Escolar Documentos
Profissional Documentos
Cultura Documentos
MEMBRANE
SCIENCE
Journal of Membrane Science 100 ( 1995 ) 259-272
ELSEVIER
Abstract
Several constant-flux filtration experiments for yeast cell suspensions, yeast cell debris, and dodecane-water emulsion were
performed at various operating conditions in both fiat-sheet and tubular-membrane systems. The aim of the paper is two-fold.
Firstly the relationship between constant-flux behaviour and membrane fouling is discussed. In some cases constant-flux filtration
was realized at a constant transmembrane pressure which was below a critical value. In general constant-flux filtration was
obtained with moderately increasing transmembrane pressure, and this approach is shown to have some advantages over normal
constant-pressure filtration because it clearly provides for the possibility of avoiding over-fouling and so reduces the severity of
fouling. Secondly, the concept of critical flux is introduced. Whilst it has long been recognised that low-pressure microfiltration
is much more effective than high-pressure microfiltration, the emphasis in this work is upon the possible existence of a critical
flux and the desirability of starting filtration operations at a low flux. The critical-flux hypothesis is that on start-up there exists
a flux below which a decline of flux with time does not occur. Equations which may enable identification of the appropriate flux
level are included.
Keywords: Critical flux; Fouling; Membrane filtration;Dodecane-water emulsion; Yeast suspensions
1. I n t r o d u c t i o n
One of the major problems in pressure-driven
membrane processes is the reduction of the flux to far
below the theoretical capacity of the membrane. The
typical variation o f the flux with time is that of an initial
rapid decrease followed by a long and gradual flux
decline. It is well known that membrane fouling is one
of the main phenomena responsible for this. The occurrence of fouling affects the performance of the
membrane either by deposition of a layer onto the
membrane surface or by blockage or partial blockage
of the pores. This changes the effective pore size distribution.
* Corresponding author.
0376-7388/95/$09.50 1995 Elsevier Science B.V. All rights reserved
SSD10376-7388(94)00265-7
TMP
TMP
/z(Rm+Rf)
tz(Rm+Rir+Rr)
(1)
260
modes. Rep is the peak Reynolds number of the superimposed oscillatory flow. X and f are respectively
amplitude (in mm) and frequency (in Hz).
(Rep = p. Vp. D./.t- 1, where p is the density of the feed
suspension, D is the I.D. of membrane tube or the
equivalent diameter of a channel,/z is the viscosity of
the suspension and Vp = 21r.f.X is the peak velocity of
oscillatory flow.)
A summary of the experiments is given in Table 2,
and the details can be found in the related description
below.
2.2. Materials
Dodecane-water emulsion
Dodecane (Fluka, Switzerland, >98% GC) was
added to distilled water and was mixed at high shear in
a blender (Waring, USA) for 6 min. The emulsion of
0.1 wt.-% was stable during experiments.
2.3. Yeast cell suspension
Fresh bakers yeast (Saccharomyces cerevisiae) was
obtained from British Fermentation Products as compressed cake. The yeast was diluted down for use with
a solution of l0 mM phosphate buffer (pH 8.8) and 1
g l - 1 bacteriological peptone. Further details are given
elsewhere [4].
Table 1
Terminology to describe the various operating modes
Abbreviation Description
ET
B
BP
ETP
2. Experimental
2.1. Terminology
The terminology used to describe the various operating modes is defined in Table i.
Two Reynolds numbers were defined. Re is the
Reynolds number of the net forward flow in the BP
mode or cross-flow Reynolds number in the B and ET
Table 2
Summary of experimental set-up
Process fluid
Flat-sheet system
Tubular system
miD.-
261
Fig. 1. Helical baffle used for mode B runs with the Carbosep
membrane.
262
6LI
,o
tb
t~
ab
2~
3b
a~
4~
50
Fig. 2. Comparisonof flux versustime behaviourfor flow in a baffledsystem (BP), pulsedflow in an unbaffledsystem (ETP) and normalflow
in an unbaffledsystem (ET) as given in Ref. 14]. Flat-sheetsystemat C=5% dry weight yeast; TMP= 1.05 bar.
