You are on page 1of 15

Assessment of mass transfer coefcients in coalescing slug ow in vertical

pipes and applications to tubular airlift membrane bioreactors


N. Ratkovich
a,n
, P.R. Berube
b
, I. Nopens
a
a
BIOMATH, Department of Applied Mathematics, Biometrics and Process Control. Ghent University, Coupure Links 653, B-9000, Ghent, Belgium
b
Department of Civil Engineering, The University of British Columbia, 6250 Applied Science Lane, Vancouver, BC, Canada V6T 1Z4
a r t i c l e i n f o
Article history:
Received 8 September 2010
Received in revised form
15 December 2010
Accepted 20 December 2010
Available online 28 December 2010
Keywords:
Side-stream membrane bioreactor
Gasliquid slug ow
Mass transfer coefcient
Le v eque relationship
Non-Newtonian liquid
Laminar regime
a b s t r a c t
One of the operational challenges associated with membrane bioreactors (MBRs) is the fouling of the
membranes. In tubular side-stream MBRs, fouling reduction can be achieved through controlling the
hydrodynamics of the two-phase slug ow near the membrane surface. The two-phase slug ow
induces higher shear stresses near the membrane surface, which generate high mass transfer
coefcients from the surface to the bulk region. However, measuring the mass transfer coefcient is
difcult in complex heterogeneous mixtures like activated sludge and existing techniques (e.g.
electrochemical methods) cannot be applied directly. As an alternative, in this work, a multidisciplinary
approach was selected, by exploiting dimensionless analysis using the Sherwood number. Mass transfer
coefcients were measured at various supercial velocities of gas and liquid ow in a tubular system.
Due to the variability of the mass transfer coefcient obtained for each experimental condition, the
results were compiled into, mass transfer coefcient histograms (MTH) for analysis. A bimodal MTH
was observed, with one peak corresponding to the mass transfer induced by the liquid ow, and the
other peak induced by the gas ow. It was noted that coalescence of bubbles affects the MTH.
Coalescence increased the width of the peaks (i.e. the estimate of the variability of the mass transfer
coefcient) and the height of the peak (i.e. amount of time that a mass transfer coefcient of a given
value is maintained). A semi-empirical relationship based on the Le v eque relationship for the Sherwood
number (mass transfer coefcient) was formulated for the laminar regime. A test case comparison
between water and activated sludge was performed based on full-scale airlift MBR operational
conditions. It was found that the Sherwood number in the non-Newtonian case is 8% higher than that
in the Newtonian case.
& 2010 Elsevier Ltd. All rights reserved.
1. Introduction
1.1. Background of wastewater treatment (water-sludge separation)
Biological wastewater treatment processes consist of two main
steps: the biological removal of organic substances and nutrients
followed by solids-liquid separation. The latter step is either
achieved through gravity (conventional activated sludgeCAS) as
shown in Fig. 1 or by membrane ltration (membrane bioreactors
MBR) as depicted in Fig. 2a and b. Because of the ability of
membranes to effectively remove particulate material, the efuent
quality of MBRs is typically superior to that of CAS. Two types of
MBRs exist: a rst one has the membrane inside the bioreactor,
commonly called immersed membrane bioreactor (iMBR) (Fig. 2a);
the other has the membrane outside the bioreactor, referred to as
side-stream membrane bioreactor (sMBR) (Fig. 2b). The present
study focuses on the latter.
A common problem encountered with MBR systems is the
fouling of the membrane, resulting in a need for frequent membrane
cleaning and replacement (Judd, 2006). As a result, membrane
fouling is often considered to be the main bottleneck of full-scale
application of MBRs, reducing productivity (i.e. the actual provi-
ded puried water ux) and increasing maintenance and opera-
tional cost.
In search for better control of fouling, research to date has
mainly focused on the determination of the components of the
activated sludge that are responsible for fouling. However, it has
been shown that the hydrodynamics near the membrane surface
can play an important role in fouling control. Air is often intro-
duced in the activated sludge ow at the base of the membrane,
to create a gasliquid two-phase ow (Cui et al., 2003). The two-
phase ow that results from air addition scours the membrane
and helps to maintain high permeate uxes and improves mem-
brane rejection characteristics of solids (reduction of fouling),
Contents lists available at ScienceDirect
journal homepage: www.elsevier.com/locate/ces
Chemical Engineering Science
0009-2509/$ - see front matter & 2010 Elsevier Ltd. All rights reserved.
doi:10.1016/j.ces.2010.12.034
n
Corresponding author. Tel.: +32 9 264 59 35; fax: +32 9 264 62 20.
E-mail address: nriosrat@biomath.ugent.be (N. Ratkovich).
Chemical Engineering Science 66 (2011) 12541268
due to (i) the early transition from laminar to turbulent ow,
(ii) an increase in surface shear stress to remove already depos-
ited foulants and (iii) increases the mass transfer between the
cake layer region and the bulk region. However, the detailed
mechanisms that govern fouling control resulting from air addi-
tion are not yet completely understood. As a result, a trial and
error approach is typically used to identify the optimal air
addition conditions for fouling control.
In tubular systems with multiphase ow, such as with airlift
tubular membrane systems, different ow patterns can be found
(i.e. stratied, slug and annular ow), each of which induce
different mass transfer rates at the surface of the tube. Multiphase
ow models have been developed to predict the ow pattern and
how the hydrodynamic properties (i.e. velocities of gas and liquid,
lm thickness, void fraction, etc.) affect the mass transfer coef-
cients at surfaces (Adsani et al., 2006; Chen et al., 2006; Shirazi
et al., 2004; Drew, 1983; Cognet et al., 1978). A number of studies
have focused on characterizing mass transfer induced by single
slug ow in tubular systems for Newtonian liquids (Zheng and
Che, 2007; Ghosh and Cui, 1999; Esteves and De Carvalho, 1993;
Nakoryakov et al., 1987). However, limited information is avail-
able for more complex systems, characterized by coalescing slugs
(Pinto et al., 1998) and non-Newtonian liquids (Xu et al., 2007;
Rosehart et al., 1975), such as sMBR. The mass transfer is important
since has been linked to surface shear stress, which has been
previously linked to fouling control (Laborie et al., 1998).
Experimental tools, such as electrochemical probes that can be
mounted ush to a surface can be used to measure the mass trans-
fer induced by two-phase ow at a surface (Rehimi et al., 2008).
Unfortunately, the electrochemical approach can only be used in
systems containing a specic electrochemical solution, and therefore
cannot be used to estimate mass transfer in systems containing more
complex solutions, such as activated sludge. However, it should be
possible extrapolate the mass transfer measurements obtained using
a specic electrochemical solution, to estimate mass transfer in
more complex solutions using non-dimensional relationships (i.e.
Sherwood number).
The present paper focuses on: (i) measuring the mass transfer
coefcient of slug ow in vertical pipes with a diameter similar to
those of the airlift tubular membranes of interest, (ii) character-
izing the effect of bubble coalescence on the mass transfer coe-
fcient and (iii) extrapolating the mass transfer measurements
obtained using a specic electrochemical solution to estimate the
mass transfer in more complex solutions (i.e. non-Newtonian)
that are characteristic of the activated sludge in an sMBR. All the
above will be achieved using non-dimensional numbers.
1.2. Background of the mass transfer coefcient
1.2.1. Single-phase ow
The Sherwood number (Sh), which is the ratio of convective to
diffusive mass transport, for single-phase ow in a pipe depends
on the ow regime (Oliveira et al., 2001; De and Bhattacharya,
1997). For laminar and turbulent conditions, the Sherwood
number is dened by Eqs. (1) and (2), respectively:
Laminar Rer2000 Sh 1:62 Re Sc
d
L
_ _
1=3
Lev eque relationship
1
Turbulent Re44000 Sh 0:023Re
0:8
Sc
1=3
2
where d is the equivalent hydraulic diameter (m) and L is the pipe
length (m), Re is the Reynolds number, which is the ratio of
inertial forces to viscous forces, Sc is the Schmidt number, which
is dened as the ratio of momentum and mass diffusivity. These
numbers are dened by Eqs. (3)(5), respectively:
Re
r
l
u
i
d
m
l
3
Sh
k
m
d
D
f
4
Sc
m
l
r
l
D
f
5
where r
l
is the density of the liquid (kg m
3
), u
i
is the velocity
(m s
1
), m
l
is the viscosity of the liquid (Pa s), D
f
is the diffusion
coefcient (m
2
s
1
) and k
m
is the mass transfer coefcient (m s
1
).
It is important to highlight that Eq. (2) is only valid for smooth pipes
(Adsani et al., 2006).
Re-writing Eqs. (1) and (2) in terms of the mass transfer coef-
cient using Eq. (4) yields Eqs. (6) and (7) for laminar and turbulent
ow, respectively:
Laminar Rer2000 k
m
1:62 ReSc
d
L
_ _
1=3
D
f
d
_ _
6
Turbulent Re44000 k
m
0:023Re
0:8
Sc
1=3
D
f
d
_ _
7
1.2.2. Specicities of a two-phase ow
In a slug ow pattern, three distinctive zones can be distin-
guished (Fig. 3a): (i) the falling lm zone (ff), where the bubble is
passing, (ii) the wake zone (w), which is just behind the bubble
(here, mixing between the liquid and the gas takes place) and
(iii) the liquid zone (ls) (Ghosh and Cui, 1999). Each zone has a
distinct mass transfer coefcient (Fig. 3b).
As presented in Fig. 3b, for the liquid zone, the magnitude of
the mass transfer coefcient is lower than that for the falling lm
and wake zones. Ghosh and Cui (1999) found that the mass
transfer coefcient values in the falling lm and wake zones were
approximately 30 times higher than those in the liquid zone.
These results indicate that the overall mass transfer induced by
two-phase ow is higher than that for single-phase (i.e. liquid)
ow. Ghosh and Cui (1999) proposed equations to estimate the
mass transfer coefcient for each zone, based on hydrodynamic
models and mass balances of slug ow. For the different zones in
Fig. 1. Conventional Activated Sludge (CAS) systems.
Fig. 2. (a) Immersed MBR, (b) side-stream MBR (Judd, 2006).
N. Ratkovich et al. / Chemical Engineering Science 66 (2011) 12541268 1255
the slug ow, the mass transfer coefcient is dened by Eq. (8):
k
m_i
a
1
Re
a
2
i
Sc
a
3
d
i
L
_ _
a
4
D
f
d
i
_ _
8
where the subindex i is for each slug zone (falling lm, wake and
liquid slug). The list of parameters for Eq. (8) can be found in Table 1.
In Table 1 symbols a
TB
, a
LS
, u
m
, g, r
L
and r
G
are the void fraction of
the slug ow, the void fraction of the gas in the liquid slug, the
mixture velocity (m s
1
), the gravity acceleration (9.81 m
2
s
1
)
and the liquid and gas densities, respectively. It is assumed that a
LS
is 0 as there are no bubbles on the liquid slug (Chang and Fane,
2000). The parameters a
1
, a
2
, a
3
and a
4
correspond to the coefcients
and exponents presented in Eqs. (6) and (7). A comprehensive dis-
cussion and the derivation of the axial velocity (u
axial
) is presented
elsewhere (Ghosh and Cui, 1999).
Eq. (8) does not consider buoyancy forces. Zheng and Che (2007),
Zheng and Che (2006) proposed the addition of a semi-empirical
correction factor (C) to Eq. (8), to take into consideration the buoy-
ancy forces (i.e. gas rises faster than liquid) and the slug length. The
resulting relationship is presented in Eqs. (9) and (10):
k
m_i
C a
1
Re
a
2
i
Sc
a3
d
i
L
_ _
a4
D
f
d
i
_ _
9
C a 1
Fr
m
b
_ _
b
10
where a and b are constants with a value of 0.27 and 0.5, res-
pectively (Zheng and Che, 2006), b is the ratio of the gas slug length
(Fig. 5) to the sum of the gas and liquid slug lengths (b L
TB
=
L
TB
L
ls
) and Fr
m
is the mixture Froude number, representing the
turbulence intensity and is dened in Eq. (11):
Fr
m

