Accepted Manuscript

Slug and bubble flows in a flat sheet ultrafiltration module:

experiments and numerical simulation

Seyyed Morteza Javid , Mohammad Passandideh-Fard ,

Ali Faezian , Masoud Goharimanesh

PII: S0301-9322(16)30251-8
DOI: 10.1016/j.ijmultiphaseflow.2016.12.006
Reference: IJMF 2517

To appear in: International Journal of Multiphase Flow

Received date: 28 April 2016

Revised date: 18 August 2016
Accepted date: 20 December 2016

Please cite this article as: Seyyed Morteza Javid , Mohammad Passandideh-Fard , Ali Faezian ,
Masoud Goharimanesh , Slug and bubble flows in a flat sheet ultrafiltration module: ex-
periments and numerical simulation, International Journal of Multiphase Flow (2017), doi:
10.1016/j.ijmultiphaseflow.2016.12.006

This is a PDF file of an unedited manuscript that has been accepted for publication. As a service
to our customers we are providing this early version of the manuscript. The manuscript will undergo
copyediting, typesetting, and review of the resulting proof before it is published in its final form. Please
note that during the production process errors may be discovered which could affect the content, and
all legal disclaimers that apply to the journal pertain.
 ACCEPTED MANUSCRIPT

Highlights

 The characteristics of bubbles in flow such as their size and distribution are analyzed
based on image processing techniques.

 Bubble distribution of two phase flows indicate that the average diameter of the slugs
is much larger than that of the bubbles.

T
 Successful 3D simulation of two phase flow patterns in a flat sheet module.

IP
 The enhanced shear stress is due to two factors: increased velocity gradient of liquid

CR
near the bubbles and induced shear stress by wake region

 The slug flow induces considerably larger shear stress than bubble flow.

US
AN
M
ED
PT
CE
AC
 ACCEPTED MANUSCRIPT

Slug and bubble flows in a flat sheet ultrafiltration

module: experiments and numerical simulation

a a,1 b
Seyyed Morteza Javid , Mohammad Passandideh-Fard , Ali Faezian , Masoud

T
a

IP
Goharimanesh

CR
Department of Mechanical Eengineering, Ferdowsi University of Mashhad, Mashhad, Iran

b
Department of Food Machinery Design, Research Institute of Food Science & Technology, Mashhad, Iran

US
Abstract. In this paper, the effect of gas bubbling including slug and bubble flows on
AN
enhancing shear force in an ultrafiltration (UF) process in a flat sheet module is investigated
experimentally by image processing and numerically using the OpenFOAM software. In
order to study characteristics of bubbles in the slug and bubble flows, the flat sheet module is
M

analyzed by a video system facilitated with a high speed camera. The experimental results
show that the average diameter of the slug flow is much larger than that of the bubble flow.
ED

The permeate flux for the slug and bubble flows is increased by 78% and 30%, respectively,
compared to the case with no gas bubbling. The numerical results are shown to be in good
PT

agreement with those of the measurements both qualitatively and quantitatively. The results
of simulations also demonstrate that although both flow patterns increase the shear stress by
CE

increasing the velocity gradient and/or vorticity, the shear stress induced by the slug flow is
considerably larger. Therefore, the slug flow with a higher induced shear stress is more
AC

effective on the enhancement of the permeate flux in a UF process.

Keywords: gas bubbling, image processing, VOF technique, membrane fouling,

ultrafiltration

1
corresponding author

E-mail: mpfard@um.ac.ir
 ACCEPTED MANUSCRIPT

1 Introduction

UF is a pressure driven filtration process used in a variety of productions ranging from

dairy to pharmaceutical industry. One of the most significant problems of advanced UF
technology is membrane fouling which not only increases the operational costs but also
causes delays in filtration process for membrane cleaning purposes. The main reason of
fouling is the Concentration Polarization (CP) (Matthiasson and Sivik, 1980) which is the
tendency of material to accumulate at the surface of the membrane (Hamann, 2010). Any

T
process that limits the CP is a suitable way to ameliorate the membrane fouling. The CP can

IP
be reduced by several methods such as turbulence promoters (Najarian and Bellhouse, 1996,
Qaisrani and Samhaber, 2011), pulsatile flow (Finnigan and Howell, 1989), vortex generation

CR
(Millward et al., 1995), gas injection or bubbling (Willems et al., 2009), and ultrasound
(Kyllönen et al., 2005).

US
Gas injection is a fouling reduction treatment which has been used considerably for both
submerged and non-submerged membrane processes in the past decades (Wibisono et al.,
AN
2014). The high efficiency in fouling reduction, on one hand, and less detrimental effects on
the environment, on the other, has led to the extensive use of gas injection. Gas injection
decreases the membrane fouling dramatically by enhancing the shear force (Cui et al., 2003).
M

As a matter of fact, bubbles, injected to generate two-phase flow, increase turbulence

intensity which in turn increase the shear force. The gas bubbling performance depends upon
ED

the using two phase flow characteristics (Cui et al., 2003), the type of module (Wibisono et
al., 2014) and type of membrane process (Wibisono, 2014). Among all different types of two-
PT

phase flows, the slug and bubble flows are the most used patterns because of their relatively
low gas flow rates (Cui et al., 2003).
CE

