Journal of Membrane Science 366 (2011) 295303

Contents lists available at ScienceDirect

Journal of Membrane Science

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

A generalized model for clarication of fruit juice during ultraltration under

total recycle and batch mode
Sourav Mondal a , Alfredo Cassano b , Franco Tasselli b , Sirshendu De c,
a
Department of Chemical Engineering, Jadavpur University, Kolkata 700032, India
b
Institute on Membrane Technology, ITM-CNR, c/o University of Calabria, via P. Bucci, Cubo 17/C, I-87030 Rende, Cosenza, Italy
c
Department of Chemical Engineering, Indian Institute of Technology Kharagpur, Kharagpur 721302, India

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

Article history: A generalized gel layer controlled ltration model is presented in this work. The model was developed
Received 6 May 2010 using an integral method for analysing the concentration boundary layer and it was successfully applied to
Received in revised form 5 October 2010 experimental ultraltration data obtained in the clarication of kiwi fruit juice by ultraltration (UF). Both
Accepted 7 October 2010
total recycle and batch concentration modes were included in the model. Most model parameters, such as,
Available online 14 October 2010
gel layer concentration, viscosity, parameter in Sherwood number, etc., were obtained by minimizing the
error involved between calculated values and experimental data. This generalized model can be utilized
Keywords:
for ultraltration of any fruit juice.
Kiwi fruit juice
Gel layer 2010 Elsevier B.V. All rights reserved.
Sherwood number
Ultraltration
Concentration boundary layer

1. Introduction
A typical fruit juice processing unit comprises of various unit
operations, namely, depectinization, centrifugation, application of
nning agents, ltration agents, ltration by diatomaceous earth,
polished ltration, etc. [1]. The processing time may vary between
24 and 36 h [1]. Membrane based separation process can be an alter-
native in this regard. Not only it reduces the time of processing, it
also precludes use of some unit operations of conventional process,
like, nning treatment, polish ltration, etc. Ultraltration (UF) is
widely used for removal of pectin and microorganisms from the
juice [2]. The claried juice, being devoid of pectin, does not form
haze and it has reasonably high shelf life [3]. There are plenty of
studies available for membrane based processes involved in the
clarication of fruit juices. For example, UF has been successfully
used for processing of kiwi fruit [4], mosambi [5], apple [6,7], orange
[8], passion fruit [9], pineapple [10], acerola [11], cactus pear [12],
and carrot juice [13].
For scaling up, modelling of UF process is essential. Most models
involved in the clarication of fruit juice, are gel layer controlling
type [14,15]. This is due to the fact that pectin is a well-known
gel forming agent [16]. Apart from pectin, in a real juice, there are
other high molecular weight compounds like cellulose, hemicellu-
Corresponding author. Tel.: +91 3222 283926; fax: +91 3222 255303.
E-mail addresses: sde@che.iitkgp.ernet.in, sirshendu kgp@yahoo.com (S. De).
lose, cell debris, etc. that contribute to the formation of gel over the
membrane surface.
In the present work, a model is proposed for ultraltration dur-
ing the clarication of kiwi fruit juice under the framework of gel
controlling ltration. Both operational modes, namely, steady state
total recycle and batch concentration have been included in the
model. The model is successfully applied to the experimental data
obtained in the ultraltration of kiwi fruit juice [17].
The model presented here is a generalized one. In most of the
fruit juice processing, a full proof compact theoretical analysis is
often lacking. The need for growing fruit juice industries demands
a very predictive and practically applicable model to estimate the
steady state ux accounting most of the physical factors at a time.
A close look to the modelling approaches of ux decline during
clarication of fruit juice using ultraltration and/or microltration
reveals that the resistance in series and pore blocking models are
the most popular ones [4,5,7,9,10,18]. Resistance in series model
are used by Cassano et al. [4] for UF of kiwi fruit juice, Vladis-
avljevic et al. [7] for UF of depectinized apple juice, Jiraratananon
and Chanachai for UF of passion fruit juice [9], Rai et al. [18] for
UF of mosambi juice. It may be noted here that the resistance in
series model lacks exibility and generality as far as the predic-
tion capability of ux decline for various kinds of raw materials
(juices of different sources). These models are entirely empirical in
nature.
Pore blocking models are another class of models to quantify the
decline of permeate ux under steady state or transient mode of

0376-7388/$ see front matter 2010 Elsevier B.V. All rights reserved.
doi:10.1016/j.memsci.2010.10.015
296 S. Mondal et al. / Journal of Membrane Science 366 (2011) 295303