50BP moOe, V = 0.038 m/e
45-
40-
35-
=,
Vp=,0,38 fft/S
'
10.07
II
Incui~tKI YCD
151050
Y1CD
2o
eo
loo
12o
TInm (mlnuhm)
Fig. 3. Comparisonof filtrationperformance for 3% dry weight yeast cell debris and incubatedyeast cell debris at various TMPs with a V of
0.038 m s- ~and Vpof 0.38 or 0.74 m s- 1. The baffled/pulsedmode of the fiat-sheet system was used with a Ceramesh 0.2-/xm membrane.
Details in Ref. [5].
The filtration results for the non-incubated YCD at
low TMP (i.e. low initial flux) indicate a relatively
small degree of fouling. The reasons for the existence
of the constant flux with the incubated YCD are open
to conjecture. A m o n g the several elements responsible
for it, such as the characteristics of the feed, operating
mode and TMP, a low initial flux (i.e. very low TMP)
is definitely essential.
3.2. T u b u l a r s y s t e m
263
200-
~(15o;
Run
_
1
m
~-Run2
~J too;
U.
Run
50
OI
6b
eo
loo
1~
laO
1~o
18o
Time (minutee)
Fig. 4. Three sets of flux data of three constant-fluxfiltrationsfor a 1000-ppmdodecane-wateremulsionand with a Carbosep 0.14-/zmtubular
membrane.The data for the three runs were acquired at a Vof 0.84 m s- ~and in the baffled mode.
the flux constant for all three runs. This means that
gradual fouling occurred. From the authors' experience
of this mode of operation the chosen flux is of crucial
importance in order to obtain the best compromise
between flux and TMP. If the starting flux was too low
as shown by run 3, the flux was easy to be maintained
but at a low level. If too high, over-fouling would occur
and it would be impossible to increase T M P sufficiently
in order to maintain the flux constant. Fig. 6 compares
a constant-flux profile with two flux profiles for constant TMPs of 0.44 and 0.70 bar. The individual data
points for the constant T M P runs were obtained at the
same frequency as for the constant flux run but are not
0.8
0.7Run
0.60.5"
" o.4:
0.30.2-
,ooop p ~ , = , = ~ , s , , ~ . v = o 8, r,V,
0.10
2b
80
100
1~
140
180
180
Time (minutee)
2O0
264
-0.6
250D=,,-
0.5
200-
0.4
,~.
i
--
0.3
OA.4I ~
100-
I-0.2
50-
0.1
20
40
60
80
100
120
140
160
180
200
1'1rile (minute)
0.8
"0.7
"0.6
<
-0.5 "1~
150-
lil
o
-0.40.
-0.3 ~
0.~?I~"
-0.2
looo ~
~b
16o
==,~,
I~O
B,,~,
26o
-0.1
v = 1 ol . ~
~,o
36o
4oo
0
45O
250-
265
0.45
i 0.:35
!
* ~
~t.-~-t
-~1+" .~--~/~
0.3
io25
0.2
loo
rl
I--
0.15
i1
0.1
50
0,32_ bar
~)
4()
I~)
"
0.3.2 bar
-
130
100
120
0.05
, -140
J-O
160
Time (minutee)
Fig. 8. Comparison of the performance between constant-flux filtration and constant-pressure filtration for a 1000-ppm dodecane-water emulsion
and with a Carbosep 0.14-/xm tubular membrane. All data were obtained at a V of 0.71 m s t and in the conventional mode. For the constantpressure filtration, the T M P was varied between 0.32 and 0.80 bar.
266
"0.7
140-
a %o'wyearn.,B mode
-0.6
120-
100-
-0.5
I V = 0.84 rt~
-0.4 ~.
<
Is
-0.3 I-
r.