u
m
gd
0:5
11
An increase in the Froude number indicates more gas entrain-
ment from the gas slug nose and more gas shedding off the wake
zone (turbulent eddies), further increasing the mass transfer in
the falling lm zone (Zheng and Che, 2006), which will be reec-
ted by C41.
2. Materials and methods
2.1. Description of the setup
A schematic of the setup is presented in Fig. 4. A plexiglas pipe
with a length of 2 m and an inner diameter of 0.0099 m was used.
This inner diameter was similar in geometry to those of the airlift
tubular membranes of interest. A co-centric ow cell of similar
inner diameter to that of the pipe is located in the middle of the
plexiglas pipe (1 m). Two electrochemical shear probes, which are
used to measure surface shear stresses, are located in the ow
cell. A temperature controlled water bath (20 1C) is used to
maintain the temperature of the electrolyte solution owing
through the system constant. A peristaltic pump (Masterex LS,
USA) is used to recirculate the electrolytic solution from the gas
liquid separator tank to the plexiglass pipe at controlled liquid
ow rates. Two ow metres (Cole-Parmer, N082-03, USA) are
used to monitor the liquid and gas ow rates. Five liquid ow
rates (0.1, 0.2, 0.3, 0.4 and 0.5 L min
1
) and three gas ow rates
(0.1, 0.2 and 0.3 L min
1
) were investigated, resulting in a total of
15 combinations. From now on, the ow rates will be expressed
as supercial liquid (0.022, 0.043, 0.065, 0.087 and 0.108 m s
1
)
and gas (0.022, 0.043 and 0.065 m s
1
) velocities. In a full-scale
airlift system, the sludge and air ow rates typically range
between 1220 and 510 m
3
h
1
, respectively. The sMBR mod-
ules have 700 pipes with an internal diameter of 0.0052 m
Fig. 3. Different (a) zones and (b) mass transfer coefcients of a slug ow (Ghosh
and Cui, 1999).
Table 1
Proposed parameters expression by Ghosh and Cui (1999) model.
Falling lm Wake Liquid slug
Laminar
(Rer2000)
Turbulent
(Re44000)
a
1
0.023 0.023 1.62 0.023
a
2
0.8 0.8 0.33 0.8
a
3
0.33 0.33 0.33 0.33
a
4
0 0 0.33 0
Re
i
Re
ff
(Eq. (28))
r
L
uaxial d
m
L
Re
m
(Eq. (29))
u
i
(m s
1
)
9:916gd1a
TB

0:5

0:5
u
axial
u
m
u
SL
+u
SG
d
i
(m) 0:5d1a
0:5
TB
d d
r
i
(kg m
3
) r
L
r
L
(1a
LS
)
+r
G
a
LS
r
L
(1a
LS
)
+r
G
a
LS
m
i
(Pa s) m
L
m
L
(1a
LS
)
+m
G
a
LS
m
L
(1a
LS
)
+m
G
a
LS
N. Ratkovich et al. / Chemical Engineering Science 66 (2011) 12541268 1256
(Futselaar et al., 2007). Therefore for each pipe, the supercial
liquid and gas velocities typically range between 0.220.37 and
0.090.17 m s
1
, respectively. If each pipe has a diameter of
0.0099 m instead of 0.0052 m, as it was the case of the experimental
setup presented in this work, the liquid and gas supercial velocities
will range between 0.060.10 and 0.03 and 0.05 m s
1
, respectively.
For each experimental condition, surface shear stresses are mea-
sured for a period of 10 s, and recorded at a frequency of 1000 Hz
(Chan et al., 2007). All experimental conditions are replicated six
times. Of these, three were used to calibrate the model and the
remaining three to validate the model. The validation measurements
were performed 1 week after the calibration measurements and a
new electrolyte solution was used.
The electrochemical probes are made from two platinum
wires, imbedded ush to the inside surface of ow cell. A detailed
description of the directional electrochemical probes is presented
in Chan et al. (2007).
2.2. Mass transfer coefcient and shear stress measurements
Measurements from the directional electrochemical probes are
in volts and converted to mass transfer coefcients using Eqs. (12)
(Rosant, 1994):
k
m_probe

