Journal of the Taiwan Institute of Chemical Engineers 69 (2016) 14–24

Contents lists available at ScienceDirect

Journal of the Taiwan Institute of Chemical Engineers

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

Effect of ring sparger diameters on hydrodynamics in bubble column:

A numerical investigation
M. Pourtousi a,∗, P. Ganesan b, Shanti C. Sandaran c, J.N. Sahu d,e,f,∗∗
a
Bernal Institute, Synthesis and Solid State Pharmaceutical Centre (SSPC) & School of Engineering, University of Limerick, Ireland
b
Department of Mechanical Engineering, Faculty of Engineering, University of Malaya, 50603 Kuala Lumpur, Malaysia
c
University Technology Malaysia, Skudai, Johor, Malaysia
d
Department for Management of Science and Technology Development, Ton Duc Thang University, Ho Chi Minh City, Vietnam
e
Faculty of Applied Sciences, Ton Duc Thang University, Ho Chi Minh City, Vietnam
f
Petroleum and Chemical Engineering, Faculty of Engineering, Universiti Teknologi Brunei, Tungku Gadong, P.O. Box 2909, Brunei Darussalam

a r t i c l e i n f o a b s t r a c t

Article history: This paper applies the Eulerian–Eulerian approach to study the effect of ring sparger diameter on the
Received 26 September 2015 gas and liquid dynamics in a cylindrical bubble column reactor. The distribution of the gas hold-up and
Revised 20 September 2016
liquid flow pattern using different ring sparger diameters (i.e., 0.07, 0.14 and 0.20 m) are investigated. In
Accepted 7 October 2016
addition, the influence of different bubble sizes and interfacial force models on the accuracy of numerical
Available online 28 October 2016
results are examined. The results show that the size of ring sparger affects the amount of gas in the
Keywords: column and the position of the liquid phase recirculation (upward and downward liquid flow direction).
CFD Furthermore, the appropriate selection of bubble size, and interfacial force models results in accurate
Bubble column liquid velocity and gas hold-up in the bubble column reactor.
Multiphase flow © 2016 Taiwan Institute of Chemical Engineers. Published by Elsevier B.V. All rights reserved.
Eulerian–Eulerian model
Ring sparger

1. Introduction ciency [4,5]. Therefore detailed understanding of the macroscopic

Bubble columns are used as multiphase reactors to produce
different types of physical and chemical multiphase reactions
in several industries such as chemical, petrochemical, biochem-
ical, wastewater treatment and metallurgical industries [1–14].
Due to simple construction and shape, they contain low main-
tenance, low operating costs and simple operating (cleaning re-
actor and running experiment) in industries. They also provide
high rate of heat and mass transfer, as well as durability of cat-
alysts [15–17]. They are usually manufactured in cylindrical shape
and fitted with a gas distributor (i.e., sparger) at the bottom. The
sparger produces bubbles inside the bubble column reactor, hav-
ing either a liquid phase (water, etc.,) or a liquid–solid suspen-
sion [18,19]. Production of small bubbles with spherical shape
(having higher interfacial area) without coalescence, particularly
near the sparger region results in uniform gas hold-up profile
and homogeneous flow regime and eventually higher reactor effi-
∗
Corresponding author. Bernal Institute, Synthesis and Solid State Pharmaceutical
Centre (SSPC) & School of Engineering, University of Limerick, Ireland.
∗∗
Corresponding author at: Department for Management of Science and Technol-
ogy Development, Ton Duc Thang University, Ho Chi Minh City, Vietnam.
E-mail addresses: mohammad.pourtousi@ul.ie (M. Pourtousi),
jay_sahu@yahoo.co.in, jayanarayansahu@tdt.edu.vn, jnsahu@um.edu.my (J.N. Sahu).
(gas and liquid flow pattern) and microscopic (bubble coales-
cence and break-up) interaction between phases, will greatly as-
sists efficient design of reactors and spargers for several industrial
applications [19].
There are two main approaches for the modeling of multiphase
flows in bubble column that account the interactions between
phases; They are Eulerian–Eulerian and the Eulerian–Lagrangian
approaches [3]. The former approach is popular and also a suitable
option for an industrial bubble column reactor since the volume
fraction of dispersed phase is often not small and is distributed
using a sparger rather than a single nozzle. In addition, the for-
mer approach requires relatively lesser computational effort than
that for Eulerian–Lagrangian approach. However, correct selections
of interphase forces (e.g., drag force, lift, turbulent dispersion and
virtual mass) and its models and turbulence models (e.g., standard
k–ε model, Reynolds Stress Model (RSM), Large Eddy Simulation
(LES)) are required for an accurate flow prediction in Eulerian–
Eulerian model [3,20,21]. Recently, Pourtousi, Sahu [3] reviewed
the influence of different interfacial forces and turbulence mod-
els on the accuracy of numerical results. They suggested appro-
priate interfacial forces and turbulence models at different oper-
ational conditions for correct modeling of bubble column reactors.
Due to complex behavior of multiphase flow in the bubble column
reactor, [Pourtousi et al. [3], Pourtousi et al. [4]] recommended

http://dx.doi.org/10.1016/j.jtice.2016.10.006
1876-1070/© 2016 Taiwan Institute of Chemical Engineers. Published by Elsevier B.V. All rights reserved.
 M. Pourtousi et al. / Journal of the Taiwan Institute of Chemical Engineers 69 (2016) 14–24 15

