You are on page 1of 22

Flow field behind a fixed bluff body in a

vertical pipe simulating a wake of a Taylor


bubble
Cite as: Phys. Fluids 25, 105103 (2013); https://doi.org/10.1063/1.4823731
Submitted: 20 May 2013 . Accepted: 15 September 2013 . Published Online: 04 October 2013

V. Babin, D. Barnea, and L. Shemer

ARTICLES YOU MAY BE INTERESTED IN

On shapes and motion of an elongated bubble in downward liquid pipe flow


Physics of Fluids 29, 112103 (2017); https://doi.org/10.1063/1.4996444

On the turbulent structure in the wake of Taylor bubbles rising in vertical pipes
Physics of Fluids 19, 035108 (2007); https://doi.org/10.1063/1.2711478

Three dimensional flow around a circular cylinder confined in a plane channel


Physics of Fluids 23, 064106 (2011); https://doi.org/10.1063/1.3599703

Phys. Fluids 25, 105103 (2013); https://doi.org/10.1063/1.4823731 25, 105103

© 2013 AIP Publishing LLC.


PHYSICS OF FLUIDS 25, 105103 (2013)

Flow field behind a fixed bluff body in a vertical pipe


simulating a wake of a Taylor bubble
V. Babin, D. Barnea, and L. Shemer
School of Mechanical Engineering, Faculty of Engineering, Tel-Aviv University,
Tel-Aviv 69978, Israel
(Received 20 May 2013; accepted 15 September 2013; published online 4 October 2013)

The hydrodynamic structure of the wake of an axisymmetric bluff body that simulates
the shape of a moving elongated (Taylor) bubble in a vertical pipe was studied
using Particle Image Velocimetry in laminar and turbulent background flows. The
distribution of the mean axial and radial velocity components in the wake, as well
as the spatial variation of the normal and shear stresses are presented and compared
with the corresponding quantities in the wake of the gas bubble. The accumulated
results enable estimates of the spatial variation of turbulent energy production term in
the undeveloped separated flow in the wake of the bluff body.  C 2013 AIP Publishing

LLC. [http://dx.doi.org/10.1063/1.4823731]

I. INTRODUCTION
Information on the distribution of the mean and fluctuation velocity components in the wake of a
large elongated (Taylor) bubble rising in a vertical pipe is essential for understanding the movement
of a trailing elongated bubble and the mechanism of coalescence between consecutive bubbles in
undeveloped liquid-gas slug flow in pipes.1–6 A comprehensive study of the hydrodynamic structure
of the wake behind a Taylor bubble was carried out in both laminar and turbulent background
flows using Particle Image Velocimetry (PIV).7–10 Ensemble averaged velocities in the frame of
references moving with the Taylor bubble, as well as the characteristic turbulent quantities were
presented.
Beyond that, the movement of an elongated bubble in a pipe and the velocity field in the liquid
around the bubble constitutes an extremely interesting hydrodynamic problem. It can be seen as an
example of the so-called complex flows11 characterized by a combination of free mixing layers and
wall-bounded flows such as wall jets in confined ducts.12 A close relation exists between the flow
in the wake of a Taylor bubble, and the wakes of axisymmetric bluff bodies in a pipe.13 The Taylor
bubble wake is particularly complicated; it occupies most of the pipe cross-section and in the frame
of reference attached to the pipe, the origin of the confined wall jet does not remain stationary but
rather moves up with the translational velocity of the Taylor bubble, Utr . Moreover, while a bluff
body has a permanent shape, the Taylor bubble bottom oscillates14 thus introducing an additional
unsteady contribution to the flow in the wake.
The complexities related to carrying out detailed experiments on hydrodynamics of flow around
a Taylor bubble prompted attempts to simulate the moving Taylor bubble by a solid model placed
within a pipe. Moissis and Griffith1 used a rigid model that moved axially upward in a pipe with
downward flowing water. Tudose and Kawaji15 performed measurements of drag force on stationary
solid models of both axially symmetric and asymmetric elongated bubbles placed in downward
flowing liquid in a vertical pipe. Tokuhiro et al.16 presented results of PIV measurements of turbulent
downward flow around an oscillating bubble and compared them with the flow about a solid ellipsoid
with identical equivalent diameter in a large rectangular channel. Sotiriadis and Thorpe17 simulated
a Taylor bubble by a ventilated cavity and by a cylindrical bluff body in turbulent downward
flow; in both cases the blockage ratio was about 0.8. They performed laser Doppler velocimetry
measurements in the wake region and found significant similarities between the two cases, as well
as with the experimental results of van Hout et al.8

1070-6631/2013/25(10)/105103/21/$30.00 25, 105103-1 


C 2013 AIP Publishing LLC
105103-2 Babin, Barnea, and Shemer Phys. Fluids 25, 105103 (2013)

FIG. 1. Sketch of the experimental facility.

Comparison of results of numerical simulations between flow in the wake of a Taylor bubble and
its solid model carried out by Lertnuwat and Bunyajitradulya18 indicated that there are no significant
difference between the wakes of these two cases provided the trailing corner radius of the model is
sufficiently small.
In the present work, we attempt to model a Taylor bubble by a bluff body with a realistic
bubble’s shape and perform PIV measurements of the velocity field in the body’s wake for flow
parameters corresponding to those in our previous experimental studies on Taylor bubbles.4, 8–10 The
accumulated results also contribute to a better understanding of the flow around a bluff body in a
confined conduit.

II. EXPERIMENTAL FACILITY AND DATA PROCESSING


Experiments are carried out in a facility shown schematically in Fig. 1. The facility consists of
a water supply system and a test section made of 6 m long vertical Perspex pipe with an internal
diameter of 0.026 m. Filtered tap water flowing in a closed loop is used as the working fluid. Special
care was taken to ensure symmetric and smooth entrance of water to the pipe. The inlet section of
the pipe, shown in the inset of Fig. 1, consists of a large settling chamber, a honeycomb, a number
of screens, and a converging nozzle. The water flow rate is monitored continuously by a set of
rotameters. The accuracy of each rotameter is 1.6% of the full scale. The test section is surrounded
by a rectangular transparent box filled with water to reduce image distortion.
A solid model of a Taylor bubble shown in Fig. 2 is inserted into the pipe at the bottom part of
the measuring station at a distance of 5 m from the inlet, see Fig. 1. The model’s shape was based
on the film thickness calculation19 for a 5D length Taylor bubble moving in stagnant liquid. Since
the liquid is flowing upwards, the solid model of the Taylor bubble is fixed with its nose pointed
downward. In this way the flow field in the wake of the model can be measured and compared to
that of a real gas Taylor bubble moving upward. More details on the experimental facility can be
found in Shemer et al.10
Axial and radial liquid velocities in the solid body wake are determined using Particle Image
Velocimetry (PIV). The PIV system uses a Kodak ES 1.0 CCD camera with a maximum frame rate
of 14 frame pairs per second. The camera resolution is 1008 × 1008 pixels. The camera is operated
105103-3 Babin, Barnea, and Shemer Phys. Fluids 25, 105103 (2013)

