Abstract
Coiled and straight hollow-bre modules have been built and tested; the permeate ux obtained in ultraltration with these
two geometries is compared for two feeds: a colloidal bentonite suspension and a dextran solution. In the case of colloidal
suspensions, the secondary ows induced by the coiled geometry allow fouling to be reduced and the permeate ux is
multiplied by a factor of up to 2. An empirical relationship is proposed to express the limiting ux of permeate as a function of
both the velocity and some geometrical parameters of the coiled modules. Analogous results are obtained during the
ultraltration of dextran. It is also shown that under certain conditions almost no deposit was formed; the permeate ux under
these conditions is three times higher for coiled modules than for straight ones. For a given energy expenditure and
ultraltration process, the gain in permeate ux can reach a factor of 1.8. # 1999 Elsevier Science B.V. All rights reserved.
Keywords: Ultraltration; Hollow-bre modules; Membrane fouling; Secondary ows; Dean vortices
1. Introduction
Over recent years, much research work has been
devoted to the improvement of both the design and
operation of membrane separation modules [14]. It
has been found that Dean vortices, generated by a
coiled geometry, can be used to modify the hydrodynamics near the membrane surface. In the case of
two membrane processes using gasliquid contactors
[5,6], it was shown that the vortices improved mass
*Corresponding author. Present address: Laboratoire d'Etudes et
d'Applications de Procedes Separatifs (LEAPS), Universite d'AixMarseille, Avenue Escadrille Normandie Niemen, 13397 Marseille
Cedex 20, France. Tel.: +33-491-027776; fax: +33-491-288600;
e-mail: philippe.moulin@leaps.u-3mrs.fr
0376-7388/99/$ see front matter # 1999 Elsevier Science B.V. All rights reserved.
PII: S0376-7388(98)00333-0
110
Ptm
Lp0 Ptm :
0 Rm
(1)
Ptm
Lp Ptm :
0 Rm Ra
(2)
0.57
0.48
0.24
0.57
0.48
0.24
2S9
2C9
2c9
2S7
2C7
2c7
0.47
0.47
0.23
0.47
0.47
0.23
Active
length l (m)
20
20
5
9
10
3
Number of
fibres n
0.70
0.70
0.70
0.93
0.93
0.93
Int. diam.
di (mm)
2.05
2.05
0.25
1.22
1.36
0.21
Surface area
(m2)100
1
1
1
1
No. of
coil
11
4.1
11
4.1
Coil diameter
dc (cm)
1
10
6
1
10
6
Pitch b
(mm)
6.1
5.8
4.1
6.1
6.2
5.3
Numbering: the first number indicates the module series; S: straight; C: coiled (11 cm diameter); c: coiled (4.1 cm diameter); the second number indicates the hollow-fibre
diameter: 7 for 0.70 mm and 9 for 0.93 mm.
Total length
ltot 0 (m)
Name
Table 1
Characteristics of ultrafiltration hollow-fibre modules
112
di P
:
2v2 lT
(6)
(7)
:
(8)
0 J J0
The ux at the zero-deposit critical point is obtained
by plotting Ra vs. Ptm and extrapolating the curve to
the abscissa.
3. Experimental
3.1. Ultrafiltration of a colloidal suspension of
bentonite
The experimental apparatus is represented in Fig. 2.
The suspension (4 l) contained in a tank (A) is maintained under agitation (I) and at constant temperature.
The colloidal suspension used in this study is a
bentonite suspension whose the characteristics are
given in Table 2. A feed screw pump (PCM
P2MGI, Moineau, Vanves, France) (B) ensures the
circulation and a by-pass (G) allows the feed suspension to be pumped at low velocity. The permeate and
the retentate leaving the module (D) are recycled back
to the tank so that the composition in the tank can be
assumed constant if the quantity of matter retained on
the membrane is considered as invariant. The pressure
Fig. 2. Schematic diagram of experimental apparatus: tank (A), screw pump (B), flowmeter (C, C0 ), module (D), pressure indicators (E, E0 ),
retentate valve (F), by-pass (G).
Values
Manufacturer
Bentonite reference
Concentration
Solvent
Size
pH
Zeta potential
Conductivity
Turbidity
Kinematic viscosity at 208C
113
The ultraltration modules were built using cellulose-acetate hollow bres (BCDA 48 and BCDA 68,
Aquasource, Toulouse, France) having inner diameters of 0.93 and 0.70 mm, respectively. Two module geometries were used: straight and coiled. The
hollow bres are placed in a exible PVC tube which
is then wound around a rigid cylindrical tube so as to
form a coil with touching pitch. The bres are potted
in an epoxy plug at each end of the modules.