4V
v
e
Fpd
2
e
C
o
RG
12
where k
m_probe
is the probe mass transfer coefcient (m s
1
), V is
the voltage signal (V), R is the resistance (100 O), G is the
amplier gain (1000), v
e
is the number of electrons involved in
the reaction (1), F is the Faraday constant (96 500 C mol
1
),
d
e
is the diameter of the probe (m) and C
o
is the bulk concentra-
tion of ferricyanide (3 mol m
3
), The mean wall velocity gra-
dient (or shear rate, _ g (s
1
)) and the shear stress (t
w
(Pa)) can be
estimated using Eqs. (13) and (14), respectively (Rosant, 1994):
_ g
d
e
D
2
f
k
m_probe
0:862
_ _
3
13
t
w
m
l
_ g 14
where D
f
is the diffusion coefcient of ferricyanide (7.1410
10
m
2
s
1
(Rosant, 1994)) and m
l
is the dynamic viscosity of the
solution (0.001 Pa s). It is important to note that the following
assumptions were made in deriving Eqs. (12) and (13): (i) the
axial diffusion is negligible, (ii) the velocity gradient normal to
surface is negligible relative to that parallel to the surface, (iii) a
linear velocity gradient exists in the region of the probe surface,
and (iv) a quasi-steady-state condition exists near probe surface
(Rosant, 1994).
2.3. Viscosity of non-Newtonian liquids
Viscosity (m) is a property that inuences the hydraulic regime
and transport phenomena. It is dened as the ratio between shear
stress (t) and shear rate ( _ g). The viscosity of Newtonian liquids
(e.g. water) exhibits a linear shear stress and shear rate relation-
ship and hence, a constant viscosity. However, some particulate
suspensions (e.g. activated sludge in sMBR), exhibit pseudoplastic
(non-Newtonian) behaviour (Laera et al., 2007; Pollice et al.,
2007; Rosenberger et al., 2006), where the shear stress can be
related to the shear rate according to a power-law relationship as
presented in Eq. (15):
t
w
m_ g
n
15
where m is the ow consistency index (Pa s
n
) and n is the ow
behaviour index (). For activated sludge in MBR, Rosenberger
et al. (2006) proposed empirical models for m and n as function of
the total suspended solids (TSS), as presented in Eqs. (16) and (17)
respectively:
m0:001exp2TSS
0:41
16
n 10:23TSS
0:37
17
2.4. Wall friction (shear stress)
The friction factor (f) is used in internal ow calculations to
expresses the linear relationship between mean ow velocity and
shear stress at the wall of a pipe (White, 2002), as presented in
Eq. (18):
t
w

f r
L
u
2
i
8

d
i
DP
4DL
18
where r
L
is the density of the liquid, u
i
is the velocity, d
i
is
hydraulic equivalent diameter, DP is the pressure drop (Pa) and
DL is the section length (m). The friction factor for a uid owing
in a pipe is dened as function of the Reynolds number and the
pipe roughness (e (m)). Depending on the ow regime, laminar
(f
lam
) or turbulent (f
tur
), the friction factor is dened using
Eqs. (19)(21):
Laminar Reo2000 : f
lam
64Re
1
19
f
tur
0:316Re
0:25
Blasius relationship 20
Turbulent Reo4000 : f
tur
0:25 log
10
e
3:7d

5:74
Re
0:9
_ _ _ _
2
21
Eq. (20) is used for smooth pipes (eo110
6
m) such as the
Plexiglas pipes used in the present study (Perry and Green, 1999),
whereas Eq. (21) is valid for rough pipes. Note, that no relation-
ships exists to accurately estimate the friction factor in the
transition regime (2000oReo4000). However, a weighting fac-
tor can be used to determine the friction factor in the transition
regime as suggested by Cheng (2008).
Fig. 4. Description of the electrochemical shear measurement setup (Ratkovich
et al., 2009).
N. Ratkovich et al. / Chemical Engineering Science 66 (2011) 12541268 1257
2.5. Slug ow hydrodynamics
The gas slug rising velocity (u
TB
) is important because it
signicantly affects the velocity eld of the slug unit due to the
slippage of the phases and pressure drop (Fig. 5).
A great deal of research has been performed to determine this
parameter (Sousa et al., 2007; Nogueira et al., 2006a, 2006b;
Sousa et al., 2005; Dziubinski, 2003; Bugg and Saad, 2002; Pinto
et al., 2001, 1998; Collins et al., 1978). Typically, Eq. (22) can be
used to estimate the velocity of rising gas slugs in vertical tubes
(Polonsky et al., 1999; van Hout et al., 2002):
u
TB
C
TB
u
m
k
TB
gd
r
l
r
g

r
l
_ _
0:5
22
where the rst term on the right hand side is related to the bubble
rising in a moving liquid and the second term is related to the
bubble rising in a stagnant liquid. In the rst term, u
m
is the
mixture velocity, which is the sum of the liquid (u
SL
) and gas (u
SG
)
supercial velocities. The parameter C
TB
depends on the velocity
prole of the rising bubble. In the second term, k
TB
is function of
the Froude number.
In previous work (Ratkovich et al., 2010), the values of C
TB
and
k
TB
were estimated to be 1.07770.014 and 0.27770.068, respec-
tively (error range corresponds to the 95% condence interval
of the estimated parameters) using a high speed camera (HSC)
for water, activated sludge and carboxymethyl cellulose (non-
Newtonian liquid).
Hydrodynamic models of slug ow based on mass and momen-
tumconservation have been described in literature (Chang and Fane,
2000; Fabre and Line, 1992; Mao and Dukler, 1985; Fernandes et al.,
1983). These models assume that the slug ow is axisymmetric,
one-dimensional and steady. They are based on equations for mass
balance and liquid-gas velocities as described by Chang and Fane
(2000): the velocity of the falling lm (u
ff
), assuming that it does not
contain gas and the bubble has a cylindrical shape, can be estimated
using Eq. (23); the gas/liquid volume ratio entering the system per
unit of time can be estimated using Eq. (24); for the liquid phase, the
rate of liquid ow approaching the nose of the slug equals that
entering the falling lm and can be estimated by Eq. (25); the
thickness of the falling liquid lm (d
L
) can be estimated using
Eq. (26) (Zheng and Che, 2007):
u
ff
9:916gd1a
0:5
TB

0:5
23
u
SG
u
TB
ba
TB
24
u
TB
u
m
u
TB
u
ff
1a
TB
25
d
L
0:5 d1a
0:5
TB
26
It has been suggested that the void fraction of the gas in the
liquid slug (a
LS
) can be neglected (Section 1.2.2). Therefore, this
term does not appear in Eqs. (23)(26).
Once the supercial liquid and gas velocities are known for a
given slug ow unit, it is possible to determine the other
velocities of the system (gas slug rising velocity, mixture velocity,
falling lm velocity, etc.) by simultaneously solving Eqs. (23)(26)
which contain four unknowns (u
ff
, b, a
TB
, d
L
). Once all velocities
are determined, the Reynolds numbers of the falling lm (Re
ff
)
and the mixture (Re
m
) can be calculated using Eqs. (27) and (28),
respectively:
Re
ff