FIG. 2. Solid model of the Taylor bubble in the pipe.

in a triggered double-exposure mode that allows capturing a pair of single-pulsed frames at a short
time interval. The images are digitized and saved in the RAM of a computer. Pulsed double-head
Nd: YAG MiniLase III PIV-15 laser that generates a pair of short (5–7 ns) pulses at a wavelength of
532 nm (green light) with a total energy of 50 mJ/ pulse is used. The light sheet thickness is about
1 mm at the measurement position.
Almost neutrally buoyant fluorescent particles (density ρ = 1.006 g/cm3 ) with a diameter less
than 80 μm are used as tracers. When illuminated by green laser light, the particles emit orange light
with wavelength of about 580 nm. A filter with a cut-off wavelength of about 550 nm is placed on
the PIV camera objective lens to filter out the green laser light from the recorded frames.
An external synchronizer unit is used to synchronize all the components of the measuring
system. Each illumination pulse in the pair is captured by the camera in a different frame and the
resulting image contains a pair of frames. The instantaneous velocity magnitude in the flow field
under investigation ranges from few mm/s to about 1.5 m/s. The size of the interrogation window
in the PIV measurements is 32 × 32 pixels. Subpixel resolution processing is applied. The time
delay between the laser pulses is selected so that the maximum particle displacement between the
frames in each pair does not fall below 1 pixel and does not exceed about 16 pixels. The time delay
thus varies from 0.2 ms for measurements in the near wake region characterized by relatively high
instantaneous velocities, and may increase up to 5 ms to determine liquid flow velocities far away
from the model bottom.
The camera and laser were moved along the test section and at each location at least 1000
pairs of images were captured. The experiments were carried out for the following Re numbers
(Re = ρUL D/μ): 820, 7500, 10 500, and 13 500, where UL is the mean liquid velocity, and D is the
pipe diameter. The lowest flow rate corresponds to laminar undisturbed flow in the pipe, while at all
other flow rates the flow is turbulent.

III. RESULTS
A. Mean flow field
Since the mean flow structure in the wake region is qualitatively similar for all flow rates
corresponding to turbulent background flow, only results for Re = 7500 (turbulent flow regime) and
Re = 820 (laminar flow regime) are shown here.
The liquid flow passes around the model and enters the liquid in the model wake as a circum-
ferential wall jet, creating a toroidal vortex extending to approximately one pipe diameter behind
the model tail. The mean flow field is axially symmetric. Due to flow symmetry, it was decided to
perform the majority of measurements at higher spatial resolution, with the horizontal scale covering
slightly more than the pipe radius. The image resolution was 0.028 mm/pixel for full width images
or 0.018 mm/pixel for radius-wide images. The time-averaged 1.2D by 0.5D maps presenting the
distribution of the velocity vectors in the vertical axial plane that were obtained immediately down-
stream the model are presented in Figs. 3(a) and 3(b) for Re = 820 and Re = 7500. Both maps
indicate the existence of a toroidal vortex in the bluff body’s wake.
105103-4 Babin, Barnea, and Shemer Phys. Fluids 25, 105103 (2013)

FIG. 3. Map of the mean flow field in the near wake region with streamlines sketched: (a) Re = 820; (b) Re = 7500.

For turbulent flow regime this toroidal vortex is about 1.4D long with the center of the vortex
located 0.7–0.8D downstream of the model. Similar vortex geometries were obtained at two higher
mean flow rates as well, with the vortex center shifting somewhat away from the body with increase
in the flow rate. These results agree well with previous studies of a Taylor bubble wake and compare
well with the results of Sotiriadis and Thorpe17 behind a bluff body. For laminar flow regime,
Fig. 3(a), the toroidal vortex behind the model has a wider and shorter shape than in the turbulent
case. It is followed by another weaker toroidal vortex which has an opposite sense of rotation and
cannot be clearly identified for the turbulent case.
The radial distributions of the mean axial velocity component U(r) for two Reynolds numbers
are presented in Fig. 4 for various distances z downstream the solid body. In both cases immediately
behind the model (z/D = 0.1) the velocity profile is nearly flat in the central region of the pipe, with
the velocity magnitude being close to zero because of proximity of the solid surface. A strong wall
jet can be clearly identified in both laminar and turbulent background flow regimes. The flow in the
near wake region is characterized by a mixing layer between the up flowing wall jet and the much
slower flow in the central region of the pipe.
The wall jet at Re = 820 decays fast with z, with the location of the maximum velocity moving
away from the wall and attaining the centerline of the pipe already at about 1D from the body. For
turbulent background flow, the maximum velocity in the wall jet decays notably slower; the jet is
stronger, with negative velocities at the central part of the pipe extending to a distance of few pipe
diameters from the body. The velocity profile in the far wake in turbulent flow is very close to the
fully developed profile already at z/D = 10. When the flow is laminar, the velocity profile in the far
wake approaches the fully developed parabolic shape only at about 20D behind the body.
The axial distributions of the axial velocities U normalized by the mean liquid velocity for
various flow rates are shown in Fig. 5 at three radial locations. For all flow rates corresponding to
turbulent background flow, the normalized mean axial velocity profiles collapse on a single curve at
each radial position. The shape of the axial velocity profiles in the laminar case is quite different.
Close to the pipe’s wall at r/D = 0.45 (Fig. 5(a)), the velocity is high close to the body in both laminar
and turbulent background flows, the normalized peak value being higher for the laminar case. For
both flow regimes, the velocity decreases with distance, attaining a local minimum, at about z/D
= 1–2, then increases again towards a weaker local maximum, before tending to a constant value
corresponding to developed pipe flow. For the laminar background flow, the mean velocity variation
in Fig. 5(a) is consistent with the short double vortex pattern shown in Fig. 3(a) exhibiting a negative
maximum at about 1D from the body. At r/D = 0.25, Fig. 5(b), for Re = 820 the axial velocity
remain close to zero up to about z/D = 0.5, the flow at this radial location then becomes directly
105103-5 Babin, Barnea, and Shemer Phys. Fluids 25, 105103 (2013)