-0.2
V = 0.7e nVs
-0.1
OJ
20
40
60
80
100
120
140
160
Time (minutee)
Fig. 9. Comparison of two constant-flux filtrations for 5% dry weight yeast cell suspension and a Carbosep 0.14-/zm tubular membrane. The
two sets of data were obtained at a V of 0.76 and 0.84 m s- t, respectively, using the baffled mode.
membrane resistance. However, the water flux obtained
after rinsing the membrane at the end o f the experiment
indicated that none o f the fouling material had been
removed. The pre-experiment water flux was restored
by cleaning. This indicates that the small increase in
resistance during the 401 m - 2 h - l constant-flux experiment was not due to a flowing cake. The water flux
obtained after rinsing the membrane at the end of the
57 1 m - 2 h - i constant-flux experiment, during which
period the overall resistance increased four-fold, indi-
160-
-0.7
140.
-0.6
120-
-0.5
<~ 100-0.4
-0.3 iI,-
6o-
-0.2
40-
0.38 ~
201
0 - - -
-0.1
20
40
80
Time
8()
100
(mlnutee)
120
140
ld
Fig. 10. Comparison of the performance between a constant-flux filtration and a constant-pressure filtration for 5% dry weight yeast cell
suspension and with a Carbosep 0.14-~m tubular membrane. All data were obtained at a V of 0.84 m s- ~and in the baffled mode. For the
constant-pressurefiltration, the TMP was varied between 0.38 and 1.04 bar as shown.
267
250
Meuured ~ Q-I~O mVmin, TMP-0.4 ~
and 30 C
~I0"
15010050-
8
1'0
1;~
1'4
Number of exp~mente
16
18
20
Fig. l 1. Illustration of the membrane condition at the end of experiments involving yeast cell suspensions in the tubular system with and without
baffles. No baffle was used for experiments numbered 1, 2, 7 and 8. The helical baffle as shown in Fig. 1 was u ~ d for experiments 3~5 and 1118. A simple symmetrically located rod insert was used for experiments 9 and 10.
268
flux at the same TMP. The weak form is that on startup a constant flux is rapidly established and maintained.
The determination of critical flux is difficult to determine a priori because surface interactions of the species
involved can vary significantly from system to system.
Nevertheless it is suggested that the concept may be of
value, as a guide, when optimising a system. Theoretical calculations for idealised particles have been made
[ 8 ]. For small particles of the order of 0.1/~m StokeEinstein diffusion away from the surface will be important and calculations suggest a critical flux of 1-10/zm
s-1 depending on the significance of charge effects.
For particles of the order of 1/zm charge is less important and shear-induced diffusion alone, for non-interacting particles, should give a critical flux of the order
of 10/zm s - 1 (equates to 36 1 m - 2 h - ~). This is close
to the values of 25 and 401 m -2 h - ~determined experimentally for incubated YCD and 5% dry weight yeast.
The critical flux should not be confused with the
limiting flux found in ultrafiltration which occurs independently of fouling and is due to concentration polarisation. This phenomenon induces both osmotic effects
and reduced mass transfer (due to increases in the viscosity at the membrane surface). The physical phenomena governing the critical flux may be in doubt but
evidence for the existence of critical-flux effects is now
quite extensive. Two experimental observations will be
mentioned. Firstly, Benkahla et al. [9] have observed
that as TMP is increased the flux increases linearly and
provided a critical value of flux is not exceeded the
behaviour is reversible, i.e. pressure can be reduced
and the same fluxes are again observed. However, if
the critical flux value is exceeded then reducing TMP
does not restore the original flux but a lower one. The
cycling of pressure then leads to a hysteresis effect. The
value of the critical flux below which no hysteresis is
observed can be increased by increasing the cross-flow
velocity. Secondly, the work of Hodgson et al. [ 10] is
also relevant; non-intrusive observation of the surface
of a 0.02- and a 0.2-/zm Anopore membrane filtering
0.1 wt.-% yeast indicated that at low flux cells were not
observed to approach the surface.