4r
l
u
ff
d
L
m
l
27
Re
m

r
l
u
m
d
m
l
28
where the factor 4 in Eq. (27) is an estimation of the equivalent
hydraulic diameter for the falling lm zone (Zheng and Che,
2007). To ensure that the slug ow is stable (bubbles completely
developed), the entry length must be sufciently long. This length
can be estimated using Eq. (29) (Ghosh and Cui, 1999):
L
e
40:6 d
u
m
gd
0:5
0:22
_ _
29
2.6. Pressure drop
2.6.1. Single-phase ow
The total pressure drop (DP
total
) in a pipe can be estimated
using Eq. (30) (White, 2002):
DP
total
DP
static
DP
frict
30
where DP
static
is the static pressure drop due to the head of the
liquid (r
l
gDL) and DP
frict
is the frictional pressure drop calcu-
lated from Eq. (18).
2.6.2. Two-phase ow
For two-phase ow, Eq. (30) can be used as well to calculate
the total pressure drop. However, the static pressure drop is
calculated based on the relationship proposed by Thome (2008),
which is exclusively a function of the density of the liquid and gas,
void fraction and surface tension (Thome, 2008). This relationship
Fig. 5. Illustration of the different velocities in a slug ow unit.
N. Ratkovich et al. / Chemical Engineering Science 66 (2011) 12541268 1258
does not depend on viscosity. Therefore, it can be used for both
Newtonian and non-Newtonian liquids.
The frictional pressure drop for Newtonian liquids can be
calculated based on the method proposed by Shannak (2008).
For non-Newtonian liquids, it can be calculated based on the
method proposed by Fidos et al. (2008) and Dziubinski (1995).
A comprehensive discussion on the derivation of the frictional
pressure drop for two-phase ow is presented elsewhere
(Shannak, 2008; Fidos et al., 2008; Dziubinski, 1995).
2.7. Mass transfer coefcient
The mass transfer coefcient for Newtonian liquids (e.g. water)
was introduced earlier in Section 1.2. A short overview of the
mass transfer coefcient for non-Newtonian liquid is given here.
2.7.1. Single-phase ow
For non-Newtonian uids (Ranjan et al., 2004a), the mass
transfer coefcient for laminar regime can be estimated using
Eq. (31):
k
m
1:016
3n1
n
_ _
1=3
Reu Scu
d
L
_ _
1=3
D
f
d
_ _
31
where Re
0
is the Metzner and Reed Reynolds (Reu r
l
u
i
d=m
a
) and
Sc
0
is the Schmidt number (Scu m
a
=r
l
D
f
) for non-Newtonian
uids. For turbulent regime, a comprehensive discussion on the
derivation of the mass transfer coefcient is presented elsewhere
(Ranjan et al., 2004b).
The apparent viscosity (m
a
) used to estimate the mass transfer
coefcient for non-Newtonian uids can be estimated using
Eq. (32):
m
a
m _ g
n1
k
3n1
4n
_ _
n=n1
8u
i
d
_ _
_ _
n1
m
3n1
4n
_ _
n
8u
i
d
_ _
n1
32
The product of the Reynolds and Schmidt numbers is the ratio
of the rate of advection by the ow to the rate of diffusion. This is
also known as the Pe clet number (Pe), as presented in Eq. (33):
Pe Reu Scu
u
i
d
D
f
33
It is possible to observe that in Eq. (33), there is no viscosity
term, due to the fact that it cancels out in the product of the
Reynolds and Schmidt numbers. However, in Eq. (31), the ow
consistency index (n) should be dened to include the non-
Newtonian behaviour in the calculation of the mass transfer
coefcient. It is important to highlight that when n1 (New-
tonian case), Eq. (31) reduces to Eq. (6).
The ratio between the mass transfer coefcient in laminar
regime for non-Newtonian (Eq. (31)) and Newtonian (Eq. (6))
equals 3n1=4n
1=3
. Hence, the mass transfer coefcient in a
non-Newtonian uid will typically be higher than that in a
Newtonian uid. For two-phase ow, the same approach pre-
sented in Section 1.2.2 can be used.
3. Results and discussion
3.1. Flow regimes
The ow regime for single-phase and two-phase ow for
water-N
2
based on the Reynolds number was rst determined.
For single-phase ow, the Reynolds number (Eq. (3)) was between
200 and 1200, implying the ow was in laminar regime. For the
two-phase ow, two Reynolds numbers need to be calculated,
based on the falling lm and mixture Reynolds numbers (Eqs. (27)
and (28)). For the gas slug zone (falling lm) the Reynolds number
was between 720 and 780 implying laminar ow. For the liquid
slug zone (mixture) the Reynolds number was between 400 and
1800, again implying laminar regime. Based on these ndings, a
laminar approach was used. It is important to highlight that the
model proposed by Zheng and Che (2006) was developed for the
turbulent regime. However, a similar approach can be followed
for laminar regime based on the experimental data that was
collected in the present study.
Regarding the entry length, which occurs after 0.35 m based
on Eq. (29) for the experimental setup and conditions investi-
gated, the gas slugs were fully developed before reaching the
probes located at 1 m. This also proves that the relationships in
Section 2 can be used as they are designed for slug ow which is
completely developed.
3.2. Mass transfer coefcient in single-phase ow
It is important to highlight that the mass transfer estimated
using the electrochemical approach (Eq. (12)) corresponds to that
of an incompletely developed mass transfer boundary layer for
the equivalent hydraulic diameter of the shear probe and not of
the pipe diameter. The mass transfer for a fully developed mass
transfer boundary layer can be estimated from the value obtained
with the electrochemical approach by considering that the Sh
tu-
be
Sh
probe
yielding Eq. (34) (Wang et al., 2002):
k
m
k
m_probe
d
e
d
34
Since the ow is laminar, the relationship for laminar regime
was used (Eq. (6)). Measurements were rst performed in single-
phase ow in order to determine whether the experimental data
could be modelled using Eq. (6) (Fig. 6). For the experimental
conditions that were used, the voltage was obtained from the
experimental setup and converted to a mass transfer coefcient
using Eq. (12).
From Fig. 6, it is possible to observe that the mass transfer
estimated using Eq. (6) slightly under-predicted the experimental
data (8%). Adding the correction factor C, that was introduced in
Eq. (9), a corrected mass transfer estimate was developed as presen-
ted in Eq. (35):
k
m
C 1:62 ReSc
d
L
_ _
1=3
D
f
d
i
_ _
35
Using SPSS v15 (SPSS, USA) Eq. (35) was tted to the experi-
mental data and the coefcient C was estimated to be 1.0707
0.013 (95% condence interval of the estimated parameter) (Fig. 6).
Fig. 6. Mass transfer coefcient vs. liquid velocity comparison for the experi-
mental data, Le v eque correlation (Eqs. (6) and (35)) for single-phase ow.
N. Ratkovich et al. / Chemical Engineering Science 66 (2011) 12541268 1259
3.3. Mass transfer coefcient in two-phase vertical pipe ow
Typical voltage results obtained using the electrochemical
approach, and the corresponding mass transfer coefcient, are
presented in Fig. 7.
In Fig. 7, the valleys in the mass transfer coefcients are caused
by the passage of gas slugs whereas the plateaus are caused by the
passage of liquid slugs. Gas slugs rising in the vertical pipes were
observed to periodically coalesce when trailing slugs reached the
wake of the leading slugs, accelerating the trailing slugs to nally
coalesce with the leading slug. For this reason, the mass transfer
coefcients induced by successive slugs are never exactly the same.
As a result, the prole of mass transfer coefcients in successive
shear events, induced by rising gas slugs, varied considerably over
time, making comparison among different proles difcult.
3.4. Mass transfer histograms
Mass transfer coefcient histograms (MTH) were introduced to
explore the effect of the different experimental conditions inves-
tigated on the distribution of mass transfer coefcients. The MTHs
summarize two important features: (i) the range of mass transfer
coefcients that the membrane surface experiences and (ii) the
relative frequency of occurrence of a particular mass transfer
coefcient. MTH for the different experimental conditions inves-
tigated are shown in Fig. 8af.
The selection of the bin size for the histograms was based on
the resolution necessary to clearly separate the mass transfer
coefcients measured for the ve single-phase liquid velocities
considered. It was found that a bin size of 210
7
m s
1
could
be used to effectively separate the peaks for the different experi-
ments in single-phase ow (Fig. 8a). Larger bin sizes causes peaks
to join together in a single peak making it difcult to distinguish
the individual peaks.
For single-phase ow (Fig. 8a), it was possible to observe a
single peak for each liquid velocity. Theoretically, each peak
should have a frequency of 1, but this was not the case, especially
for the lowest liquid velocity investigated (0.022 m s
1
). This
was attributed to the pulsating ow generated by the peristaltic
pump used in the present study. For Fig. 8bf, it is possible to
clearly distinguish two peaks in the MTH (bimodal distribution): a
rst peak occurred around 2.510
6
m s
1
, which was caused
by the passage of liquid slugs and a second peak, occurred at
510
6
m s
1
, which was caused by the passage of gas slugs.
The magnitude of the frequency for both peaks was different for
the different gasliquid velocities investigated. In Fig. 8b, it is
possible to observe that low liquid velocities (increasing the gas
velocity) tend to decrease the frequency of the rst peak (liquid
slug) and increase the frequency of the second peak. In Fig. 8c, the
liquid velocity was twice as large as that presented in Fig. 8a, but
there was just an increase of 15% in the rst peak in terms of
frequency. However, the second peak decreased by 40%. Fig. 8df
reveals that the rst peak further increases due to the further
increase as the liquid velocity increases, whereas the second peak
further decreases.
3.5. Bimodal MTH
Ochoa et al. (2007) have suggested that fouling control can be
correlated to the variability in the mass transfer coefcient.
Therefore, conditions that maximize the magnitude of both peaks
(liquid and gas slugs) are likely to promote better fouling control
(Ochoa et al., 2007). To identify the conditions that equilibrate the
magnitude of these two peaks, an empirical relationship was used
to model the bimodal MTH for the different experimental condi-
tions investigated based on the Gaussian distributions presented
in Eq. (36):
f k
m

A
l
w
l

0:5p
p exp
2k
m
k
m_1

2
w
2
l
_ _
..
Liquid slug

A
g
w
g

0:5p
p exp
2k
m
k
m_g

2
w
2
g
_ _
..
Gas slug
36
where the rst term on the right hand side describes the rst peak
in the MTH, which is induced by the liquid slug, and the second
term describes the second peak in the MTH, which is induced by
the gas slug. The parameters A
i
=w
i