0.2
z/D=0.1
(a)
z/D=0.6
0.15 z/D=1
z/D=2
z/D=4

U [m/s]
0.1
z/D=10
z/D=20
0.05 z/D=30

−0.05
0 0.1 0.2 0.3 0.4 0.5
r/D
1.5
(b)

1
U [m/s]

0.5

−0.5
0 0.1 0.2 0.3 0.4 0.5
r/D

FIG. 4. Mean axial velocity U(r) as a function of the distance z from the solid body: (a) Re = 820; (b) Re = 7500.

affected by the expanding wall jet (see Fig. 3(a)), so that the velocity increases significantly attaining
maximum at about z/D = 0.8. The weaker local minimum at about z/D = 1.5 can be attributed to
the secondary counter-rotating vortex; the axial velocity then gradually increases towards the value
corresponding to fully developed profile. For the turbulent background flow, the mean axial velocity
is initially negative, reaches its minimum value at a distance corresponding to the vortex center at
z/D ≈ 0.7 (Fig. 3(b)), the velocity then increases and changes sign at the downstream location of
about z/D = 1.2. The weak local maximum attained at about 1.8D from the model is indicative of
existence of an additional weaker toroidal vortex with an opposite sense of rotation.
At the centerline, Fig. 5(c), axial distributions of U/UL for both laminar and turbulent background
flows have a certain similarity in the near wake region, where they both are negative, although the
minimum value for Re = 820 is attained closer to the body than for Re = 7500. Farther away,
the axial profiles exhibit similarity to those in Fig. 5(b). Note that the normalized axial velocity
in laminar background flow at z/D = 20 still remains below the value of 2 expected for a fully
developed Poiseuille flow, whereas for turbulent background flow, the fully developed velocity
profile is attained. In all turbulent cases presented in Fig. 5(c), the Ucl normalized by the mean
velocity value is close to 1.2 as expected.
The distribution of the mean radial velocity component V(r, z) in the model wake shown in Fig. 6
is again consistent with the mean flow field pattern as plotted in Fig. 3 for both laminar and turbulent
cases. The spatial variation of signs of the radial velocity component in both cases indeed indicates
that the mean flow in the near wake of the body is characterized by a system of counter-rotating
vortices. At z/D > 3 the mean radial velocity component essentially vanishes.

B. Normal and shear Reynolds stresses


In order to characterize the fluctuating velocity field in the body’s wake, spatial distributions of
the averaged velocity fluctuations are calculated. The radial distributions of the r.m.s. values of the
105103-6 Babin, Barnea, and Shemer Phys. Fluids 25, 105103 (2013)

FIG. 5. Axial distribution of the axial velocity normalized by the mean liquid velocity at the wake: (a) r/D = 0.45; (b) r/D
= 0.25; (c) centerline.

axial and radial velocity component fluctuations u and v  normalized by the mean axial velocity UL
are presented in Fig. 7 for laminar background flow. Strong velocity fluctuations exist in the near
wake region as a result of the mixing between the annular wall jet and the quiet background flow.
Both velocity fluctuations components vanish at z/D > 10 thus indicating the end of the mixing
region. The high values of u are observed in the wall jet region immediately downstream the solid
model, at z/D = 0.1, Fig. 7(a). The local maximum in the radial distribution shifts from the pipe
wall vicinity to the center. Two peaks can be identified at z/D = 0.6. The peak near the pipe wall is
due to turbulence in the boundary layer flow, the second peak at r/D = 0.3–0.4, at the outer edge of
the wall jet, is in the mixing layer between the annular wall jet and the quiescent flow closer to the
pipe axis.20 For z/D > 2, the values of u decrease gradually, and the local maximum again shifts
towards the wall until full decay of fluctuations is attained. The r.m.s. values of the radial velocity
component fluctuations v  are significant in the near wake of the model (z/D < 4), and vanish at about
z/D = 20, Fig. 7(b).
105103-7 Babin, Barnea, and Shemer Phys. Fluids 25, 105103 (2013)

0.02
(a) z/D=0.1
z/D=0.4
0 z/D=0.6
z/D=1

V [m/s]
−0.02 z/D=1.7
z/D=2
z/D=4
−0.04

−0.06
0 0.1 0.2 0.3 0.4 0.5
r/D
0.1
(b) z/D=0.1
z/D=0.4
z/D=0.6
0
z/D=1
V [m/s]

z/D=1.5
z/D=2
−0.1 z/D=4

−0.2
0 0.1 0.2 0.3 0.4 0.5
r/D

FIG. 6. Distribution of the local mean radial velocity component V(r, z) in the Taylor bubble model wake: (a) Re = 820;
(b) Re = 7500.

FIG. 7. Radial distributions of the r.m.s. values of the velocity fluctuations for Re = 820: (a) Axial component; (b) Radial
component.
105103-8 Babin, Barnea, and Shemer Phys. Fluids 25, 105103 (2013)

FIG. 8. Radial distributions of the r.m.s. values of the velocity fluctuations for Re = 7500: (a) Axial component; (b) Radial
component.

The r.m.s. values of the velocity fluctuations for turbulent background flow are presented in
Fig. 8. In the near wake region, there is a notable similarity between the absolute values and the radial
distributions of both components of the velocity fluctuations with those measured in the laminar
background flow case. Starting from z/D > 4, the radial distributions of u and v  approach the shape
characteristic for developed turbulent pipe flow. For z/D > 20, the normalized turbulent fluctuations
have close values for all turbulent flow rates and do not vary notably with distance. At each axial
cross-section v  values are of the same order of magnitude as the corresponding values of u , in
agreement with Taylor and Whitelaw13 and Shemer et al.10
The radial variation of the Reynolds stresses at various distances from the bottom of the solid
model is presented in Figs. 9 and 10. The corresponding cross-correlation coefficients are plotted in
Figs. 11 and 12 for laminar and turbulent background flow. In all cases, in the immediate vicinity
of the bottom, at z/D = 0.1, the Reynolds shear stresses and the corresponding cross-correlation

−4
x 10
2
z/D=0.1
z/D=0.6
z/D=1
u′ v′ [m2/s2]

0 z/D=2
z/D=4

−2

−4
0 0.1 0.2 0.3 0.4
r/D
FIG. 9. Radial distributions of the Reynolds shear stresses for Re = 820.
105103-9 Babin, Barnea, and Shemer Phys. Fluids 25, 105103 (2013)