ity), temperature and pressure conditions and sparger specifica-

Nomenclature tions, changes the hydrodynamics in a bubble column.
Sparger designs have significant influence on the distribution of
CD drag force coefficient (dimensionless) the dispersed gas (gas hold-up), bubble size and bubble shape and
CTD turbulent dispersion coefficient (dimensionless) the distribution of liquid velocity [5,6,21,26]. Various types of gas
Cε 1 model parameter in turbulent dissipation energy sparger designs such as perforated plate [15,21], porous plate [28],
equation (dimensionless) single point/multipoint sparger [5,22], ring [2,4,27,29] are used in
Cε 2 model parameter in turbulent dissipation energy bubble columns. For example, Li et al. [26] studied the effect of
equation (dimensionless) number of spargers and their arrangements on the liquid velocity,
Cμ constant in k-ε model (dimensionless) gas hold-up, mixing characteristics and bubble size distribution in
Cμ, BI constant in bubble induced turbulence model (di- a cylindrical column. The gas holdup rises with the increase of the
mensionless) number of spargers. It was found that the arrangement of sparger
db bubble diameter (m) positions have influence on the uniformity of gas distribution and
d0 sparger hole diameter (m) liquid flow pattern. Dhotre et al. [22] studied the effect of a single
g(ρ −ρ )d 2 point and multipoint spargers on the distribution of the gas hold-
EO eotvos number (= L
σ
G B
) (dimensionless)
D diameter of the column (m) up in bubble column of different heights. For multipoint spargers,
g gravitational constant (m/s2 ) the profile of the gas hold-up is relatively flat near the sparger
G generation term (kg/m s2 ) region (at the bottom). In contrast, the single point sparger pro-
H height (m) duces a steep profile of gas hold up adjacent to the sparger region
k turbulent kinetic energy per unit mass (m2 /s2 ) but this profile becomes flatter with the increase of the vertical
MI total interfacial force acting between two phases distance from the bottom. The sparger orifice diameter and posi-
(N/m3 ) tion has an effect on the bubble distributions and sizes in bub-
MD drag force (N/m3 ) ble column [30]. Sal et al. [15] investigated the effect of the orifice
P pressure (N/m2 ) diameter of perforated plate spargers on gas holdup and regime
r radial distance (m) transition in bubble column. Orifice diameter is found to have a
R column radius (m) significant effect on the total gas hold-up in homogeneous flow
ReB reynolds number ( = dB VS /v) (dimensionless) regime but not in heterogeneous regime. Bhole et al. [28] com-
VG superficial gas velocity (m/s) pared the gas distribution from a porous plate sparger and from
VT terminal velocity (m2 /s) a perforated plate sparger. The former is found to produce smaller
bubbles, higher gas holdup and small liquid circulation in the col-
Greek Symbols umn.
ε turbulent energy dissipation rate per unit mass Although few studies have concentrated on the effect of
(m2 /s3 ) sparger design parameters on the bubble column hydrodynam-
∈ fractional phase hold-up (dimensionless) ics [5,6,15,22,23], there are still many aspects on sparger design
∈¯ average fractional phase hold-up (dimensionless) and configuration that requires further studies. Therefore, the main
μ molecular viscosity (Pa s) purpose of the present study is to examine the effect of ring
μBI bubble induced viscosity (Pa s) sparger diameter on the gas hold-up and liquid velocity in the
μeff effective viscosity (Pa s) cylindrical bubble column. Furthermore, effects of using differ-
ρ density (kg/m3 ) ent bubble sizes and the interfacial force models (such as, drag
μT turbulent viscosity (Pa s) and turbulent dispersion models) on the accuracy of Euler–Euler
σ surface tension (N/m) method are studied. In this study, the commercial CFD package of
σε prandtl number for turbulent energy dissipation ANSYS-CFX is used for all simulation cases.
rate (dimensionless)
σk prandtl number for turbulent kinetic energy (di- 2. Methodologies
mensionless)
τk shear stress of phase k (Pa) 2.1. Governing equations

Subscripts
The two-phase model based on the Eulerian–Eulerian approach
b bubble
is applied to simulate the gas and liquid interaction. Each phase
G gas phase
is treated as a continuum in the domain under consideration in
L liquid phase
this approach. The phases share this domain and interpenetrate as
they move within it. The Eulerian modelling framework is based
on ensemble-averaged mass and momentum transport equations
to considered the sensitivity study of different bubble sizes and for each phase and is written as following:
interfacial force models, as well as information in the literature to Continuity equation:
find appropriate CFD models for bubble columns. ∂
In addition to CFD modeling, many studies have focused on the ( ρ ∈ ) + ∇ ( ρ k∈k uk ) = 0 (1)
∂t k k
influence of operating conditions on the liquid and gas flow pat-
tern and detailed understanding of phase interactions [5,22–27]. Momentum transfer equation:
The gas and liquid phase flow pattern and the interaction between ∂   
phases in a bubble column are dependent on bubble column hy- ρk ∈k uk + ∇ ρk ∈k uk uk = −∇ (∈k τk ) − ∈k ∇ p + ∈k ρk g + MI,k
∂t
drodynamics parameters (i.e., gas hold-up and gas and liquid ve-
(2)
locity, etc.). The variation in phase interaction enhances the mix-
ing time, heat and mass transfer rate and this improves the over- The terms on the right hand side of Eq. (2) represents the
all efficiency of reactors. Changing of operational conditions, such stress, the pressure gradient, the gravity and the ensemble-
as bubble column dimensions, inlet velocity (superficial gas veloc- averaged momentum exchange between the phases due to
16 M. Pourtousi et al. / Journal of the Taiwan Institute of Chemical Engineers 69 (2016) 14–24

Table 1
Drag models used in present numerical study.

Author Model

CD = R24
eB
(1 + 0.15ReB 0.687 ), if ReB < 10 0 0
Schiller and Naumann [40]
CD = 0.44 i f ReB >10 0 0
4 gdB ρL −ρG
Grace et al [41] CD = 3 VT 2 ρL
Ishii and Zuber [42] CD = 2
3
EO 0.5

