Experimental Thermal and Fluid Science 94 (2018) 34–48

Contents lists available at ScienceDirect

Experimental Thermal and Fluid Science

journal homepage: www.elsevier.com/locate/etfs

Investigating bubble dynamics in a bubble column containing shear thinning T

liquid using a dual-tip probe
⁎
Niloufar Jamshidi, Navid Mostoufi
Multiphase Systems Research Lab., School of Chemical Engineering, College of Engineering, University of Tehran, P.O. Box11155/4563, Tehran, Iran

A R T I C L E I N F O A B S T R A C T

Keywords: Bubble dynamics was characterized experimentally in non-Newtonian shear thinning fluids with different
Viscosity viscosities by a dual-tip resistivity probe. The experiments were carried out in a 9 cm diameter column while the
Bubble dynamics superficial gas velocity varied between 0.5 and 7 cm/s. Air, injected through a diffuser, was the dispersed phase
Shear thinning and various solutions of carboxy-methyl cellulose were used as the continuous liquid phase. An algorithm was
Electro-resistivity probe
developed to detect each individual contact of bubbles with the probe needles. It was shown that increasing the
Bubble column
viscosity produces larger bubbles which make the homogeneous flow regime instable at liquid viscosities greater
than 7 cP. It was found that the axial profiles of bubble chord length and velocity are affected by both shear
thinning behavior of the solutions and the viscosity of the solution. The shear thinning behavior causes a des-
cending trend of axial distribution of bubble rise velocity. Moreover, gas holdup reaches a maximum in the
middle of the column in solutions studied. The probability distribution of bubble chord length was well fitted by
a log-normal distribution. However, the bubble rise velocity distribution is Gaussian in the homogeneous regime
at low gas flow rates. Correlations are given for evaluating bubble chord length and bubble rise velocity.

1. Introduction Various measurement techniques have been employed to investigate

Bubble columns are extensively used in a wide range of industrial
processes, including industrial waste treatment [1,2], Fischer-Tropsch
synthesis [3,4], enhanced oil recovery [5], evaporation [6] and fer-
mentation [7,8], in which the continuous phase is often a non-New-
tonian liquid [9,10]. These columns have many advantages as reactors
which have promoted their uses in industry. For instance, large dif-
ference between densities of gas and liquid causes phases to mix well.
Also, in catalytic reactions, the age of the solid catalyst particles would
be prolonged, since the reaction heat is easily released to liquid from
the particles because of high heat transfer coefficient of the surrounding
liquid. Simple construction of bubble columns easifies the operation
and it also reduces the relative operational costs [11]. However, the
complex hydrodynamics of bubble column makes the design and scale-
up of these reactors challenging. Therefore, further research on various
hydrodynamic parameters as well as bubble dynamics in bubble col-
umns is needed. The bubble characteristics is a key factor affecting the
hydrodynamics of bubble columns and their scale-up [12–14]. The
bubble velocity dictates the gas phase residence time in liquid. Thus, it
affects the conversion in a reactor [12]. Therefore, it is important to
investigate the bubble dynamics in more details in non-Newtonian
viscous liquids which are widely used in industrial reactors.
the behavior of bubbles in bubble columns. Bubble frequency is an
important parameter which directly affects the gas holdup [15]. Also,
bubble chord length is directly related to the bubble size [16]. These
two parameters are important because based on them interfacial mass
transfer area in a bubble column reactor can be estimated. The bubble
rise velocity and bubble chord length can be calculated by both single-
tip and dual-tip probes. The way of estimation of rise time of bubbles
differs in signal processing for single-tip [17–19] and dual-tip probes
[20–24]. The resistivity or conductivity probes and fiber optic probes
are the most applied methods of probing measurement techniques
[17,18,23,25–27]. Although, the bubble rise velocity can be determined
using a single-tip probe, a dual-tip probe can do the same with more
accuracy. Measuring the bubble rise velocity by a single-tip probe
contains two stages, (i) selecting two thresholds in the voltage signal (at
rise of the voltage when the bubble is piercing the tip) and (ii) evalu-
ating the rise time from rise of the voltage considering the chosen
thresholds and estimating the bubble rise velocity by multiplying the
rise time by a coefficient of proportionality obtained from the probe
calibration [19,26]. Assessing the rise time from the voltage signal is
affected by the level of thresholds as well as the liquid film drainage
time of the probe tip [28–30]. Furthermore, an accurate and specific
calibration procedure is needed to perform measurements, taking into

⁎
Corresponding author.
E-mail address: mostoufi@ut.ac.ir (N. Mostoufi).

https://doi.org/10.1016/j.expthermflusci.2018.01.034
Received 28 March 2017; Received in revised form 6 December 2017; Accepted 27 January 2018
Available online 31 January 2018
0894-1777/ © 2018 Elsevier Inc. All rights reserved.
N. Jamshidi, N. Mostoufi
Nomenclature
dc column diameter, (mm)
F bubble frequency, (Hz)
Fs1,s2 cross correlation of S1 and S2
H column height, (m)
h height of the probe position, (m)
hfalling height in which the probe was released, (m)
db bubble diameter, (mm)
d b' bubble chord length, (mm)
K consistency index, (Pa.sn)
n flow index
Np number of peaks in signal
vth threshold level voltage, (V)
T test time, (s)
t time, (s)
ti peak width, (s)
S1 response signal of the downstream needle, (V)
account both single-tip probe design and liquid-gas system [29,31]. In
the dual-tip probing, the bubble rise time is the lag time between the
corresponding peaks appearing in response signals of the first and the
second tips. It is reported that despite proper calibration of single-tip
probe, still its measurements are associated with uncertainties [26].
Therefore, the rise time can be estimated directly by a dual-tip probe
which is more accurate compared to a single-tip probe. In the present
work, the dual-tip resistivity probe was utilized to measure bubble
dynamics.
Magaud et al. [32] considered four probable arrangements for
bubble crossing a dual-tip probe and accepted one of these layouts by a
simple statistical analysis. Several researchers have considered the
temporal distance between two successive peaks in the signals of
downstream and the upstream wired mesh sensors to be the shortest
bubble rise time in each event [23,33,34]. Other methods, such as
cross-correlation and multichannel methods, also can be employed to
calculate the mean bubble rise velocity [35]. A major drawback of these
methods is that it is not easy to detect the distribution of bubble rise
velocity and chord length. In the presented study, bubble rise time
distribution was obtained by implementing cross-correlation method
applied to small segments of two pulsed signal pairs simultaneously for
detecting individual bubble.
Many researchers have studied the effect of liquid viscosity on the
hydrodynamic parameters of bubble column, mainly gas holdup.
However, a few researches have reported the effect of liquid viscosity
on bubble dynamic parameters, like bubble size and bubble velocity.
The effect of liquid viscosity on the column hydrodynamics also has
been widely studied [21,22,36–42]. In most of these works, velocity
and the size of bubbles and the corresponding distributions were not
investigated. Some researchers studied the bubbly flow in non-New-
tonian solutions [21,43–46] and investigated the effect of viscosity on
gas holdup without focusing on bubble dynamics. Most of these re-
searchers reported the negative effect of liquid viscosity on the gas
holdup. Nevertheless, there are researches reporting an initially slight
increase and a subsequent decrease of the gas holdup with increasing
the liquid viscosity [22,40,47]. More recently, researchers have in-
vestigated the effect of liquid rheology on gas holdup and bubble size
[18]. though; the bubble rise velocity and relative distributions are not
studied. In major of the studies, it is confirmed that as the liquid visc-
osity increases, large bubble fractions and relative distribution width
increases as gas holdup decreases [21,22,41,48]. Though, for bubble
rise velocity which is less investigated, diverse results are reported. Wu
et al. [22] proposed that for CMC solutions with viscosities less than
10.39 cP, bubble rise velocity increases as the viscosity increases and
then remains constant for viscosities above 10.39 cP. Although, in
35
Experimental Thermal and Fluid Science 94 (2018) 34–48
S2 response signal of the upstream needle, (V)
vg mean gas voltage, (m/s)
vl mean liquid voltage, (m/s)
V voltage amplitude, (V)
v impact velocity, (m/s)
ugs superficial gas velocity, (m/s)
Greek letters
γ shear rate, (s−1)
ɛg gas holdup
µapp apparent viscosity, (Pa s)
ν dynamic viscosity, (m2 s−1)
σ surface tension, (N m−1)
ρl liquid density, (kg m−3)
Φg(ti) phase density function
Ω electrical resistance, (Ω)
another study with CMC solutions the increasing trend of bubble rise
velocity with viscosity is confirmed up to 23 cP [21]. Therefore, the
systematically investigating methods for direct measuring of the bubble
dynamics, especially bubble rise velocity and its distribution need to be
provided.
In this work, a non-Newtonian shear thinning liquid (aqueous so-
lutions of carboxy methyl cellulose or CMC) was used for investigating
the effect of viscosity on the bubble dynamics in a bubble column.
Bubble velocity, bubble chord length, gas holdup, and bubble fre-
quencies were measured by a dual-tip resistivity probe. The effects of
CMC concentration, axial position of the probe and superficial gas ve-
locity on the bubble dynamics were investigated. Also, the impact of the
shear thinning behavior of liquid on bubble dynamics along the column
was discussed. An algorithm was also developed to detect individual
bubbles not deviated after touching the first needle.
2. Experiments
2.1. Experimental setup and materials
A cylindrical lab-scale bubble column made of Plexiglas was used to
conduct experiments. The schematic of the experimental setup is sket-
ched in Fig. 1. A perforated distributor was employed with about 0.3%
open area. The column was initially filled with liquid, roughly up to L/
D = 13. The gas flow rate was controlled using a mass flow controller
(Alicat – MC series 50 SLPM). Details of the experimental setup are
provided in Table 1.
Solutions of CMC in tap water with different concentrations (0.05,
0.1, 0.2, 0.6%wt, with corresponding viscosities of 1.4, 4.1, 7 and 21
cP, respectively) were used in the experiments. The solutions were
prepared by gradual dissolving of small amounts of CMC powder to a
stirred tank containing tap water. Physical properties of these liquids
are given in Table 2. Surface tension of each solution was measured
with a Krüss K6 tensiometer by the platinum ring detachment method
with a precision of 1 mN/m. The simple shear study was also carried out
using modular compact rheometer (MCR300 SN621205, Anton Paar) in
the range of 1–1000 s−1. The solutions exhibited a stronger shear
thinning behavior as the CMC concentration was increased. The shear
thinning behavior of CMC solutions is also reported in other in-
vestigations [21,49]. The power-law parameters evaluated based on the
experimental data are reported in Table 2. Note that the values reported
in this table are representative for the apparent viscosity at
ugs = 0.05 cm/s. The apparent viscosity in this table was estimated
from [12]:
N. Jamshidi, N. Mostoufi Experimental Thermal and Fluid Science 94 (2018) 34–48