0:5p
p
correspond to the height
of each peak (i.e. frequency), k
m_i
corresponds to the position of
the centre of each peak (i.e. average mass transfer coefcient),
and w
i
corresponds to the width of each peak (i.e. standard
deviation of the mass transfer coefcient).
From Eq. (36), there are six empirical constants that need to be
identied (A
l
, A
g
, w
l
, w
g
, k
m_l
and k
m_g
) from the experimental data.
Eq. (36) was tted to the MTH using SPSS v15 (IBM, USA) for each
of the 15 experimental conditions investigated. The estimated
values of the Gaussian distribution parameters for each experi-
mental condition investigated are presented in Table 2. The R
2
was more than 0.84 for all experimental conditions, indicating
that Eq. (36) could be used to effectively model the MTH.
3.6. Semi-empirical model based on MTH
Zheng and Che (2006) presented a semi-empirical model to
determine the mass transfer coefcient based on the different
Fig. 7. Typical raw voltage signal for the two probes (top), conversion to shear
stress (Eq. (3)) (middle) and mass transfer coefcient (Eq. (39)) (bottom) for a
water-N
2
combination of 0.0220.022 m s
1
during a period of 2 s. (Note that for
clarity, schematics of the slugs observed to generate the signals are presented in
the top of the gure.)
N. Ratkovich et al. / Chemical Engineering Science 66 (2011) 12541268 1260
zones of the slug ow (Fig. 3). In the present study, only two
zones could be identied based on the MTH. Therefore, a similar
analysis to the one performed by Zheng and Che (2006) was
made; however, only two zones were considered: (i) gas slug
zone (i.e. falling lm and wake zone) and (ii) liquid slug zone
based on the Le v eque relationship (laminar regime based on the
Fig. 8. MTH for (a) single-phase ow and for ve different liquid velocities, (b) 0.022, (c) 0.043, (d) 0.065, (e) 0.087 and (f) 0.108 m s
1
with three gas velocities
combinations.
Table 2
Gaussian distribution for the bimodal distributions for each experimental condition investigated with the standard deviation.
u
SL
(m s
1
) u
SG (
m s
1
) k
ml
10
6
(m s
1
) w
l
10
6
(m s
1
) A
l
10
6
(m s
1
) k
mg
10
6
(m s
1
) w
g
10
6
(m s
1
) A
g
10
6
(m s
1
) R
2
0.022 0.000 1.51772.4E04 0.13379.7E04 0.24872.1E03 0.863
0.043 0.000 1.82876.3E04 0.12075.3E04 0.16572.6E04 0.961
0.065 0.000 2.05573.9E04 0.13175.7E04 0.20777.6E04 0.807
0.087 0.000 2.22471.3E04 0.12471.1E04 0.16473.2E05 0.960
0.108 0.000 2.40573.6E05 0.13572.4E05 0.16672.9E05 0.950
0.022 0.022 2.11472.5E02 0.67076.8E02 0.09471.4E02 3.97472.7E01 2.67577.6E01 0.11874.1E02 0.923
0.043 0.022 2.15572.2E02 0.59075.9E02 0.09071.3E02 3.46572.4E01 2.36175.0E01 0.11473.2E02 0.937
0.065 0.022 2.23871.4E02 0.49373.9E02 0.09371.1E02 3.06471.5E01 1.60672.3E01 0.09871.8E02 0.970
0.087 0.022 2.46079.3E03 0.60273.5E02 0.13571.8E02 3.24872.6E01 1.36573.2E01 0.06172.1E02 0.990
0.108 0.022 2.46978.6E03 0.56973.4E02 0.12972.0E02 3.14472.1E01 1.15972.3E01 0.06872.3E02 0.994
0.022 0.043 2.81276.9E02 1.40671.4E01 0.13071.1E02 5.38572.0E02 0.33173.7E02 0.05275.3E03 0.849
0.043 0.043 2.56674.3E02 0.76079.1E02 0.09771.3E02 4.69271.9E01 2.07474.3E01 0.11071.8E02 0.843
0.065 0.043 2.63771.9E02 0.76275.4E02 0.09871.1E02 4.22371.7E01 2.19073.0E01 0.10771.4E02 0.971
0.087 0.043 2.63471.4E02 0.65674.5E02 0.09671.2E02 3.70071.5E01 1.82472.0E01 0.10471.4E02 0.982
0.108 0.043 2.66671.2E02 0.66374.1E02 0.11671.4E02 3.67472.0E01 1.66572.5E01 0.08371.7E02 0.986
0.022 0.065 2.87776.3E02 1.21271.3E01 0.10479.3E03 5.51871.4E02 0.35172.8E02 0.07575.1E03 0.906
0.043 0.065 2.89275.6E02 1.21471.1E01 0.12479.9E03 5.45272.6E02 0.45375.2E02 0.06176.0E03 0.867
0.065 0.065 2.96074.3E02 1.07379.1E02 0.12979.3E03 5.21378.8E02 1.20871.9E01 0.07679.8E03 0.882
0.087 0.065 2.80872.0E02 0.74375.5E02 0.09971.2E02 4.39571.7E01 2.08273.0E01 0.10671.5E02 0.965
0.108 0.065 2.83671.7E02 0.76275.4E02 0.10671.4E02 4.15672.0E01 2.01173.0E01 0.09771.7E02 0.978
N. Ratkovich et al. / Chemical Engineering Science 66 (2011) 12541268 1261
Reynolds number calculated in Section 3.1) using Eq. (37):
k
m
1:62a 1
Fr
m
b
_ _
b
Re
i
Sc
d
i
L
_ _
1=3
D
f
d
i
_ _
37
The semi-empirical model (Eq. (37)) was calibrated based on
the data presented in Table 2 for the average mass transfer
coefcient for the liquid (k
m_l
) and for the gas (k
m_g
) slugs. The
resulting relationships to estimate the average mass transfer
coefcient for the liquid and gas slugs are presented in Eqs. (38)
and (39), respectively:
k
m_l
1:62a 1
Fr
m
b
_ _
b
Re
m
Sc
d
L
_ _
1=3
D
f
d
_ _
38
k
m_g
1:62a 1
Fr
m
b
_ _
b
Re
ff
Sc
4d
L
_ _
1=3
D
f
4d
_ _
39
The parameters a and b were identied using SPSS v15 based
on the experimental data. The estimated values are presented in
Table 3.
Eqs. (38) and (39), with the estimated parameters (Table 3) were
plotted along with the tted data as presented in Fig. 9a and b.
As presented in Fig. 9a, when the liquid velocity increased,
the mass transfer coefcient for the liquid slug increased, due to
the fact that the mixture Reynolds number increased (Fig. 9c). The
mixture Reynolds number increased because the mixture velocity
increased (Fig. 9e) and the mass transfer coefcient is proportional
to the mixture Reynolds number (Eq. (38)). On the other hand, as
presented in Fig. 9b, when the liquid velocity increased, the mass
transfer decreased for the gas slug, which is due to the fact that the
falling lm Reynolds number decreased (Fig. 9d). This falling lm
Reynolds number decreases because the falling lm velocity decre-
ases (Fig. 9f). Also, it was found that the thickness of the falling lm
was about 2.6310
4
72.5710
6
m for all experimental condi-
tions. The mass transfer coefcient is proportional to the falling lm
Reynolds number (Eq. (39)). These results indicate that a high gas
velocity and a low liquid velocity result in conditions that equili-
brate the magnitude of both peaks in the MTH.
From Fig. 9a, it is possible to observe that the t of the semi-
empirical model to the experimental data is relatively good for the
mass transfer coefcient induced by the liquid slug. However, the t
of the semi-empirical to model the experimental data is not as good