interface forces. The pressure is shared by the both phases. The
stress term of phase k is described as follows:
 2
 k2
τk = −μeff,k ∇ uk + (∇ uk )T − I (∇ uk )
3 ε
where μeff , k is the effective viscosity. The effective viscosity of the
liquid phase is composed of three contributions: the molecular vis-
cosity, the turbulent viscosity and an extra term due to bubble in-
duced turbulence as given below;
μeff,L = μL + μT,L + μBI,L
The model proposed by Sato and Sekoguchi [31] was used to
take account the turbulence due to the movement of bubbles,
which can be written as:
μBI,L = ρLCμ,BI εG dB |uG − uL |
with the model constant of Cμ,BI is set to 0.6 as recommended in
ref [31]. The calculation of the effective gas viscosity is based on
the effective liquid viscosity which is:
ρG
μeff,G = μ (6)
ρL eff,L
The total interfacial force acting between the two phases are
based on the interphase drag force and turbulent dispersion force,
which can be written as follows:
MI,L = −MI,G = MD,L + MT D,L
The effect of drag force is significantly higher than added mass,
lift and turbulent dispersion effects when it comes to predicting
the flow pattern inside bubble column [2,3,29,32–34]. The inter-
phase momentum transfer between the gas and liquid due to the
drag force is given by
3 CD
MD,L = − ∈G ρL |uG − uL |(uG − uL )
4 dB
where CD is the drag coefficient taking into account the charac-
teristic of the flow around bubble, and dB is the bubble diame-
ter. In homogeneous regime, bubbles have approximately a same
shape, diameter size and terminal velocity [4,29,35–37]. In present
study a constant drag coefficient of 0.44 is used which is recom-
mended in references [2,3,27,29,38,39]. In addition, the sensitivity
of the use of Schiller & Naumann [40] Grace et al. [41], Ishii & Zu-
ber [42] drag models are examined and CD of the respective mod-
els is given in Table 1. The turbulent dispersion force derived by
Lopez de Bertodano [43] is used, which is
MT D,L = −MT D,G = −CT D ρL k∇ ∈L
turbulent kinetic energy (k) and its energy dissipation rate (ε ) are
described as follows:
(3) μT,L = ρLCμ (10)
∂  μ 
(ρL ∈L k ) + ∇ (ρL ∈L uL k ) = −∇ ∈L eff,L ∇ k + ∈L (G − ρL ε )
∂t σk
(11)
(4)
∂
( ρ ∈ ε ) + ∇ ( ρL ∈ L u L ε )
∂ t L L 
μL,e f f ε
= −∇ ∈L ∇ε + ∈L (Cε1 G − Cε2 ρL ε ) (12)
σε k
(5)
The model constants are Cμ = 0.09; σ k = 1.00; σ ε = 1.00;
Cε1 = 1.44, Cε2 = 1.92. The term G in above equation is the produc-
tion of turbulent kinetic energy and described by:
G = τL : ∇ u L (13)
As k–ε is used for turbulence modelling, the turbulent eddy
viscosity is described using the standard k–ε turbulence model,
where k shows the turbulent kinetic energy and ε its dissipation
rate in the continuous phase. k and ε describe the energy in tur-
bulence and the scale of the turbulence, respectively. The turbulent
(7) kinetic energy significantly changes as the bubble size changes in
the column. For example, Pourtousi et al. [4] reported that as a re-
sult of exciting smaller bubble in the bubble column reactor the
value of turbulent kinetic energy is high due to higher amount of
gas in the column and higher upward and downward axial liquid
velocity near the column centre and walls.
2.2. Geometrical structure and simulation cases
(8)
A cylindrical bubble column reactor with a height and diame-
ter of 2.6 m and 0.288 m, respectively is used, and this is the same
as that used in Pfleger and Becker [27]. Air bubbles are sparged in
the quiescent water in the column using a ring sparger. The gas has
a superficial velocity of 0.005 m/s at the ambient condition. Three
types of gas ring spargers with diameters of 0.07 m, 0.14 m and
0.20 m, namely Cases 1, 2 and 3, respectively are simulated. Each
sparger has 20 nozzles of a diameter of 0.0 0 07 m. For comparison
and benchmarking, the sparger of Case 2 is the same as that used
in Pfleger and Becker [27]. However, the inlet boundary condition
for the sparger in the present study is not the same as that used
(9) in Pfleger and Becker [27] and this will be discussed later.