Fig. 2. Geometric configuration of the dual-tip resistivity probe.

Fig. 1. Schematic of the experimental setup.

μapp = Kγ n − 1 (1)
Table 1
Experimental apparatus and conditions.
where the mean shear rate in bubble column can be obtained from [12]:
Column dc = 90 mm
H = 175 cm ugs gρ 1/(n + 1)
Gas distributor (perforated design) Number of orifices = 100 γ = ⎛10.3n−0.63 ⎞
⎝ K ⎠ (2)
Orifice diameter = 0.5 mm
Mass flow controller Alicat (MC series)
A/D convertor Advantech (USB – 4716)
Tests were conducted by placing the probe at various altitudes such
Probe needle spacing 1.8 mm that its tips were placed at the central radial position of the column. The
Static liquid height 130 cm gas flow rate was then changed in every individual liquid mixture. Axial
Superficial gas velocity 0.5–7 cm/s positions of the probe were 20, 75, and 100 cm above the distributor
Operating temperature 22 °C
level. The superficial gas velocity was varied between 0.5 and 7 cm/s.

Table 2
2.2. Dual-tip resistivity probe
Physical properties of CMC solutions.

Concentration, wt% ρ, kg/m3 σ, N/m K, Pa sn n µapp, Pa s (γ from Schematic of the dual-tip resistivity probe is presented in Fig. 2. The
Eq. (2), @ principle of phase recognition in this probe is based upon the difference
ugs = 0.05 m/s) in electrical conductivity of the phases (gas/liquid). As demonstrated in
0 (water) 1049 0.073 0.001 1 0.001
Fig. 2, the sensing segments of the probe were beveled tips of two 27-G
0.05 1049.4 0.0757 0.00158 0.984 0.0014 spinal needles. These needles were electrically insulated by polyester
0.1 1049.6 0.0652 0.0049 0.976 0.004138 resin, except at the tips. Then, these fine needles were placed in a
0.2 1050.1 0.054 0.0106 0.948 0.00746 cannula (16-G angiocath) to hold them close. Afterwards, the needles
0.6 1051.3 0.052 0.0481 0.870 0.02132
were stained with polyester resin again to maintain the suitable cover of
insulating resin. Finally, this assembly was placed in another cannula
(14-G angiocath) which was acting as the ground electrode. Vertical
distance between the two tips was 1.8 mm.

