Andreussi - 1999 - Dispersed Bubble Flow in Horizontal Pipes
Andreussi - 1999 - Dispersed Bubble Flow in Horizontal Pipes
Andreussi - 1999 - Dispersed Bubble Flow in Horizontal Pipes
Abstract
A set of experiments has been performed to analyse gas-phase distribution in horizontal bubble flow. The experimental data have
been obtained by means of conductance probes and include measurements of the local void fraction and the bubble diameter and
velocity. Present observations show large variations of the local void fraction and the bubble size moving from the top to the bottom
of the pipe, with local maxima located close to the upper pipe wall. Measurements also show that, at the boundary of the dispersed
bubble flow regime, the bubble size, and therefore the surface available for heat and mass transfer, changes abruptly for small
variations of the liquid or gas flowrate. This change allows the transition between the dispersed and the elongated bubble regimes to
be identified with good precision. Present data make it possible to develop a new correlation for the maximum bubble size in dispersed
bubble flow. 1999 Elsevier Science Ltd. All rights reserved.
0009-2509/99/$ — see front matter 1999 Elsevier Science Ltd. All rights reserved.
PII: S 0 0 0 9 - 2 5 0 9 ( 9 8 ) 0 0 2 8 9 - 9
1102 P. Andreussi et al. /Chemical Engineering Science 54 (1999) 1101—1107
q
¼e" A , (1) Fig. 1. Local void fraction along the vertical diameter.
p/d
where q is a characteristic pressure or shear stress acting
A mounted on the bench by precision supports. Liquid and
on the bubble, p the interfacial tension and d the
gas enter the pipe through a T-section. At the inlet,
maximum bubble size, exceeds a critical value, ¼e ,
stratified flow conditions are created by a thin diaphragm
a bubble of diameter larger than d breaks up due to
which separates the liquid from the gas inlet.
the effect of turbulent velocity fluctuations. The tube is equipped with differential pressure trans-
The present work proposes a new correlation of ducers to measure pressure drops, and with an absolute
bubble size based on the Kolmogoroff—Hinze theory. The transducer to determine the gas density. The cross-sec-
experimental observations make it possible both to pre- tion mean void fraction has been measured by the con-
dict the maximum size of bubbles flowing in dispersed ductance probe method (two-ring probe) described by
bubbles regime and to identify the range of existence of Andreussi et al. (1988). The test section used for the
this flow pattern. measurement of the local void fraction and bubble size is
The present investigation has been carried out using located 12 m far from the inlet section and about 5 m
conductance probes (Andreussi et al., 1991) for the from the outlet. These distances make it possible to
measurements of bubble size, diameter and concentra- obtain a fully developed dispersed bubble regime.
tion. This technique, also adopted by Kocamustafaogul- Measurements of the local bubble concentration have
lari and Wang (1991), offers many advantages over other been performed by means of the conductance probe,
methods, such as photography, light attenuation, ultra- schematically shown in Fig. 3a. The probe is made of
sonic attenuation or chemical absorption. To give an a thin copper wire, 0.3 mm in diameter, mounted inside
example, Fig. 1 illustrates the local void fraction profile a stainless steel tube, 1.0 mm OD. The copper wire is
along the vertical diameter determined with a local con- electrically insulated, except for the tip, and faces the flow
ductance probe. As shown in this figure bubbles tend to with a sharpened section 0.5 mm long. The presence of
migrate toward the upper wall because of the influence of gas bubbles at the tip can be detected by the drop of the
buoyancy and a well-defined peak of the local void frac- electrical conductance between the copper wire and the
tion can be observed at a distance of 2—3 mm from the stainless-steel tube. The distance between the tip and
upper pipe wall. the tube is less than 1 mm. Considering the size of the
probe, it can be assumed that measurements of bubble size
and local void become reliable when the typical bubble
2. Experimental set-up size is above 2 mm. In this range of bubble sizes, a com-
parison between the mean void fractions, measured with
The experiments have been performed in the atmo- the two-ring probe (Andreussi et al., 1988), and evaluated
spheric loop schematically shown in Fig. 2. Working from the measurements taken with the local probe, gave
fluids are water and air. The liquid is circulated by two satisfactory results, with a difference between the two
centrifugal pumps of different size, while air is supplied values of less than 20%. Considering that local probe
from a high-pressure line. Both streams are metered by measurements, on a pipe cross-section, were only taken
two sets of rotameters placed before the mixing section. along four diameters (see Fig. 4) and that in this example
The pipe is mounted on an inclinable bench 17 m long, a significant fraction of small bubbles was present, the
the slope of which can be varied continuously in a range agreement between these values appears to be acceptable.