ltration. These models are used by Rai et al. [5] for UF of mosambi During ultraltration of juice, high molecular weight solutes are
juice, De Barros et al. [10] for UF of pineapple juice, Chayya et al. transported towards the membrane wall, forming a gel/cake type
[19] for MF of water melon juice. Again, these models are semi- of layer over the membrane surface. It may be pointed out here
empirical in nature and lacks much theoretical understanding. The that in membrane literature, generally cake refers to an assembly
gel/cake controlling model is a variant of pore blocking models. of deformable particles, like polymers, polysaccharides, etc. On the
Moreover, often it is observed that various mechanisms, namely, other hand, gel indicates a collection of non-deformable materials,
complete, intermediate, standard pore blocking, cake ltration can like, protein, silica, etc. However, in the present study, it is assumed
individually t same set of experimental ux decline data closely that the cake and gel are equivalent in nature without the loss of
[5]. The suitability of the models is determined by analysing the generality of the physics involved in this system. The gel concentra-
correlation coefcient values of the t. tion is assumed to be constant within this layer. Therefore, there
Therefore, a model having more fundamental basis is war- exists a concentration boundary layer next to the gel layer with
ranted to quantify the ux decline during UF of fruit juice. The a variation in concentration from bulk to gel concentration. The
present model is based on gel/cake layer controlling ltration under convectivediffusive ux equation for gel forming solute within
the framework of boundary layer analysis. The model includes concentration boundary layer is given for a tubular module as
the developing mass transfer boundary layer over the gel layer,  
which is the most realistic situation. Thus, this model overcomes C C C
u +v = D (1)
the shortcoming of conventional lm theory [20] that considers x r r r
a constant thickness of mass transfer boundary layer. Secondly,
Since, the thickness of the concentration boundary layer is much
solution viscosity is a strong function of solute concentration and
less than the radius of the tube, the curvature effect is entirely lost
it varies signicantly within the mass transfer boundary layer, as
and Eq. (1) is reduced to a planar coordinate by dening, y = R r,
the solute concentration increases from bulk to gel layer concen-
where y is the distance from the membrane surface (tube wall).
tration. Gel layer concentration is quite often 45 times of bulk
Under this coordinate system the above equation becomes
concentration. This variation of viscosity as a function of concen-
tration is included in the present model. Thirdly, variation of feed C C 2 C
temperature affects solution viscosity and solute diffusivity signif- u +v =D 2 (2)
x y y
icantly. This phenomenon has been incorporated in the present
analysis using the StokesEinstein equation. The model parameters It may be noted that the cross ow velocity in a membrane mod-
involved are gel layer concentration, solute diffusivity and viscosity ule is generally 5 to 6 order of magnitude higher compared to the
variation parameter. permeation velocity. Thus, it can be assumed that the parabolic
A fruit juice being a complex mixture of various components, it velocity prole (under laminar ow regime) inside the tubular
is difcult to estimate these parameters independently. Thus, these module remain undisturbed due to permeation in the wall. The
parameters can be easily evaluated from the steady state perme- velocity in the tube is given as [21]:
ate ux data at various operating conditions. The present model   r 2 
is also extended to quantify the ux decline as well as the vol- u = 2u0 1 (3)
ume reduction factor (VRF) during batch mode of operation from R
the rst principles by writing down the overall material balance,
Fixing the co-ordinate system on the wall of the tube, y = R r,
overall solute balance and solute balance within the mass transfer
within the thin mass transfer boundary layer the velocity prole
boundary layer. A numerical solution of these balance equations
becomes linear [22]. Since, (y/R)  1, the x-component velocity pro-
leads to the ux decline and VRF prole. Specic gel layer resistance
le simplies to
is the only parameter that has to be optimized using experimental
data. The parameters obtained from the steady state model are used y
u = 4u0 (4)
in the batch mode ltration model. Therefore, the present model R
is a comprehensive one including various fundamental transport Since, the concentration boundary layer thickness is extremely
aspects. The extensive analytical treatment makes the model easy small, it can be assumed that the y-component velocity is equal
to estimate the steady state ux values in either of the operation to the permeation velocity at the wall [23]. Therefore, the y-
modes by simple computational techniques. component velocity becomes
The designers may nd it extremely useful in studying the model
results while selecting the processing equipment, the mode of oper- v = vw (5)
ation, optimum operational time, etc. The model parameters are
evaluated for this particular juice and may be used for the purpose Inserting the velocity proles into Eq. (2) the following equation is
of scaling up. obtained:
y C C 2 C
4u0 vw =D 2 (6)
R x y y
2. Theoretical development
Using x* = x/L; y* = y/R; C* = C/C0 , the above equation is non-
2.1. Total recycle mode dimensionalized as
u0 d2 C vw d C 2 C
In this mode of operation, both the permeate and retentate y = (7)
DL x 2D y y2
streams are recycled back to the feed tank so that a steady state
is attained with xed concentration of the feed. This is a popu- where di = 2R. It is to be noted here that u0 di2 /DL = Re Sc(d/L)
lar mode of operation in any membrane based process in order to (denoted as A, henceforth) and vw di /D is the non-dimensional ux
evaluate the effects of operating parameters (feed concentration, (denoted as Pew ).
cross ow velocity, transmembrane pressure drop) on steady state So, Eq. (7) can be rearranged as
permeate ux and permeate quality [5]. In order to quantify the
C Pew C 2 C
permeate ux, mass transfer coefcient has to be estimated. In this Ay

= (8)
regard, an analysis in a tubular module is presented in this section. x 2 y y2
 S. Mondal et al. / Journal of Membrane Science 366 (2011) 295303 297

A quadratic concentration prole within concentration boundary 2.1.1. Estimation of the mass transfer coefcient
prole is assumed The denition of mass transfer coefcient can be written as

C C
C = = a1 + a2 y + a3 y2 (9) k(Cg C0 ) = D (22)
C0 y
y=0
where a1 , a2 and a3 are the constant coefcients.
Non-dimensionalizing the above equation and substituting
Using the following boundary conditions, the constants of Eq.
(C /y )|y =0 from Eq. (13) leads to the following expression of
(9) are evaluated
Sherwood number
at y = 0, C = Cg = Cg /C0 (10) 4
Sh = (23)

at y = , C = C0 ; C =1 (11)
Substituting the prole of * from Eq. (21), the expression of Sher-
C wood number becomes
at y = , =0 (12)
y  1/3