36
N. Jamshidi, N. Mostoufi
2.3. Data acquisition
To do the measurements by the electro-resistivity probe, a signal of
a certain voltage was applied to the probe needles. In this way, these
needles were converted to variable resistors. Meanwhile, the response
voltage of the needles (variable resistors) in the experimental condition
was then collected. Fig. 3 shows a schematic of the electrical circuit
used for both applying input signal to the probe and scanning the
output signal of each of the needles. A sinusoidal input signal with
5 kHz frequency and 3 V amplitude was applied to the circuit. The al-
ternative current was used to prevent needle corrosion. The voltage of
the needles was sent through the electrical circuit to the data acquisi-
tion system. Data of the response signals of needles were collected via a
data acquisition system comprised of an analogue to digital converter
card (Advantech USB – 4716) connected to a computer. The sampling
frequency for each channel was set to 100 kHz. The sampling time in
each test was 100 s which was the shortest time after which the time
averaged gas holdup becomes stable. Each bubble sensed by the probe
makes a peak in response signals of both needles due to increasing the
relative resistivity of needles. Characteristics of bubbles can be ex-
tracted by processing the fluctuations of these signals.
3. Data analysis procedure
3.1. Bubble detection
Steps of the contact of a bubble with a dual-tip of an electro-re-
sistivity probe are shown schematically in Fig. 4. When the bubble
touches the tip of the probe at time t1, electrical resistivity of the
medium between the needle tip (positive pole) and the cannula (ground
pole) increases. According to Ohm’s law, so does the voltage between
the needles. The bubble then reaches the second needle at t2. This is
when the voltage of the second needle jumps for the same reason. The
lag time can be calculated from pulsed signals forms as it is described in
the following sections. Knowing the distance between tips of the two
needles, the calculated lag time can be used to estimate the bubble rise
velocity. When the bubble continues its upward movement, it passes
over the probe and the liquid gradually replaces the gas between
electrode poles. In this situation, the electrical resistance of the probe
decreases and the voltage drops significantly. This phenomenon occurs
at times t3 and t4 for the first and the second needle, respectively. The
time when the bubble rises through the probe and the needle is sur-
rounded by the gas phase is called the bubble flight time. In the present
study, the flight time of the first needle (t3-t1) was considered since
after passing through the first needle the bubble path may diverge a
little due to intrusive effect of the first needle. One can estimate the
bubble chord length from the flight time of the bubble if the bubble rise
velocity is known. Hence, after acquiring the raw response signals, the
fluctuations of the output voltage signal (which are caused by the
contacts between needles and bubbles) can be used firstly for detecting
bubble-needle contacts (through sudden rise and fall of the voltage) and
then estimating both the lag time and bubble flight time (from which
the bubble rise velocity and bubble chord length can be calculated)
according to the algorithm described above.
3.2. Signal processing
Fig. 5 demonstrates the signal processing steps applied to a short
interval of the response signal. An example of the raw sampled signal is
illustrated in Fig. 5(a). As mentioned above, step-like fluctuations are
generated when the needle is touched by bubbles. Since an alternative
voltage was applied to the needle, each change in the voltage amplitude
mirrors in both positive and negative values of the signal. Therefore,
the information in each signal can be found in the absolute value of any
change in the voltage amplitude, as shown in Fig. 5(b). The relative
maxima of each signal demonstrate passage of bubbles. These maxima
37
Experimental Thermal and Fluid Science 94 (2018) 34–48
were picked up, among all data points of the raw signal to obtain the
fluctuations of the signal. In this way, the new signal of fluctuations has
much less data points reducing further computational expenses. This
signal was then smoothed by discrete wavelet transform (DWT). The
DWT was adopted for de-noising based on the physical form of the
signal. The signal has numerous abrupt changes in a short time which
results in pulses in the frequency domain. For de-noising such signals,
DWT is desirable since it keeps the signal information in the same time
interval after applying de-noising analysis. In the present study, a
nearly asymmetric and orthogonal mother wavelet (sym-5) was used
[23,27]. Number of decomposition levels was considered to be 5. The
smoothed signal is shown in Fig. 5(c) illustrating the passage of bubbles
more clearly.
In order to detect phases (gas/liquid) from the amplitude signal, the
phase density square wave was formed by defining a threshold for the
amplitude signal. In the present work, the threshold was considered to
be the liquid phase voltage value (vl) plus 15% of the voltage gap be-
tween the mean voltage in air and in liquid (vg - vl). The phase density
square wave is then defined as:
1 if:S (ti ) ⩾ vth = vl + 0.15(vg−vl )
φ (ti ) = ⎧
⎨
⎩ 0 else (3)
where S(ti) is the voltage of the signal S at time ti and νth is the threshold
voltage. With this definition, when the needle touches a bubble and
voltage of the signal exceeds the threshold value, corresponding value
of the phase density function becomes 1. This value of the phase density
function indicates the presence of gas phase at the specified time. On
the other hand, if the voltage does not change enough to reach the
threshold voltage, the phase density function would be 0 reporting the
presence of liquid at the tip of the probe. Such a phase density function
is shown in Fig. 5(d).
Each peak in the phase density function is associated with a bubble
passage from the probe. Width of the square peak in the phase density
function is proportional to the time the bubbles are present at the tip of
the probe. The summation of the bubbles residence time, ti, during the
test time, T, is the time that the needle was surrounded with the gas
phase (bubble). Moreover, the number of bubbles pierced by the probe
corresponds to the number of peaks in the phase density function, Np. In
this study all quantities were measured for the bubbles with positive
velocity, i.e., those moving upward along the column. These values
were used for calculating the gas holdup and bubble frequency, re-
spectively, as follows [21,23,50]:
∑i ti
εg =
T (4)
Np
f=
T (5)
In previous researches, constructing the phase density signal was
considered as the last step of signal processing for calculating velocity
and chord length of bubbles [23,27]. In the present work, however, the
Fig. 3. Electrical circuit used for applying the voltage to each probe needle and gathering
the data of the response voltage of the needle.
N. Jamshidi, N. Mostoufi
signal processing was continued in order to estimate the bubble rise
velocity and chord length more accurately by transferring the square
waves in the phase density into pulsed signals, as shown in Fig. 5(e).
Each square peak in the phase density signal can be specified with two
parameters: (i) its start time and (ii) its length. In the present work,
each square peak in the phase density signal was represented by a pulse
at the middle of the corresponding square wave. In the pulsed signal,
each bubble contact is transformed into a unique pulse. Thus, the
maximum of the cross correlation of pulsed sub-signals provide the
most probable delay of the most resembled pulses. Another advantage
of this transformation is further simplifying the calculations.
4. Calculating bubble dynamics parameters
Bubble velocity and bubble chord length are two key parameters of
bubble dynamics. In the present study, the method of signal processing
is proposed for evaluating these parameters. Bubble velocity was cal-
culated using the lag time between two successive peaks in signals of
the upstream and downstream needles of the probe. In the next step,
bubble chord length was estimated using bubble velocity and bubble
flight time. The flow chart of the algorithm used in this work for
evaluating bubble characteristics is presented in Fig. 6.
4.1. Bubble velocity
In previous similar works, bubble velocity was determined using the
lag time between two successive peaks in probe signals. The lag time is
the time between the first contact of bubble with the first needle and
the first contact of bubble with the second needle (see Fig. 4). Thus,
bubble velocity can be estimated from:
L needles. The first scheme is the case in which a bubble rises vertically
VB = 0
tlag (6)
where VB is the bubble velocity, L0 is the needle spacing and tlag is the
lag time. The method of assessing the lag time significantly affects the
estimated value of the bubble velocity. Magaud et al. [32] considered
four probable patterns in the signal to show how a bubble may hit the
probe. A major drawback of their method is that every two succeeding
peaks in the two signals might not be emerged from the same bubble. It
38
Experimental Thermal and Fluid Science 94 (2018) 34–48
Fig. 4. Schematic diagram of bubble contact with
dual-tip resistivity probe.
was found in the experiments of this work that the bubble frequency
based on the upstream needle signal is always greater than that of the
downstream one. This means that some bubbles either have deviated
from the vertical direction hitting the first needle such that it misses the
second needle or their paths were not straight upward. Thus, signal
patterns, other than those proposed by Magaud et al. [32], can be
proposed to represent all other modes of bubble contact with the probe.
The method proposed in the present study for estimating the lag
time improves the measurement accuracy by computing the lag time for
only those bubbles not deviated from their vertical direction. In this
method only consequential peaks in both signals which are generated
by the same bubble are identified. Cross correlation function was used
for locating the same peaks in two signals. The time of maximum cross
correlation indicates the lag time between the same peaks in the two
signals. The cross correlation function for two discrete signals S1 and S2
is:
n−1
Fj (n) = ∑ s2 (k ) s1 (j + k ) j= −n + 1,…,−1,0,1,…,n−1
k=0 (7)
where n is the length of signals and j is the lag time. The cross corre-
lation function can reveal the time lag between the two consequent
pulses but does not guarantee that these two pulses are originated from
the same bubble. Therefore, a time window with limited length was
considered to ensure that the contacts are related to the same bubble.
The length of the window was selected short enough to detect only a
single bubble contact.
4.2. Possible schemes of bubble-probe contacts
Fig. 7 depicts possible contact schemes between a bubble and the
after contacting the needles. This bubble first hits the first needle and
then passes through the second needle (see pattern 1 in Fig. 7). This
situation is the desirable base case which is illustrated as pattern 1 in
Fig. 7. In this case, two successive similar peaks occurring in the first
and the second needle signals are located in a time window. This pat-
tern has a positive time lag in the cross correlation and is therefor ac-
ceptable. Pattern 2 in Fig. 7 is encountered when a bubble hits both
needles simultaneously. This means that the direction of the movement
N. Jamshidi, N. Mostoufi
Fig. 5. (a) Original signal; (b) amplitude of the original signal; (c) de-noised amplitude
signal; (d) phase density function; (e) pulsed signal.
of the bubble is not vertical or two different bubbles have hit either of
the needles. As a result, the maximum of the cross correlation in this
scheme occurs at time zero and it should be rejected at the time of data
processing. The contact of a bubble moving downward is illustrated in
pattern 3. In this case, the bubble touches the second needle first and
then the first one. Considering the location of the probe at the center of
the column, most bubbles have positive velocity. In fact, there would be
more erroneous results than real bubbles with negative velocity. Thus,
this scheme was also rejected in the processing since the pattern of the
peaks does not satisfy the time sequence of contacts with needles and it
produces the cross correlation peak at a negative time. Another scheme,
which was not considered by previous researchers, can be encountered
when the bubble deviates from its main direction after it is sensed by
the first needle. In this scheme, which is shown as pattern 4 in Fig. 7,
the bubble touches the first needle and changes its path without
touching the second one. The second peak does not exist in the second
signal and calculation of the cross correlation is impossible. Pattern 5 in
Fig. 7 demonstrates a case in which a bubble is deviated from its path
39
Experimental Thermal and Fluid Science 94 (2018) 34–48
after its contact with the first needle and another bubble reaches the
second needle in the same time window. This is most likely to happen
when the gas velocity is high and the bubble frequency measured by the
first needle is considerably greater than that of the second one. In the
pattern 5 shown in Fig. 7, the signal corresponding to contacts of two
bubbles with the first needle is shown. One of the bubbles deviates after
it hits the first needle and another passes through the second needle.
Therefore, there will be two peaks for this scheme in the cross corre-
lation calculations. This case was accepted during the processing by
obtaining the delay time based on the maximum value of the cross
correlation. It is worth mentioning that pattern 5 was not treated cor-
rectly by previous researchers. Applying transformation of the square
wave to the pulse signal in this work creates clear peaks in cross cor-
relation of the most similar peaks relating to a specific bubble. This
approach results in rejection of the false contact detection which can be
encountered frequently at high gas velocities.
In the present study, all patterns in Fig. 7 were taken into account
during analysis of signals and filtered (accepted/rejected) by the afor-
mentioned approach for evaluating bubble dynamics parameters. By
implementing this approach, the influence of false contacts provided by
deviated bubbles were eliminated. This error was not trivial because it
was found in the experiments of the present work that the bubble fre-
quency obtained by the first needle in all tests is greater than for the
second one. This shows that there are considerable numbers of bubbles
that deviate after hitting the first needle. This problem is more serious
at higher superficial gas velocities.
4.3. Signal discretization
The signal patterns of bubbles touching the probe were searched
and identified in many sub-signals in the time domain. After validation
of the right pattern according to the criteria described in the previous
section, the bubble lag time was calculated and the bubble rise velocity
was obtained from Eq. (6) for each sub-signal (Fig. 8, left side). The best
length of the sub-signals was determined by checking the histogram of
bubble rise velocity distribution obtained from searching all the sub-
signal (Fig. 8, right side). If the segment length is too short, slow
bubbles would not be found since the relative peaks would appear at a
longer time. This situation is presented in Fig. 8(a) in which 0.01 s was
considered as the length of sub-signals (corresponding to 1000 data
points with sampling frequency of 100 kHz). It can be seen in this figure
that velocities less than around 0.18 m/s were not detected and are not
seen in the velocity distribution. Slower bubbles can be detected as the
length of the sub-signals is increased. This can be seen in Fig. 8((b) and
(c)) in which slower bubbles are counted by increasing the segment
length and are shown in the velocity distribution. By further increasing
the length of sub-signals and considering very long sub-signals, number
of detected slow bubbles increases and even unrealistic slow bubbles
are detected, as can be seen (see Fig. 8(d)). In fact, longer segments
include more peaks and results of the cross-correlation become erro-
neous due to detection of fake slow bubbles. The length of sub-signals
increased to the limit that the fake slow bubbles do not appear with
high frequency since it is known that not many slow bubbles can be
found in the center of the column. Therefore, length of the time window
was considered such that the number of bubbles with velocity less than
10 cm/s do not show a peak in the distribution. This means that only
the fake slow bubbles will be excluded. For instance, in the case shown
in Fig. 8(c), the proper segment length is 0.05 s.
4.4. Bubble chord length
After determination of the bubble rise velocity, the bubble chord
length can be estimated from:
dB′ = VB t f lg (8)
N. Jamshidi, N. Mostoufi Experimental Thermal and Fluid Science 94 (2018) 34–48