One of the early studies on the effect of gas injection on the flat membrane modules was
carried out by Li et al. (1998) which showed that the gas bubbling can significantly
AC

ameliorate the filtration performance. In term of gas injection properties, Li et al. (1997)
reported significant effects of both bubble size and frequency on the permeate flux of cross-
flow UF using tubular membranes. Derradji et al. (2000) considered the effect of gas
injection combined with static mixer on the membrane fouling. They reported that the gas
bubbling in the presence of turbulence promoter can significantly increase the permeate flux.
In order to investigate the effect of using gas type on gas bubbling, Wang et al. (2004) used
different types of gas (compressed air and n-hexadecane) in a UF membrane made of
 ACCEPTED MANUSCRIPT

regenerated cellulose. The permeate flux for n-hexadecane/water and air/water two-phase
flows was increased by 25% and 17%, respectively, compared to the single-phase flow. The
module geometry is another effective factor. Cheng et al. (2007) studied the effect of flow
channel height on gas bubbling cross-flow UF in a flat sheet module. It was found that shear
stress induced by gas bubbling was lower in the large channel height compared to the narrow
channel, although the gas velocity was high enough in the both channels. The shear stress
induced by two phase flow was measured experimentally by Ducom et al. (2003). They used
an electrochemical method in different points of a flat sheet nanofiltration module and for

T
several operating conditions. The results showed that gas bubbling can considerably increase

IP
the local shear stress.

CR
With the advances in computational fluid mechanics, the basis for further insight into the
dynamics of multiphase flows has been provided. The multi-phase numerical simulation

US
approaches are currently divided into two categories: the Euler-Lagrange approach and the
Euler-Euler approach. The Euler-Lagrange approach is appropriate when the volume fraction
of the second phase is negligible (Sokolichin et al., 1997, Laın et al., 2002). As opposed to
AN
the Euler-Lagrange approach, the Euler-Euler approach is suitable for any application where
the volume fraction of the second phase is considerable such as gas injection in the UF
M

process. In the Euler-Euler approach, a well-known method for tracking the free surface of a
liquid is the volume-of-fluid (VOF) technique (Farhangi et al., 2010). The VOF formulation
ED

is designed for two or more immiscible fluids where the position of the interface between the
fluids is of interest. This technique relies the volume fraction function. The fields for all
PT

variables and properties are shared by the phases and represent volume-averaged values, as
long as the volume fraction of each of the phases is known at each location. Applications of
CE

the VOF model include free-surface flows (Mirzaii and Passandideh-Fard, 2012), the motion
of bubbles in a liquid (Krishna and Van Baten, 1999), the motion of liquid after a dam break
(Biscarini et al., 2010), and the steady or transient tracking of any liquid-gas interface (Reddy
AC

and Banerjee, 2015). Therefore, most of the CFD modeling of two-phase flows in various
module geometries is based on the VOF technique (Taha and Cui, 2002b, Ndinisa et al.,
2005, Taha et al., 2006, Khalili-Garakani et al., 2011, Yang et al., 2012). One of the most
significant CFD modeling was carried out by Taha and Cui (2002a). The slug flow UF
process in tubular membranes was modeled with the commercial software FLUENT. First,
the properties of the slug and the distribution of local wall shear stress in the membrane tube
were calculated; next, a connection between the permeate flux and the slug size and
 ACCEPTED MANUSCRIPT

frequency was established. However, they did not examine the relationship between various
gas and liquid rates and the shear stress. Ratkovich et al. (2009) showed that the frequency of
shear events was decreased at higher gas flow rates and lower liquid flow rates. Compared to
the extensive studies performed on tubular and hollow fiber membrane systems, a few studies
are available in the literature for the flat sheet systems. The shape and the motion of a
spherical cap bubble in flat sheet module was simulated by Essemiani et al. (2001). The wall
shear stress and pressure distributions in the flat sheet module, however, were not calculated.
Wei et al. (2013) simulated a single 3D slug bubble rising in a flat sheet module. The results

T
showed that increasing the bubble size enhanced the induced shear stress.

IP
The missing piece of the above-referred studies is an investigation on the bubbles

CR
characteristics in various two-phase flow patterns especially the slug and bubble flows, and
their effect on the membrane fouling and permeate flow. In this study, the flat sheet module is

US
analyzed with a video system facilitated with a high speed camera in order to examine
bubbles characteristics of the patterns. The experimental results show that the average
diameter of the slug flow is much larger than that of the bubble flow. Numerical results
AN
obtained by the OpenFOAM code using the VOF technique are compared to those of the
experiments. The results show that although both flow patterns increase the shear stress by
M

increasing the velocity gradient and/or vorticity, the shear stress induced by the slug flow is
considerably larger.
ED

2 Experiment
PT

2.1 Setup
CE

The experimental setup used in this study is shown in Fig. 1. The system consists of a feed
tank, a peristaltic pump (Deng Yuan, Taiwan), a gas flow meter, a rotameter for the liquid
AC

flow, a digital scale (A&D, Japan) for measuring permeate flux and a flow cell. The flow cell,
shown in Fig. 2, is made from 2cm thick PMMA plates and has a rectangular flow channel of
0.6cm×10cm. Liquid and air enters through 0.5cm single pipe from the bottom of the
channel, and exits via a similar pipe from the top of the channel. The flow cell is positioned
vertically for all the experiments. Polyethersulfone flat sheet membrane (Sepro, USA) with
10 KD molecular weight cutoff with an effective membrane area of 300 is placed
between the PMMA thick plates. All experiments last for about 30 minutes in a fixed room
temperature after which the membrane is renewed for the next experiment. The setup is a
 ACCEPTED MANUSCRIPT

closed cycle in which a small percentage of the liquid flow is filtered and the rest is recycled
to the feed tank to maintain the feed concentration. The gas flow is injected directly to the
liquid flow at the entrance of the flow cell. For a two-phase mixture of a gas and a liquid
flowing together in a channel, different internal flow patterns can occur depending on the size
of the flow channel, the magnitudes of the gas and liquid velocity, and on the fluid properties
of the two phases at the entrance. In order to generate the slug and bubble flows (Taitel et al.,
1980), the liquid flow rates are selected ⁄ and ⁄ , with a gas flow rate kept