The characteristics of the modules are shown in
Table 1. The effective bre length is taken as the total
length minus the length of the resin plug. As the feed
ow is inside the bres, the hydraulic diameter used to
calculate the characteristic parameters (Reynolds
number, pressure drop) is the inner bre diameter.
Before beginning ultraltration, the reference state
of the membrane is measured by the pure water
permeability coefcient Lp0 at 208C; it can be estimated with an uncertainty of 5%.
The ultraltration of a bentonite suspension is carried
out at both constant temperature and constant cross-ow
velocity.Foreachvelocity,thepermeateuxismeasured
as a function of the transmembrane pressure. The pressure is initially set at a low value, then is increased by
successive stages until a limiting ux is reached (Fig. 3).
These values of limiting ux were corrected to 208C
Fig. 3. Variation of permeate flux and transmembrane pressure with time (module 2C9, bentonite suspension, C03.27 g kg1, di0.93 mm,
v1.70 m s1, Re1428, De0 131, T208C).
114
Values
Dextran
Manufacturer
Concentration
pH
Conductivity
Kinematic viscosity (T208C)
DT 2000
Pharmacia
1 g kg1
5.6
4.1 mS cm1
1.071106 m2 s1,
water1.005106 m2 s1
Table 4
Variation of permeability coefficient before and during the
ultrafiltration of dextran solution
Module
1010Lp0
(m Pa1 s1)
1010Lp
(m Pa1 s1)
2S9
2C9
2c9
2S7
2c7
2C7
6.1
6.2
5.3
6.1
4.1
5.8
2.9
3.6
3.5
3.5
3.1
3.5
Lp0 and Lp are the membrane permeability before and after dextran
adsorption, respectively.
115
Fig. 4. Variation of permeability coefficient before, during and after pre-adsorption of dextran (module 2C9, dextran solution, T208C).
Fig. 5. Comparison of predicted pressure drop with experimental values (module 2C7, water, di0.70 mm, dc11 cm, Re660v, De0 58v,
T208C).
straight module 2S9 on the one hand and for the coiled
module 2c7 on the other. Beyond a certain value of
cross-ow velocity, the values of the limiting ux
obtained in a coiled module are always higher than
those obtained in the straight one. The higher the ow
velocity, the greater is the improvement in mass
transfer in the presence of Dean vortices. The higher
116
Fig. 6. Effect of the velocity on permeate flux (module 2S9, bentonite suspension, C03.27 g kg1, di0.93 mm, Re840v, T208C).
Fig. 7. Effect of the velocity on permeate flux (module 2c7, bentonite suspension, C03.27 g kg1, di0.70 mm, Re660v, De0 99v,
T208C).
(9)
117
Table 5
In a straight module, relation between the limiting flux of permeate and the velocity (bentonite suspension)
Reference
[20]
[21]
[21]a
[22]
Present study
Suspensions
Bentonite
Bentonite
Bentonite
Bentonite
Bentonite
[0.033 g l ]
[1 g l1]
[3 g l1]
[1.431035103 g l1]
[3.27 g l1]
Configuration
Regime
Laminar
Laminar
Laminar
Laminar
Laminar
a
(Re<1600)
(Re<600)
(Re<200)
(Re<2100)
0.73
0.60
0.80
0.55
0.6
As Wandelt's work is not aimed at determining the limiting flux as a function of velocity, the two exponents were obtained from a small
number of measurements. This could explain the higher value of the exponent.
(10)
R2 0:998:
(11)
b105
R2
2S9
2C9
2c9
2S7
2c7
2C7
2.71
3.93
4.62
3.24
5.85
4.29
0.60
0.85
0.96
0.60
0.88
0.84
0.998
0.998
0.998
0.997
0.982
0.990
R2 0:98:
(12)
118
Fig. 8. In laminar regime and for straight modules, limiting flux of permeate vs. shear stress (module 2S9 and 2S7, bentonite suspension,
C03.27 g kg1, T208C).
Fig. 9. Permeate flux vs. transmembrane pressure at a given velocity for different geometry of modules (bentonite suspension,
C03.27 g kg1, di0.93 mm, v0.80 m s1, Re672, T208C).
number of 1400, and over the whole range of transmembrane pressure, the 2c7 module gives a permeate
ux for the ultraltration of a bentonite suspension
119
Fig. 10. Permeate flux vs. transmembrane pressure at a given velocity for different geometry of modules (bentonite suspension,
C03.27 g kg1, di0.70 mm, v0.80 m s1, Re525, T208C).