Fig. 6. Algorithm of data analysis.

where VB is the bubble velocity and tflg is the bubble flight time. Bubble
flight time is the time when the bubble passes through the first needle
(t2−t1 in Fig. 4). The second needle has much less invasive effect on
bubbles, thus the data from the downstream needle were selected for
estimating the bubble flight time.
5. Results and discussion
5.1. Validation of the probe data
Before using the experimental data acquired in this work, the ac-
curacy of the measurements was validated. Probe data were compared

Fig. 7. Probable patterns of collision between a bubble and needles of the probe.

40
N. Jamshidi, N. Mostoufi
Fig. 8. Cross-correlation of small signal pairs (left) and corresponding bubble rise velocity distribution (right).
with two reference data set for both measured velocity and bubble
chord length. Experiments were conducted to find the bubble velocity
estimated by probe and the velocity was compared with the actual
velocity. For this purpose, free falling of the probe from air to water at a
certain height above the liquid level was considered. The velocity of
falling probe at the interface of air-water (where the phase changes),
which is known theoretically from the kinematic equation of free fall,
was considered as the reference for velocity measurement. In this test,
the ground pole was connected to the water container and the probe
was released from a specified height above the water surface. Once each
of the first and the second needle touched the air-water interface, the
voltage of each tip decreases in order to increase in the conductivity at
the air-water interface. The delay time between the peaks in the probe
signals were used to estimate the velocity as described above (Eq. (6)).
41
Experimental Thermal and Fluid Science 94 (2018) 34–48
Velocities were measured at three initial probe heights.
Fig. 9(a) shows the signals of the probe when released into the water
from 1 cm above the water surface. The theoretical falling velocity at
the contact of the probe with water surface changes between 44.3 cm/s
(when the first needle touches the surface) and 48.1 cm/s (when the
second needle touches the surface). The theoretic delay time between
these two impacts is 3.9 ms (v = gt = 2ghfalling ). This is the time when
the first and the second needles with 1.8 mm spacing in between, dip
into the water. Fig. 9(a) demonstrates that the delay time detected by
the probe is 4.2 ms (8% error) which indicates that the dual-tip electro
resistivity probe can measure the velocity of phase change properly.
This test was repeated for three impact velocities (three initial heights)
and a comparison between theoretical velocities and those measured by
the probe are shown in Fig. 9(b). The error could be attributed to the
N. Jamshidi, N. Mostoufi Experimental Thermal and Fluid Science 94 (2018) 34–48

Fig. 9. (a) Signals of the probe needles at the moment the probe dip from air into water of 1 cm height. (b) Comparison between calculated velocity and the measured velocity by probe.
(error bars show relative error percentage).

measurement of the needle spacing and/or measuring the initial height

of the probe. This figure suggest that the probe is an appropriate device
to detect the velocity with an acceptable precision. Error bars in
Fig. 9(b) are relative error of the measured veloity.
To examine the accuracy of measuring chord length by the probe,
the measured chord length was compared with the average bubble
diameter achieved by the digital images of the bubbles. For this pur-
pose, photos of the column containing water at three superficial gas
velocities up to 2 cm/s were taken when the flow regime was homo-
geneous to allow visualization of discrete bubbles. In each case, more
than 100 bubbles were detected by image processing for estimating the
average bubble diameter. The steps of image processing and a sample
picture at the superficial gas velocity of 1 cm/s are shown in Fig. 10.
Calculated average bubble diameter in this case is 2.63 mm. The
average bubble diameter at three superficial gas velocities are reported
in Fig. 11. This figure illustrates that the chord length measurements
are very close to the bubble size evaluated by image processing. Fig. 11
shows that the chord length determined by the probe is less than the Fig. 11. Comparison between average bubble diameter estimated by image analysis and
bubble size calculated by signal processing since many bubbles might average chord length estimated by probe (error bars show relative error).
have reached the probe from the corner giving a chord length smaller
than the bubble diameter. According to [51], bubble diameter is 33% dynamics has been studied extensively. Most researchers have reported
greater than the measured chord length. Therefore, it can be concluded the descending trend of gas holdup versus increment of liquid viscosity.
that the probe measures the chord length with an appropriate accuracy. This trend can be explained by the change of the size of bubbles at
various operating conditions. The decrease of gas holdup with in-
creasing the liquid viscosity can be attributed to the higher rate of
5.2. Effect of liquid apparent viscosity
bubble coalescence in viscous solutions as well as born of larger bubbles
form the distributor. Energy dissipation rate of turbulence eddies is
One of the main hydrodynamic parameters in a bubble column is
directly correlated to the liquid viscosity. When the viscosity increases,
gas holdup. This parameter is a function of bubble dynamics, column
the dissipation becomes faster and the energy of eddies become lower.
geometry, operational conditions and physical properties of phases
Thus, the less energetic eddies with equal or smaller size of the bubbles
(gas/liquid). Effect of liquid viscosity on gas holdup and bubble

Fig. 10. A sample photo of bubbles in the

column at ugs = 1 cm/s. (a) raw image, (b)
threshold image, (c) output of particle ana-
lysis tool.