T
constant at ⁄ . A high speed camera (NEX-FS 700, Sony, Japan) with a Zoom lens

IP
(Canon EF 24-70mm, Canon, Japan) is used in order to observe two phase flows. The camera
at a working distance of 50 cm is connected to a PC unit. This system is recorded at a frame

CR
rate of 400 fps at a resolution of pixels. In order to enhance the quality of
recorded movie, a constant light intensity with in the flow cell is obtained by a light source
(Fomex, South Korea).
US
AN
M
ED
PT
CE
AC

Figure 1: Experimental setup of cross flow ultrafiltration with gas bubbling injector
 ACCEPTED MANUSCRIPT

T
IP
CR
Figure 2: Schematic representation of the flow cell a) components b) flow channel

2.2 Material
US
Nitrogen with a purity of 99.97% is used as the gas phase and whey protein concentrate
AN
(WPC) solution with 1% wt solid content concentration is used as the liquid feed for the UF
process. For image processing purposes, however, the demineralized water is used as the
M

liquid phase in order to improve the identification of bubbles. It should be noted that the
physical properties of demineralized water and WPC solution with 1% wt are very similar as
ED

shown in Table 1 (GONZÁLEZ‐TELLO et al., 2009).

Table1: Physical properties

PT

material Density Viscosity Surface Tension

( ⁄ ) ( ) ( ⁄ )
CE

WPC solution with 1% wt 1000.2 10.64 0.066

demineralized water 997.1 8.9 0.0719
AC

2.3 Image processing

In order to validate the numerical model, the simulation results are compared to those of
the experiments. The validation process is conducted utilizing the image processing toolbox
in the MATLAB software where the edge detection methods and strategies are employed.
The edge is the boundary between the object to be extracted and the background. This
technique in image processing is applied in numerous applications such as face recognition,
 ACCEPTED MANUSCRIPT

image segmentation, and image enhancement. Consequently, it is vital to improve the

accuracy of edge detection. In order to detect the bubbles accurately in a background of fluid
flow, it is necessary to detect and extract an accurate edge. Traditional edge detection
methods are conducted using Roberts operator, Sobel operator, Prewitt operator and LOG
operator. All of these operators are local window gradient operators which are interesting
because of their simple methodology and fast detection. Nonetheless, they are more affected
by noise compared to other edge-detection techniques (Basu, 2002). Consequently, these
methods are not completely suitable in experimental studies. In 1986, John Canny (Canny,

T
1986) suggested a new criterion based on which he could find an optimal edge detection

IP
operator, Canny operator. In this study, therefore, the Canny operator is used for all

CR
photographs.

Canny operator can be expressed by the following steps:

Step 1:
US
Basically, the edge detection algorithm is based on the first derivative and second
AN
derivative of the image intensity. However, the derivatives are generally sensitive to the
noise, therefore, the performance of the edge detector with the noise related to the filter must
M

be improved. The commonly used filtering method is Gaussian filtering, firstly using the
discrete filter function to generate a set of normalized Gaussian kernels, then calculating the
ED

weighed sum of each point in the gray matrix image based on Gaussian kernel function.

Eq. (1) is the 2D discrete Gaussian function, and the vector of the 2D kernel can be
PT

obtained by substituting with specific parameters.

1  x2y2
G x , y   
CE

exp  (1)
2 2  2 2 

where,  is a control parameter for level of smoothness. Assume g  x , y  is the smoothed

AC

image, then the smoothness of f  x , y  by G  x , y  could be formulated as:

g x , y   G x , y  f x , y  (2)
 ACCEPTED MANUSCRIPT

Step 2: In this step, gradient of the smoothed image, g  x , y  , is calculated by the

approximate expression of the finite difference of the first partial derivative in the 2D
domain.

Step 3: As the identification of the edges by the overall gradient itself is not a sufficient
condition. The maximum point must be kept in the partial gradient and the non-maximums
must be suppressed. At each point, comparing the center pixel M of the neighboring regions
with two pixels along with the gradient line, if the gradient value of M is smaller than the

T
gradient value of the two pixels along with the gradient line, then M=0, otherwise, M=1.

IP
Step 4: In the dual thresholds algorithm, set the two thresholds as T h and T l for the image

CR
with suppression on non-maximums, and T l  0.4T h . Based on the high threshold, an edge
image will be obtained, which rarely has false edge. Using the high threshold to obtain the

US
profile of the edge, then at each breaking point, the algorithm will search for a point meeting
the low threshold among the neighboring eight points, then start from that point to search for
AN
the new edge. This process will be repeated frequently until the whole image edge is closed.
M

3 Numerical Method

Motion of bubbles in a flowing liquid is simulated by the CFD software OpenFOAM

ED

(Release 2.3.1, 2014) whose accuracy in simulating two-phase flows has been shown in many
studies in the literature (Higuera et al., 2013, Bannari et al., 2008). The VOF technique, a
PT

popular method for gas–liquid interface tracking, is used to calculate bubbles shape and
velocity, and shear stress on the membrane surface.
CE