ACg
Using the boundary conditions (Eqs. (10)(12)) the constants a1 , a2 Sh(x ) = 4 x1/3 (24)
36
and a3 are evaluated and Eq. (9) is rewritten as
 y   y 2 The length averaged Sherwood number thus, becomes
C = Cg 2(Cg 1) + (Cg 1) (13)   1/3
1
d 1/3
ShL = Sh(x ) dx = 1.816 Re Sc Cg (25)
Now C*/ x*, C*/ y* and 2 C*/ y*2 in Eq. (8) are evaluated using 0
L
Eq. (13). These partial derivatives are established in Eq. (8) and after
It is to be noted that the co-efcient in Sherwood number relation
simplication the following parameter is obtained
using classical Leveque solution is 1.86 for rectangular and 1.62 for
   y  tubular geometry [24]. It may also be noted that these co-efcients
y2 y3 d Pew 1 1
A 3 = (14) remain unaltered in case of incorporation of SiederTate correc-
2 dx 2 2 2
tion factor [25]. However, the coefcient increases if one includes
Taking the zeroth moment of above equation by multiplying both the developing mass transfer boundary layer. In case of rectangular
sides by dy* and integrating across the boundary layer thickness geometry, the co-efcient increases from 1.85 to 2.10 [26]. In case
from 0 to * of tubular geometry, the Sherwood number co-efcient increases
      y  from 1.62 to 1.816 as derived in the present work.
d y2 y3 Pew 1
A 3 dy dy
dx 0
2 2 0
2 2.1.2. Introduction of the wall viscosity effect on Sherwood

number
1 Since, the gel layer concentration is several order of magnitude
= dy (15)
2 0 higher than the bulk concentration it is obvious that the viscos-
ity variation within concentration boundary layer is signicant.
On solving the above integral the following equation is arrived
Therefore, a SiederTate viscosity correction factor is included in
  the expression of Sherwood number (Eq. (25)) as [14,25]:
A2 d Pew
+ =1 (16)
12 dx 4  d
1/3   0.14
1/3 m
ShL = 1.816 Re Sc Cg (26)
Now considering a steady-state mass balance over the membrane L w
surface (y = 0), the following equation is obtained: where m is the viscosity at the mean concentration of the solution
within concentration boundary layer and w is the viscosity at the
dC
vw Cg + D =0 (17) wall concentration (Cg ).
dy
Following exponential variation of viscosity is assumed
The non-dimensional version of the above equation is
 = 0 eC (27)
1 dC
Pew Cg + =0 (18) Therefore, the viscosity at the wall becomes
2 dy
C*/ y* is evaluated from Eq. (13) and is substituted in the above w = 0 eCg (28)
equation to obtain In evaluating m we have to rst calculate the mean concentration
  Cm in the concentration boundary layer.
Cg 1
Pew = 4 (19) is
The expression of Cm
Cg

1
Replacing this value of Pew * in Eq. (16), results in the following Cm = C dy (29a)
governing equation of concentration boundary layer thickness 0

2 d 1 Substituting concentration prole for Eq. (13) the following value

A = (20) is obtained
of Cm
12 dx Cg
1
Integration of the above equation leads to the prole of concentra- Cm = (C + 2) (29b)
3 g
tion boundary layer thickness with x* as
 1/3 Hence using Eqs. (28) and (29) the following result is obtained:
36 1/3 m (2/3)C0 (Cg 1)
= x (21) =e (30)
ACg w
298 S. Mondal et al. / Journal of Membrane Science 366 (2011) 295303

So, substituting the value of viscosity correction factor m /w in where f and p are densities in feed and permeate streams; V is
Eq. (26), the Sherwood number relation is modied to the feed volume and Am is the effective membrane area. Assuming
 1/3 f  p the above equation is modied as
d (2/3)C0 (Cg 1) 0.14 1/3
ShL = 1.816 Re Sc (e ) Cg (31) dV
L = vw Am (41)
dt
2.1.3. Introduction of temperature effect on Sherwood number Using overall species balance of gel forming component, the fol-
According to the StokesEinstein equation, the solute diffusivity lowing equation is obtained:
is directly proportional to temperature and inversely proportional
d
to viscosity as [27]: (C V ) = vw Am Cp (42)
dt b
T Since, concentration of the gel forming material in the permeate is
D (32)
 zero (Cp = 0) [28], the above equation reduces to a simple algebraic
As viscosity varies with temperature inversely, so equation