0.01
z/D=0.1
0 z/D=0.6
z/D=1

u′ v′ [m2/s2]
−0.01 z/D=2
z/D=4
−0.02

−0.03
(a)
−0.04
0 0.1 0.2 0.3 0.4 0.5
r/D
−4
x 10
(b) z/D=10
z/D=20
u′ v′ [m2/s2]

2 z/D=30

0 0.1 0.2 0.3 0.4


r/D
FIG. 10. Radial distributions of the Reynolds shear stresses for Re = 7500: (a) Near wake; (b) Far wake.

coefficients are close to zero, i.e., the velocity fluctuations in the axial and the radial directions
remain largely uncorrelated. Effective Reynolds stresses appear only at some distance downstream.
For laminar background flow, the Reynolds shear stresses plotted in Fig. 9 are mostly negative
at about 0.5 < z/D < 1, their absolute values then decrease with distance. The Reynolds stresses
then become positive attaining a maximum at about z/D = 2 and then vanishing as expected for
the laminar flow regime at z/D = 4. Fig. 11 demonstrates that the correlation coefficient between
the velocity fluctuations in the axial and the radial velocity in the near wake may be very high,
attaining values approaching unity and thus indicating the existence of coherent structures. To gain
a better understanding of the reason for appearance of apparent Reynolds stresses in a basically
laminar flow, instantaneous values of the velocity fluctuations at a location corresponding to a high
cross correlation coefficient are plotted in Fig. 13(a). This plot shows that the instantaneous absolute
values of both u and v  are mostly quite small, they however grow intermittently and randomly,
with the fluctuations in both directions having opposite signs. The map of measured fluctuations
at a single location plotted in Fig. 13(b) indeed largely collapses into vicinity of a straight line.
PIV velocity maps corresponding to instants with high instantaneous fluctuations are presented in

FIG. 11. Radial distributions of the cross-correlation coefficient values in the near wake for Re = 820.
105103-10 Babin, Barnea, and Shemer Phys. Fluids 25, 105103 (2013)

FIG. 12. Radial distributions of the cross-correlation coefficient values for Re = 7500: (a) Near wake; (b) Far wake.

Fig. 14. Well-defined vortices with opposite sense of rotation can be identified in Figs. 14(a) and
14(b). The apparent Reynolds shear stresses in the near wake region of an axisymmetric bluff body
in laminar background flow can thus be associated with random vortex shedding from the body’s
rear.
For turbulent background flow, the Reynolds shear stresses distributions in the near wake
are qualitatively similar to those observed in the laminar background flow, being negative with
relatively large values until about z/D < 2 and then changing sign at larger distances (Figs. 10(a) and
10(b)). The Reynolds stresses values in the wake decrease with distance up to about 20D and then
remain constant with the maximum close to the wall exhibiting radial distribution characteristic for
developed turbulent pipe flow.21 The observed spatial variation of the Reynolds stresses in the wake
is in agreement with previous studies of the flow in wakes of bluff bodies13 and of Taylor bubbles in
pipes.10
The absolute values of the cross-correlation coefficient in the near wake region for turbulent
background flow (Fig. 12(a)) exhibit certain similarities to those observed in the laminar case.
The high negative values of the Reynolds shear stresses and of the cross-correlation coefficient at
distances z/D < 2 can probably be associated with vortices shed at the model rear as discussed above.
At Re = 7500, however, the cross-correlation coefficients in the near wake region do not exceed
the maximum negative values of about −0.5, notably lower than for the Re = 820 case, apparently
due to significant contribution of non-coherent turbulent velocity fluctuations. The cross-correlation
coefficients nearly vanish at about z = 2D and then increase up to a maximum of about 0.4 at a
distance of 4D. While this maximum value does not alter significantly with further increase in z, the
shape of the radial distribution of the cross-correlation coefficients varies significantly (Fig. 12(b))
and the location of the maximum gradually shifts towards the pipe wall. At about z/D = 20 shapes
similar to those observed for fully developed pipe flow are attained.

C. Turbulence production in the wake of the model


It should be stressed that only limited information is available on energy production in undevel-
oped turbulent flow in general and in undeveloped pipe flow in particular (see, e.g., Piirto et al.22 ).
This information is essential for understanding the characteristics of the flow in the Taylor bubble’s
wake. Computations of energy production require accurate data and differentiation of experimentally
obtained profiles. The inaccuracies in the experimental results in Shemer et al.10 due to the limited
ensemble size (about 200 events) and inevitable jitter in determination of the frame location relative
105103-11 Babin, Barnea, and Shemer Phys. Fluids 25, 105103 (2013)

0.1
(a) u’
v’

Fluctuations [m/s]
0.05

−0.05

−0.1
0 10 20 30 40 50
Time [s]

0.15
(b)

0.1
u’ [m/s]

0.05

−0.05
−0.06 −0.04 −0.02 0 0.02 0.04
v’ [m/s]
FIG. 13. (a) and (b) Radial and axial velocity fluctuations at a location with high cross correlation coefficient (about 0.9) for
Re = 820.

to the Taylor bubble in the presence of bubble bottom oscillations did not allow obtaining reliable
results on turbulent energy production in that study. In steady geometry of the present experiments,
accumulating large body of accurate data is much easier than for a moving Taylor bubble.
The turbulent kinetic energy in cylindrical coordinates system (r, θ , z) is defined as

1 2
q2 = (u + v 2 + w 2 ). (1)
2

(a) (b)
1 1

0.8 0.8

0.6 0.6
z/D

z/D

0.4 0.4

0.2 0.2

0 0
−0.5 −0.25 0 0.25 0.5 −0.5 −0.25 0 0.25 0.5
r/D r/D

FIG. 14. (a) and (b) Instantaneous maps of velocity fluctuations for Re = 820 at two instances.
105103-12 Babin, Barnea, and Shemer Phys. Fluids 25, 105103 (2013)

The flow entrance conditions and the axisymmetric model geometry in the present study assure that
the mean circumferential velocity component W̄ is zero everywhere. In the present study, distribu-
tions of the mean axial Ū and radial V̄ velocity components are determined with resolution that
allows computation of their spatial derivatives. This information, together with the measured spatial
distributions of the normal, u 2 and v 2 , and shear u  v  Reynolds stresses enables reliable estimates
of the turbulent kinetic energy production in the model wake. For axisymmetric undeveloped flow,
the expression for the turbulent energy production by the Reynolds stresses has the following form:
⎡ ⎤
⎢ ∂U ∂U ∂V ⎥
⎢ V ⎥
 = − ⎢(u 2 − v 2 ) − v 2 + u  v  ( + )⎥ . (2)
⎣  ∂z  r  ∂r ∂z ⎦
(I ) (I I ) (I I I a,b)