3.1 Governing equations

The flow governing equations are the continuity equation as:

AC


 .( V )  0 (3)
t

and the momentum equation for incompressible two phase flows as:

( V )
 .( VV )  P  . (   t )(V  (V )T )    g  Fb (4)
t

where Fb is the body force acting on the fluid per unit volume.
 ACCEPTED MANUSCRIPT

To track the liquid/gas interface, the VOF technique is used where a volume fraction α is
defined whose value depends on the fraction of the cell occupied by liquid as:

0

  0 , 1 (5)
1


The advection of volume fraction function through the computational domain is calculated
according to:

T


IP
 .(V )  0 (6)
t

CR
To limit the solution of  equation between zero and one, the OpenFOAM uses an extra
artificial compression (Berberović et al., 2009):


t
 .(V )  .(V r  (1   ))  0
US (7)
AN
where V r V l V g is the vector of relative velocity.

The properties  and  are dependent on the volume fraction and they are defined in:
M

  l .   g .(1   ) (8)
ED

  l .  g .(1   ) (9)

Using the continuum surface force model (CSF) (Bussmann et al., 1999), the surface tension
PT

force is modeled as a body force expressed as:

CE

Fb   (10)

where  ,  and  are respectively the surface tension, curvature of the interface, and the
AC

volume fraction function. Curvature of the phase interface can be calculated from:


  .( ) (11)


3.2 The Turbulence model

The standard k   model is chosen for the bulk solution of each phase. In this model, k is
the turbulence kinetic energy and  is the turbulence dissipation rate, the standard k  
 ACCEPTED MANUSCRIPT

model equations are equations for isothermal flows expressed as (Launder and Spalding,
1972):

(  k ) 
 .(  kV )  .[(   t )k ]  t S 2   (12)
t k

(  )  C  2
 .( V )  .[(   t ) ]  1 t S 2  C 2  (13)
t  k k

where

T
IP
k2
t  C  


CR
1 u i u j
S  2S ij S ji S ij  (  )
2 x j x i

US
C   0.09 C 1  1.44 C 2  1.92    1.3  k  1
AN
It should be noted that in the present study, the PIMPLE algorithm (OpenCFD, 2009) is used
to compute the coupled flow and turbulence equations. The PIMPLE is a merged PISO-
SIMPLE algorithm in which the PISO is an inner-corrector and the SIMPLE is an outer-
M

corrector. The PIMPLE repeats an iterative process until convergence is reached for
velocities and pressures. In the present work, the residuals for pressure and velocity are set at
ED

and , respectively.

3.3 Finite Volume discretization

PT

Integral over a conjectural control volume V P for a general scalar transport equation leads
CE

to(Robertson et al., 2015):

 
  dV  V .(V  )dV  V .( )dV   S (14)
AC

 t V P P P


where  ,  and S  are respectively the dependent variable, the diffusion coefficient, and the

source term. Table 2 provides a list of the numerical methods in OpenFOAM which is used in
the simulation.
 ACCEPTED MANUSCRIPT

Table2: The numerical methods in OpenFOAM

Discretization Schemes(in-text notation) Description

Time schemes Euler 1st order bounded implicit

Gradient schemes Linear Linear interpolation, 2nd central differencing

Divergence schemes Linear upwind Upwind differencing, 2nd order unbounded

T
IP
4 Results and Discussion

CR
4.1 Model geometry and boundary condition

The 3D flow cell geometry and computational boundary conditions are displayed in Fig. 3.

US
The computational domain is large enough to encompass the entire module up to the
membrane. The liquid and gas flow intermittently enters the module at the inlet. This means
that depending on the desired liquid and gas flow rates, the liquid flows through the inlet up
AN
to a certain time after which the liquid flow is disconnected and the gas is permitted to flow
through the inlet. The liquid and gas flow cycle is then repeated intermittently until a periodic
M

flow structure is obtained in the module. The velocity profile at the inlet of the computational
domain is considered as a fully developed whose distributed value for each phase is set based
ED

on its corresponding flow rate. Since the velocity at the outlet of the domain is not known, a
zero gradient boundary condition is used at the outlet. The Right plate is considered with no
slip condition because the liquid comes in contact with the module wall. For the membrane, a
PT

no slip condition is also applied. The zero velocity for the tangential velocity at the
membrane is valid. Since the permeate flux compared to that of the main flow is very small
CE

(less than 1% of the main flow rate), neglecting the normal velocity to the membrane is a
valid assumption applied in many filtration process simulations reported in the literature (Wei
AC

et al., 2013). The wall shear stress induced by the main flow is calculated based on the
tangent velocity to the membrane. As explained above, this velocity is not altered
significantly because of the small permeate flux, therefore, the wall shear stress and vorticity
calculations performed in this study are not expected to be affected considerably by
neglecting the permeate flux. The time step for the simulations is of the order of s, as
the Courant number is set to be below 0.2. The total simulation time is around 0.2s, by which
time, the bubbles at the inlet complete one period in their trajectory towards the exit.
 ACCEPTED MANUSCRIPT

T
IP
CR
US
AN
M
ED
PT

Figure 3: Computational domain and boundary conditions of the problem

4.2 Grid sensitivity and model verification

CE

A grid independency test is performed for the computational results. Fig. 4 displays the
variation of the magnitude of the shear stress on the horizontal mid-plane
AC

for four computational meshes consisting of 414000, 658800, 946240, 1072952 number of
cells. No significant changes in the results are observed for two mesh sizes of 946240 and
1072952. Hence, a mesh size corresponding to 946240 cells is considered in this study.
 ACCEPTED MANUSCRIPT