1 Cb V = C0 V0 (43)
 (33)
Tn with initial boundary condition as C = C0 and V = V0 at t = 0.
Hence, using the above two equations, the variation of diffusivity Now, following the material balance for the gel-forming com-
with temperature is quantied as ponent in the concentration boundary layer results in the following
equation [29]:
D
 T n+1
= (34) dH dc1
D0 T0 for 0 < y < , j1 = mass ux = g = vw C1 D (44)
dt dy
Therefore, using Eq. (31) and Eq. (34), the following equation for
The pertinent boundary conditions are:
average Sherwood number is obtained.
 1/3  T (n+1)/3 C 1 = C b (t) at y = 0 (45)
d (2/3)C0 (Cg 1) 0.14 0 1/3
ShL = 1.816 Re Sc (e ) Cg C1 = Cg at y = (46)
L T
(35) The solution of Eq. (42) using the above stated boundary conditions
represents the variation of the gel layer thickness (H) with time
dH Cg Cb exp(vw /k)
g = vw (47)
2.1.4. Estimation of the non-dimensional ux Pew dt 1 exp(vw /k)
Combining Eqs. (17) and (22), the following equation of ux is
where k is the mass transfer coefcient dened as D/. In this case,
obtained:
the expression of mass transfer coefcient is not same as Eq. (35).
vw Cg = k(Cg C0 ) (36) This is because the boundary condition of the concentration prole
within concentration boundary layer at the edge is no longer ini-
Non-dimensional form of the above equation is tial feed concentration (C0 ). It becomes bulk concentration that is a
  function of time, Cb (t). The non-dimensional solute balance equa-
Cg 1
Pew = Sh (37) tion within concentration boundary layer in this case can be written
Cg as
Combining Eqs. (37) and (25) the length averaged permeate ux 1 C C Pew C 2 C
+ Ay = (48)
becomes 4  x 2 y y2
(T1 ) (T2 )
 d
1/3  T (n+1)/3 (T3 ) (T4 )
(2/3)C0 (Cg 1) 0.14 0
Pew = 1.816 Re Sc (e ) where the non-dimensional time is dened as,  = tD/d2 . Next, an
L T
order of magnitude analysis of Eq. (48) is carried out termwise.
1/3 2/3
(Cg Cg ) (38) O(x* ) is 1; order of y is same as that of thickness of concen-
tration boundary layer, D/k = 1011 /106 = 105 . Thus, O(y* ) is
1/3 2/3
For Cg < e3 , that is Cg  20, (Cg Cg ), is reduced to ln Cg , 105 /103 = 102 . O(A) is u0 d2 /DL = 1 106 /(1011 101 ) = 106 .
thus O(Pew ) is vw d/D = (106 103 )/1011 = 102 . Therefore, order of
 d
1/3  T (n+1)/3 the terms, T2 , T3 and T4 is 104 . Thus, it may be noted that T1 has
(2/3)C0 (Cg 1) 0.14
ln Cg
0
Pew = 1.816 Re Sc (e ) signicant magnitude compared to other three terms up to a time
L T of operation of 100 s. Beyond 100 s, it is reduced in order of magni-
(39) tude. Hence, comparing the full operation time in this experiment
(460 min), T1 is small enough to be ignored. Therefore, we can take
recourse to a quasi steady state analysis for estimation of concen-
2.2. Batch concentration mode
tration boundary layer prole. The governing equation of solute
mass balance is same as Eq. (8). The concentration prole can be
In this mode of operation, retentate stream is recycled back to
approximated as Eq. (9), along with the boundary conditions (Eqs.
the feed tank but the permeate is continuously taken out. This
(10) and (12)). Eq. (11) now becomes
results in an increase in feed concentration accompanied by a
reduction in feed volume with time of operation. at y = , C = Cb ; C = Cb (49)
Considering an overall material balance, the following equation
is obtained: The concentration prole within the boundary layer now becomes
 y   y 2
d
( V ) = vw Am p (40) C = Cg 2(Cg Cb ) + (Cg Cb ) (50)
dt f
 S. Mondal et al. / Journal of Membrane Science 366 (2011) 295303 299

Proceeding exactly like previous case as described in Section 2.1, It may be noted that as time of operation proceeds, the effective
the mean concentration within boundary layer is bre diameter decreases by deposition of gel layer and it is quan-
1 tied as

Cm = (C + 2Cb ) (51)
3 g d(t) = di 2H(t) (63)
and
Consequently, the cross ow velocity u0 (t) inside the bre changes
m (2/3)C0 (Cg C )
=e b (52) as
w
4Q
The Sherwood number relation is modied as u0 (t) = 2
(64)
 1/3 
1/3 [d(t)]
d (2/3)C0 (Cg C ) 0.14 Cg
ShL = 1.816 Re Sc (e b ) (53) The above expressions of bre diameter and cross ow velocity
L Cb
within bre have been utilized to evaluate the mass transfer coef-
Including temperature correction, the average Sherwood number cient in Eq. (54). The governing equation of volume at any time
is point is given by Eq. (41).
 d
1/3  T (n+1)/3  C 1/3 Eqs. (41), (60), (62), (63) and (64) present a system of
(2/3)C0 (Cg C ) 0.14 0 g differential-algebraic equations. These are solved numerically
ShL =1.816 Re Sc (e b )
L T Cb using fourth order RungeKutta method. Thus, the time proles of
volume reduction factor (VRF), retentate concentration, gel layer
(54)
thickness, etc., are obtained.
Average dimensionless permeate ux becomes
 d
1/3  T (n+1)/3 3. Materials and methods
(2/3)C0 (Cg C ) 0.14 0
Pew = 1.816 Re Sc (e b )
L T
Modied poly(ether ether ketone) (PEEKWC) hollow bre ultra-