of $3° by means of a motorised support. The test The analog signal from the local conductance
sections consist of transparent Plexiglas tubes with inner probe was converted into a digital signal and stored in
diameters of 18 and 50 mm, and the maximum length the transient recorder of a Lecroy 9400 Digital Oscil-
equal to the length of the bench. The tubes are made of loscope. Each sequence of data was transferred to a PC and
carefully flanged, 2 m long, interchangeable sections analysed. This procedure was followed in order to use
P. Andreussi et al. /Chemical Engineering Science 54 (1999) 1101—1107 1103
Fig. 3. Conductance probes for local void measurements (a), and for bubble velocity measurements (b).
large sampling frequencies. Considering that the transit be given by the 1/7 power law relative to single phase
time of a bubble 2 mm in diameter moving at 5 m/s is turbulent flow in a pipe. In order to check this assump-
0.4 ms, an acceptable sampling frequency in these experi- tion bubble velocity has also been measured with the
ments has been estimated to be of the order of 25 kHz. probe shown in Fig. 3b. The distance between the two
An even larger sampling frequency on two channels is tips is 4.1 mm, and as can be seen, the two-tip probe is
required in order to obtain a good measurement of the normal to the flow direction rather than parallel, as the
transit time of a bubble between the two-tip probe used single-tip probe. Considering the short distance be-
for velocity measurements. Bubble size has been derived tween the electrodes and the large sampling frequency,
from the transit time, as the product between transit time the measurements of the local velocity are fairly accu-
and bubble velocity. Bubble velocity has been assumed to rate. Andreussi et al. (1991) have shown that these
1104 P. Andreussi et al. /Chemical Engineering Science 54 (1999) 1101—1107
Fig. 4. Local void along on different pipe diameters. Inclination meas- Fig. 5. Experimental value of d plotted against the superficial mix-
ured with respect to the horizontal. (D"50 mm, v "0.2 m/s,
ture velocity.
v "4.4 m/s).
measurements are consistent with other experimental stable bubble size in the DB regime considering only data
observations made at the University of Pisa and that, as relative to d (6 mm.
a first approximation, the bubble velocity profile is close According to the Kolmogoroff—Hinze theory, d can
to the turbulent velocity profile relative to single phase be determined from Eq. (1), where the value of ¼e has
flow in a pipe. an order of magnitude
1. In the analysis of present
In the following, the maximum bubble size has been measurements it is necessary to assess the effect of coales-
defined, as suggested by Hesketh et al. (1987), as the cence on d , as the Kolmogoroff—Hinze theory was
diameter that is larger than 99% of all the diameters in developed for single bubble flow in a turbulent liquid
the distribution. This approach allows reliable data on stream. It is also necessary to define the characteristic
the maximum bubble size to be obtained. shear stress to be used in the definition of ¼e . A few
authors, Barnea (1986), Thomas (1981) and Hesketh et al.
(1987), following the Kolmogoroff—Hinze theory, as-
3. Analysis of measurements sumed that q scales as
A
q " o v , (2)
3.1. Maximum bubble size in the DB regime A J A
where the characteristic velocity v is defined as
A
It is assumed that dispersed bubble (DB) flow is char-
acterised by the presence of spherical bubbles of small v "(2(ed ). (3)
A
size (a few millimeters) flowing in the pipe. When bubbles In this equation, d is the maximum bubble size that
lose their spherical shape, coalesce with other bubbles
can exist when a single bubble flows in a full liquid pipe,
sharply increases and elongated bubble (EB) flow is ap- e is the energy dissipation due to turbulence
proached. This hypothesis is confirmed by present
measurements of bubble size. Fig. 5 represents the max- 2fv
e" AN , (4)
imum bubble size, d (evaluated as d ), as a function of D
the superficial mixture velocity. This figure shows that
and v is the velocity of the continuous phase, equal to
two regions exist and that the transition between them AN
the mixture velocity, v .
occurs at an approximately constant value, d ,
6 mm,
K
According to this hypothesis, d can be estimated as
of the maximum bubble size, independently of the super-
ficial gas or liquid velocity and of the pipe diameter. At p
the transition between DB and EB regimes, the max- d "¼e e\. (5)
o
imum bubble size and therefore, the area available for J
heat and mass transfer change abruptly, for instance, at Eq. (1) has been modified by Levich (1962), who included
v "0.2 m/s, D"50 mm, d is 6 mm for v "3.8 m/s in the definition of the critical Weber number the density
K
and is 30 mm for v "3.4 m/s. ratio o /o raised to 1/3 power. Recently, Hesketh et al.