T
IP
CR
Figure 4: A grid independence study for a) slug flow b) bubble flow

US
To validate the simulation, a qualitative and quantitative comparison of experimental and
numerical results are displayed in Figs. 5 and 6. The shape and distribution of different size
AN
bubbles along with the velocity of a selected bubble over a period of 0.2 seconds are shown
in these figures. It is seen in the two figures that the bubbles in both flows move in an
oscillating trajectory. This is in contrast to a single rising bubble where the trajectory is a
M

straight line (Wei et al., 2013). A good agreement is found between the results of simulations
and measurements. The average discrepancy between the measured and calculated velocities
ED

for the slug and bubble flows is less than 10.6% and 12.1%, respectively. As it is expected,
the bubble distribution for both numerical and experimental results is log normal where the
PT

distribution is symmetric with respect to the mean diameter. More specifically, the bubbles
diameter in the slug flow ranges from 0.094 to 1.989 cm with 0.547 cm mean diameter from
CE

the experiments (see Fig. 5c). For numerical results, however, the range is between 0.118 and
2.152 cm with a mean diameter of 0.572 cm. The discrepancy of the two results for the mean
diameter is, therefore, less than 5%. Fig. 6 shows that the bubble flow bubble distributions
AC

are different from those of the slug flow. The size of bubbles diameter in the bubble flow for
experimental and numerical results ranges from 0.088 to 0.888 cm with 0.376 cm mean
diameter and from 0.031 to 1.236 cm with 0.314 cm, respectively. A quantitate agreement
between the measurements and simulations is again observed with an error of less than 16%
for the mean diameter. These results indicate a larger range of bubble distribution and a
bigger bubble size for the slug flow compared to those of the bubble flow.
 ACCEPTED MANUSCRIPT

T
IP
CR
US
AN
M
ED
PT
CE
AC

Figure 5: Qualitative and quantitative comparisons of numerical result of slug flow with those of the
experiments over a period of 0.2 seconds. For quantitative comparison, the velocity of the marked bubble from
experiments and simulations are compared with each other. For qualitative comparison, the bubble distribution
from experiments and simulations are compared with each other.
 ACCEPTED MANUSCRIPT

T
IP
CR
US
AN
M
ED
PT
CE
AC

Figure 6: Qualitative and quantitative comparisons of numerical results of bubble flow with those of the
experiments over a period of 0.2 seconds. For quantitative comparison, the velocity of the marked bubble from
experiments and simulations are compared with each other. For qualitative comparison, the bubble distribution
from experiments and simulations are compared with each other.
 ACCEPTED MANUSCRIPT

4.3 Experimental results

In the present study, the slug and bubble patterns are experimentally investigated to study
their effects on the permeate flux. For this purpose, the permeate flux is measured in different
cases and the bubbles distribution in the flow is analyzed by using image processing with the

MATLAB software. The permeate flux (J ) is measured by using following equation:

m
J  (15)
  S  t

T
IP
where m is the mass permeated through the membrane at a t time interval. The time
interval considered for the experiments is 10 s and the permeated mass is measured using a

CR
digital scale.

In order to examine the effectiveness of two-phase flow patterns, the permeate flux is

US
shown in non-dimensional form ( J J ) where J o is the initial permeate flux. As can be seen
o
AN
in Fig. 7, the permeate flux is enhanced significantly by the bubbling treatment. The permeate
flux is increased by nearly 78% and 30% through the slug flow and the bubble flow,
respectively. The slug flow is more effective than the bubble flow.
M
ED
PT
CE
AC

Figure 7: Permeate flux under different two-phase flow patterns

 ACCEPTED MANUSCRIPT

To investigate this phenomenon more clearly, both qualitative and quantitative

comparisons between the results for the slug and bubble flows are displayed in Fig. 8. The
bubble distributions given in Figs.8a and 8b from the experiments helps in a better
understanding of the difference between slug and bubble flows with regard to bubble size and
its distribution in flat sheet modules.

Using the image processing technique, the size and number of bubbles in each flow pattern
are determined and the results of bubble size are plotted in Fig. 8c. A close inspection of this

T
figure reveals that the number of bubbles in the bubble flow (330) is more than three times as

IP
that of the slug flow (107). However, the average diameter of the slug flow (0.547 cm) is
much larger than that of the bubble flow (0.376 cm). As there is a direct relationship between

CR
the size of gas bubbles and the induced wake (Miyahara et al., 1988), the slug flow provide
more turbulence. Furthermore, due to a larger size and a higher buoyancy force, slug bubbles

US
in general have a higher average velocity which in turn results in more collision of bubbles
leading to more turbulence.
AN
M
ED
PT
CE
AC

Figure 8: a) slug flow b) bubble flow c) Bubbles distribution: slug flow; bubble flow
 ACCEPTED MANUSCRIPT

4.4 Numerical results

The numerical model can be used to provide more information regarding various flow
patterns. Fig. 9 shows the shear stress distribution on the whole membrane surface with
respect to bubbles location and their induced vorticity distribution at t=0 and 0.2 seconds for
the slug flow pattern. The same numerical results for the bubble flow pattern are displayed in
Fig. 10. It should be mentioned that the liquid flow rates for the slug and bubble flow are
⁄ and ⁄ , respectively; the gas flow rate, however, is kept constant

T
at ⁄ for both cases. As observed from Figs. 9 and 10, the effects of gas bubbling on

IP
shear stress are significant. The enhanced shear stress is due to two factors: increased velocity