 1/3  2/3 
Cg Cg ltration membranes were prepared according to the dry-wet
(55) spinning process [17]. Membrane modules were prepared by
Cb Cb
embedding three to ve hollow bre membranes inside a 20 cm
1/3 2/3
long glass tube (effective membrane length 18 cm). Kiwi fruit juice
For Cg /Cb < e3 , that is Cg  20Cb , (Cg /Cb ) (Cg /Cb ) , is was claried by using a laboratory bench plant (DSS LabUnit M10,

reduced to ln(Cg /Cb ). Under this condition, the nal expression of Danish Separation System AS, Denmark) equipped with a mem-
length averaged permeate ux is brane module having an effective membrane area of 20 cm2 . The
 1/3  T (n+1)/3  
d (2/3)C0 (Cg C ) 0.14 0 Cg inside and outside diameter of the bre are 1.41 mm and 1.96 mm,
Pew =1.816 Re Sc (e b ) ln respectively. UF experiments were performed according to the total
L T Cb
recycle and the batch concentration mode. The effect of transmem-
(56) brane pressure (P), feed ow rate (Qf ) and temperature (T) on the
The ux vw can be expressed using the phenomenological equation permeate ux was evaluated in UF experiments performed accord-
ing to the total recycle conguration in which both permeate and
P retentate streams were recycled back to the feed reservoir in order
vw = (57)
(Rm + Rg ) to maintain the volume and the composition of the feed constant.
where Rm is the hydraulic membrane resistance determined exper- P was varied in the range of 0125 kPa; Qf was in the range of
imentally and Rg is the gel layer resistance. The gel layer is assumed 1050 l/h corresponding to a mean velocity of 0.592.96 m/s; T was
to be a deposit of porous cake, using the ltration concept. Hence, varied between 15 and 35 C (Fig. 1).
the gel layer resistance and its characteristics are described with the UF experiments were also performed in selected operating con-
platform of traditional cake ltration theory [30]. So, Rg is expressed ditions (P = 75 kPa, Qf = 40 l/h, T = 25 C) according to the batch
as concentration mode (the permeate was collected separately and
the retentate was recycled back to the feed reservoir) up to a nal
Rg = (1 g )g H (58) volume reduction factor (VRF) of 2.05.
where is the specic cake resistance, g is the porosity of the cake,
and g is the density of the cake layer. Since g , g are all constants
during the experiment, the product (1 g )g is clubbed together
into a single parameter and is treated as another constant () during
the course of the simulation.
Thus
Rg = H (59)
Combining Eqs. (47) and (59), the governing equation of gel layer
thickness becomes
dH Cg Cb exp(P/(k(Rm + Rg )))
g = vw (60)
dt 1 exp(P/(k(Rm + Rg )))
The initial condition of the above equation is
H= 0 at t = 0 (61)
From Eq. (43) the bulk concentration becomes
Fig. 1. Scheme of the UF pilot laboratory plant (1, feed tank; 2, feed pump; 3 and 6,
C0 V0
Cb = (62) manometers; 4, membrane module; 5, thermometer; 7, heat exchanger; 8, pressure
V valve; 9, owmeter; 10, permeate; 11, digital balance).
300 S. Mondal et al. / Journal of Membrane Science 366 (2011) 295303

5 40

n
viscosity = B/T

Steady-state permeate flux (l/m h)

Values of B and n are estimated to be 35
4

2
B= 17 0.12; n =0.6 0.05
Viscosity (mPa.s)

correlation coefficient r1 = 0.998 30

3
experimental points
25 simulated results
2
20 correlation coefficient r1 = 0.985

1
15

0
15 20 25 30 35 40 10
0
14 16 18 20 22 24 26 28 30 32 34 36
T ( C) 0
T ( C)
Fig. 2. Variation of kiwi fruit juice viscosity with temperature (suspended solids Fig. 4. Effect of temperature on steady state permeate ux (operating conditions:
concentration = 10%). P = 50 kPa; Qf = 40 l/h).

4. Results and discussions
The viscosity of kiwi fruit juice having suspended solids of 10%
was analyzed by increasing the temperature (Fig. 2) of the juice
from 16 to 40 C. The value of the exponent n in Eq. (33) was
obtained by curve tting of the experimental data points and it
was found to be 0.6 0.05. Viscosity decreased by increasing the
temperature and also by decreasing the suspended solids concen-
tration.
4.1. Total recycle mode
Fig. 3 shows the effect of the ow rate on the steady state
permeate ux at a transmembrane pressure drop of 50 kPa and a
temperature of 25 C. The ow rate is varied from 10 to 50 l/h while
the steady state permeate ux correspondingly increases from 16 to
32 l/(m2 h). With increase in cross ow velocity, the concentration
polarization decreases due to enhanced forced convection, leading
to an enhancement in mass transfer coefcient and consequently
to the permeate ux.
The values of the parameters D, and Cg in Eq. (39) are estimated
by minimizing the sum of squares between the calculated and
36
34
Steady-state permeate flux (l/m h)
experimental data of steady state permeate ux at various oper-
ating conditions. The estimated values of these three parameters
are 3.6 1011 m2 /s, 0.007 0.0008 and 4.2 0.06, respectively. Cg
is the ratio of the gel concentration to the initial bulk concentra-
tion. The value of Cg is found to be 4.2 which signies that the
gel layer adds a considerable resistance to the driving force. The
results show that the value of Cg is well below the limit of 20.08.
It is to be noted here that the SiederTate viscosity correction fac-
0 g (2/3)C (C 1) 0.14
tor, (e ) has a signicant value of around 0.81. This
signies that the effect of viscosity has a strong inuence on the
ux decline mechanism. In the SiederTate correction factor, the
parameter includes the effect of viscosity. Physically, higher val-
ues of indicates the dominance of the viscosity variation on the
overall ux decline mechanism. It is clear from Fig. 3 that all the
experimental data, except that at Q = 10 l/h, are within 5% of the
calculated values.
In Fig. 4 the effects of temperature on the steady state perme-
ate ux are described at a constant ow rate of 40 l/h and P of
50 kPa. Temperature in this case is varied from 17 to 32 C while
the steady state permeate ux increases from 20 to 38 l/(m2 h). It is
quite evident from the gure that on increasing the temperature,
the permeate ux increases. This is because on increasing the tem-
perature, the mass transfer coefcient kT increases according to Eq.
(35). It may be noted that the variation of mass transfer coefcient
with temperature can be derived from Eq. (35) and the concept
presented in Section 2.1.3 and the dependence of mass transfer
2(n+1)/3
coefcient with temperature is kT = kT0 (T/T0 )