Fig. 11. Limiting permeate flux vs. velocity for different module geometries (bentonite suspension, C03.27 g kg1, di0.93 mm, Re840v,
T208C).
obtained:
J 2C9=J 2S9 0:5De00:25 ;
R2 0:996;
(13a)
0:36
R2 0:996;
(13b)
120
Fig. 12. Limiting permeate flux vs. velocity for different module geometries (bentonite suspension, C03.27 g kg1, di0.70 mm, Re660v,
T208C).
Fig. 13. For coiled modules, experimental values of limiting flux of permeate vs. values calculated by Eq. (12) (bentonite suspension,
C03.27 g kg1, T208C).
0:28
0 0:24
R2 0:979;
R 0:987:
(13d)
121
Fig. 14. Variation with the transmembrane pressure of additional resistance due to the filter cake for different module geometries (bentonite
suspension, C03.27 g kg1, di0.93 mm, v0.80 m s1, Re672, T208C).
Table 7
Characteristics of large hollow-fibre modules
Characteristics
Large straight
module
Large coiled
module
0.85
0.701.26
450
0.84
4.71010
0.85
0.701.26
436
0.82
4.41010
R2 0:986:
(14)
122
Fig. 15. For large modules, permeate flux vs. transmembrane pressure (bentonite suspension, C03.27 g kg1, di0.70 mm, v0.35 m s1,
Re231, T208C).
Fig. 16. Comparison of experimental limiting permeate flux with calculated values (Eq. (12)) (bentonite suspension, C03.27 g kg1,
di0.70 mm, v0.35 m s1, Re231, T208C).
Relations
Reference
Dextran
Straight
Straight
2S9
2S7
Jlim2.3105v0.4
Jlim1.6105v0.4
Jlim1.7105v0.4
Jlim1.9105v0.4
[27]
[28]
This study
This study
DT
DT
DT
DT
70
70
2000
2000
0:5
(15)
For modules 2c9 and 2c7 (dc4.1 cm) this correlation does not t as well. This is probably due to the
123
(16a)
0:1
(16b)
Fig. 17. Permeate flux vs. transmembrane pressure at a given velocity for different module geometries (dextran solution, C01 g kg1,
di0.70 mm, v0.40 m s1, Re294, De0 38 (2c7); 23 (2C7), T208C).
124
Fig. 18. Limiting permeate flux vs. velocity for different module geometries (dextran solution, C01 g kg1, di0.93 mm, Re840v,
T208C).
Fig. 19. Limiting permeate flux vs. velocity for different module geometries (dextran solution, C01 g kg1, di0.70 mm, Re660v,
T208C).
125
Fig. 20. Variation of additional resistance due to the concentration polarisation with the transmembrane pressure and the module geometries
(dextran solution, C01 g kg1, di0.70 mm, v0.40 m s1, Re294, De38 (2c7); 23 (2C7), T208C.
6. Discussion
Table 9
Critical Dean number for various processes
Process
Critical Dean
number
23.4
25
25
19
1628
ques are used, together with optical and magneticresonance imaging methods to characterise Dean vortices in a curved channel. The authors use a critical
Dean number Dec to characterise that ow rate above
which the ow pattern changes and vortices are
detected. We can dene the critical Dean number in
membrane operations as the value of the Dean number
beyond which an improvement in mass transfer
occurs: Table 9 summarises the critical Dean numbers
found in various studies. There is an agreement
between all these values obtained with coiled geometries for different applications. When the Dean
number is below 20, the secondary ows do not have
a measurable effect.
126
Fig. 21. Energy consumption vs. transmembrane pressure (module 2C9, bentonite suspension, C03.27 g kg1, di0.93 mm, v0.5 m s1,
Re420, T208C).
(17)
The dissipated power is expressed in W, the volumetric ow rate in m3 s1 and pressures in Pa. The
energy expenditure per unit volume of permeate (in
kW h m3) is
Pd
QP
E
Ptm
3:6 106 :
3:6 106 Qp
Qp
(18)
127
Fig. 22. Ratio of limiting permeate flux between a coiled and a straight module vs. energy consumption (bentonite suspension,
C03.27 g kg1, di0.93 mm, Re840v, T208C).
Fig. 23. Ratio of limiting permeate flux between a coiled and a straight module vs. energy consumption (dextran solution, C01 g kg1,
di0.70 mm, Re660v, T208C).
128
Fig. 24. Variation of Sherwood number with energy consumption (water, modules 1S-1D, di3.2 mm and 2S-2D di2.4 mm).
Pd QP:
(19)
129
8. Nomenclature
References
b
d
De
De0
E
f
J
lT
Lp
P
P
Ptm
Pd
Q
Re
Ra
Rm
Sh
t
T
v
Greek letters
Subscripts
a
c
i
lim
m
p
s
tm
0
additional
coil, coiled
internal
limiting
membrane
permeate
straight
transmembrane
solvent
130