effect of buoyancy
on void fraction and mean velocity can be
seen in Figs. 9 to 12. However, the computations were conducted only for inclination
of 10. This is chosen since at the said angle there is sufficient buoyancy to distort the distribution
of bubbles and at the same time the angle is not far from the vertical. Whereas inclination lower than this the buoyancy will not be strong enough to compete with the mean shear turbulence.
The computed void fractions for bubbly flow with 2-mm bubbles (Fig. 9) do not show
the equivalent wall buoyancy as can be seen in the measured voidage profile, rather their average value across the flow is much lower than the actual values. Whereas, the computational
results for bubbly flow with 46-mm bubbles (Fig. 10) do show the near wall buoyancy layer,
this indicates to the fact that the buoyancy enhances with increase in bubble size. Also obvious
in Fig. 10, that the results obtained by use of Eulerian model, show more realistic void fraction
distribution than the mixture model. However, the computed values are lower than the measured void fraction values.

Fig. 9. Void fraction with 2-mm bubble at 10 inclination.

Experimental results (1), CFD results (Mixture model) (2), and CFD results (Euler model) (3).

470

Thermophysics and Aeromechanics, 2015, Vol. 22, No. 4

Fig. 10. Void fraction with 46-mm bubble at 10 degrees inclination.

See notations in Fig. 9.

Fig. 11. Velocity profile with 2-mm bubbles at 10 inclination.

Figures 11 and 12 show the computed velocity profiles at inclination of 10. Compared to
computed void fraction profiles, the calculated velocity profiles show more convincingly
the distortion of the symmetry of mean velocity due to bubble buoyancy. However, the calculated
velocity profiles with bubble size of 46 mm show slightly more peaked near the upper wall as
compared to the bubbly flow with 2-mm bubble size.
Conclusion
A potentially important practical connection suggested by our study links to mixed convection in (nearly) vertical flows of uniphase fluids where axial buoyancy forces due to thermally induced density reduction can attenuate the wall zone shear stress sufficiently to cause
major reduction in the turbulence energy and thus also the heat flux. Comparison of our profiles with those found in mixed convection flows indicate that the buoyant layer in our flows is
471

K. Sanaullah, M. Arshad, A. Khan, and I.R. Chughtai

Fig. 12. Velocity profile with 46-mm bubbles at 10 inclination.

Experimental (1) and CFD (2) results.