Terms (I) and (II) represent the contribution of normal stresses that not necessarily vanish in
undeveloped flow, while term (III) is the contribution of shear stresses. In the developed pipe flow,
only the 1st part of the third term (IIIa) remains in (2).
Radial distributions of the production terms in (2) are plotted in Figs. 15 and 16 at a number
of axial locations for laminar and turbulent background flows, respectively. Term (II) in (2) was
calculated and found to be negligibly small everywhere in the model wake and under all experimental
conditions, so it is not plotted in those figures. In view of the fast variation with the distance z/D of

3 40
z/D=0.1 z/D=0.6
2
Π/((U )3/D)

Π/((UL) /D)

1 20
3
L

0
−1 normal stresses 0
shear stresses
−2
total
−3 −20
0 0.1 0.2 0.3 0.4 0.5 0 0.1 0.2 0.3 0.4 0.5
r/D r/D

3 1.5
z/D=1 z/D=2
2
1
Π/((U )3/D)

Π/((U )3/D)

1
L

0.5
0

−1 0

−2 −0.5
0 0.1 0.2 0.3 0.4 0.5 0 0.1 0.2 0.3 0.4 0.5
r/D r/D
−3
x 10
0.08 10
z/D=4 z/D=10
0.06
Π/((UL)3/D)

Π/((U )3/D)

5
0.04
L

0.02
0
0

−0.02 −5
0 0.1 0.2 0.3 0.4 0.5 0 0.1 0.2 0.3 0.4 0.5
r/D r/D

FIG. 15. Radial distributions of the energy production term at various distances in the wake of the body for Re = 820.
105103-13 Babin, Barnea, and Shemer Phys. Fluids 25, 105103 (2013)

2 15
z/D=0.1 z/D=0.6
1 10
Π/((U ) /D)

Π/((UL)3/D)
3
L
0 5
normal stresses
−1 shear stresses 0
total
−2 −5
0 0.1 0.2 0.3 0.4 0.5 0 0.1 0.2 0.3 0.4 0.5
r/D r/D

3 0.05
z/D=1 z/D=2
2 0
Π/((U ) /D)

Π/((U )3/D)
3

1
L

−0.05

L
0 −0.1

−1 −0.15
0 0.1 0.2 0.3 0.4 0.5 0 0.1 0.2 0.3 0.4 0.5
r/D r/D
−3
x 10
0.04 20
z/D=3.9 z/D=10
0.03 15 normal stresses
Π/((UL)3/D)
Π/((UL) /D)

shear stresses
0.02 10
3

total
0.01 5 Laufer(1954)

0 0

−0.01 −5
0 0.1 0.2 0.3 0.4 0.5 0 0.1 0.2 0.3 0.4 0.5
r/D r/D

FIG. 16. Radial distributions of the energy production term for Re = 7500.

various terms in , different wake regions are presented in separate panels that may have vertical
scale that differ by orders of magnitude. The turbulence production is rendered dimensionless using
the pipe diameter D and the mean background flow velocity UL since these quantities do not vary
along the wake region.
For laminar background flow, Fig. 15, immediately downstream the model’s rear at z/D = 0.1,
all contributions to the turbulent kinetic energy production are small for r/D < 0.35. In the wall jet
region closer to the wall, the production terms become significant. Note that the contributions of
the normal and the shear stresses have different signs. The maximum value of the dimensionless
energy production attains a value close to unity at r/D ≈ 0.48. The production term then increases
by an order of magnitude and at z/D = 0.6 attains a maximum that exceeds 10 in the mixing
region, at about r/D = 0.4. The contributions due to normal and shear stresses again have opposite
signs, however, the total production is mostly positive; the domain of significant turbulent energy
production at z/D = 0.6 is notably wider (r/D > 0.25) than at z/D = 0.1. Note that closer to the wall
the normal stresses have a dominant contribution to the production of the turbulent kinetic energy.
The domain of turbulent energy production moves closer to the pipe axis (0.1 < r/D < 0.35) at z/D
= 1, with negative production closer to the axis and positive in the outer part of this domain. The
total production values  decrease sharply and the total maximum value is close to unity; become
small and positive, with normal and shear contributions having opposite signs roughly balance each
other. Approximately, the same maximum value and the location of  are retained at z/D = 2,
105103-14 Babin, Barnea, and Shemer Phys. Fluids 25, 105103 (2013)

however, the turbulent energy production at this axial location is mainly due to the shear stresses and
is spread over the whole pipe cross-section. Further downstream (z/D = 4) only the shear stresses
contribution remains essential; the maximum values of  decrease by an order of magnitude and
the production domain moves towards the wall. At z/D = 10 the maximum production term values
decrease further and the turbulent energy production becomes negligible at distances exceeding
about 20D, as expected for the laminar background flow.
For turbulent background flow (Re = 7500), the radial distributions of the normalized turbulence
energy production terms are presented in Fig. 16. Immediately downstream the body (z/D = 0.1)
the radial distribution  bears certain similarity to that measured in the laminar background flow,
see the 1st panel of Fig. 15. The notable difference between the two cases is the non-zero positive
turbulence production in the central part of the pipe, exclusively due to normal stresses. In both
Figs. 15 and 16 for z/D = 0.1, a domain exists with negative turbulent production. Farther away
from the body the peak values of  are attained at a similar location as in the laminar background
flow in the near wake region (z/D = 0.6, r/D = 0.4), however, contrary to the pattern observed in
Fig. 15, the dominant contribution to  in Fig. 16 at this axial location is due to shear stresses.
Similar to the production term distribution in the laminar background flow, the values of  at
z/D = 1 decrease considerably; the spread over the whole cross-section is somewhat more uniform,
with contribution of normal stresses being essential. At z/D = 2 the values of  decrease sharply
by an order of magnitude and become negative in the central part of the cross-section due to
the dominant and mostly negative contribution of terms associated with the normal stresses. At
z/D = 4, the contribution of the normal stresses becomes negligible, whereas the total production
is relatively large in the central part of the pipe, with a secondary smaller peak emerging near
the pipe wall. Farther away in the wake, z/D = 10, the radial distribution of  attains the shape
characteristic for the fully developed turbulent pipe flow. To validate the accuracy of the present
results, quantitative comparison with the classical results of Laufer21 (see also Hinze23 ) is carried
out in this panel. To enable the comparison, Laufer’s results are rescaled as in the present study.
The measured radial profiles of  at larger downstream distances from the body do not change
significantly.
The spatial evolution of the turbulent kinetic energy production in laminar and turbulent back-
ground flows thus exhibits significant similarities, in particular in the near wake region. At all axial
distances in the range 0.1 ≤ z/D ≤ 2, the normalized maximum values of  are nearly identical for
Re = 820 and Re = 7500, although the cross-sectional distributions and the relative contribution of
normal and shear stresses may differ somewhat. For both laminar and turbulent background flows,
the normalized production term attains maximum of about 15 at short distance from the body within
the mixing layer between the relatively stagnant liquid downstream of the body and the wall jet. This
maximum value is by 3 orders of magnitude higher than the maximum normalized  in a developed
turbulent pipe flow, cf. Fig. 16. The similarity in the values of normalized turbulent energy production
in the near wake region in spite of different background flow regimes in Figs. 15 and 16 stems from
the dominant contribution of the essentially turbulent mixing in this region that is not particularly
sensitive to the Reynolds number and the prevailing flow regime. For the Re = 820 case, at z/D ≥ 4 the
production term in the central part of the pipe effectively vanishes; the decay of the production in the
near wall region, however, is quite slow, with the peak value of the normalized  falling below 0.02
characteristic for the turbulent pipe flow only at z/D approaching 10. For Re = 7500 at z/D = 10, the
distribution of the turbulent energy production attains the shape typical for the fully developed pipe
flow.