32 . Here, T0 indi-
2

cates a reference temperature of 21 C and kT0 is the mass transfer

30
coefcient at this temperature, kT corresponds to mass transfer
28 coefcient at a different feed temperature T. As presented earlier,
26 the value of n for kiwi fruit juice is 0.6. Thus, the variation of mass
experimental points transfer coefcient with temperature is 1.067. Therefore, the varia-
24
simulated results tion of permeate ux with temperature is almost linear as shown in
22 this gure. For example, as temperature changes from 21 to 29 C,
correlation coefficient r1 = 0.986 the steady state ux increases from 24 to 33 l/(m2 h).
20
As observed from the experimental data, the steady state per-
18 meate ux shows a weak variation of ux with transmembrane
16 pressure drop, beyond 60 kPa [17]. This is true for a gel layer con-
trolled ltration. However, it is reported in the literature that even
14 in the gel layer controlled ltration, the Sherwood number may
0 10 20 30 40 50 60
Qf (l/h) vary with transmembrane pressure drop [31,32]. In Fig. 5, the
variation of Sherwood number with transmembrane pressure is
Fig. 3. Effect of feed ow rate on steady state permeate ux (operating conditions: presented at a temperature of 25 C, a feed ow rate of 40 l/h and
P = 50 kPa; T = 25 C). suspended solids concentration of 10%. The experimental ux data
 S. Mondal et al. / Journal of Membrane Science 366 (2011) 295303 301

400 200
correlation coefficient r1 = 0. 9 4 3
2.0

350
150

Permeate flux (l/m h)

Sherwood number

1.5

2
300
*
Shexp = Pw/ ln Cg
VRF

VRF
100
Shtheory= Sh limit(1 - exp(- P)) 1.0
250
o permeate flux
value of = 0.05 0.01 and = 1.4 0.08
= (1.5 0.3) x10 m
16 -2
*
Cg = 4.2 0.06 already determined
50
200 0.5
correlation coefficient r1 = 0.932 correlation coefficient r = 0. 9 9 3

150 0 0.0
0 20 40 60 80 100 120 140 0 100 200 300 400 500
P (kPa) time (min)

Fig. 5. Effect of transmembrane pressure on the Sherwood number (operating con- Fig. 6. Time course of permeate ux and volume reduction factor (VRF) in the UF
ditions: T = 25 C; Qf = 40 l/h; suspended solids concentration = 10%). of kiwifruit juice according to the batch concentration mode (operating conditions:
P = 75 kPa; T = 25 C; Qf = 40 l/h).

points [17] are used to calculate the experimental Sherwood num-

ber. The transmembrane pressure is increased from 25 to 125 kPa 0.7
while the Sherwood number increases from 223 to 382. It is evident solid line - gel layer
from the gure that there exists a limiting value of the pressure 0.6 dotted line - retentate concentration 200

concentration of the retentate (kg/m )

3
beyond which the Sherwood number (hence, permeate ux) does
Height of the gel layer (mm)
16 -2
not increase. The value of the limiting pressure is about 75 kPa = (1.5 0.3) x10 m
0.5 180
and the corresponding limiting Sherwood number is 382. Variation
of Sherwood number with transmembrane pressure drop is tted
0.4
as 160

ShP = Shlimit (1
exp( P)) (65) 0.3
140
where the parameters
and  are found to be 0.05 0.01 and
0.2
1.4 0.08 kPa1 , respectively.
120
4.2. Batch concentration mode 0.1

The kiwi fruit juice, with a suspended solids concentration
of 10%, was claried according to the batch concentration con-
guration in optimal operating conditions (P = 75 kPa, T = 25 C,
Qf = 40 l/h) previously identied [17]. The intrinsic membrane resis-
tance was determined to be 1.12 1012 m1 , while the initial feed
concentration C0 was 100 kg/m3 .
The transient simulation of the ux decline mechanism is
obtained by simultaneously solving Eqs. (41), (60), (62), (63) and
(64). The value of the gel concentration and effective diffusivity (D)
are already evaluated earlier. Thus, in this case, only one parameter
() has to be evaluated. This parameter is evaluated by minimiz-
ing the sum of the least squares between the experimental and
the calculated ux decline data. The value of  is estimated as
(1.5 0.3) 1016 m2 . Fig. 6 shows the decline of permeate ux
as well as VRF with time. The decline of ux is very rapid at the
initial few minutes. Beyond 30 min, the ux continues to decrease
gradually with time due to enhanced concentration polarization.
Fig. 5 shows remarkable agreement between the calculated and
the experimental observations. As the experiment progresses, the
volume of the feed (or retentate) decreases, leading to an increase
in VRF as shown in this gure. Thus, to maintain the mass con-
servation, the concentration of the feed increases with time as
represented by Eq. (43) and this variation is illustrated in Fig. 6.
For example, at 464 min, the VRF is 2.1 and the bulk concentration
is 210 kg/m3 . The variation of the thickness of the gel layer with
time is also presented in Fig. 7. The gel layer thickness increases
with time non-linearly and it goes up to 0.6 mm at the end of the
experiment. It may be mentioned here that the Sherwood number
0.0 100
0 100 200 300 400 500
time (min)
Fig. 7. Time course of gel layer thickness and retentate concentration in the UF of
kiwi fruit juice.
decreases from 307 to 258 during the entire course of the experi-
ment.
5. Conclusion
A gel layer controlled model has been proposed to quantify the
ux decline behaviour during ultraltration of kiwi fruit juice. The
model is developed from the rst principles. The model is proposed
both for total recycle mode and batch concentration mode. The
model parameters are evaluated by comparing the experimental
data with calculated values. In the steady state model, effects of
temperature, pressure, viscosity variation with concentration have
been incorporated, successfully. Parameters obtained from steady
state model are: gel layer concentration, viscosity correction factor,
effective solute diffusivity. These parameters are system specic
and for kiwi fruit juice, these are 4.2 0.06, 0.007 0.0008 and
3.6 1011 m2 /s, respectively. For the batch concentration mode,
the theory is suitably modied, considering the time variation of
bulk concentration, gel layer thickness, etc. With one adjustable
parameter,  (1.5 0.3 1016 m2 ) and using previously estimated
parameters, the model successfully predicts the permeate ux
302 S. Mondal et al. / Journal of Membrane Science 366 (2011) 295303