sufficiently confined as to sufficiently inhibit the production of wall shear stress. Whilst there
is no room here to incorporate established and emerging fundamental literature on the behavior
and effects of bubbles in turbulent wall-layers, we must mention that bubble size appears to be
a key factor insofar as microbubbles (100s microns) seemingly can reduce the wall shear stress
whereas macrobubbles (mm sizes as used here) apparently act to increase it.
References
1. Y. Taitel, D. Barnea, and A.E. Dukler, Modelling flow pattern transitions for steady upward gas-liquid
flow in vertical tubes, AICheE J., 1980, Vol. 26, P. 345354.
2. D. Barnea, O. Shoham, Y. Taitel, and A.E. Dukler, Gas-liquid flows in inclined tubes: flow pattern transition for
upward flow, Chem. Engng Sci., 1985, Vol. 40, . 735740.
3. D. Barnea, O. Shoham, Y. Taitel, and A.E. Dukler, Flow pattern transitions for gas-liquid flow in horizontal and
inclined pipes: comparison of experimental data with theory, Int. J. Multiphase Flow, 1980, Vol. 6, P. 265272.
4. H. Mukherjee and J.P. Brill, Pressure drop correlations for inclined two-phase flow, J. Energy Res. Tech., 1985,
Vol. 107, P. 549554.
5. Y. Taitel and A.E. Dukler, A model for prediction flow regime transition in horizontal and near horizontal gasliquid flow, AIChE J., 1976, Vol. 22, P. 4755.
6. D. Barnea, Transitions from annular flow and from dispersed bubble flow-unified models for the whole range
of pipe inclinations, Int. J. Multiphase Flow, 1986, Vol. 12, No. 5, P. 733744.
7. S.D. Mobbs and G.P. Lucas, A turbulence model for inclined, bubbly flows, Appl. Sci. Res., 1993, Vol. 51, No. 1, 2,
P. 263268.
8. O.N. Kashinsky, V.V. Randin, and A.V. Chinak, The effect of channel orientation on heat transfer and wall
shear stress in the bubbly flow, Thermophysics and Aeromechanics, 2013, Vol. 20, No. 4, P. 391398.
9. O.N. Kashinsky, V.V. Randin, and A.V. Chinak, Heat transfer and shear stress in gas liquid flow in an inclined flat channel, J. Engng Thermophys., 2014, Vol. 23, No. 1, P. 3946.
10. K. Sanaullah and N.H. Thomas, Velocity and voidage profiles for steeply inclined bubbly flows for segregateddispersed regime, Experimental and Computational Aspects of Validation of Multiphase Flow CFD Codes, 1994,
FED, Vol. 180, P. 119127.
11. K. Sanaullah, X. Yang, and N.H. Thomas, Buoyancy profile effects in inclined bubbly shear flow, Advances in Turbulence VI: Fluid Mechanics and its Applications, 1996, Vol. 36, P. 567570.
12. J.D. Jackson and W.B. Hall, Influences of buoyancy on heat transfer to fluids flowing in vertical tubes under
turbulent conditions, in: S. Kakac and D.B. Spalding (Eds.), Turbulent Forced Convection in Channels and Bundles.
Hemisphere Publ. Corporation, Washington, 1979, Vol. 2, P. 613640.
13. H. Tanaka, S. Marugama, and S. Hatano, Combined forced and natural convection heat transfer for upward flow
in a uniformly heated vertical pipe, Int. J. Heat Mass Transfer, 1987, Vol. 30, No. 1, P. 165174.

472

Thermophysics and Aeromechanics, 2015, Vol. 22, No. 4

14. M.A. Cotton and J.D. Jackson, Vertical tube air flows in the turbulent mixed convection regime calculated using
a low-Reynolds-number k- model, Int. J. Heat & Mass Transfer, 1990, Vol. 33, No. 2, P. 275285.
15. S. Torii and Y. Wen-Jei, Thermal-fluid transport phenomena in strongly heated channel flows, Int. J. Num.
Methods Heat Fluid Flow, 2000, Vol. 10, No. 8, P. 802823.
16. J.D. Jackson, Mixed forced and free convection the influence of buoyancy on heat transfer in vertical pipes,
EuroMech-72 Conf. Salford, U.K, 1976.
17. K. Sanaullah, Buoyancy effects in two-phase inclined bubbly flows. A Ph.D thesis submitted to the University
of Birmingham, U.K, 1995.
18. W.G. Nicholas and R.G. Rice, Bubble size prediction for rigid and flexible spargers, AIChE J., 1991, Vol. 37,
No. 2, P. 161168.
19. G. Keitel and U. Onken, The effect of solutes on bubble size in air-water dispersions, Chem. Engng Commun.,
1982, Vol. 17, P. 8598.
20. A.D. Carr, M.A. Connor, and H.O. Buhr, Velocity, temperature and turbulence measurements in air for pipe
flow with combined free and forced convection, Trans. ASME. J. Heat Transfer, 1973, Vol. 95, P. 445452.
21. A. Steiner, On the reverse transition of a turbulent flow under the action of buoyancy forces, J. Fluid Mech., 1971,
Vol. 47, No. 3, P. 503512.
22. J.E. Byrne and E. Ejiogu, Combined free and forced convection heat transfer in a vertical pipe, I. Mech. E Symp.
on Heat and Mass Transfer by Combined Forced and Natural Convection, Manchester, 1975, P. C118/71.

473