where k and CTD are the liquid turbulent kinetic energy per unit of
mass and turbulent dispersion coefficient respectively. In this CFD
study, values of 0.2–0.5 are used for the turbulent dispersion coef-
ficient as suggested in prior studies [2–4,21,29,44].
2.1.1. Turbulence modelling
For turbulence modelling the standard k–ε model is used for
all simulation cases. This model has been intensively used to pre-
dict the phase flow pattern in the bubble columns over the past
two decades [2,3,18,21,27,29,33]. The turbulent eddy viscosity, the
2.3. Grid
A structured grid based on hexahedral grid as shown in Fig. 1
is used throughout the domain. The type of grid adopted herein is
similar to that used in Boutet et al. [32]. The axial length of the
domain is divided into 60 grid elements. The grid namely Grid 1,
typically has about 40,500 elements. On average, the grid has an
aspect ratio of 3.1, skewness of 0.62 and orthogonal quality of 0.67
and these values are within the acceptable range [45]. In addition,
two denser girds namely Grid 2 and 3, were created and has a total
 M. Pourtousi et al. / Journal of the Taiwan Institute of Chemical Engineers 69 (2016) 14–24
Fig. 1. (a) Schematic of bubble column reactor, having ring sparger; (b) Grid intensity of the cylindrical bubble column reactor consisting of 40,500 structural elements.
of 59,0 0 0 and 82,320 elements, respectively for a grid dependency
study. The grid study was limited by the processing capability of
the computer used and will be discussed in detail in Section 3.
2.4. Boundary conditions and numerical methods
The gas velocity from each nozzle of the ring spargers is cal-
culated based on a superficial gas velocity of 0.005 m/s. The ring
nozzles are not directly modelled but instead a mass source was
applied at the exact locations of the nozzles to reduce a total num-
ber of mesh elements [21,26]. Hence, the ‘mass sources’ are used
as the inlet condition at the bottom of the domain. The top surface
of the bubble column is treated as the ‘degassing’ boundary condi-
tion which allows only the dispersed air bubbles to escape but not
the liquid phase [26,44,46]. On the side walls, a no-slip boundary
condition is used for the liquid phase and a free-slip condition for
the gas phase [36,46–48].
Simulations were carried out using the commercial CFD soft-
ware package of ANSYS-CFX 13.0 [49]. The conservation equations
were discretized using the control volume technique. The equa-
tion system was solved using the SIMPLEC procedure. The high or-
der differencing schemes of the total variation diminishing (TVD)
is used, which is suggested for the Eulerian–Eulerian multiphase
models to reduce numerical diffusion [18,27,50]. The flow was sim-
ulated for a total of 1400 s and most of results presented in the fol-
lowing section are time averaged over the last 1300 s. The effect of
time-step on the liquid velocity and gas hold-up results is exam-
ined. To study the size of time step, the Courant–Friedrichs–Levy
(CFL) number, is used. The CFL order number of one is necessary
to resolve multiphase flow inside the bubble column reactor. The
17
CFL number can be formulated such as follows:
y
t ≤ (14)
|v|
where |v| ,y and  t are the magnitude of the velocity vector’s
vertical component, characteristic dimension of the cell and time
step, respectively. Different time step (i.e., 0.1–0.01) sizes are ex-
amined for sensitivity study in this research. Small different was
found between numerical results for various time step sizes. There-
fore, in order to minimize computational cost of this study, the
time step size of 0.1 has been selected for all simulation cases. The
selection of size of time step is also based on CFL number. As the
CFL number is smaller than one, the stability of numerical results
increases. Prior studies showed that as the CFL number is less than
one, the Eulerian method can accurately predict liquid flow pat-
tern and amount of gas and further refining of time step does not
improve the numerical results. However, when the CFL number in-
creases more than 1, the stability and accuracy of numerical results
reduces [51–57].
3. Results and discussion
3.1. Mesh sensitivity and validation of CFD method
The CFD results of Case 2 based on three types of grids as
given in Section 2.3 are benchmarked against that from Pfleger and
Becker [27]. Figs. 2(a) and (b) shows the liquid velocity and time
averaged gas hold-up respectively, versus the normalized radial co-
ordinate for Grids 1, 2 and 3 at a column height of 1.6 m. Referring
to Fig. 2(a), in general, the liquid has an up-flow near the column
centre and a down-flow near the column wall for all the Grids. The
18 M. Pourtousi et al. / Journal of the Taiwan Institute of Chemical Engineers 69 (2016) 14–24
Fig. 2. (a) Comparison between the average of axial liquid velocity from CFD (Grids
1,2 and 3) and that from experimental and numerical from Pfleger and Becker
[27] at height 1.6 m. (b). Comparison between the average of gas hold-up from
CFD (Grids 1,2 and 3) and that from experimental and numerical from Pfleger and
Becker [27] at height 1.6 m. (c). Comparison of local gas hold-up obtained from
present CFD results (Grids 1,2 and 3) with experimental and numerical data from
Pfleger and Becker [27] at various heights.
radial profile of the axial liquid velocity of Grid 1 is symmetric at
the centre. However, this is not the case for Grids 2 and 3 where
the peak velocity is skewed to the right and to the left, respec-
tively, having asymmetric velocity profiles. Among the three grids,
the distribution of the liquid velocity of Grid 1 is very close to the
numerical results of Pfleger and Becker [27] with small differences
near the column axis.
Now refer to Fig. 2(b), which shows the average gas hold-up at
1.6 m height of Grids 1, 2 and 3 of Case 2. The profiles of grids
are quite non-uniform near the core of the flow (column centre).
Clearly, two peaks profiles can be seen, which is quite obvious for
Grid 2. The formation of the two peaks is suspected due to the
effect of a ring sparger, which is truthfully spatially modelled us-
ing the inlet source term. This will be further discussed in the fol-
lowing section. For a similar case setup, such result from the nu-
merical investigations of Pfleger and Becker [27] has a symmet-
ric profile and slightly a higher value. However, it should be noted
that in Pfleger and Becker [27] a simplified model of the gas in-
let boundary conditions is used and this has been identified to be
one of reasons for an over prediction of the gas hold-up. There-
fore, a direct comparison may not possible and instead planar aver-
age gas hold-ups at various column heights were studied and pre-
sented in Fig. 2(c) which includes the numerical and experimen-
tal results from Pfleger and Becker [27]. Our results using Grid 1
and 3, on overall, from the bottom to the top of the column have a
much closer prediction to Pfleger and Becker’s experimental results
in comparison to Pfleger and Becker’s numerical results which are
based on two types of computational Grids. The maximum differ-
ence of our results is less than 15% everywhere in the column. Grid
1, which is less denser than Grids 2 and 3, has been selected for
various investigations in the current study based on the reason-
able predictions of the axial liquid velocity and gas hold-up in the
column as presented above.
3.2. Effect of the diameter of ring sparger
Cases 1, 2 and 3 are used to investigate the effect of the ring
sparger diameters on gas hold-up and liquid velocity in the col-
umn reactor. Fig. 3(a, b and c) show an average profile of the axial
liquid velocity as a function of the radial position in the column at
axial heights of 1, 1.6 and 2.4 m for Cases 1 to 3. The velocity pro-
file of Case 2 (i.e., our benchmark case) is the result of the use of
the ring sparger with a diameter 0.14 m. The maximum axial liquid
velocity is about 0.13 m/s near the column centre. The velocity pro-
file of Case 1 is steeper at the column centre in comparison to that
of Case 2 with the maximum axial liquid velocity of about 0.18 m/s,
resulting a high flow at this location. This is understandable due to
the concentrated type of the ring sparger used in Case 1 which has
a diameter of 0.07 m and located very close to the column axis.
Unlike Cases 1 and 2, the velocity profile of Case 3 has a down-
ward flow at the column central region, whereas the upwards flow
only occurs at a small circumferential region, i.e., near r/R = ±0.75.
The downward flow dominates having the downward peak axial
velocity of about 0.1 m/s as a result of using a ring sparger with a
diameter that is close to the column diameter. However, the use of
this type of sparger increases the axial velocity near the wall.
Fig. 4 (a, b and c) show the comparison of time average profiles
of the gas holdup as a function of the column radial position at
axial heights of 1, 1.6 and 2.4 m for Cases 1 to 3. Clearly, the fig-
ure shows a significant profile different between these three cases.
The peaks of the gas hold-up move from the column centre to-
wards the column wall with the increase of the diameter of the
ring sparger (i.e., from Case 1 to 3). The gas hold-up of Case 1 has
a steep profile with a peak near the column centre. Due the con-
centrated sparger type in Case 1, the gas flow is maximum at the
column axis and minimum at the wall. The profile is steeper near
the sparger, i.e., 1 m axial height, due to a strong sparger effect
and less steep as the gas flow up and away from the sparger. This
is quite clear as the peak of the gas hold-up decreases with the in-
crease of the axial height, e.g., the peak value is about 0.048, 0.41
and 0.0351 at 1, 1.6 and 2.4 m column heights, respectively for Case
1. However, as the profile becomes less steep at a higher column
height, the gas hold-up near the wall increases, e.g., the value near
the column wall is the highest at 2.4 m axial height for Case 1(see
Fig. 4(c)).
The gas hold-up of Case 2 has a blunter profile in comparison
to that of Case 1 at all axial heights. Most of the gas flows in the
central column region of r/R = ±0.5. In addition, the profile seems
 M. Pourtousi et al. / Journal of the Taiwan Institute of Chemical Engineers 69 (2016) 14–24
Fig. 3. The averaged axial liquid velocity versus normalized radial coordinate for Case 1, 2 and 3 at height: (a) 1 m; (b) 1.6 m; (c) 2.4 m.
to have two weakly developed peaks, see near r/R = ±0.3–0.5 due
the effect of the use of the ring sparger of 0.14 m diameter in the
0.288 m column diameter, which is located in between the axis
and the wall. The value of the gas hold-up near the wall is higher
in Case 2 than in Case 1, e.g., see r/R<−0.4 for clear differences.
This is expected due to the blunter profile which often increases
the value near the wall because of the mass conservation. As de-
scribed for Case 1, the value of the maximum gas hold-up of Case
2 is higher at the axial height of 1 m, i.e., near sparger, but lower
at a higher axial height. The gas hold-up of Case 3 has a similar
profile as its axial liquid velocity (Fig. 3) where the gas holdup has
two peaks near the wall region. The gas hold-up is almost zero
near the column centre.
The cross-section plane average of the gas holdup at various
column heights is presented in Fig. 5 for Cases 1, 2 and 3. Note
that Case 2 is our benchmark case and the results from this case
have been validated and discussed in Section 3.1. The results of
Cases 1 and 3 are lower by about 5% and 21% maximum, respec-
tively than that of Case 2 at most of the heights except at the inlet.
Given the same superficial gas velocity, a ring sparger located near
the wall as that in Case 3 will result in a lower value of a planar
averaged gas holdup as well as the total overall gas hold-up in the
19
column. It is also clear that Case 2 sparger is able to produce the
highest average gas hold-up everywhere in the column in compar-
ison to other cases. This may have resulted due to a lower bubble
flow resistance by using such sparger.
Comparison of profiles of the time-averaged turbulent kinetic
energy (k) versus the normalized radial coordinate are illustrated
in Fig. 6 (a, b and c) for Cases 1 to 3 at axial heights of 1, 1.6
and 2.4 m. The radial profiles of k are significantly influenced by
the use the different types of gas spargers. The distribution of k,
in general, is higher at the column central region and it reduces to
a zero value moving toward the column wall in all cases. Case 1
has the highest k value in the central region followed Case 2 and
then Case 3 which has the least k value. For example, at r/R = 0, k
is about 0.0251, 0.0141 and 0.0076 for Cases 1, 2 and 3 respectively
at 1 m axial height (see Fig. 6(a)). For each case, k at the central
region reduces the increase of the axial height. The reduction is
greater for Case 1 and least reduction occurs for Case 3 based on
differences of the absolute k value at different heights. However,
in percentage, the reduction is constant for all cases, e.g., k value
at 1.6 m and 2.4 m heights is a reduction about 20% and 40%, re-
spectively from that at 1 m height. The effect of the different types
of spargers is again reflected in the k radial profile, i.e., the steep
20 M. Pourtousi et al. / Journal of the Taiwan Institute of Chemical Engineers 69 (2016) 14–24