CR
gradient of liquid near the bubbles and induced shear stress by wake region (high vorticity
region). Although the effect of increased velocity gradient is more significant, the effect of
wake region covers a wider area. The figures also show clearly the direct relationship

US
between the size of gas bubbles and induced wake; i.e. the larger the bubble, the more
extensive the wake is. For each flow pattern, it can also be seen that the shear stress
AN
significantly changes in a small period of time. For the bubble flow, a widespread distribution
of smaller size bubbles result in a widespread shear stress in the entire computational domain.
M
ED
PT
CE
AC

Figure 9: The calculated distribution of (a) shear stress on the entire membrane surface (b) bubbles in the flow
cell and (c) vorticity in slug flow
 ACCEPTED MANUSCRIPT

T
IP
CR
US
AN
Figure 10: The calculated distribution of (a) shear stress on the entire membrane surface (b) bubbles in the flow
cell and (c) vorticity in bubble flow
M

The velocity distribution of slug and bubble flows in the middle section of the module (half
way between the membrane and the opposite wall) at t=0 and 0.2 seconds is shown in Fig. 11.
ED

For the slug flow (Fig. 11a), the velocity starts with around zero value at the side walls and
increases to a maximum of 0.2 m/s at the location of slug bubbles. For the bubble flow,
however, a widespread distribution of smaller size bubbles results in a more distributed
PT

velocity across the module with a lower velocity at the location of the bubbles (around 0.1
m/s). Due to buoyancy force, the bubbles at both patterns have higher velocity compared to
CE

the liquid. This velocity gradient causes the vorticity which increases the shear stress. As slug
flow in general has larger bubbles, the vorticty induced by slug bubbles is higher. The
AC

bubbles higher velocity also rises velocity gradient of liquid near the bubbles which increases
shear stress locally.
 ACCEPTED MANUSCRIPT

T
IP
CR
US
AN
Figure 11: The velocity distribution of (a) slug flow (b) bubble flow at t=0 and 0.2 seconds in the middle section
of the module
M

The velocity profile along the x-line (flow direction) and related shear stress induced at the
membrane for slug and bubble flows are shown in Figs. 12 and 13. The tangential velocity at
ED

the wall is zero, as a no slip condition is applied. The fluid velocity is considerably affected
by the bubbles in both flow regimes. The velocity distribution profile is highly influenced
whether the phase is gas or liquid. As the gas phase has a higher velocity (see Fig. 11), the
PT

maximum velocity occurs in this phase. It can also be observed that the velocity magnitude in
cross sections before the presence of the bubble (Figs. 12 and 13 cross sections a to c) has a
CE

lower magnitude compared to those after the bubble (Figs. 12 and 13 cross sections f to g).
Velocity profile is practically uniform across the channel before the two phase zone. Within
AC

the two phase zone, however, due to significant value of the flow vorticity, considerable
fluctuations in velocity profiles are observed. It can be seen that the fluctuation of the
velocity magnitude is larger for the slug flow compared to that of the bubble flow. The shear
stress induced at the membrane is related to velocity gradient at the membrane surface. The
higher velocity gradient in the wake region results in a higher shear stress (see Figs. 12 and
13 part c).
 ACCEPTED MANUSCRIPT

T
IP
CR
US
AN
Figure 12: The slug flow velocity profile at module: (a) coordinate system and notation used in analysis (b)
velocity profile (c) shear stress through the x-line
M
ED
PT
CE
AC

Figure 13: The bubble flow velocity profile at module: (a) coordinate system and notation used in analysis (b)
velocity profile (c) shear stress through the x-line
 ACCEPTED MANUSCRIPT

As water flow rates for the two-phase flow patterns are different, the shear stress should be
shown in a non-dimensional form for a better comparison of the two patterns. The non-

dimensional average shear stress on the membrane (  ) for both slug and bubble flow is
n

displayed in Fig. 14 where  and  n are the average shear stress for gas bubbling and no gas
bubbling cases, respectively. As can be seen in Fig. 14, gas bubbling in general can
considerably increase the average shear stress. The results also show that the slug pattern is

T
more effective in increasing the average shear stress on the membrane. The average shear

IP
stress is increased by the slug flow and the bubble flow by 268% and 152%, respectively.

CR
US
AN
M
ED
PT

.
CE

Figure 14: The relative value of the average shear stress on the entire membrane surface for slug and bubble
flows over no gas bubbling flow over a period of 0.2 seconds
AC

5 Conclusion

In this paper, the multiphase flow patterns through a flow cell were studied using both
experiments and numerical simulations. In the experiments, a video system facilitated with a
high speed camera was used to capture images of the gas bubbling in the liquid flow. Two
flow patterns of slug and bubble flow were investigated. For the numerical simulations, the
 ACCEPTED MANUSCRIPT

OpenFOAM software with the VOF technique was used. The results of numerical model for
the two flow patterns compared well with those of the experiments both qualitatively and
quantitatively. The effectiveness of the two patterns was shown by measuring the permeate
flux through the membrane. Based on the results, the slug flow was found to be more
effective than the bubble flow in increasing the permeate flux. The characteristics of bubbles
in flow such as their size and distribution were also analyzed based on image processing
techniques. The results showed that significant differences can be seen between the two
patterns. For the same gas flow rate, the size of bubbles were larger in the slug flow which in

T
turn enhanced the turbulence and shear force leading for more permeate flux. The numerical

IP
results showed that induced shear stress by wake region and increased velocity gradient are

CR
two factors which enhance the shear stress.