42
N. Jamshidi, N. Mostoufi
contribute to less effective interactions with bubbles leading to
breakage of the bubbles [52]. In a non-Newtonian liquid, as apparent
viscosity of the solution increases, the mean shear rate of the mixture
decreases resulting in decreasing the rate of bubble breakup (Eq. (2)).
In this situation, bubble break up diminishes and coalescence dom-
inates, thus, large bubbles can be found in the column. Also, the gas
holdup decreases with increasing the apparent viscosity.
Images of bubbles taken in different CMC solutions (changing the
viscosity) can be seen in Fig. 12. It can be seen in all cases that larger
bubbles appear at the center of the column as the liquid viscosity in-
creases. Also, distribution of bubbles changes from single class bubble
size in less viscous solutions (water and 0.05 wt%) to two class bubble
size in more viscous solutions. At very high viscous solutions, even slugs
were formed.
Time averaged local gas holdup as a function of gas velocity for
various CMC concentrations (various liquid viscosities) are depicted in
Fig. 13(a). It can be seen in this figure that the gas holdup increases by
increasing the liquid viscosity up to 4.1 cP (0.2 wt% CMC solution).
This trend is observed because the bubbles grow larger when increasing
the viscosity of the liquid, as discussed in the previous paragraph.
However, at very high viscosities (21 cP for 0.6 wt% CMC solution), the
gas holdup has increased to even greater than that of water. This in-
creasing trend of gas holdup is due to the change in the flow regime to
slugging which was visually observed in the experiments (see Fig. 12).
In fact, in small diameter columns with increasing the liquid viscosity,
the wall effect induces formation of slugs [38,53,54]. Slugs move
slower in the column due to the wall effect, thus, the gas holdup in-
creases in the slugging regime compared to homogeneous and hetero-
geneous regimes.
Motion of bubbles inside the bubble column induces shear to the
liquid in its vicinity. Shear rate affects the bubble frequency [55] since
it is correlated directly to destroying stresses of bubble breakup rate
[56]. Thus, bubble frequency can be counted as one of the most im-
portant parameters affected by breakup phenomena inside the bubbly
flow. The bubble frequency in various solutions at different superficial
gas velocities are shown in Fig. 13(b). Bubble frequency increases when
the superficial gas velocity is increased due to increasing the amount of
gas (i.e., number of bubbles) in the column as well as higher bubble rise
velocity due to formation of larger bubbles. Also, it can be seen in
Fig. 13(b) that the bubble frequency decreases as the liquid viscosity
increases. Increasing the liquid viscosity improves the bubble coales-
cence as explained before. In this situation number of bubbles decreases
as they merge and become larger this results in decrease in the bubble
frequency.
Mean and distribution of bubble rise velocity as a function of su-
perficial gas velocity in various solutions is shown in Fig. 14. Fig. 14(a)
illustrates that the average bubble rise velocity increases by increasing
the superficial gas velocity. Increasing the superficial gas velocity
promotes the turbulent axial eddy diffusion while the radial diffusivity
is hampered by walls of the small diameter column [57]. Thus, proper
mixing and recirculation of the liquid results in increasing the bubble
rise velocity with increasing superficial gas velocity. Increasing the
superficial gas velocity also increases both number and size of bubbles
in the column [57]. Formation of more bubbles leads to greater
Fig. 12. Bubbles photography captured in the middle of the column at ugs = 3 cm/s superficial gas velocity: (a) Water (1 cP); (b) 0.05 wt% CMC solution (1.42 cP); (c) 0.1 wt% CMC
solution (4.1 cP); (d) 0.2 wt% CMC solution (7 cP); (e) 0.6 wt% CMC solution (21 cP).
43
Experimental Thermal and Fluid Science 94 (2018) 34–48
collision rates between them [51]. Higher collision rates between
bubbles result in better circulation of the liquid and also in the for-
mation of larger bubbles. Fig. 14(a) also demonstrates that the bubble
rise velocity increases as the liquid viscosity is increased to 7 cP (0.2 wt
% CMC solution) due to formation of large, fast rising bubbles. This
trend is in agreement with what have been reported in the literature for
viscous solutions [21]. With increasing the liquid viscosity to 21 cP
(0.6 wt% CMC solution), the bubble rise velocity significantly decreases
since the wall effect hinders movement of slugs. This descending trend
of the bubble rise velocity in the slugging regime was also observed by
other researchers [53,54].
Fig. 14((b)–(d)) illustrate bubble rise velocity distributions at su-
perficial gas velocity of 2, 5 and 6.5 cm/s for various solutions studied
in this work. It can be seen in these figures that shape of the bubble rise
velocity distribution alters with increasing the liquid viscosity (or CMC
concentration). The distribution become wider, the peak height be-
comes lower and bimodal distributions appear at high liquid viscosity
as the gas superficial velocity increases. The distributions appear to be
Gaussian at 2 cm/s gas supervisal velocity (Fig. 14(b)), which shows
that no large bubbles exist in this condition. However, as shown in
Fig. 14(c), increasing the gas superficial velocity to 5 cm/s results in
developing a bimodal distribution (viscosity of 7 cP viscosity, 0.2 wt%
CMC solution). At the gas superficial velocity of 6.5 cm/s (Fig. 14(d)),
the bimodal distribution can be seen at even lower viscosities (0.05, 0.1
and 0.2 wt% CMC solutions). As discussed before, by increasing the
viscosity of solution, bubbles become larger which results in increase in
the bubble rising velocity. Consequently, frequency of bubble impacts
increase producing bubbles in a wide range of size. High viscosity of
solution expedites bubble coalescence and formation of large bubbles
after their collisions. Thus, the bubble rise velocity distribution be-
comes broader as larger bubbles appear in the solution. Therefore, two
classes of bubbles can be observed in the bubble rise velocity prob-
ability distribution with increase of viscosity which makes the dis-
tribution bimodal. The hydrodynamic regime becomes closer to the
heterogeneous regime when size/velocity of bubbles increases. In other
words, increasing the liquid viscosity makes the homogeneous flow
regime instable. It can be seen in Fig. 14 that the velocity distribution of
the viscose solutions become bimodal at lower velocities. The trend is
different than what described above in the 21 cP viscosity solution
(0.6 wt% CMC) due to formation of slugs in this case. Frequency of slugs
is low with a sharper peak. The range of observed bubble chord length
is narrower (due to existence of slugs) when the bubble column oper-
ates in the slugging regime. Accordingly, the distribution of rise velo-
city of slugs is narrow.
Experimental data of the mean bubble chord length as a function of
superficial gas velocity and the bubble chord length distributions at
three different superficial gas velocities are shown in Fig. 15. Fig. 15(a)
reveals that the growth of bubbles is faster in more viscous solutions
since bubble coalescence enhances by increasing the viscosity. How-
ever, it can be seen in this figure that at low gas velocities and low
viscosities, the bubble size remains almost constant. This can be at-
tributed to the effect of surface tension of the liquid on the bubble size
at low gas velocity where effect of viscosity (drag force) is not dominant
on the motion of bubbles. Many researchers have pointed to the direct
N. Jamshidi, N. Mostoufi Experimental Thermal and Fluid Science 94 (2018) 34–48