Fig. 4. The averaged gas hold-up versus normalized radial coordinate for Case 1, 2 and 3 at height: (a) 1 m; (b) 1.6 m; (c) 2.4 m.

profile for Case 1 especially at 1 m axial height and two weakly

developed peaks near r/R = ±0.3–0.5 for Case 2. However, this is
different for Case 3 where its k profile seems not affected by the
use of Case 3 sparger. This is despite of the significant differences
seen in the radial profiles of the axial liquid velocity (Fig. 3) and
gas hold-up (Fig. 4) of Case 3. The low values of the axial liquid ve-
locity and the effect of the column wall can be the possible causes
that have damped down the k in Case 3.
In general, to accurately predict the bubble column hydrody-
namics parameters, the sensitivity study of different single bubble
sizes and interfacial force models is necessary to find appropriate
CFD model. In this study, different drag and turbulent dispersion
force models, as well as bubble sizes are examined to find the best
condition for modelling of various sparger sizes.

3.3. Effect of drag laws

In this section, the sensitivity of the use of different types of

Fig. 5. Local average gas hold-up for Cases 1, 2, and 3 at various column heights.
drag laws in predictions of axial liquid velocity and gas hold-up in
 M. Pourtousi et al. / Journal of the Taiwan Institute of Chemical Engineers 69 (2016) 14–24
Fig. 6. The averaged turbulence kinetic energy (TKE) versus normalized radial coordinate for Case 1, 2, and 3 at height: (a) 1 m; (b) 1.6 m; (c) 2.4 m.
the bubble column have been investigated. Simulations were car-
ried out using the same settings of the Case 2 CFD model expect
for drag forces. The radial distribution of the time-averaged ax-
ial liquid velocity and the gas hold-up at column height 1.6 m is
shown in Fig. 7(a) and (b), respectively based on drag models of
Grace et al. [41], Ishii & Zuber [42]and Schiller & Naumann [40].
Such results from our Case 2 and that from experimental and nu-
merical investigations of Pfleger and Becker [27] are included in
the figures for a comparison. Referring to 7(a), all the axial liquid
velocity profiles from the drag models are similar to that of Case
2 and the differences between them are marginal. Near the central
region of the column, Grace et al. [41] and Ishii & Zuber [42] mod-
els have a slightly better prediction than Case 2 when comparing
to Pfleger and Becker’s numerical data. Referring Fig. 7(b), both
Grace and Ishii & Zuber models over predict the average gas hold-
up everywhere in the column in comparison to that of Case 2. In
addition, these models also over predict this value at central re-
gion of the column in comparison to Pfleger and Becker’s results.
However, there is no significant difference in the gas hold-up be-
tween the Schiller & Naumann [40] model and Case 2. Fig. 7(c)
shows the planar averaged gas hold-up at various column heights.
Schiller & Naumann [40] and Case 2 results are almost identical.
However, the Grace and Ishii & Zuber drag models over predict the
gas hold-up at all heights in comparison with Case 2 and Pfleger
and Becker’s experimental results. As report by Tabib et al. [21] and
Zhang et al. [58], drag force may decreases the gas phase (bub-
ble) velocity and increases the residence time of the gas phase (gas
hold-up) inside the bubble column.
3.4. Effect of bubble diameters
The sensitivity of the use of different diameters of gas bubble
in predictions of axial liquid velocity and gas hold-up in the bub-
ble column have been investigated. Simulations were carried out
using the same settings of the Case 2 CFD model expect for the
size of the bubble diameter. The radial distribution of the time-
averaged axial liquid velocity and the gas hold-up at column height
1.6 m is shown in Fig. 8(a) and (b), respectively based on bubble
diameters of 0.004 to 0.0055 m. Fig. 8(c) shows the planar aver-
aged gas hold-up at various column heights. Such results from our
Case 2 and that from the experimental investigation of Pfleger and
Becker [27], which both used a bubble diameter of 0.004 m, are
included in the figures for a comparison. The use of a bubble di-
ameter of 0.0045 m, an increase about 0.0005 m from that of Case
2, only results in small differences in the radial of the axial liq-
uid velocity (Fig. 8(a)), the gas hold-up (Fig. 8(b)) and the aver-
aged gas hold-up (Fig. 8(c)) in the column. However, this is not
the case for the bubble diameter of 0.005 m (or 0.0055 m) with an
increase of 0.001 m (or 0.0015 m ) from that of Case 2. The dif-
21
ferences are significant. For example, the radial profile of the axial
liquid is asymmetric with more positive velocity at right side of
the column. Consequently, the radial profile of the gas rises asym-
metrically in the column (Fig. 8(b)) and has a two-peak profile. The
amount of gas for case of bubble diameter 0.005 and 0.0055 m at
different heights of bubble column is lesser than smaller bubble
(i.e., 0.0 04 and 0.0 045 m). For example, at 2 m height the planner
average gas hold-up for 0.004 and 0.005 m are 0.0156 and 0.0138
respectively (see Fig. 8(c)). The increase of gas hold-up may at-
tribute to the fact that, as bubble size decreases, bubble interfa-
cial area increases. This finding is consistent with previous studies
[59–62].
3.5. Effect of turbulent dispersion
In this section, the sensitivity of the coefficient value of the tur-
bulent dispersion (CTD ) in range of 0.2–0.5 is investigated on the
results predicted. We compare the radial distribution of the time-
averaged axial liquid velocity and the gas hold-up at column height
1.6 m for turbulent dispersion coefficient 0.2–0.5 with results from
experimental and numerical investigations of Pfleger and Becker
[27]. It was observed that the predicted results of axial liquid ve-
locity and gas hold-up for different values of CTD are almost iden-
tical near the wall, whereas near the central region, these param-
eters are slightly closer to experimental results [27] for CTD = 0.3
(Case 2).
3.6. Limitations
There are some limitations in the present study. First, the grid
used is a compromise between the availability of computational re-
sources and the numerical accuracy. Despite that, the type of grid
in simulations is similar to that reported in reference [32]. The sen-
sitivity study of using a much denser grid requires further investi-
gations. However, it should be noted that it is not always the case
that a denser grid increases the accuracy of a numerical prediction