Acknowledgements

US
The authors gratefully thank the Research Institute of Food Science & Technology for
funding this research. The authors also acknowledge the Ferdowsi University of Mashhad
AN
HPC laboratory for their help through the course of simulations performed in this study.
M

Nomenclature

CP Concentration Polarization  stress

ED

J Permeate flux  Surface tension

g Gravity acceleration  Curvature of interface
PT

P Pressure  Dynamic viscosity

t
CE

V Velocity Turbulent viscosity

VOF Volume of fluid  turbulence dissipation rate

AC

S Membrane area k turbulence kinetic energy

UF Ultrafiltration  Density
Fb Body forces
Subscripts
TMP Transmembrane pressure

m Mass interval n No gas bubbling

t Time interval o Initial time

 ACCEPTED MANUSCRIPT

G x , y  Superscripts
Gaussian function

g x , y  Smoothed image

f x , y  average
Regular image

Smoothness control

parameter

T
Greek symbols

IP
 Volume fraction

CR
References

US
BANNARI, R., KERDOUSS, F., SELMA, B., BANNARI, A. & PROULX, P. 2008. Three-
dimensional mathematical modeling of dispersed two-phase flow using class method
of population balance in bubble columns. Computers & chemical engineering, 32,
AN
3224-3237.
BASU, M .2002 .Gaussian-based edge-detection methods-a survey. IEEE Transactions on
Systems, Man, and Cybernetics, Part C, 32, 252-260.
M

BERBEROVIĆ, E., VAN HINSBERG, N. P., JAKIRLIĆ, S., ROISMAN, I. V. & TROPEA,
C. 2009. Drop impact onto a liquid layer of finite thickness: Dynamics of the cavity
evolution. Physical Review E, 79, 036306.
ED

BISCARINI, C., FRANCESCO, S. D. & MANCIOLA, P. 2010. CFD modelling approach

for dam break flow studies. Hydrology and Earth System Sciences, 14, 705-718.
BUSSMANN, M., MOSTAGHIMI ,J. & CHANDRA, S. 1999. On a three-dimensional
PT

volume tracking model of droplet impact. Physics of Fluids (1994-present), 11, 1406-
1417.
CANNY, J. 1986. A computational approach to edge detection. Pattern Analysis and
CE

Machine Intelligence, IEEE Transactions on, 679-698.

CHENG, T.-W. & LI, L.-N. 2007. Gas-sparging cross-flow ultrafiltration in flat-plate
membrane module: Effects of channel height and membrane inclination. Separation
and purification technology, 55, 50-55.
AC

CUI, Z., CHANG, S. & FANE, A. 2003 .The use of gas bubbling to enhance membrane
processes. Journal of Membrane Science, 221, 1-35.
DERRADJI, A., BERNABEU-MADICO, A., TAHA, S. & DORANGE, G. 2000. The effect
of a static mixer on the ultrafiltration of a two-phase flow. Desalination, 128, 223-
.230
DUCOM, G., PUECH, F. P. & CABASSUD, C. 2003. Gas/Liquid Two‐phase Flow in a Flat
Sheet Filtration Module: Measurement of Local Wall Shear Stresses. The Canadian
Journal of Chemical Engineering, 81, 771-775.
 ACCEPTED MANUSCRIPT

ESSEMIANI, K., DUCOM, G., CABASSUD, C. & LINÉ, A. 2001. Spherical cap bubbles in
a flat sheet nanofiltration module: experiments and numerical simulation. Chemical
engineering science, 56, 6321-6327.
FARHANGI, M., PASSANDIDEH-FARD, M. & MOIN, H. 2010. Numerical study of
bubble rise and interaction in a viscous liquid. International Journal of Computational
Fluid Dynamics, 24, 13-28.
FINNIGAN, S. & HOWELL, J. 1989. The effect of pulsatile flow on ultrafiltration fluxes in
a baffled tubular membrane system. Chemical engineering research & design, 6 ,7
.282-278
GONZÁLEZ‐TELLO, P., CAMACHO, F., GUADIX, E., LUZÓN, G. & GONZÁLEZ, P.
2009. Density, viscosity and surface tension of whey protein concentrate solutions.

T
Journal of food process engineering, 32, 235-247.

IP
HAMANN, M. L. 2010. System hydrodynamics to reduce fouling of air-sparged immersed
flat-sheet microfiltration membranes. Stellenbosch: University of Stellenbosch.

CR
HIGUERA, P., LARA, J. L. & LOSADA, I. J. 2013. Simulating coastal engineering
processes with OpenFOAM®. Coastal Engineering, 71, 11.434-9
KHALILI-GARAKANI, A., MEHRNIA, M. R., MOSTOUFI, N. & SARRAFZADEH, M.
H. 2011. Analyze and control fouling in an airlift membrane bioreactor: CFD

US
simulation and experimental studies. Process Biochemistry, 46, 1138-1145.
KRISHNA, R. & VAN BATEN, J. 1 .999Simulating the motion of gas bubbles in a liquid.
Nature, 398, 208-208.
AN
KYLLÖNEN, H., PIRKONEN, P. & NYSTRÖM, M. 2005. Membrane filtration enhanced
by ultrasound: a review. Desalination, 181, 319-335.
LAıN, S., BRÖDER, D., SOMMERFELD, M. & GÖZ, M. 20 .02Modelling hydrodynamics
M

and turbulence in a bubble column using the Euler–Lagrange procedure. International

Journal of Multiphase Flow, 28, 1381-1407.
LAUNDER, B. E. & SPALDING, D. B. 1972. Lectures in mathematical models of
ED

turbulence.
LI, Q., CUI, Z & .PEPPER, D. 1997. Effect of bubble size and frequency on the permeate
flux of gas sparged ultrafiltration with tubular membranes. Chemical Engineering
PT

Journal, 67, 71-75.