K E J
Present observations have been interpreted assuming (1987) were able to interpret available data relative to the
that the maximum bubble size in dispersed bubble flow flow of gas—liquid, and also liquid—liquid dispersions on
depends on a critical Weber number and that, at the the basis of the theory developed by Levich.
transition to EB, bubble size increases without bond. Present experiments do not allow the effect of density
Present data show that this happens for d '6 mm. It ratio on bubble size to be detected, but according to the
is then possible to develop a correlation for the maximum theories developed by Kolmogoroff, Hinze and Levich, it
P. Andreussi et al. /Chemical Engineering Science 54 (1999) 1101—1107 1105
d q o
A· E "¼e "¼e (1#aaN L), (7)
p o
J
where the parameters ¼e , a and n can be evaluated by
a best fit procedure, and q is defined as in Eq. (2). In
A
Eq. (7) the Weber number has been multiplied by the
density ratio raised to 1/3 power, as proposed by Levich
(1962) in order to make a comparison between present
data and the correlation proposed by Hesketh et al.
(1987). The results obtained are reported in Fig. 7a, which
shows the experimental values of ¼e plotted against
the mean void fraction. As can be noticed from this
figure, the scatter is appreciable. On the other hand, the
value found for the critical Weber number at zero void,
¼e "1.05, is very close to the value determined by
Hesketh et al. (1987), ¼e "1.1, on the basis of a wider
set of experimental conditions. The final correlation ob-
Fig. 6. Local void fraction (a) and local d (b) along the vertical
tained for ¼e taking into account the effect of coales-
diameter (D"50 mm, v "0.15 m/s, v "4.53 m/s).
cence is
¼e "1.05(1#51.7aN
). (8)
can be expected that the Weber number, as defined in
Eq. (1) be a constant. This is approximately true for low An alternative correlation can be based on the assump-
void fractions, but is definitely incorrect at increasing tion that the characteristic shear or pressure force acting
values of the gas to liquid ratio. on a bubble, q simply scales as the wall shear stress, q .
A
The effect of void fraction on bubble size is apparent This approach can be justified considering that in hori-
from data relative to bubble size along the vertical dia- zontal flow the largest bubble is found very close to the
meter. These data are plotted in Fig. 6 along with the upper wall, at a position where the time averaged radial
measurements of the local void fraction as a function of velocity profile is far from being uniform. Under these
the vertical coordinate. From this figure, it can be noticed circumstances, the maximum stable bubble diameter may
that bubble size increases from a value of approximately depend on the mean shear stress rather than the dynamic
1.7 mm close to the bottom of the pipe, where the void is effect of turbulent velocity fluctuations.
about zero, to a maximum of about 3.7 mm at the top, The results are reported in Fig. 7b which represents the
where the void attains its maximum (a
0.13). experimental relation between ¼e and a in this case.
These data confirm that the critical Weber number or A comparison between Figs. 7a and b shows that a better
the maximum bubble size increases with the local void correlation of the experimental measurements is ob-
fraction. It is then necessary to modify the definition of tained when q is assumed to be the wall shear stress. The
A
¼e in order to take into account the effect of bubble correlation obtained for ¼e in this case is
coalescence. Calderbank (1958) and Barnea (1986) pro-
¼e "0.11(1#8.30aN
). (9)
posed that the actual maximum bubble diameter, d ,
can be related to the bubble diameter at zero void frac- The different value obtained for ¼e with respect to the
tion, d , by the equation previous correlation is due to the choice of the character-
istic shear stress q , as equal to the wall shear stress q .
d "d (1#aaL), (6) A
Using this equation the mean square error between mea-
where a and n are two adjustable parameters and a is the sured and predicted value of d
is 12%. Eq. (8) gives
local void fraction. a mean square error of 19%.
In the present work, the local void fraction has only
been determined for a limited set of flow conditions. It is Flow regime transitions
then necessary to develop a correction for bubble coales-
cence based on the mean void fraction, a, rather than the Bubble sizes reported in Fig. 5, allow the two flow
local value. As the slip between the gas and the liquid is patterns, DB and EB, to be clearly identified. In fact, for
negligible, in horizontal flow, the mean void fraction, a, each gas flow, a limiting value of the liquid velocity below
1106 P. Andreussi et al. /Chemical Engineering Science 54 (1999) 1101—1107