Fig. 13. Hydrodynamic parameters measured in the middle height of the column and in the central radial position: (a) Gas holdup; (b) bubble frequency.

effect of surface tension on the bubble size [58,59], as is a well-known
theoretical fact. The surface tension of the solution used in this work
decrease significantly with increasing the CMC concentration (see
Table 2). The bubble size is determined based on the mutual effect of
surface tension and viscous drag force. Based on the data given in
Fig. 15(a) and properties given in Table 2, it can be concluded that the
viscous drag force is dominant compared to the surface tension force.
As the viscosity of solution increases and larger bubbles appear in the
column, drag and bubble inertial forces increase accordingly while the
surface tension decreases (see Table 2). It has been reported that sur-
factant effect causes formation of smaller bubbles and stabilizing the
homogeneous flow regime [60]. On the contrary, it can be readily seen
from the data presented in this work that the bubbles size is larger in
more viscous solutions with lower surface tension. Thus, the results of
this work confirm the negligible effect of surface tension on the bubble
size at high viscosities or velocities.
Based on the experimental data obtained in this work, a correlation
can be proposed for evaluating the bubble chord length from proper
dimensionless numbers. The effective dimensionless numbers in this
problem are Galileo number, Bond number and Froude number which
incorporate all main forces presenting in this process (viscous drag,
inertia, buoyancy, surface tension) [48,61,62]. Therefore, the following
power law equation can be proposed for correlating the mean bubble
chord length to liquid properties and operational conditions:

Fig. 14. Parameters of bubble dynamics measured in the middle height of the column and in the central radial position: (a) Mean bubble rise velocity; (b) velocity distribution at
ugs = 2 cm/s; (c) velocity distribution at ugs = 5 cm/s; (d) velocity distribution at ugs = 6.5 cm/s.

44
N. Jamshidi, N. Mostoufi Experimental Thermal and Fluid Science 94 (2018) 34–48

Fig. 15. Parameters of bubble dynamics measured in the middle height of the column and in the central radial position: (a) Mean bubble chord length; (b) chord length distribution at
ugs = 2 cm/s; (c) chord length distribution at ugs = 5 cm/s; (d) chord length distribution at ugs = 6.5 cm/s.

dB VB
= CFr aGabBoc = εf (ε ) = 1.49ε −0.9
dc (10) ugs (15)

where the dimensionless numbers are defined as follows: The probability density distributions of bubble chord length at gas
ugs superficial velocities of 2, 5 and 6.5 cm/s for all tested solutions are
Fr =
gdc shown in Fig. 15((b)–(d)). As illustrated in this figure, with an increase
(11)
in the viscosity, the distribution becomes wider and the peak becomes
gdc3 less intense. In viscosities less than 4.1 cP (0.1 wt% CMC), the dis-
Ga = tribution is close to a log-normal curve at gas superficial velocities less
νL2 (12)
than 5 cm/s. In these solutions, most of the bubbles are small, thus, the
distribution is limited to a narrow range with a sharp peak. Increasing
ρl gdc2
Bo = the viscosity alters the shape of distribution addressing the presence of
δ (13) slugs and a wide variety of bubble sizes. In fact, increase in the viscosity
Using the experimental data of this work shown in Fig. 15(a), as decreases the energy of eddies such that they cannot generate enough
well as the experimental data of Deng et al. [21] and Esmaeili et al. shear stress to break bubbles. Therefore, bubbles remain intact after
[49], constants of Eq. (10) can be calculated by the least square method coalescence. Thus, large bubbles continue to rise faster through the
as follows: column. This is the reason for observing a bimodal and even multi-
modal bubble rise velocity distribution at higher viscosities (see
dB′ Fig. 14). It can be seen that the chord length distribution becomes
dc
= 3.85 × 102Fr 0.7Ga−0.2Bo−0.3 0.00139 ⩽ μ ⩽ 0.089
considerably wide with increasing the superficial gas velocity. The
0.5 ⩽ ugs ⩽ 14 (14) homogeneous regime transition occurs in higher gas superficial velo-
This correlation is well fitted to the experimental data of the present cities for less viscous samples (0.1 wt% CMC at 5 cm/s (Fig. 15 (c)) and
work and those proposed by Deng et al. [21] and Esmaeili et al. [49] 0.05 wt% CMC at 6.5 cm/s (Fig. 15(d)). For water, which has the lowest
with a correlation coefficient of 0.72. The apparent viscosity used in the viscosity, the chord length distribution remains almost log-normal, al-
correlation was calculated from Eq. (2). It is worth mentioning that the though the peak height decreases slightly as the gas superficial velocity
data obtained in the slugging regime (CMC 0.6 wt%) were not taken is increased. Considering the results reported for bubble rise velocity
into account when evaluating constants of Eq. (14). Furthermore, an- and bubble chord length distributions, the homogeneous regime be-
other Richardson-Zaki type correlation [27] for bubble rise velocity was comes instable as the liquid viscosity is increased.
fitted with a correlation coefficient of 0.77 using the same experimental
data:

45
N. Jamshidi, N. Mostoufi
5.3. Axial distribution of bubble dynamics
The experimental data of the mean bubble chord length, mean
bubble rise velocity and gas holdup as a function of superficial gas
velocity at three heights of column (0.1 wt% CMC) is presented in
Fig. 16((a)–(c)). The data were collected in three axial regions (20 cm
(bottom), 75 cm (middle) and 100 cm (top) above the distributor) as the
superficial gas velocity varied in the range of 0.5–7 cm/s. Fig. 16(c)
demonstrates that as moving up along the column, the gas holdup in-
creases and reaches the peak in the middle, while it slightly decreases
again along the column in the upper zone. This can be explained by
variation of trend of bubble size and bubble velocity which have sig-
nificant influence on the gas holdup. The mean bubble chord length and
mean bubble rise velocity are illustrated in Fig. 16((a) and (b)), re-
spectively. It can be seen that the bubble rise velocity decreases along
the column which is due to shear thinning and viscous behavior of the
liquid. Fig. 17 shows the axial variation of bubble rise velocity in so-
lution with different CMC concentrations. As the shear thinning beha-
vior of the liquid dominates, reduction in the bubble rise velocity along
the column can be better observed. The distributor zone in the bubble
column has the highest local shear rate. On the other hand, the bubble
frequency is reported to simulate the shear rate liquid experiences [18].
As moving up along the column, the local bubble frequency decreases
as a result of coalescence. The shear rate then decreases by moving up
the column. Thus, the apparent liquid viscosity, and consequently the
viscous drag force, increase along the column. This produces larger
bubbles (due to faster coalescence, as discussed earlier) which move
slower (due to greater drag force) resulting in an increasing trend of gas
holdup when moving away from the distributor Fig. 16(a) shows that
coalescence of bubbles continues when moving up the column and
bubbles continue to become larger between middle and top of the
column. However, according to Fig. 16(b), a substantial decrease in the
bubble rise velocity can be observed in this zone. As a result, the gas
Fig. 16. Hydrodynamic parameters measured at three axial locations along the column in the central radial position for 0.1 wt% CMC solution: (a) Mean bubble chord length; (b) mean
bubble rise velocity; (c) gas holdup.
46
Experimental Thermal and Fluid Science 94 (2018) 34–48
holdup remains almost unchanged in this region. Decreasing the bubble
rise velocity is a consequence of decrease in the local shear rate as
bubbles move away from the distributor. The shear thinning behavior
of the liquid increases the local viscosity along the column that can
impose a greater drag force on bubbles and restrain their movement.
6. Conclusions
Bubble dynamics in a bubble column was evaluated extensively
using a dual-tip resistivity probe. Four different concentrations of CMC
solutions were used to investigate the effect of rheological behavior of
the shear thinning solutions on the bubble dynamics. The measure-
ments were performed in three locations along the column height.
Furthermore, the effects of superficial gas velocity on hydrodynamic
parameters, including bubble chord length and bubble rise velocity,
were investigated. An algorithm was developed for signal processing in
order to improve detection of bubbles. In the presented algorithm, those
bubbles deviated from vertical direction along the probe, were rejected
in the signal processing step and were neglected in evaluating bubble
velocity and bubble chord length. It was shown that increasing the li-
quid viscosity widens the corresponding distribution of bubble rise
velocity and bubble chord length. Higher viscosity destabilize the
homogeneous flow regime. The gas holdup and bubble rise velocity
showed highly inverted trends in the liquid viscosity of 21 cP, at which
the column was operating in the slug flow regime. The descending trend
of the bubble rise velocity, as moving away from the distributor, besides
acceleration of coalescence rate across the course to the top of the
column brings on an optimum condition in the middle of the column
which maximizes the gas holdup. The signs of regime transition as the
superficial gas velocity increased was followed in the alteration of the
shape of the bubble rise velocity distribution and the bubble chord
length distribution. The Gaussian distribution of bubble velocity in the
homogeneous regime changed to a multi modal distribution after the
N. Jamshidi, N. Mostoufi
Fig. 17. Mean bubble rise velocity measured at three axial locations along the column in the central radial position for various solutions: (a) 0.05 wt% CMC; (b) 0.1 wt% CMC; (c) 0.2 wt%
CMC; (d) 0.6 wt% CMC.
regime transition, whereas the log-normal shape of the chord length
distribution showed longer tails and shorter peaks. Growth of the
bubble chord length was much greater than the bubble velocity that can
be attributed to the dominance of the drag force.
References
[1] H. Debellefontaine, J.N. Foussard, Wet air oxidation for the treatment of industrial
wastes. Chemical aspects, reactor design and industrial applications in Europe,
Waste Manage. 20 (2000) 15–25.
[2] A. Khalili, M.R. Mehrnia, N. Mostoufi, M. Sarrafzadeh, Flow characteristics in an
airlift membrane bioreactor, Chem. Prod. Process Model. 4 (2009).
[3] R. Guettel, U. Kunz, T. Turek, Reactors for Fischer-Tropsch synthesis, Chem. Eng.
Technol. 31 (2008) 746–754.
[4] M. Hillestad, Modeling the Fischer-Tropsch product distribution and model im-
plementation, Chem. Prod. Process Model. 10 (2015) 147–159.
[5] S. Kumar, A. Mandal, Investigation on stabilization of CO 2 foam by ionic and
nonionic surfactants in presence of different additives for application in enhanced
oil recovery, Appl. Surf. Sci. 420 (2017) 9–20.
[6] M. Shahid, C. Fan, R.M. Pashley, Insight into the bubble column evaporator and its
applications, Int. Rev. Phys. Chem. 35 (2016) 143–185.
[7] J.L.S. Sonego, D.A. Lemos, C.E.M. Pinto, A.J.G. Cruz, A.C. Badino, Extractive fed-
batch ethanol fermentation with co2 stripping in a bubble column bioreactor:
Experiment and modeling, Energy Fuels 30 (2016) 748–757.
[8] E. León-Becerril, R. Maya-Yescas, Axial variation of mass transfer volumetric
coefficients in bubble column bioreactors, Chem. Prod. Process Model. 5 (2010).
[9] R. Zenit, J.J. Feng, Hydrodynamic interactions among bubbles, drops, and particles
in non-newtonian liquids, Ann. Rev. Fluid Mech. 50 (2017).
[10] R.P. Chhabra, J.F. Richardson, Non-Newtonian Flow and Applied Rheology:
Engineering Applications, Butterworth-Heinemann, 2011.
[11] A.A. Youssef, M.E. Hamed, J.T. Grimes, M.H. Al-Dahhan, M.P. Duduković,
Hydrodynamics of pilot-scale bubble columns: effect of internals, Ind. Eng. Chem.
Res. 52 (2013) 43–55.
[12] J.A.S. Pérez, E.M.R. Porcel, J.L.C. López, J.M.F. Sevilla, Y. Chisti, Shear rate in
47
Experimental Thermal and Fluid Science 94 (2018) 34–48
stirred tank and bubble column bioreactors, Chem. Eng. J. 124 (2006) 1–5.
[13] M. Kagumba, M.H. Al-Dahhan, Impact of internals size and configuration on bubble
dynamics in bubble columns for alternative clean fuels production, Ind. Eng. Chem.
Res. 54 (2015) 1359–1372.
[14] A. Akhtar, V. Pareek, M. Tadé, CFD simulations for continuous flow of bubbles
through gas-liquid columns: Application of VOF method, Chem. Prod. Process
Model. 2 (2007).
[15] S.M. Son, S.H. Kang, U.Y. Kim, Y. Kang, S.D. Kim, Bubble properties in three-phase
inverse fluidized beds with viscous liquid medium, Chem. Eng. Process. Process
Intensif. 46 (2007) 736–741.
[16] W. Liu, N.N. Clark, A.I. Karamavruç, Relationship between bubble size distributions
and chord-length distribution in heterogeneously bubbling systems, Chem. Eng. Sci.
53 (1998) 1267–1276.
[17] D. Pjontek, V. Parisien, A. Macchi, Bubble characteristics measured using a
monofibre optical probe in a bubble column and freeboard region under high gas
holdup conditions, Chem. Eng. Sci. 111 (2014) 153–169.