in a bubble column reactor as discussed in Buwa et al. [30], Pfleger
and Becker [27] and Diaz et al. [46] in which the use of a coarser
grid have resulted much accurate results. Therefore a careful selec-
tion of a grid is required. Secondly, the use k–ε turbulent model is
based on isotropic assumption where the turbulent kinetic energy
(k) formulation is constructed with a limitation that all the nor-
mal components of stresses are equal to each other. This may lead
to inaccuracy in the prediction kinetic energy. Note that various
studies in literature on bubble column reactor are based on k–ε
turbulent model and this was our main benchmark for selecting
this model for the current study. In addition, as far as this study
concern, the focus herein is to demonstrate the possible hydrody-
namic differences resulted from the use of different sizes of ring
22 M. Pourtousi et al. / Journal of the Taiwan Institute of Chemical Engineers 69 (2016) 14–24
Fig. 7. (a) Comparison between the average of axial liquid velocity from drag co-
efficient = 0.44 (CFD Case 2), Grace et al. [41], Ishii & Zuber [42] and Schiller &
Naumann [40] and that from experimental and numerical from Pfleger and Becker
[27] at height 1.6 m. (b). Comparison between the average of gas hold-up from drag
coefficient = 0.44 (CFD Case 2) , Grace et al. [41], Ishii & Zuber [42] and Schiller &
Naumann [40] and that from experimental and numerical from Pfleger and Becker
[27] at height 1.6 m. (c). Local averaged gas hold-up for drag coefficient=0.44 (CFD
Case 2), Grace et al. [41], Ishii & Zuber [42] and Schiller & Naumann [40] at various
column heights.
diameters of a ring sparger and not to present an absolute value
of any parameters. Our justification is that similar results as that
from this study will be repeated by adopting either a denser grid
or a different turbulence models.
In this research, we specifically concentrate on the size of ring
sparger on the bubble column hydrodynamics parameters and find
appropriate numerical method for prediction in one size of indus-
trial bubble column. Increasing column diameter or height requires
high number of elements, as well as mesh sensitivity study. How-
Fig. 8. (a). Comparison between the average of axial liquid velocity from various
bubble diameter sizes and that from experimental and numerical from Pfleger and
Becker [27] at height 1.6 m. (b). Comparison between the average of gas hold-up
from various bubble diameter sizes and that from experimental and numerical from
Pfleger and Becker [27] at height 1.6 m. (c). Local averaged gas hold-up for various
bubble diameters at different column heights.
ever, we have limitation in computer resource to increase number
of elements for larger bubble column size.
4. Conclusions
The effect of different ring sparger diameters (i.e., 0.07, 0.14 and
0.20 m) on the hydrodynamic parameters of bubble column is in-
vestigated. Furthermore, sensitivity studies of the different drag
law models, values of turbulent dispersion coefficient and bubble
diameters are carried out to find an appropriate CFD model for ho-
mogeneous flow regime. The conclusions of this study are as fol-
lows:
 M. Pourtousi et al. / Journal of the Taiwan Institute of Chemical Engineers 69 (2016) 14–24
• For the small ring sparger diameter (0.07 m), coalescence of
bubbles is more likely to appear towards the column centre,
particularly at the sparger region, while the liquid velocity and
amount of gas are larger at the column centre. In this case, a
steeper profile of the axial liquid velocity and the gas hold-up
appeared near the column centre due to a high flow at this lo-
cation from the bottom to the top of the column. As the ring
sparger diameter reaches 0.14 m (middle size ring sparger), the
axial liquid velocity and gas hold-up at the centre decreases,
while the overall amount of gas inside the column rises. This
shows that there is an optimum ring sparger size to provide
a higher gas hold-up, which enhance the bubble column effi-
ciency. In the case of using the larger ring sparger size, (0.2 m),
the coalescence of bubbles rarely occurs near the sparger re-
gion. However, this type of sparger can provide a strong down-
ward flow at the column centre. This type of sparger increases
the axial liquid velocity and gas hold-up near the wall region
and changes the liquid circulation direction. This feature in the
larger sparger size may be further explored to enhance the liq-
uid agitation and reaction process near the column wall.
• For homogeneous flow regime, the selection of smaller bubble
size (0.004 m) in the numerical method results in improvement
of liquid velocity and gas hold-up. In addition, the constant
drag coefficient can predict liquid flow pattern and gas hold-
up towards the wall and central region. In this regime, the tur-
bulent dispersion coefficient of 0.3 is recommended to improve
the numerical results, particularly near wall region.
Acknowledgment
This research is financially supported by University of
Malaya, Ministry of Higher Education High Impact Research
(UM.C/HIR/MOHE/ENG/13) and Bernal Institute, Synthesis and
Solid State Pharmaceutical Centre (SSPC), University of Limerick,
Ireland with Grant Numbers (14/SP/2750-MOMEnTUM).
References
[1] Mousavi S, et al. Experiments and CFD simulation of ferrous biooxidation in a
bubble column bioreactor. Comput Chem Eng 2008;32(8):1681–8.
[2] Pourtousi M, et al. A combination of computational fluid dynamics (CFD) and
adaptive neuro-fuzzy system (ANFIS) for prediction of the bubble column hy-
drodynamics. Powder Technol 2015;274:466–81.
[3] Pourtousi M, Sahu JN, Ganesan P. Effect of interfacial forces and turbulence
models on predicting flow pattern inside the bubble column. Chem Eng Pro-
cess: Process Intensif 2014;75:38–47.
[4] Pourtousi M, Ganesan P, Sahu JN. Effect of bubble diameter size on predic-
tion of flow pattern in Euler–Euler simulation of homogeneous bubble column
regime.. Measurement 2015;76:255–70.
[5] Pourtousi M, et al. Methane bubble formation and dynamics in a rectangular
bubble column: a CFD study. Chemom Intell Lab Syst 2015;147:111–20.
[6] McClure DD, et al. Experimental investigation into the impact of sparger de-
sign on bubble columns at high superficial velocities. Chem Eng Res Des
2016;106:205–13.
[7] McClure DD, et al. Impact of surfactant chemistry on bubble column systems.
Chem Eng Technol 2014;37(4):652–8.
[8] McClure DD, et al. Development of a CFD model of bubble column bioreac-