D. Comparison between the model wake and that of an air bubble


The flow field in the wake of the solid model is now compared to that in the wake of a moving
gas Taylor bubble.10 To compare the far wake characteristics downstream of the model with those
behind a Taylor bubble, the same Re numbers were chosen. The flow in the far wake can be seen as
an undeveloped pipe flow affected by flow separation due to either the Taylor bubble or to the solid
fixed model.
105103-15 Babin, Barnea, and Shemer Phys. Fluids 25, 105103 (2013)

The jet velocity in the frame of reference fixed to the bubble/solid model characterizes the flow
in the near wake. Thus, in order to carry out quantitative comparison of the flow characteristics in the
near wake region for these two cases, the absolute mean flow velocity at the film exit cross-section
in this frame of reference, Ufb = |U f¯ − Utr |, should be equal. Here Uf is the exit mean film velocity.
The Taylor bubble translational velocities Utr were calculated using Nicklin et al.24 correlation Utr
= CUL + Ud with the coefficient C = 2.0 and C = 1.2 for laminar and turbulent background flows,
respectively. The mean film velocities Uf at the exit plane of the liquid film are calculated following
the model by Barnea19 for a 5D bubble length. It should be stressed that in spite of the fact that the
background flow rates and thus the Reynolds numbers in the Taylor bubble experiments by Gulitski9
vary by an order of magnitude, the resulting values of Ufb do not differ from 1.5 m/s by more than
10%. The flow in the near wake behind a Taylor bubble, studied by Gulitski9 and Shemer et al.,10
for the case with Re = 7500 is therefore compared here with the results in the near wake of the solid
model for flow with Re = 10 500 that corresponds to Ufb = 1.57 m/s.
Velocity profiles in the near wake of the model in the laminar flow regime (Re = 820) were
found to be different from those in the wake of a gas bubble. The reason is that for a Taylor
bubble case, the characteristic velocity scale in the vicinity of the bubble bottom is the bubble
translation velocity Utr . For the pipe diameter in the present experiments this velocity is sufficiently
high even in the laminar background flow,25 so that the flow close to the Taylor bubble bottom is
effectively turbulent for all Reynolds numbers considered. Similar behavior was therefore observed
in this domain by Shemer et al.10 for all background flow rates. However, for a solid fixed body,
the characteristic velocity in the near wake region for the laminar case is much lower than that
for a gas bubble resulting in significant differences in the flow characteristics between these two
cases.

1. Near wake region


The mean flow fields in the near wake of a Taylor bubble and of the solid model are presented
in Figs. 17(a) and 17(b). The vectors in the Taylor bubble wake were shifted by Utr to represent the
velocity field in the moving frame of reference. This velocity transformation renders the toroidal
vortex more visible. Both flow fields are qualitatively similar but have opposite directions. Although
the background flow in both Figs. 17(a) and 17(b) is directed upwards, the model and the Taylor
bubble point into opposite directions. It should be stressed that the axial velocity in the central part
of the pipe is directed towards the bubble/solid body in both cases.
The radial distributions of the mean axial velocities relative to the moving bubble normal-
ized by Ufb are compared with the corresponding profiles in the wake of the model in Fig. 18.
In view of Fig. 17, the sign of the normalized relative axial velocity in the near wake of the
Taylor bubble is changed. The corresponding radial profiles of the axial velocity in both cases are
very similar. However, close to the pipe wall the velocities downstream the model are somewhat
higher than those in the Taylor bubble wake, most probably due to the difference in the boundary
conditions.
Fig. 19 presents the distribution of mean axial velocity along the centerline, UCL , normalized
by Ufb . In the near wake, up to z = 2D, the values of UCL are taken relative to Utr with an opposite
sign. For z > 2D, the effect of the bubble translational velocity on the flow field in the wake becomes
negligible, the actual axial velocities in both cases normalized by Ufb are compared. Fig. 19 shows
similar distribution in both cases, with minor differences that may stem from a shift in the vortex
center location and some variation of its length.
The comparison of the mean radial velocities, normalized by Ufb , in the near wake is shown
in Fig. 20. In the vicinity of the bottom the velocities are higher for the model case. The mean
radial velocities are positive for all values of r close to the bottom and become negative at axial
locations exceeding the vortex center cross-section that is somewhat different in the two cases
causing qualitative disagreement at z/D = 0.6. The profiles become similar again at z/D = 1, farther
downstream the radial velocities vanish for both cases.
Figs. 21(a) and 21(b) present the comparison of the radial distributions of the r.m.s. values of
the axial u and of the radial v  velocity fluctuations. At z/D = 0.1, both u and v  behind the Taylor
105103-16 Babin, Barnea, and Shemer Phys. Fluids 25, 105103 (2013)

FIG. 17. Mean flow field in the near wake region: (a) Results by Gulitski9 for a Taylor bubble at Re = 7500; (b) Present
results for the solid model at Re = 10 500.

bubble are notably higher than those in the model wake. This difference may be attributed to the
effect of bubble bottom oscillations.14 For the rigid model, a much lower fluctuations level was
observed at this axial location. Farther downstream, the profiles of both u and v  become similar. At
each cross-section, the axial and radial fluctuations are of the same order, in agreement with Taylor
and Whitelow13 and Shemer et al.10

1.5
Solid model
z/D=0.1
z/D=0.6
1
z/D=1
z/D=2
U/Ufb

Taylor bubble
0.5
z/D=0.1
z/D=0.6
z/D=1
0
z/D=2

−0.5
0 0.1 0.2 0.3 0.4 0.5
r/D

FIG. 18. Comparison of the radial distribution of the axial velocity in the near wake for Ufb ≈ 1.5 m/s.
105103-17 Babin, Barnea, and Shemer Phys. Fluids 25, 105103 (2013)

FIG. 19. Comparison of the axial distribution of the centerline axial velocity in the near wake.