for the mass transfer coefcient induced by the gas slug (Fig. 9b). For
the liquid slug, absolute errors up to 10% were obtained, whereas for
the gas slug, errors up to 25% were obtained. The larger errors in the
gas slug can be attributed to: (i) the coalescence taking place in the
slug ow, (ii) that the model of Zheng and Che (2006) was developed
Table 3
Parameters of Eqs. (38) and (39).
Zone a b R
2
Liquid slug 1.5670.07 0.3170.04 0.82
Gas slaug 0.7570.09 0.5270.13 0.61
Fig. 9. Supercial liquid velocity vs. mass transfer coefcient for different supercial gas velocities for (a) liquid slug and (b) gas slug for the experiment data and the
model; supercial liquid velocity vs., (c) mixture Reynolds numbers, (d) falling lm Reynolds number, (e) mixture velocity and (f) falling lm velocity.
N. Ratkovich et al. / Chemical Engineering Science 66 (2011) 12541268 1262
based on single-phase ow plus a buoyancy term and (iii) the model
developed by Zheng and Che (2006) was for a turbulent regime;
however, the model developed in this work is for a laminar regime.
3.7. Coalescence of bubbles
Coalescing produces bubbles of different sizes over time, which
yielded different mass transfer coefcients over time for the diffe-
rent slug ow zones; even though the liquid and gas ow rates were
constant. The MTH were used to assess the effects of coalescing on
the frequency (i.e. height of peaks) and standard deviation (i.e.
width of peaks) of the mass transfer coefcients.
When considering the width of the peaks (Fig. 10a and b), it was
possible to observe that the width was higher at lowliquid velocities
(for the liquid and gas slug) and decreased at high liquid ow rates.
This phenomenon was attributed to the pulsating ow generated by
the peristaltic pump. As presented in Fig. 10a, at high liquid and gas
supercial velocities, the width of the liquid peak decreased,
whereas as presented in Fig. 10b, when the supercial gas velocity
increased, the width of the gas peak increased. This increase in the
width of the gas slug peak was related to occurrence of coalescence
(longer bubbles), but also due to a larger amount of air in the system
relative to experiments with lower gas ow rates.
When considering the height of the peaks (Fig. 11a and b), it
was possible to observe that at high liquid and low supercial gas
velocities, the height of the liquid peak increased (increase in
mass transfer in the liquid slug section). However, at high liquid
and high gas supercial velocities the height of the gas peak
increased (increase in mass transfer in the gas slug section).
3.8. Pressure drop
The pressure drop was calculated based on the empirical
relationships presented in Section 2.5. The results are presented
in Fig. 12a and b.
From Fig. 12a, it is possible to observe that having a two-phase
ow decreased the static pressure drop, compared to the decrease
for single-phase ow, by approximately 7%. On the other hand, the
frictional pressure drop (Fig. 12b) increased by having two-phase
ow. The latter is related to the wall shear stress. Comparing the
curve for single-phase ow with that of two-phase ow with a gas
velocity of 0.022 ms
1
, it is observe that in single-phase ow the
pressure drop is higher up to a liquid velocity of 0.07 ms
1
. There-
fore, having a low gas velocity for two-phase ow will decrease the
frictional pressure drop compared to single-phase ow. However, at
high gas velocities the opposite is true. Moreover, it is important to
highlight that the frictional pressure drop was approximately 0.7% of
the total pressure drop. Therefore, in a vertical pipe the static pressure
drop is more important than the frictional pressure drop.
3.9. Wall shear stress
From the mass transfer coefcient it is possible to calculate
the wall shear stress by combining Eqs. (12)(14) and (34)
Fig. 10. Supercial liquid velocities vs. the MTH width for (a) liquid and (b) gas
slug for three different gas velocities.
Fig. 11. Supercial liquid velocities vs. the MTH height for (a) liquid and (b) gas
slug for three different gas velocities.
N. Ratkovich et al. / Chemical Engineering Science 66 (2011) 12541268 1263
yielding Eq. (40):
t
w
m
L
1:25d
3=2
D
f
d
e
_ _2
k
3
m
40
Combining Eq. (40) with (38) and (39) yields the shear stress
for the liquid and gas slug, as presented in Eqs. (41) and (42),
respectively:
t
w_l
m
L
1:25d
3=2
D
f
d
e
_ _2
1:62a 1
Fr
m
b
_ _
b
Re
m
Sc
d
L
_ _
1=3
D
f
d
_ _
_ _
3
41
t
w_g
m
L
1:25d
3=2
D
f
d
e
_ _2
1:62a 1
Fr
m
b
_ _
b
Re
ff
Sc
4d
L
_ _
1=3
D
f
4d
_ _
_ _
3
42
The average wall shear stress is dened by Eq. (43) (Zheng
et al., 2008):
t
w_ave
t
w_g
bt
w_l
1b 43
The wall shear stress for the liquid and gas slug are presented
in Fig. 13ac.
From Fig. 13a, it is possible to observe that an increase in the
liquid and gas supercial velocity increases the wall shear stress.
On the other hand, an increase in the liquid and gas supercial
velocity decreases the shear stress in the gas slug (falling lm)
zone. Nevertheless, it is possible to observe that, at low liquid
supercial velocities, the ratio of the shear stress in the gas slug
zone to the liquid slug zone is 6, whereas at high liquid velocities
this ratio is 2. Therefore, the results indicate that to maximize the
shear stress, a combination of low liquid and high gas super-
cial velocities is preferred (Fig. 13c). However, this contradicts
Eq. (18), where the shear stress is proportional to pressure drop
for single-phase ow. Comparing Figs. 13c and 12, it is possible to
observe that the decrease in shear stress occurred when the
frictional pressure drop increased. Hence, based on this, using
Eq. (18) to calculate the shear stress in two-phase ow is not
Fig. 12. Supercial liquid velocity vs. (a) static and (b) frictional pressure drop for
single-phase ow and three different gas velocities.
Fig. 13. Supercial liquid velocities vs. wall shear stress for (a) liquid, (b) gas slug
and (c) average for three different gas velocities.
N. Ratkovich et al. / Chemical Engineering Science 66 (2011) 12541268 1264
recommended. Zheng and Che (2008) modied Eq. (18) presented
in Eqs. (44) and (45) for gas and liquid slugs, respectively:
t
w_g