decline, VRF, bulk concentration, gel layer thickness proles with

operating time. The experimental data obtained for the kiwi fruit
juice tted well in the developed model that can be successfully * non-dimensional thickness of the boundary layer
applied, as general model, to other pulpy fruit juices to predict the g porosity of the gel layer
permeate ux decline in real industrial applications.  adjustable parameter in Sherwood number with
pressure relation (kPa1 )

adjustable parameter in Eq. (65)
 viscosity of the solution (mPa s)
Nomenclature
m viscosity of the solution at mean concentration
A product of Reynolds number, Schmidt number and (mPa s)
d/L, u0 d2 /DL 0 viscosity of the pure liquid, i.e. without any sus-
Am effective area of the membrane surface (m2 ) pended solids (mPa s)
C concentration (kg/m3 ) w viscosity of the solution near the wall (mPa s)
C0 initial concentration (kg/m3 ) f density of the feed solution (m3 /kg)
Cb bulk concentration (kg/m3 ) g density of the gel layer (m3 /kg)
Cg gel concentration (kg/m3 ) p density of the solute particles (m3 /kg)
Cm mean solution concentration (kg/m3 )  product of (1 g )g (m2 )
Cp permeate concentration (kg/m3 )
C* non-dimensional concentration
Cg non-dimensional gel concentration References
Cm non-dimensional mean concentration
d membrane pore diameter, as a function of time (m) [1] B. Girard, L.R. Fukumoto, Membrane processing of fruit juices and beverages:
a review, Critical Reviews in Food Science and Nutrition 40 (2000) 91157.
di membrane pore diameter initially (m) [2] P. Rai, C. Rai, G.C. Majumdar, S. DasGupta, S. De, Storage study of ultral-
D effective diffusivity (m2 /s) tered mosambi (Citrus sinensis [L.] Osbeck) juice, Journal of Food Processing
D0 initial effective diffusivity (m2 /s) and Preservation 32 (2008) 923934.
[3] P. Rai, G.C. Majumdar, S. DasGupta, S. De, Effect of various pretreatment meth-
H thickness of the gel layer (m)
ods on permeate ux and quality during ultraltration of mosambi juice,
j1 mass ux per unit area (kg/(m2 s)) Journal of Food Engineering 78 (2007) 561568.
kT mass transfer coefcient at any temperature (T) [4] A. Cassano, L. Donato, E. Drioli, Ultraltration of kiwifruit juice: operating
(m/s) parameters. Juice quality and membrane fouling, Journal of Food Engineering
79 (2007) 613621.
kT0 mass transfer coefcient at any temperature (T0 ) [5] P. Rai, G.C. Majumdar, G. Sharma, S. DasGupta, S. De, Effect of various cutoff
(m/s) membranes on permeate ux and quality during ltration of mosambi (Citrus
Pew non-dimensional ux sinensis L.) Osbeck juice, Food and Bioproducts Processing 84 (2006) 213219.
[6] L.R. Fukumoto, P. Delaquis, B. Girard, Microltration and ultraltration ceramic
Q volumetric ow rate (m3 /s) membranes for apple juice clarication, Journal of Food Science 63 (1998)
r local radius of the tube (m) 845850.
r1 correlation coefcient [7] G.T. Vladisavljevic, P. Vukosavljevic, B. Bukvic, Permeate ux and fouling resis-
tance in ultraltration of depectinized apple juice using ceramic membranes,
R radius of the tube (m) Journal of Food Engineering 60 (2003) 241247.
Rg resistance of the gel layer (m1 ) [8] E. Hernandez, C.S. Chen, P.E. Shaw, R.D. Carter, S. Barros, Ultraltration of orange
Rm hydraulic resistance of the membrane (m1 ) juice: effect on soluble solids, suspended solids, and aroma, Journal of Agricul-
tural and Food Chemistry 40 (1992) 986.
Re Reynolds number [9] R Jiraratananon, A. Chanachai, A study of fouling in the ultraltration of passion
Sc Schmidt number fruit juice, Journal of Membrane Science 111 (1996) 39.
Sh local Sherwood number [10] S.T.D. De Barros, C.M.G. Andrade, E.S. Mendes, L. Peres, Study of fouling mech-
anism in pineapple juice clarication by ultraltration, Journal of Membrane
ShL average Sherwood number, averaged along the
Science 215 (2003) 213224.
length L [11] V.M. Matta, R.H. Moretti, L.M.C. Cabral, Microltration and reverse osmosis for
Shlimit limiting Sherwood number clarication and concentration of acerola juice, Journal of Food Engineering 61
ShP Sherwood number corresponding to transmem- (2004) 477482.
[12] A. Cassano, C. Conidi, R. Timpone, M. DAvella, E. Drioli, A membrane-based pro-