LI, Q., GHOSH, R., BELLARA, S., CUI, Z. & PEPPER, D. 1998. Enhancement of
ultrafiltration by gas sparging with flat sheet membrane modules. Separation and
CE

Purification Technology, 14, 79-83.

MATTHIASSON, E. & SIVIK, B. 1980. Concentration polarization and fouling.
Desalination, 35, 59-103.
AC

MILLWARD, H., BELLHOUSE, B., SOBEY, I. & LEWIS, R. 1995. Enhancement of

plasma filtration using the concept of the vortex wave. Journal of membrane science,
100, 121-129.
MIRZAII, I. & PASSANDIDEH-FARD, M. 2012. Modeling free surface flows in presence
of an arbitrary moving object. International Journal of Multiphase Flow, 39, 216-226.
MIYAHARA, T., TSUCHIYA, K. & FAN, L.-S. 1988. Wake properties of a single gas
bubble in three-dimensional liquid-solid fluidized bed. International journal of
multiphase flow, 14, 749-763.
 ACCEPTED MANUSCRIPT

NAJARIAN, S. & BELLHOUSE, B .4996 .Enhanced microfiltration of bovine blood using

a tubular membrane with a screw-threaded insert and oscillatory flow. Journal of
membrane science, 112, 249-261.
NDINISA, N., WILEY, D. & FLETCHER, D. 2005. Computational fluid dynamics
simulations of Taylor bubbles in tubular membranes: model validation and
application to laminar flow systems. Chemical Engineering Research and Design, 83,
40-49.
OPENCFD, O. 2009. The Open Source CFD Toolbox. User Guide, OpenCFD Ltd.
QAISRANI, T. & SAMHABER, W. 2011. Impact of gas bubbling and backflushing on
fouling control and membrane cleaning. Desalination, 266, 154-161.

T
RATKOVICH, N., CHAN, C., BERUBE, P. R. & NOPENS, I. 2009. Experimental study and
CFD modelling of a two-phase slug flow for an airlift tubular membrane. Chemical

IP
Engineering Science, 64, 3576-3584.
REDDY, R. & BANERJEE, R. 2015. GPU accelerated VOF based multiphase flow solver

CR
and its application to sprays. Computers & Fluids, 117, 287-303.
ROBERTSON, E., CHOUDHURY, V., BHUSHAN, S. & WALTERS, D. 20 .45Validation
of OpenFOAM numerical methods and turbulence models for incompressible bluff
body flows. Computers & Fluids, 123, 122-145.
US
SOKOLICHIN, A., EIGENBERGER, G., LAPIN, A. & LÜBERT, A. 1997. Dynamic
numerical simulation of gas-liquid two-phase flows Euler/Euler versus
Euler/Lagrange. Chemical engineering science, 52, 611-626.
AN
TAHA, T., CHEONG, W., FIELD, R. & CUI, Z. 2006. Gas-sparged ultrafiltration using
horizontal and inclined tubular membranes—A CFD study. Journal of membrane
science, 279, 48.494-7
M

TAHA, T. & CUI, Z. 2002a. CFD modelling of gas-sparged ultrafiltration in tubular

membranes. Journal of membrane science, 210, 13-27.
TAHA, T. & CUI, Z. 2002b. Hydrodynamic analysis of upward slug flow in tubular
ED

membranes. Desalination, 145, 179-1.82

TAITEL, Y., BORNEA, D. & DUKLER, A. 1980. Modelling flow pattern transitions for
steady upward gas‐liquid flow in vertical tubes. AIChE Journal, 26, 345-354.
PT

WANG, H.-M., LI, C.-Y., CHEN, S.-J., CHENG, T.-W. & CHEN, T.-L. 2004. Abatement of
concentration polarization in ultrafiltration using n-hexadecane/water two-phase flow.
Journal of membrane science, 238, 1-7.
CE

WEI, P., ZHANG, K., GAO, W., KONG, L. & FIELD, R. 2013. CFD modeling of
hydrodynamic characteristics of slug bubble flow in a flat sheet membrane bioreactor.
Journal of Membrane Science, 445, 15-24.
AC

WIBISONO, Y. 2014. Two-phase flow for fouling control in membranes, Universiteit

Twente.
WIBISONO, Y., CORNELISSEN, E., KEMPERMAN, A., VAN DER MEER, W. &
NIJMEIJER, K. 2014. Two-phase flow in membrane processes: A technology with a
future. Journal of membrane science, 453, 566-602.
WILLEMS, P., KEMPERMAN, A., LAMMERTINK, R., WESSLING, M., VAN SINT
ANNALAND, M., DEEN, N., KUIPERS, J. & VAN DER MEER, W. 2009. Bubbles
in spacers: Direct observation of bubble behavior in spacer filled membrane channels.
Journal of Membrane Science, 333, 38-44.
 ACCEPTED MANUSCRIPT

YANG, J., VEDANTAM, S., SPANJERS, H., NOPENS, I. & VAN LIER, J. B. 2012.
Analysis of mass transfer characteristics in a tubular membrane using CFD modeling.
Water research, 46, 4705-4712.

T
IP
CR
US
AN
M
ED
PT
CE
AC