[18] A. Esmaeili, C. Guy, J. Chaouki, The effects of liquid phase rheology on the hy-
drodynamics of a gas–liquid bubble column reactor, Chem. Eng. Sci. 129 (2015)
193–207.
[19] Y. Mizushima, A. Sakamoto, T. Saito, Measurement technique of bubble velocity
and diameter in a bubble column via single-tip optical-fiber probing with judgment
of the pierced position and angle, Chem. Eng. Sci. 100 (2013) 98–104.
[20] R. Kaji, D. Zhao, P. Licence, B. Azzopardi, Studies of the interaction of ionic liquid
and gas in a small-diameter bubble column, Ind. Eng. Chem. Res. 48 (2009)
7938–7944.
[21] Z. Deng, T. Wang, N. Zhang, Z. Wang, Gas holdup, bubble behavior and mass
transfer in a 5m high internal-loop airlift reactor with non-Newtonian fluid, Chem.
Eng. J. 160 (2010) 729–737.
[22] Q. Wu, X. Wang, T. Wang, M. Han, Z. Sha, J. Wang, Effect of liquid viscosity on
hydrodynamics and bubble behaviour of an external-loop airlift reactor, Can. J.
Chem. Eng. 91 (2013) 1957–1963.
[23] M. Shiea, N. Mostoufi, R. Sotudeh-Gharebagh, Comprehensive study of regime
transitions throughout a bubble column using resistivity probe, Chem. Eng. Sci. 100
(2013) 15–22.
[24] Y.F. Han, N.D. Jin, Z.Q. Yin, Y.Y. Ren, Y. Gu, Measurement of oil bubble size dis-
tribution in oil-in-water emulsions using a distributed dual-sensor probe array, Exp.
N. Jamshidi, N. Mostoufi
Therm Fluid Sci. 86 (2017) 204–223.
[25] A.M. Lali, A.S. Khare, J.B. Joshi, K.D.P. Nigam, Behaviour of solid particles in
viscous non-Newtonian solutions: settling velocity, wall effects and bed expansion
in solid-liquid fluidized beds, Powder Technol. 57 (1989) 39–50.
[26] A. Mota, A. Ferreira, A.A. Vicente, P. Sechet, J.M.F. Martins, J.A. Teixeira,
A. Cartellier, Customization of an optical probe device and validation of a signal
processing procedure to study gas–liquid–solid flows. Application to a three-phase
internal-loop gas-lift Bioreactor, Chem. Eng. Sci. 138 (2015) 814–826.
[27] N. Hooshyar, P.J. Hamersma, R.F. Mudde, J.R. van Ommen, Gas fraction and
bubble dynamics in structured slurry bubble columns, Ind. Eng. Chem. Res. 49
(2010) 10689–10697.
[28] J.M. Burgess, P.H. Calderbank, The measurement of bubble parameters in two-
phase dispersions—I, Chem. Eng. Sci. 30 (1975) 743–750.
[29] H. Chanson, Air-water flow measurements with intrusive, phase-detection probes:
Can we improve their interpretation? J. Hydraul. Eng. 128 (2002) 252–255.
[30] K. Sanaullah, S.H. Zaidi, J.H. Hills, A study of bubbly flow using resistivity probes in
a novel configuration, Chem. Eng. J. 83 (2001) 45–53.
[31] C. Boyer, A.M. Duquenne, G. Wild, Measuring techniques in gas–liquid and ga-
s–liquid–solid reactors, Chem. Eng. Sci. 57 (2002) 3185–3215.
[32] F. Magaud, M. Souhar, G. Wild, N. Boisson, Experimental study of bubble column
hydrodynamics, Chem. Eng. Sci. 56 (2001) 4597–4607.
[33] F.A. Hamad, S. He, Evaluation of hot-film, dual optical and Pitot tube probes for
liquid–liquid two-phase flow measurements, Flow Meas. Instrum. 21 (2010)
302–311.
[34] M.A.S. Richter, H.M. Prasser, R. Hampel, Approach towards spatial phase re-
construction in transient bubbly flow using a wire-mesh sensor, Int. J. Heat Mass
Transf. 45 (2002) 1063–1075.
[35] A. Serizawa, I. Kataoka, I. Michiyoshi, Turbulence structure of air-water bubbly
flow—I. Measuring techniques, Int. J. Multiphase Flow 2 (1975) 221–233.
[36] L.-S. Fan, G. Yang, D. Lee, K. Tsuchiya, X. Luo, Some aspects of high-pressure
phenomena of bubbles in liquids and liquid–solid suspensions, Chem. Eng. Sci. 54
(1999) 4681–4709.
[37] G. Yang, B. Du, L. Fan, Bubble formation and dynamics in gas–liquid–solid fluidi-
zation—a review, Chem. Eng. Sci. 62 (2007) 2–27.
[38] O.T. Kajero, B.J. Azzopardi, L. Abdulkareem, Experimental investigation of the
effect of liquid viscosity on slug flow in small diameter bubble column, EPJ Web of
Conferences 25 (2012).
[39] G. Olivieri, M. Elena Russo, M. Simeone, A. Marzocchella, P. Salatino, Effects of
viscosity and relaxation time on the hydrodynamics of gas–liquid systems, Chem.
Eng. Sci. 66 (2011) 3392–3399.
[40] M.C. Ruzicka, J. Drahoš, P.C. Mena, J.A. Teixeira, Effect of viscosity on homo-
geneous–heterogeneous flow regime transition in bubble columns, Chem. Eng. J. 96
(2003) 15–22.
[41] C. Xing, T. Wang, J. Wang, Experimental study and numerical simulation with a
coupled CFD–PBM model of the effect of liquid viscosity in a bubble column, Chem.
Eng. Sci. 95 (2013) 313–322.
[42] V. Shilapuram, P.S.T. Sai, Axial solids holdup distribution in a liquid–solid circu-
lating fluidized bed: effect of the liquid distributor, method of operation, and
viscosity of the fluidizing media, Ind. Eng. Chem. Res. 51 (2012) 16242–16250.
[43] A. Anastasiou, A. Passos, A. Mouza, Bubble columns with fine pore sparger and non-
Newtonian liquid phase: Prediction of gas holdup, Chem. Eng. Sci. 98 (2013)
48
Experimental Thermal and Fluid Science 94 (2018) 34–48
331–338.
[44] L. Böhm, M. Kraume, Fluid dynamics of bubble swarms rising in Newtonian and
non-Newtonian liquids in flat sheet membrane systems, J. Membr. Sci. 475 (2015)
533–544.
[45] J. Philip, J. Proctor, K. Niranjan, J. Davidson, Gas hold-up and liquid circulation in
internal loop reactors containing highly viscous Newtonian and non-Newtonian
liquids, Chem. Eng. Sci. 45 (1990) 651–664.
[46] Y. Kawase, S. Umeno, T. Kumagai, The prediction of gas hold-up in bubble column
reactors: Newtonian and non-newtonian fluids, Chem. Eng. J. 50 (1992) 1–7.
[47] J. Zahradník, M. Fialová, M. Růžička, J. Drahoš, F. Kaštánek, N.H. Thomas, Duality
of the gas-liquid flow regimes in bubble column reactors, Chem. Eng. Sci. 52 (1997)
3811–3826.
[48] X. Zhang, H. Dong, Y. Huang, C. Li, X. Zhang, Experimental study on gas holdup and
bubble behavior in carbon capture systems with ionic liquid, Chem. Eng. J. 209
(2012) 607–615.
[49] A. Esmaeili, C. Guy, J. Chaouki, Local hydrodynamic parameters of bubble column
reactors operating with non-Newtonian liquids: Experiments and models develop-
ment, AIChE J. 62 (2015) 1382–1396.
[50] N. Jamshidi, N. Mostoufi, Measurement of bubble size distribution in activated
sludge bubble column bioreactor, Biochem. Eng. J. 125 (2017) 212–220.
[51] D.H. Lim, J.H. Park, Y. Kang, K.W. Jun, Structure of bubble holdups in a viscous
slurry bubble column with low surface tension media, Fuel Process. Technol. 108
(2013) 2–7.
[52] M.J. Prince, H.W. Blanch, Bubble coalescence and break-up in air-sparged bubble
columns, AIChE J. 36 (1990) 1485–1499.
[53] S.P. Godbole, M.F. Honath, Y.T. Shah, Holdup structure in highly viscous
Newtonian and non-Newtonian liquids in bubble columns, Chem. Eng. Commun. 16
(1982) 119–134.
[54] M. Götz, J. Lefebvre, F. Mörs, R. Reimert, F. Graf, T. Kolb, Hydrodynamics of or-
ganic and ionic liquids in a slurry bubble column reactor operated at elevated
temperatures, Chem. Eng. J. 286 (2016) 348–360.
[55] A. Esmaeili, S. Farag, C. Guy, J. Chaouki, Effect of elevated pressure on the hy-
drodynamic aspects of a pilot-scale bubble column reactor operating with non-
Newtonian liquids, Chem. Eng. J. 288 (2016) 377–389.
[56] Y. Liao, R. Rzehak, D. Lucas, E. Krepper, Baseline closure model for dispersed
bubbly flow: Bubble coalescence and breakup, Chem. Eng. Sci. 122 (2015)
336–349.
[57] M. Hamed, Hydrodynamics, Mixing, and Mass Transfer in Bubble Columns with
Internals, Washington University, USA, 2012.
[58] E.E. Zukoski, Influence of viscosity, surface tension, and inclination angle on mo-
tion of long bubbles in closed tubes, J. Fluid Mech. 25 (1966) 821–837.
[59] S.d.P. Eugénie, D. Fabrice, C. Gérard, M. Samir, Effect of bulk viscosity and surface
tension kinetics on structure of foam generated at the pilot scale, Food
Hydrocolloids 34 (2014) 104–111.
[60] X. Li, C. Yang, S. Yang, G. Li, Fiber-optical sensors: basics and applications in
multiphase reactors, Sensors 12 (2012) 12519–12544.
[61] K. Akita, F. Yoshida, Bubble size, interfacial area, and liquid-phase mass transfer
coefficient in bubble columns, Ind. Eng. Chem. Res. 13 (1974) 84–91.
[62] P.M. Wilkinson, H. Haringa, L.L. Van Dierendonck, Mass transfer and bubble size in
a bubble column under pressure, Chem. Eng. Sci. 49 (1994) 1417–1427.