f
ff
r
l
u
2
ff
8
44
t
w_l

f
m
r
l
u
2
m
8
45
where f
ff
and f
m
are the friction factors based on the falling lm
and the mixture Reynolds numbers and they depend on the ow
regime (Eqs. (19)(21)). The comparison of the two previous
equations with the results from the measurements (Fig. 12) is
presented in Fig. 14.
From Fig. 14a, it can be observed that the equations proposed
by Zheng and Che (2008) cannot be used to accurately model the
shear stress. The reason for the large discrepancies between the
proposed equations and the experimental measurements resides
in the use of single-phase ow expressions for two-phase ows,
which is clearly not the case in this study. Also, Eqs. (44) and (45)
originate from Eq. (18), which is a relationship used for ows in
pipes, where the main assumption is that the pipe is completely
lled with the uid being transported (pipe ow). This is not the
case in two-phase ow, where part of the pipe is lled with
gas slugs. Therefore an open-channel ow should be considered.
However, this is beyond the scope of the present.
3.10. Extrapolation of the semi-empirical model to activated sludge
The use of dimensionless numbers (i.e. Sherwood number) was
considered to extrapolate the results obtained in the present study
to estimate mass transfer coefcients in commercial airlift sMBRs.
Fig. 14. Supercial liquid velocities vs. wall shear stress for (a) liquid and (b) gas
slug for three different gas velocities comparing the experimental data and the
model proposed by Zheng and Che (2008).
Fig. 15. Mixture Reynolds number for (a) water and (b) sludge. Falling lm
Reynolds numbers for (c) water and (d) sludge vs. ve different liquid velocity and
three different gas velocities.
N. Ratkovich et al. / Chemical Engineering Science 66 (2011) 12541268 1265
The commercial airlift sMBR module of interest has pipes with
a diameter of 0.0052 m (Futselaar et al., 2007). Five liquid (0.220,
0.258, 0.295, 0.333 and 0.370 m s
1
) and three gas (0.09, 0.14 and
0.19 m s
1
) supercial velocities are considered in the present
analysis (range were discussed in Section 2.1). Assuming a TSS of
10 g l
1
, the apparent viscosity can be estimated using Eqs. (16)
and (17), resulting in a ow behaviour index and ow consistency
index of 0.461 and 0.171 Pa s, respectively. The resulting Reynolds
numbers for the gas and liquid ows considered are presented in
Fig. 15ad.
From Fig. 15a and b, it is possible to observe that the single- and
two-phase ow conditions are in the laminar-transition regime for
water (3000) and in the laminar regime for sludge based on the
mixture Reynolds number. On the other hand, in Fig. 15c and d the
falling lm Reynolds numbers are in the laminar regime for water
and sludge. It is important to highlight that when the velocity
increased, the falling lm Reynolds number increased for water
(Newtonian uid), whereas, it decreased for activated sludge. The
decrease is due to the effect of viscosity on the Reynolds number.
Based on the results of the Reynolds numbers (laminar-transi-
tion regime), the semi-empirical relationship developed in the
present study (Eq. (37)) can be used to determine the mass transfer
coefcient under the assumption that this is valid slightly above the
laminar regime. Based on this assumption, the mass transfer coe-
fcient for non-Newtonian uids can be estimated using Eq. (46).
This is derived from Eqs. (31) and (37):
k
m
1:016a 1
Fr
m
b
_ _
b
3n1
n
_ _
1=3
Reu Scu
d
L
_ _
1=3
D
f
d
_ _
46
To facilitate the comparison between the results obtained for
water (Sections 3.13.9) and those obtained for activated sludge,
the mass transfer relationships are presented in terms of the
Sherwood number, which were estimated using Eq. (47):
Shu 1:016a 1
Fr
m
b
_ _
b
3n1
n
_ _
1=3
Reu Scu
d
L
_ _
1=3
47
The average diffusivity in activated sludge was assumed to be
7.410
10
m
2
s
1
(Smith and Coackley, 1984), which is similar
to the diffusivity of the electrolyte solution used in the present
study. Therefore, the Sherwood number for the liquid and gas slug
can be estimated using Eq. (47) with the empirical constants presen-
ted in Table 3. Results of the Sherwood numbers are presented in
Fig. 16ad.
Comparing Fig. 16a and b, it is possible to observe that the
liquid slug Sherwood number increased by 8% for the non-New-
tonian uid (activated sludge) compared to that for the New-
tonian uid (water). The 8% can be explained by the factor,
3n1=4n
1=3
1:089 (which was introduced in Section 2.7,
and it is function of the ow consistency index). From Fig. 16c
and d, it is possible to observe that the gas slug Sherwood number
increased by 8% between the Newtonian and non-Newtonian
uids. However, for the case of water, the Sherwood number
increased as the Reynolds number increased, whereas for sludge,
the Sherwood number decreased (due to viscosity). As the shear
stress is proportional to mass transfer (or Sherwood number), the
same results as presented above will apply for the shear stress
over the membrane surface.
The dimensionless analysis presented above, even if it not
validated, based on experimental data demonstrates that the slug
ow is of great importance to generate high mass transfer
coefcients. High mass transfer coefcients are essential for the
effective removal of particles that are attached to the membrane
surface and decrease and/or control fouling. In addition, the
hydrodynamic behaviour of activated sludge in sMBR is affected
Fig. 16. Mixture Reynolds number vs. the liquid slug Sherwood number for
(a) water and (b) activated sludge and falling lm Reynolds number vs. the gas
slug Sherwood number for (a).
N. Ratkovich et al. / Chemical Engineering Science 66 (2011) 12541268 1266
by the non-Newtonian behaviour of the solution and should be
considered in the design of these systems.
4. Conclusions
A setup with shear probes and an electrolytic solution was
used to measure the mass transfer coefcient in slug ow.
Mass transfer coefcient proles were found to be highly
variable due to the occurrence of coalescence of gas slugs.
Histograms (MTHs) were used to characterize the mass trans-
fer to which the membrane surface was exposed. The MTHs
exhibited two peaks, the rst peak was due to the mass
transfer induced by liquid ow and a second peak was due
to the mass transfer induced by gas slugs. Average values of
mass transfer coefcients were determined for the entire
range of operational conditions tested, based on a bimodal
distribution.
Based on the analysis, the mass transfer is optimized (bimodal
distribution is balanced in terms of the shape (height and width)
of the two peaks) at low liquid velocities, which decrease the
magnitude of the rst peak and at high gas velocities, which
increase the magnitude of the second peak.
Semi-empirical relationships were developed based on the
existing Le v eque relationship for mass transfer to determine
the Sherwood number for gas and liquid slug ow in a pipe.
The latter was calibrated and validated using experimental
measurements.
A comparative case, using the operational conditions of a full-
scale airlift sMBR was considered for solutions consisting of
water (Newtonian uid) and activated sludge (non-Newtonian
uid), using the developed semi-empirical relationships. Higher
Sherwood numbers were found for the non-Newtonian case (8%)
compared to the Newtonian case. This difference was attributed
to the viscosity. Therefore, viscosity plays an important role,
especially in the falling lm zone (gas slug) and should be deter-
mined thoroughly.
The analysis performed in this study suggests that dimension-
less analysis (i.e. Sherwood number) can be used to correlate
mass transfer coefcient with the liquid and gas supercial
velocities.
Nomenclature
C
o
bulk concentration of ferricyanide (3 mol m
3
)
C
TB
slope
d pipe diameter (m)
d
e
diameter of the probe (m)
D
f
diffusion coefcient of ferricyanide
(7.1410
10
m
2
s
1
)
f friction factor ()
F Faraday constant (96 500 C mol
1
)
Fr
m
mixture Froude number ()
g gravity acceleration (9.81 m s
2
)
G amplier gain (1000)
k
i
mass transfer coefcient of i (m s
1
)
k
m
mass transfer coefcient (m s
1
)
k
m_probe
probe mass transfer coefcient (m s
1
)
k
TB
intercept
L length of the pipe (m)
L
e
entry length (m)
m ow consistency index (Pa s
n
)
n ow behaviour index ()
Pe Pe clet number ()
WP pressure drop along the pipe (Pa)
WP
frict
frictional pressure drop (Pa)
WP
static
static pressure drop (Pa)
WP
total
total pressure drop (Pa)
Re
i
Reynolds number of i ()
R resistance (100 O)
Sc Schmidt number ()
Sh Sherwood number ()
TSS Total Suspended Solids (g L
1
)
u
axial
axial velocity (m s
1
)
u
i
velocity of i (m s
1
)
V voltage (V)
v
e
number of electrons involved in the reaction (1)
Subindices
ff falling lm
g gas
l liquid
lam laminar
ls liquid slug
m mixture
SG supercial gas
SL supercial liquid
TB gas slug
tur turbulent
w wake
Greek
a
TB
void fraction of the slug ow ()
a
LS
void fraction of liquid slug ()
b ratio of the gas slug length to the sum of the gas and
liquid slugs length ()
d
L
thickness of the falling liquid lm (m)
e pipe roughness (m)
_ g mean wall velocity gradient or shear rate (s
1
)
m
i
dynamic viscosity of i (Pa s)
r
i
density of i (kg m
3
)
t
w
wall shear stress (Pa)
Acknowledgments
This research project has been supported by a Marie Curie
Early Stage Research Training Fellowship of the European Com-
munitys Sixth Framework Programme under contract number
MEST-CT-2005-021050. Funding for the infrastructure used to
measure surface shear stress was provided by the Natural Science
and Engineering Research Council of Canada (NSERC).