tors: part two–comparison of experimental data and CFD predictions. Chem
Eng Technol 2014;37(1):131–40.
[9] McClure DD, et al. Development of a CFD model of bubble column
bioreactors: part one–a detailed experimental study. Chem Eng Technol
2013;36(12):2065–70.
[10] McClure DD, et al. Mixing in bubble column reactors: experimental study and
CFD modeling. Chem Eng J 2015;264:291–301.
[11] McClure DD, et al. Validation of a computationally efficient computational
fluid dynamics (CFD) model for industrial bubble column bioreactors.. Ind Eng
Chem Res 2014;53(37):14526–43.
[12] McClure DD, et al. Impact of surfactant addition on oxygen mass transfer in a
bubble column. Chem Eng Technol 2015;38(1):44–52.
[13] McClure DD, et al. Towards a CFD model of bubble columns containing signif-
icant surfactant levels. Chem Eng Sci 2015;127:189–201.
[14] Pareek V, et al. Particle residence time distribution (RTD) in three-phase an-
nular bubble column reactor. Chem Eng Sci 2001;56(21):6063–71.
23
[15] Şal S, Gül ÖF, Özdemir M. The effect of sparger geometry on gas holdup and
regime transition points in a bubble column equipped with perforated plate
spargers.. Chem Eng Process: Process Intensif 2013;70:259–66.
[16] Upadhyay RK, Pant HJ, Roy S. Liquid flow patterns in rectangular air-water
bubble column investigated with Radioactive Particle Tracking.. Chem Eng Sci
2013;96:152–64.
[17] Salierno GL, et al. Discrete axial motion of a radioactive tracer reconstructed
from the response of axially aligned detectors: application to the analysis of a
bubble column dynamics. Chem Eng Sci 2013;100:402–12.
[18] Pfleger D, et al. Hydrodynamic simulations of laboratory scale bubble columns
fundamental studies of the Eulerian–Eulerian modelling approach.. Chem Eng
Sci 1999;54(21):5091–9.
[19] Kantarci N, Borak F, Ulgen KO. Bubble column reactors. Process Biochem
2005;40(7):2263–83.
[20] Gupta A, Roy S. Euler–Euler simulation of bubbly flow in a rectangular bubble
column: experimental validation with radioactive particle tracking.. Chem Eng
J 2013;225:818–36.
[21] Tabib MV, Roy SA, Joshi JB. CFD simulation of bubble column—an analysis of
interphase forces and turbulence models.. Chem Eng J 2008;139(3):589–614.
[22] Dhotre MT, Ekambara K, Joshi JB. CFD simulation of sparger design and height
to diameter ratio on gas hold-up profiles in bubble column reactors. Exp
Therm Fluid Sci 2004;28(5):407–21.
[23] Dhotre MT, Joshi JB. Design of a gas distributor: Three-dimensional CFD simu-
lation of a coupled system consisting of a gas chamber and a bubble column.
Chem Eng J 2007;125(3):149–63.
[24] Di Bari S, Lakehal D, Robinson AJ. A numerical study of quasi-static gas
injected bubble growth: Some aspects of gravity. Int J Heat Mass Transf
2013;64:468–82.
[25] Di Marco P, Grassi W. Effect of force fields on pool boiling flow patterns in
normal and reduced gravity. Heat Mass Transf 2009;45(7):959–66.
[26] Li G, Yang X, Dai G. CFD simulation of effects of the configuration of gas dis-
tributors on gas–liquid flow and mixing in a bubble column. Chem Eng Sci
2009;64(24):5104–16.
[27] Pfleger D, Becker S. Modelling and simulation of the dynamic flow behavior in
a bubble column. Chem Eng Sci 2001;56(4):1737–47.
[28] Bhole MR, Roy S, Joshi JB. Laser Doppler anemometer measurements in bubble
column: effect of sparger. Ind Eng Chem Res 2006;45(26):9201–7.
[29] Pourtousi M, et al. Prediction of multiphase flow pattern inside a 3D
bubble column reactor using a combination of CFD and ANFIS. RSC Adv
2015;5:85652–72.
[30] Buwa VV, Ranade VV. Dynamics of gas–liquid flow in a rectangular bubble
column: experiments and single/multi-group CFD simulations. Chem Eng Sci
2002;57(22):4715–36.
[31] Sato Y, Sekoguchi K. Liquid velocity distribution in two-phase bubble flow. Int
J Multiph Flow 1975;2:79–95.
[32] Laborde-Boutet C, et al. CFD simulation of bubble column flows: In-
vestigations on turbulence models in RANS approach. Chem Eng Sci
2009;64(21):4399–413.
[33] Chen P, Duduković MP, Sanyal J. Three-dimensional simulation of bubble col-
umn flows with bubble coalescence and breakup. AIChE J 2005;51(3):696–712.
[34] Chen P, Sanyal J, Duduković MP. Numerical simulation of bubble columns
flows: effect of different breakup and coalescence closures. Chem Eng Sci
2005;60(4):1085–101.
[35] Lemoine R, et al. An algorithm for predicting the hydrodynamic and mass
transfer parameters in bubble column and slurry bubble column reactors. Fuel
Process Technol 2008;89(4):322–43.
[36] Monahan SM, Vitankar VS, Fox RO. CFD predictions for flow-regime transitions
in bubble columns. AIChE J 2005;51(7):1897–923.
[37] Ruzicka MC, et al. Effect of surfactant on homogeneous regime stability in bub-
ble column. Chem Eng Sci 2008;63(4):951–67.
[38] Dhotre MT, Smith BL. CFD simulation of large-scale bubble plumes: compar-
isons against experiments. Chem Eng Sci 2007;62(23):6615–30.
[39] Smith BL. On the modelling of bubble plumes in a liquid pool. Appl Math
Model 1998;22(10):773–97.
[40] Schiller L, Naumann Z. A drag coefficient correlation. Vdi Zeitung
1935;77:318–20.
[41] Grace JR, Wairegi T, Nguyen TH. Shapes and velocities of single drops and
bubbles moving freely through immiscible liquids. Trans Inst Chem Eng
1976;54(3):167–73.
[42] Ishii M, Zuber N. Drag coefficient and relative velocity in bubbly, droplet or
particulate flows. AIChE J 1979;25(5):843–55.
[43] Lopez de Bertodano M, Lahey RT Jr, Jones OC. Turbulent bubbly two-phase
flow data in a triangular duct. Nucl Eng Des 1994;146:43–52.
[44] Silva MK, d’Ávila MA, Mori M. Study of the interfacial forces and turbulence
models in a bubble column. Comput Chem Eng 2012;44:34–44.
[45] Kerdouss F, Bannari A, Proulx P. CFD modeling of gas dispersion and bubble
size in a double turbine stirred tank. Chem Eng Sci 2006;61(10):3313–22.
[46] Díaz ME, et al. Numerical simulation of the gas–liquid flow in a laboratory
scale bubble column: influence of bubble size distribution and non-drag forces.
Chem Eng J 2008;139(2):363–79.
[47] Wang H, et al. CFD modeling of hydrodynamic characteristics of a gas–liquid
two-phase stirred tank. Appl Math Model 2014;38(1):63–92.
[48] Zarei A, Hosseini SH, Rahimi R. CFD study of weeping rate in the rectangular
sieve trays. J Taiwan Inst Chem Eng 2013;44(1):27–33.
[49] ANSYS Inc. ANSYS CFX-13.0 User Manual. 2011.
[50] Deen NG, Solberg T, Hjertager BH. Numerical simulation of the gas–liq-
24 M. Pourtousi et al. / Journal of the Taiwan Institute of Chemical Engineers 69 (2016) 14–24