Now the rate of convection of material towards a
membrane surface under low TMPs is much lower than
that which occurs during the initial period of conventional operation. The rate of back-diffusion (or tangential removal) of rejected material will depend upon its
concentration, material properties and the fluid
Finally the concept of critical flux immediately suggests that there may be advantages in operating below
it so as to minimise cleaning. In fact this has been
exploited by Kubota. Ishida et al. [ 16] reported on their
membrane-activated sludge process in which membranes are submerged in the aeration tank and operated
at constant flux. Under normal conditions flux is maintained at 0.5 m 3 m - 2 per day with a steady pressure of
about 30 kPa. Means to cope with storm overloads is
important and tests at 70 kPa showed that the flux would
initially increase to 1.05 m 3 m -2 per day but decline
gradually to 0.94 m 3 m - 2 per day due to fouling. Thus,
the new increased flux can be said to have been greater
than the critical flux for this system. On returning to
the original flux, which was below the critical flux, the
pressure required was one third higher than that originally required but the extent of fouling gradually diminished and after 4 h the pressure was about 35 kPa. The
advantages of operating in such a simple mode over
extended periods in unmanned systems is clear. Lastly
a second Japanese paper presented at ICOM '93 will
be mentioned. Whilst Yamamura et al. [ 17] attribute
the performance of the Toray "superfiltration
membrane" to low absorption of fouling materials and
the pore size range of 0.01-0.03/.tm, the operation at
moderate fluxes is also considered, by us, to be highly
relevant. The corresponding TMP was around 50 kPa.
5. Conclusions
For three different systems constant-flux filtration
was realized by gradually increasing TMP from an
appropriate starting value. Under certain circumstances
a constant flux can be achieved at a constant TMP as
shown in the filtration of yeast cell debris and 5% dry
weight yeast suspension. The critical and generally low
value of TMP below which constant-flux filtration can
be realized depends upon the hydrodynamics. Below
this critical TMP, there will be little or even no irreversible surface fouling.
Fundamentally the process is best understood in
terms of a critical flux; the value of TMP is not of
physical significance per se but is crucial in that for a
given membrane resistance it is related to flux and is
often the controlled parameter. By correctly selecting
the initial TMP the rate of fouling is greatly reduced
because a critical flux is not exceeded. Thus ideally a
269
Appendix
d2t
k( dt ~n
t791
(A1)
dZt
d( l/AJ)
dV
dV 2 =
1 dJ
1 dJdt
AJ 2 dV
AJ 2 dt dV
1 dJ
A2J 3
dt
1 dJ k
A2J 3
dt
(A2)
270
da.
-~lJo=-~lao
(A3)
(A4)
a filter resistance, which is assumed to remain unchanged, and a cake resistance. For dead-end filtration of
incompressible particles the cake thickness is proportional to the volume filtered and the overall resistance
is
R = Ro + am = Ro + ak V/A
dJ
(A5)
- ~ = erAJ - Bao
dt
(A6)
(A9)
l=~ofJdt
ak '
Jo
(A10)
R = Ro + a(k~V/A - f S d t )
(All)
dJ
dt
(A7)
(A8)
Jo
Jdt-JoRo
(A12)
Thus
l dJ
j2 dt
akJ
aS
JoRo JoRo
(A13)
(A14)
J Jo-Js~
G t = ~1 [ l n ( -j s k kJo
(1
1~]
(A15)
271
2. Summary
Intermediate blocking filtration law
Each particle does not necessarily block a pore; the
probability of landing upon particles already on the
surface is taken into account. The physical model leads
to
1
....
J Jo
trt
(A16)
trJ
(A17)
o-J-B'
(A18)
(A19)
(A20)
-~tJ~- Z=k(J-J )
(A21)
272
defined by Eq. A 1
V
a
(units depend on
mechanism)
cake filtration constant (kg m-3)
mass of cake per unit (kg m-2)
area
index defined by Eq.
(units depend on
A1
mechanism)
hydraulic resistance
(m - t )
rate of cake erosion per ( k g m - 2 s -1)
unit area
filtrate volume
(m 3)
specific resistance of
(m kg- 1)
cake
membrane surface
k
m
n
R
S
porosity
blocked area per unit ( m - ~)
volume of filtrate
Subscript
0
condition of time t = 0
tr
Acknowledgement
One of us (RWF) is grateful to the Royal Academy
of Engineering and Shell Research for financial support
and the award of a Senior Research Fellowship.