brane pressure drop P cess for the clarication and the concentration of the cactus pear juice, Journal
T temperature of the feed ( C) of Food Engineering 80 (2007) 914921.
T0 reference standard temperature ( C) [13] A. Cassano, E. Drioli, G. Galaverna, R. Marchelli, G. Di Silvestro, P. Cagnasso, Clar-
ication and concentration of citrus and carrot juices by integrated membrane
u velocity in x-direction (m/s) processes, Journal of Food Engineering 57 (2003) 153163.
u0 maximum radial velocity (m/s) [14] M. Pritchard, J.A. Howell, R.W. Field, The ultraltration of viscous uids, Journal
v velocity in y-direction (m/s) of Membrane Science 102 (1995) 223235.
[15] M.C. Vincent-Vela, E. Bergantinos-Rodrguez, S. lvarez-Blanco, J. Lora-
V volume of the retentate at any time t during the Garca, Ultraltration permeate ux decline prediction for gel layer forming
experiment (m3 ) solutes using monotubular ceramic membranes, Desalination 240 (2009) 89
V0 initial feed volume (m3 ) 93.
[16] B.R. Thakur, R.K. Singh, A.K. Handa, Chemistry and usage of pectin, Critical
vw permeate ux (m3 /(m2 s)) Reviews in Food Science and Nutrition 37 (1997) 4773.
x axial coordinate (m) [17] F. Tasselli, A. Cassano, E. Drioli, Ultraltration of kiwifruit juice using modied
y transverse coordinate (m) poly(ether ether ketone) hollow bre membranes, Separation and Purication
Technology 57 (2007) 94102.
x* non-dimensional axial distance
[18] P. Rai, C. Rai, G.C. Majumdar, S. DasGupta, S. De, Resistance in series
y* non-dimensional transverse distance model for ultraltration of mosambi (Citrus sinensis (L.) Osbeck) juice in
a stirred continuous mode, Journal of Membrane Science 283 (2006) 116
Greek symbols 122.
[19] P. Chayya, G.C. Rai, S. Majumdar, S. De, DasGupta, Mechanism of permeate ux
an adjustable parameter in SiederTate correction decline during microltration of water melon (Citrullus lanatus) juice, Food and
factor (m3 /kg) Bioprocess Technology 3 (4) (2010) 545553.
product of and C0 [20] W.F. Blatt, A. Dravid, A.S. Michaels, L. Nelson, Solute polarization and cake
formation in membrane ultraltration: causes, consequences and control tech-
P transmembrane pressure (kPa) niques, in: J.E. Flinn (Ed.), Membrane Science & Technology, Plenum Press, New
York, 1970, pp. 4775.
 S. Mondal et al. / Journal of Membrane Science 366 (2011) 295303
[21] R. Bird, E. Stewart, E.N. Lightfoot, Transport Phenomena, McGraw-Hill, New
York, 1960.
[22] A. Lvque, Les lois de la transmission de chaleur par convection, Ann. Mines
13 (1928) 201299.
[23] S. De, S. Bhattacharya, A. Sharma, P.K. Bhattacharya, Generalized integral and
similarity solutions for concentration proles for osmotic pressure controlled
ultraltration, Journal of Membrane Science 130 (1997) 99121.
[24] G.B. van den Berg, I.G. Rcz, C.A. Smolders, Mass transfer coefcients in cross-
ow ultraltration, Journal of Membrane Science 47 (1989) 2551.
[25] F.P. Incropera, D.P. DeWitt, Fundamentals of Heat and Mass Transfer, John Wiley
& Sons, Singapore, 1996.
[26] R.F. Probstein, J.S. Shen, W.F. Leung, Ultraltration of macromolecular solutions
at high polarization in laminar channel ow, Desalination 24 (1978) 116.
[27] E.L. Cussler, Diffusion: Mass Transfer in Fluid System, Cambridge University
Press, Cambridge, UK, 1998.
303
[28] M. Cheryan, Ultraltration and Microltration Handbook, Technomic Publish-
ing Company, Lancaster, PN, USA, 1998.
[29] S. De, P.K Bhattacharya, Modeling of ultraltration process for a two compo-
nent aqueous solution of low and high (gel-forming) molecular weight solute,
Journal of Membrane Science 136 (1997) 5769.
[30] S. Bhattacharjee, A. Sharma, P.K. Bhattacharya, A unied model for ux predic-
tion during batch cell ultraltration, Journal of Membrane Science 111 (1996)
243258.
[31] S.T. Johnston, W.M. Deen, Hindered convection of proteins in agarose gels,
Journal of Membrane Science 159 (1999) 271279.
[32] P. Rai, G.C. Majumdar, S. DasGupta, S. De, Modeling of permeate ux decline
of synthetic fruit juice and mosambi juice (Citrus sinensis (L.) Osbeck) in
stirred continuous ultraltration, LWT Food Science and Technology 40 (2007)
17651773.