uid flow in a square cross-sectioned bubble column.. In: Proceeding of [57] Dhotre MT, et al. Large-eddy simulation (LES) of the large scale bubble plume.
the 14th Int. Congress of Chemical and Process Engineering, Praha-Czech Chem Eng Sci 2009;64(11):2692–704.
Republic; 20 0 0. [58] Zhang D, Deen NG, Kuipers JAM. Numerical simulation of the dynamic flow
[51] Ma T, et al. Scale-adaptive simulation of a square cross-sectional bubble col- behavior in a bubble column: a study of closures for turbulence and interface
umn. Chem Eng Sci 2015;131:101–8. forces. Chem Eng Sci 2006;61(23):7593–608.
[52] Ziegenhein T, Rzehak R, Lucas D. Transient simulation for large scale flow in [59] Bouaifi M, et al. A comparative study of gas hold-up, bubble size, interfacial
bubble columns. Chem Eng Sci 2015;122:1–13. area and mass transfer coefficients in stirred gas–liquid reactors and bubble
[53] Laín S. Dynamic three-dimensional simulation of gas-liquid flow in a cylindri- columns. Chem Eng Process: Process Intensif 2001;40(2):97–111.
cal bubble column. Lat Am Appl Res 2009;39(4):317–26. [60] Lau Y, et al. Experimental study of the bubble size distribution in a pseudo-2D
[54] Buffo A, et al. Simulation of polydisperse multiphase systems using popula- bubble column. Chem Eng Sci 2013;98:203–11.
tion balances and example application to bubbly flows. Chem Eng Res Des [61] Hebrard G, Bastoul D, Roustan M. Influence of the gas sparger on the hydro-
2013;91(10):1859–75. dynamic behavior of bubble columns. Chem Eng Res Des 1996;74(3):406–14.
[55] Dhotre MT, et al. Large eddy simulation for dispersed bubbly flows: a review. [62] Ravinath M, Kasat GR, Pandit AB. Mixing time in a short bubble column. Can J
Int J Chem Eng 2013;2013:1–22. Chem Eng 2003;81(2):185–95.
[56] Laín S. Large eddy simulation of gas–liquid flow in a bubble column reactor. El
Hombre y la Máquina 2009;32:108.