References
Adsani, E., Shirazi, S.A., Shadley, J.R., Rybicki, E.F., 2006. Validation of mass transfer
coefcient models used in predicting CO
2
corrosion in vertical two-phase ow
in the oil and gas production, Corrosion 2006, September 10, 2006September
14, 2006, pp. 65731657312.
Bugg, J.D., Saad, G.A., 2002. The velocity eld around a Taylor bubble rising in a
stagnant viscous uid: numerical and experimental results. International
Journal of Multiphase Flow 28 (5), 791803.
Chan, C.C.V., Berube, P.R., Hall, E.R., 2007. Shear proles inside gas sparged
submerged hollow ber membrane modules. Journal of Membrane Science
297 (12), 104120.
Chang, S., Fane, A.G., 2000. Filtration of biomass with axial inter-bre upward slug
ow: performance and mechanisms. Journal of Membrane Science 180 (1),
5768.
N. Ratkovich et al. / Chemical Engineering Science 66 (2011) 12541268 1267
Chen, L., Tian, Y.S., Karayiannis, T.G., 2006. The effect of tube diameter on vertical
two-phase ow regimes in small tubes. International Journal of Heat and Mass
Transfer 49 (2122), 42204230.
Cheng, N.S., 2008. Formulas for friction factor in transitional regimes. Journal of
Hydraulic Engineering 134 (9), 13571362.
Cognet, G., Lebouche, M., Souhar, M., 1978. Utilisation des techniques electro-
chimiques pour la mesure du frottement parietal dans les ecoulements
diphasiques. Left bracket Electrochemical Methods for Measurement of Wall
Shear Stress in Two-Phase Flow right bracket, Houille Blanche 33 (5), 319322.
Collins, R., De Moraes, F.F., Davidson, J.F., Harrison, D., 1978. Motion of a large gas
bubble rising through liquid owing in a tube. Journal of Fluid Mechanics 89,
497514.
Cui, Z.F., Chang, S., Fane, A.G., 2003. The use of gas bubbling to enhance membrane
processes. Journal of Membrane Science 221 (12), 135.
De, S., Bhattacharya, P.K., 1997. Prediction of mass-transfer coefcient with
suction in the applications of reverse osmosis and ultraltration. Journal of
Membrane Science 128 (2), 119131.
Drew, D.A., 1983. Mathematical modeling of two-phase ow. Annual Review of
Fluid Mechanics 15, 261291.
Dziubinski, M., 2003. Comments on bubble rising velocity in non-Newtonian
liquids. Chemical Engineering Science 58 (11), 24412443.
Dziubinski, M., 1995. General correlation for two-phase pressure drop in inter-
mittent ow of gas and non-Newtonian liquid mixtures in a pipe. Chemical
Engineering Research and Design 73, 528534.
Esteves, M.T.S., De Carvalho, J.R.F., 1993. Liquid-side mass transfer coefcient for
gas slugs rising in liquids. Chemical Engineering Science 48 (20), 34973506.
Fabre, J., Line, A., 1992. Modelling of two phase slug ow. Annual Review of Fluid
Mechanics 24, 2146.
Fernandes, R.C., Semial, R., Dukler, A.E., 1983. Hydrodynamic model for gasliquid
slug ow in vertical tubes. AICHE Journal 29 (6), 981989.
Fidos, H., Sowinski, J., Dziubinski, M., Krokos, R., 2008. Pressure drop in the slug
ow of two-phase of gasnon Newtonian liquid mixtures. Przemysl Che-
miczny 83 (2), 111115.
Futselaar, H., Schonewille, H., De Vente, D., Broens, L., 2007. NORIT AirLift MBR: side-
stream system for municipal waste water treatment, Desalination 204 (1-3), 17.
Ghosh, R., Cui, Z.F., 1999. Mass transfer in gas-sparged ultraltration: upward slug
ow in tubular membranes. Journal of Membrane Science 162 (12), 91102.
Judd, S., 2006. The MBR Book. Elsevier.
Laborie, S., Cabassud, C., Durand-Bourlier, L., Laine, J.M., 1998. Fouling control by
air sparging inside hollow bre membraneseffects on energy consumption.
Desalination 118 (13), 189196.
Laera, G., Giordano, C., Pollice, A., Saturno, D., Mininni, G., 2007. Membrane
bioreactor sludge rheology at different solid retention times. Water Research
41 (18), 41974203.
Mao, Z.S., Dukler, A.E., 1985. Rise velocity of a Taylor bubble in a train of such
bubbles in a owing liquid. Chemical Engineering Science 40 (11), 21582160.
Nakoryakov, V.E., Pokusaev, B.G., Petukhov, A.V., 1987. Absorption during slug ow
in a vertical pipe. Journal of Engineering Physics (English Translation of
Inzhenerno-Fizicheskii Zhurnal) 52 (4), 404408.
Nogueira, S., Riethmuler, M.L., Campos, J.B.L.M., Pinto, A.M.F.R., 2006a. Flow in the
nose region and annular lm around a Taylor bubble rising through vertical
columns of stagnant and owing Newtonian liquids. Chemical Engineering
Science 61 (2), 845857.
Nogueira, S., Riethmuller, M.L., Campos, J.B.L.M., Pinto, A.M.F.R., 2006b. Flow
patterns in the wake of a Taylor bubble rising through vertical columns of
stagnant and owing Newtonian liquids: an experimental study. Chemical
Engineering Science 61 (22), 71997212.
Ochoa, J.C., Coufort, C., Escudie, R., Line, A., Paul, E., 2007. Inuence of non-uniform
distribution of shear stress on aerobic biolms. Chemical Engineering Science
62 (14), 36723684.
Oliveira, T.A.C., Cocchini, U., Scarpello, J.T., Livingston, A.G., 2001. Pervaporation
mass transfer with liquid ow in the transition regime. Journal of Membrane
Science 183 (1), 119133.
Perry, R.H., Green, D.W., 1999. Perrys Chemical Engineers Handbook. McGraw-Hill.
Pinto, A.M.F.R., Coelho Pinheiro, M.N., Campos, J.B.L., 2001. On the interaction of
Taylor bubbles rising in two-phase co-current slug ow in vertical columns:
turbulent wakes. Experiments in Fluids 31 (6), 643652.
Pinto, A.M.F.R., Pinheiro, M.N.C., Campos, J.B.L.M., 1998. Coalescence of two gas
slugs rising in a co-current owing liquid in vertical tubes. Chemical
Engineering Science 53 (16), 29732983.
Pollice, A., Giordano, C., Laera, G., Saturno, D., Mininni, G., 2007. Rheology of sludge
in a complete retention membrane bioreactor. Environmental Technology 27,
723732.
Polonsky, S., Shemer, L., Barnea, D., 1999. The relation between the Taylor bubble
motion and the velocity eld ahead of it. International Journal of Multiphase
Flow 25 (67), 957975.
Ranjan, R., DasGupta, S., De, S., 2004a. Mass transfer coefcient with suction for
laminar non-Newtonian ow in application to membrane separations. Journal
of Food Engineering 64 (1), 5361.
Ranjan, R., DasGupta, S., De, S., 2004b. Mass transfer coefcient with suction for
turbulent non-Newtonian ow in application to membrane separations.
Journal of Food Engineering 65 (4), 533541.
Ratkovich, N., Chan, C.C.V., Berube, P.R., Nopens, I., 2010. Investigation of the effect
of viscosity on slug ow in airlift tubular membranes in search for a sludge
surrogate. Water Science and Technology 61 (7), 18011809.
Ratkovich, N., Chan, C.C.V., Berube, P.R., Nopens, I., 2009. Experimental study and
CFD modelling of a two-phase slug ow for an airlift tubular membrane.
Chemical Engineering Science 64 (16), 35763584.
Rehimi, F., Legrand, J., Aloui, F., 2008. Electrochemical method for precise
determination of wall shear rate. Russian Journal of Electrochemistry 44 (4),
434444.
Rosant, J.M., 1994. Liquid-wall shear stress in stratied liquid/gas ow. Journal of
Applied Electrochemistry 24 (7), 612618.
Rosehart, R.G., Rhodes, E., Scott, D.S., 1975. Studies of gasliquid (non-Newtonian)
slug ow: void fraction meter, void fraction and slug characteristics. Chemical
Engineering Journal and the Biochemical Engineering Journal 10 (1), 5764.
Rosenberger, R., Kubin, K., Kraume, M., 2006. Rheology of activated sludge in
membrane bioreactors. Engineering in Life Sciences 2 (9), 269275.
Shannak, B.A., 2008. Frictional pressure drop of gas liquid two-phase ow in pipes.
Nuclear Engineering and Design 238 (12), 32773284.
Shirazi, S.A., Al-Adsani, E., Shadley, J.R., Rybicki, E.F., 2004. A mechanistic model for
predicting heat and mass transfer in vertical two-phase ow. In: Proceedings
of the ASME Heat Transfer/Fluids Engineering Summer Conference 2004, HT/
FED 2004, July 1115, 2004, vol. 3, pp. 685693.
Smith, P.G., Coackley, P., 1984. Diffusivity, tortuosity and pore structure or
activated sludge. Water research 18 (1), 117122.
Sousa, R.G., Pinto, A.M.F.R., Campos, J.B.L.M., 2007. Interaction between Taylor
bubbles rising in stagnant non-Newtonian uids. International Journal of
Multiphase Flow 33 (9), 970986.
Sousa, R.G., Riethmuller, M.L., Pinto, A.M.F.R., Campos, J.B.L.M., 2005. Flow around
individual Taylor bubbles rising in stagnant CMC solutions: PIV measure-
ments. Chemical Engineering Science 60 (7), 18591873.
Thome, J.R., 2008. Wolverine Heat Transfer Engineering Data Book III. Wolverine
Tube Inc..
van Hout, R., Barnea, D., Shemer, L., 2002. Translational velocities of elongated
bubbles in continuous slug ow. International Journal of Multiphase Flow 28
(8), 13331350.
Wang, H., Jepson, W.P., Hong, T., Cai, J., Bosch, C., 2002. Enhanced mass transfer
and wall shear stress in multiphase slug ow. Corrosion, 2002.
White, F., 2002. Fluid Mechanics. McGraw-Hill.
Xu, J.Y., Wu, Y.X., hong Shi, Z., yun Lao, L., hui Li, D., 2007. Studies on two-phase co-
current air/non-Newtonian shear-thinning uid ows in inclined smooth
pipes. International Journal of Multiphase Flow 33 (9), 948969.
Zheng, D., Che, D., 2007. An investigation on near wall transport characteristics in
an adiabatic upward gasliquid two-phase slug ow. Heat and Mass Transfer/
Waerme- und Stoffuebertragung 43 (10), 10191036.
Zheng, D., Che, D., 2006. Experimental study on hydrodynamic characteristics
of upward gasliquid slug ow. International Journal of Multiphase Flow 32
(1011), 11911218.
Zheng, D., Che, D., Liu, Y., 2008. Experimental investigation on gasliquid two-
phase slug ow enhanced carbon dioxide corrosion in vertical upward pipe-
line. Corrosion Science 50 (11), 30053020.
N. Ratkovich et al. / Chemical Engineering Science 66 (2011) 12541268 1268

You might also like