0.04
Solid model
z/D=0.1
0.02
z/D=0.6
z/D=1
0
z/D=2
V/Ufb

Taylor bubble
−0.02 z/D=0.1
z/D=0.6
−0.04 z/D=1
z/D=2
−0.06

−0.08
0 0.1 0.2 0.3 0.4 0.5
r/D

FIG. 20. Comparison of the radial distribution of the radial velocity in the near wake.

FIG. 21. Comparison of the radial distributions of the r.m.s. values of the velocity fluctuations in the near wake: (a) Axial
component; (b) Radial component.
105103-18 Babin, Barnea, and Shemer Phys. Fluids 25, 105103 (2013)

FIG. 22. Comparison of the radial distributions of the cross-correlation coefficient values.

Comparison of the radial distributions of the cross-correlation coefficients is plotted in Fig. 22 in


the near wake region for various downstream locations. For reasons specified above, the sign of the
cross-correlation coefficient in the Taylor bubble wake was changed. In all cases, in the immediate
vicinity of the Taylor bubble bottom, at z/D = 0.1, the cross-correlation coefficients are close to
zero, i.e., the velocity fluctuations in the axial and the radial directions remain largely uncorrelated
and the effective Reynolds stresses appear only at some distance downstream. The maximum values
of the cross-correlation coefficient are attained at about z/D = 1 and effectively vanish at about z/D
= 2 in both cases. It can thus be concluded that the comparison of the turbulent parameters in the
near wake shows significant qualitative similarity between a real air bubble and the model cases
except for the bottom vicinity region where the effect of the Taylor bubble bottom oscillations is
strong.

2. Far wake region


The flow fields in the far wake region of the solid model and of a Taylor bubble are compared for
identical flow rates. For the laminar background flow, the mean axial velocity behaves different in
the two cases. The flow behind the Taylor bubble attains a developed shape at much longer distances
from the bottom than in the solid model case; in both cases the limit value of 2 corresponding to
fully developed Poiseuille velocity profile was not attained. The axial variation of the mean axial
centerline velocity (normalized by UL ) presented in Fig. 23 clearly shows this. Contrary to the
laminar case, for the turbulent background flow the mean axial velocity variation in the far wake
is similar for both the solid model and the Taylor bubble. The fully developed velocity profiles are
attained with Ucl /UL ≈ 1.2. Normalized turbulent velocities were almost equal for all turbulent flows
(Fig. 5); therefore, comparison is only presented here for Re = 7500. Mean radial velocities in the
far wake are negligible.

FIG. 23. Comparison of the mean axial centerline velocity normalized by the corresponding mean liquid velocity at the far
wake.
105103-19 Babin, Barnea, and Shemer Phys. Fluids 25, 105103 (2013)

FIG. 24. Radial distributions of the r.m.s. values of the velocity fluctuations in the far wake for Re = 7500: (a) Axial
component; (b) Radial component.

Radial distributions of the r.m.s. values of the axial and radial velocity component fluctuations
u and v  in the far wake for Re = 7500 are presented in Fig. 24. Only a qualitative similarity between
the both cases exists. The values of u gradually increase in the radial direction attaining a maximum
close to the pipe wall, whereas the values of v  remain almost constant within the pipe’s cross-section
and decrease with the distance from the bottom. The turbulence level is higher for the real bubble
case; even at 70D from the bubble bottom the values of u continued to vary with distance, in contrast
to the model case, where for z/D > 20 the values of the velocity fluctuations remained constant.
Comparison of the radial distributions of the cross-correlation coefficient between the fluctu-
ations in the axial and the radial directions in the far wake is presented in Fig. 25. In the near
wake region certain similarity between the model and the bubble cases was observed. In the far
wake, however, the downstream evolution of the cross-correlation coefficient is quite different. The
cross-correlation between u and v  in the wake of the Taylor bubble practically vanished in the range
of 2 < z/D < 30 and then gradually attained distribution characteristic for turbulent pipe flow at z/D
> 50. For the solid model, the cross-correlation which is initially negative in the near wake (see Fig.
22) becomes small at end of the near wake domain at about z/D = 2, and then increases changes
the shape of radial distribution, approaching the characteristic radial distribution for fully developed
flow already at about z/D = 10.

FIG. 25. Radial distributions of the cross-correlation coefficient values in the far wake for Re = 7500.
105103-20 Babin, Barnea, and Shemer Phys. Fluids 25, 105103 (2013)

IV. SUMMARY AND CONCLUSIONS


The present study was motivated by difficulties in obtaining detailed and accurate information
on the flow in the wake of an elongated (Taylor) bubble rising in a vertical pipe in either stagnant or
co-flowing liquid.8–10 To mitigate complications associated with the bubble’s movement along the
pipe and with its unsteady shape that manifests itself in bubble bottom oscillations, a solid model of
the bubble that represents the mean shape of a Taylor bubble was constructed. Beyond its importance
for two-phase pipe flows, investigation of the wake flow downstream a Taylor bubble-shaped solid
body placed within a pipe can serve as an example of a separated flow around a bluff body in a
confined conduit and of transition from the highly disturbed flow in the near wake to fully developed
flow farther downstream.
Experimental study of the flow in the model’s wake offers significant advantages as compared to
carrying out measurements for a real gas bubble. There are, however, non-negligible dissimilarities
between these two flows that stem from different boundary conditions. For the pipe diameter in the
present experiments, the gas bubble motion due to buoyancy results in an effectively turbulent flow
even when the bubble propagates in a stagnant liquid.5 Thus the bubble motion has been taken into
account in determining the experimental conditions. The comparison of the developing flow in the
far wake of the bubble and the solid body is carried out for identical Reynolds numbers, whereas in
the near wake region the jet velocity in the frame of references fixed to the Taylor bubble or its model,
respectively, served as the velocity scale. The effect of the remaining differences in flow between
the two cases has been investigated by carrying out as detailed as possible comparison between the
present results and those obtained in the previous studies of the flow in the Taylor bubble’s wake.
PIV measurements of two velocity components in the vertical axial plane of the wake of the
model were carried out for laminar and turbulent background flows. For each flow condition and
each location in the wake of the model, results are presented based on an ensemble of at least 1000
independent recordings, thus allowing obtaining reliable mean flow parameters. The flow pattern in
the model wake is characterized by a toroidal vortex and has a strong similarity to that measured in
the near wake of a Taylor bubble. The spatial distributions of mean axial and radial velocities, as
well as of the normal and shear turbulent stresses are presented; the similarities and dissimilarities
between the wake flow structures in laminar and turbulent background flows are discussed.
The advantages in carrying out measurements in the wake of a fixed axisymmetric model
enable to carry out calculations of turbulent energy production terms and to distinguish between the
contributions of the normal and shear stresses to the total production.
For laminar background flow, unexpectedly high values (approaching unity) of cross-correlation
coefficient between axial and radial velocity fluctuation were obtained in the near wake of the model.
These cross-correlation coefficient values are notably higher than those measured in the turbulent
background flow. This phenomenon may be attributed to the existence of highly coherent structures
in the separated flow downstream the model.