References
[ 1 ] P. Aimar and J.A. Howell, Effects of concentration boundary
layer development on the flux limitations in ultrafiltration,
Chem. Eng. Res. Des., 67 (1989) 255-261.
[2] J.A. Howell, Fouling and process design, Plenary Lecture, in
P. Aimar and P. Aptel (Eds.), Euromembrane 92, Pads, 1992
(Recents Progres en Genie des Procedes, Voi. 6, No. 22),
Groupe Francais de Genie des Procedes, Nancy, 1992, pp. 195207.
[ 3 ] M. Turker and J. Hubble, Membrane fouling in a constant flux
ultraliltration ceil, J. Membrane Sci., 34 (1987) 267-281.
[4] J.A. Howell, R.W. Field and D. Wu, Yeast cell microfiltration:
flux enhancement in baffled and pulsatile flow systems, J.
Membrane Sci., 80 (1993) 59-71.
[5l D. Wu, J.A. Howell and R.W. Field, Pulsatile flow filtration of
yeast cell debris: influence of pre-incubation on performance,
Biotechnol. Bioeng., 41 (1993) 998-1002.
[6] C.A. Romero and R.H. Davis, Global model of crossflow
microfiltration based on hydrodynamic particle diffusion, J.
Membrane Sci., 39(1988) 157-185.
[7] P H. Hodgson, G.L. Leslie, R.P. Schneider, A.G. Fane, C.J.D.
Fell and K.C. Marshall, Cake resistance and solute rejection in
bacterial microfiitration: the role of the extracellular matrix, J.
Membrane Sci., 79 (1993) 35-53.
[8] P. Bacchin, P. Aimar and V. Sanchez, A model for colloidal
fouling of membranes, AIChE J., in press.
[9] Y.K. Beakahla, A. Ould-Dfis, M.Y. Jaffrin, Hydrodynamic
anti-fouling mechanism in crossflow microfiltration, Poster
Presentation at ICOM '93, Aug. 30-Sept. 3, 1993, Heidelberg.
[ 10] P.H. Hodgson, V.L. Pillay and A.G. Fane, Visual studies of
crossfow microfiltration with inorganic membranes: resistance
of biomass and particulate cakes, 6th World Filtration
Congress, Nagoya, 1993, pp. 607--610.
[11] J. Hermia, Constant pressure blocking filtration laws.
Application to power-law non-newtonian fuids, Trans. Ind.
Chem. Eng., 60 (1982) 183-187.
[ 12] A. Koltuniewicz, Predicting permeate flux in ultrafiltration on
the basis of the surface renewal concept, J. Membrane Sci., 68
(1992) 107-118.
[13] M.H. Lojkine, R.W. Field and J.A. Howell, Crossflow
microfiltration of cell suspensions: a review of models with
emphasis on particle size effects, Trans. Ind. Chem. Eng., 70C
(1992).
[ 14] E.S. Tarleton and R.J. Wakerman, Understanding flux decline
in crossflow microfiltration. Part 1. Effects of particle and pore
size, Trans. Ind. Chem.Eng., 71A (1993).
[ 15] R.J. Baker, A.G. Fane, C.J.D. Fell and B.H. Yoo, Factors
affecting flux in crossflow filtration, Desalination, 53 (1985)
81.
[ 16] H. lshida, Y. Yamada, M. Tsuboi and S. Matsumura, Kubota
submerged membrane activated sludge process. Its application
into activated sludge process with high concentrations of
MLSS, Presentation at ICOM'93, Aug. 30~Sept. 3, 1993,
Heidelberg.
[ 17] H. Yamamura, Y. Fujii and T. Uemura, Advanced feed outside
hollow fibre superfiltration membrane with excellent
characteristics to remove the fouling components, Presentation
at ICOM'93, Aug. 30-Sept. 3, 1993, Heidelberg.