ACKNOWLEDGMENTS
This study was supported by Grant No. 1444/09 from Israel Science Foundation.
1 D. Moissis and P. Griffith, “Entrance effects in a two-phase slug flow,” Trans. ASME, Ser. C: J. Heat Transfer 84, 29
(1962).
2 C. Aladjem Talvy, L. Shemer, and D. Barnea, “On the interaction between two consecutive elongated bubbles in a vertical

pipe,” Int. J. Multiphase Flow 26, 1905 (2000).


3 A. M. F. R. Pinto, M. N. Coelho Pinheiro, and J. B. L. M. Campos, “On the interaction of Taylor bubbles rising in two-phase

co-current slug flow in vertical columns: Turbulent wakes,” Exp. Fluids 31, 643 (2001).
4 L. Shemer, A. Gulitski, and D. Barnea, “Movement of two consecutive Taylor bubbles in vertical pipes,” Multiphase Sci.

Technol. 19, 99 (2007).


5 S. Quan, “Co-current flow effects on a rising Taylor bubble,” Int. J. Multiphase Flow 37, 888 (2011).
6 J. D. P. Araújo, J. M. Miranda, A. M. F. R. Pinto, and J. B. L. M. Campos, “Wide-ranging survey on the laminar flow of

individual Taylor bubbles rising through stagnant Newtonian liquids,” Int. J. Multiphase Flow 43, 131 (2012).
7 S. Nogueira, M. L. Reithmuller, J. B. L. M. Campos, and A. M. F. R. Pinto, “Flow patterns in the wake of a Taylor bubble

rising through vertical columns of stagnant and flowing Newtonian liquids: An experimental study,” Chem. Eng. Sci. 61,
7199 (2006).
105103-21 Babin, Barnea, and Shemer Phys. Fluids 25, 105103 (2013)

8 R. van Hout, A. Gulitski, D. Barnea, and L. Shemer, “Experimental investigation of the velocity field induced by a Taylor
bubble rising in stagnant water,” Int. J. Multiphase Flow 28, 579 (2002).
9 A. Gulitski, “Experimental investigation of the flow field in the wake of a Taylor bubble and its effect on the motion of a

consecutive bubble in vertical slug flow,” Ph.D. thesis (Tel-Aviv University, 2005).
10 L. Shemer, A. Gulitski, and D. Barnea, “Experiments on the turbulent structure and void fraction distribution in the Taylor

bubble wake,” Multiphase Sci. Technol. 17(1–2), 1 (2005); “On the turbulent structure in the wake of Taylor bubbles rising
in vertical pipes,” Phys. Fluids 19(3), 035108 (2007).
11 P. Bradshaw, “Review: Complex turbulent flows,” ASME Trans. J. Fluids Eng. 97, 146 (1975).
12 D. H. Wood and P. Bradshaw, “A turbulent mixing layer constrained by a solid surface. Part 2. Measurements in the

wall-bounded flow,” J. Fluid Mech. 139, 347 (1984).


13 A. M. K. P. Taylor and J. H. Whitelaw, “Velocity characteristics in the turbulent wakes of confined axisymmetric bluff

bodies,” J. Fluid Mech. 139, 391 (1984).


14 S. Polonsky, L. Shemer, and D. Barnea, “Averaged and time-depended characteristics of the motion of an elongated bubble

in a vertical pipe,” Int. J. Multiphase Flow 25, 795 (1999).


15 E. T. Tudose and M. Kawaji, “Experimental investigation of Taylor bubble acceleration mechanism in slug flow,” Chem.

Eng. Sci. 54, 5761 (1999).


16 A. Tokuhiro, M. Maekawa, K. Iizuka, K. Hishida, and M. Maeda, “Turbulent flow past a bubble and an ellipsoid using

shadow-image and PIV techniques,” Int. J. Multiphase Flow 24, 1383 (1998).
17 A. A. Sotiriadis and R. B. Thorpe, “Liquid re-circulation in turbulent vertical pipe flow behind a cylindrical bluff body and

a ventilated cavity attached to a sparger,” Chem. Eng. Sci. 60, 981 (2005).
18 B. Lertnuwat and A. Bunyajitradulya, “Effects of interfacial shear condition and tailing – Corner radius on the wake vortex

of a bubble,” Nucl. Eng. Des. 237, 1526 (2007).


19 D. Barnea, “Effect of bubble shape on pressure drop calculations in vertical slug flow,” Int. J. Multiphase Flow 16, 79

(1990).
20 T. van Hooff, B. Blocken, T. Defraeye, J. Carmeliet, and G. J. F. van Heijst, “PIV measurements of a plane wall jet in a

confined space at transitional slot Reynolds numbers,” Exp. Fluids 53, 499 (2012).
21 J. Laufer, “The structure of turbulence in fully developed pipe flow,” NACA Report, 1174, 1954.
22 M. Piirto, P. Saarenrinne, H. Eloranta, and R. Karvinen, “Measuring turbulence energy with PIV in a backward-facing step

flow,” Exp. Fluids 35, 219 (2003).


23 J. O. Hinze, Turbulence, 2nd ed. (McGraw-Hill, New York, 1975).
24 D. J. Nicklin, J. O. Wilkes, and J. F. Davidson, “Two-phase flow in vertical tubes,” Trans. Inst. Chem. Eng. 40, 61 (1962).
25 S. Polonsky, D. Barnea, and L. Shemer, “The relation between the Taylor bubble motion and the velocity field ahead of

it,” Int. J. Multiphase Flow 25, 957 (1999).

You might also like