Professional Documents
Culture Documents
Abstract
We have investigated the performance of two rotating disk prototypes for dynamic filtration designed in our
laboratory for the treatment of various industrial effluents by micro- (MF), ultra- (UF) and nanofiltration (NF). The
larger unit was built in stainless steel and could be equipped with two circular membranes of460 cm* area. Permeate
fluxes were compared with those obtained on the same effluents using a vibrating shear-enhanced filtration system
(VSEP) equipped with the same membrane or with regular crossflow filtration in tubular membranes. Permeate
fluxes obtained with the rotating disk fitted with vanes were consistently higher than those of the VSEP, by 80% in
MF and by up to 45% in NF at high pressures. Anionic effluents from a detergent plant, with 81,000 mg/l initial
chemical oxygen demand (COD), were treated using a 150-D NF membrane (DSS-DL). The permeate flux from the
disk unit increased linearly with transmembrane pressure to reach 200 I h-l m-* at 40 bar with a COD of 730 mg/l vs.
130 1h-’m-’for the VSEP equipped with the same membrane at its maximum frequency. These good performances
in UF and NF can be attributed to the very high shear rates, up to 220,000 s-’ at the membrane rim at 1500 rpm,
which reduce concentration polarization to much lower levels than in conventional crossflow filtration.
00 I l-91 64/03/$- See front matter 0 2003 Elsevier Science B.V. All rights reserved
PII: SOOll-9164(03)00436-3
80 Ii. Bourerar et al. I Desalination 158 (2003) 7945
;;‘
800
c 600 ) [Fe3’]=9g/L
5 T=25”C
500
2
; 400
g 300
z
p 200
Fig. 2. Variation of permeate flux with
100 gap peripheral pressure in MF of ferric
-D-18mm + 10mm A 5mm hydroxide suspension in saline medium
0 ‘-
using a smooth disk rotating at
20 30 40 50 60 ‘, ”
Peripheral pressure (kPa)
1500 ‘pm.
82 R. Bouzerar et al. ,, Desalination 1.58 (2003) 7945
1600
CaC03 50 pm
Nylon O.Z~m,1500 rpm 0
1400 L
1200
;;‘ 1000
E
2
800
a
5
600
concentration was restored to the initial value of the end ofthe test, there is an important hysteresis.
200 g/l by dilution and the final permeate flux was
at 1I40 I/hm’ lower than the initial one (I 400 I/hmz).
3.3. Concentration of silica suspension by UF at
Corresponding results obtained with a VSEP
150 kD
equipped with the same membrane are also shown
in Fig. 3 for comparison. VSEP fluxes, obtained The variation of permeate flux and inlet
at the resonant frequency of 60.75 Hz, are much pressure with silica concentration in percent of
lower (600 I/hm2) but drop very little at the mass is presented in Fig. 4, using a disk equipped
maximum concentration due to an increase in with vanes and rotating at 1500 rpm. The permeate
TMP from 1.4 to 2.3 bar caused by the increase fluxes, although high, are lower than in Fig. 3
in density and viscosity. However, when the because of the lower membrane permeability and
concentration is restored to 200 g/l by dilution at the much lower particle size (50 nm), which
350 1.6
1.4
300
I.2 f
;;‘ 250
.C 1 k
._
f 200 0.
0.8
‘5 150
0.6 Fig. 4. Variation of permeate flux
results in a more compact layer on the membrane. to a more favorable internal flow. Since the flow
Here again, the reduction in permeate flux, when induced by the pump is directed from the lower
silica concentration increases from 5 to 32%, is compartment to the upper one, it will be, on the
relatively small, from 300 to 245 l/hm2. Also shown lower membrane, opposite to the inward radial
for comparison are the permeate fluxes collected velocity induced by the disk rotation on the mem-
in a similar test using a ceramic tubular membrane brane due to the recirculation between the disk
with same cut-off and an axial fluid velocity of and the membrane. On the contrary, on the upper
5 m/s. These permeate fluxes are much lower than membrane, the radial velocity induced by the
for the rotating disk and decay by SO%, from 125 pump, and that of the recirculation induced by
to 60 l/hm? during the concentration test. the disk rotation, will be in the same direction.
Thus, the radial velocity component will be higher
3.4. Concentration by UF at 50 kD of silica on the upper membrane than on the lower one,
suspended in a Klebosol solution resulting in a slightly higher radial shear stress
and a higher permeate flux.
In these experiments dealing with a typical
effluent of a semiconductor plant (silica particles
suspended in a 2% ammoniac solution), the stain- 3.5. Depollution of effluents from detergent
less steel rotating disk module was fitted with a industry by UF and NF
second membrane identical to the first on the
3.5. I. Ultrafiltration at 5 and 5 kD
lower side of the disk, and the permeate fluxes
on each membrane were measured separately. A These effluents have large chemical oxygen
flat disk, smooth on both sides, was used in these demand (COD) contents, generally in excess of
tests. The variation of these permeate fluxes, 15,000 mg/l, and the purpose of the membrane
together with that of inlet pressure, is plotted as a filtration is to reduce the COD in the permeate to
function of silica percent mass concentration in acceptable levels. Tests were conducted on cationic
Fig. 5. The flux collected from the upper effluents with an initial COD of 17,300 mg/l and
membrane remains slightly higher than that from a 5 kD PES membrane at constant concentration,
the lower one. Since the hydraulic permeabilities recycling both permeate and retentate and in-
of these two membranes were measured to be the creasing inlet pressures to 30 bar. Results of these
same before the test, and transmembrane pressure tests conducted with a disk equipped with vanes
was not higher on the upper membrane, we can and rotating at 1500 rpm are shown in Fig. 6. The
attribute the higher flux on the upper membrane permeate flux keeps increasing with pressure to
1~ 1.8
140
I 1.6
120 1
I.4
g 100 ’ 1.2T
z2 80 1, 3
5 60 L 0.8 h
Concentration 1 : 12% -> 30%
40
0.6 Fig. 5. Comparison between permeate
, 0.4 fluxes of lower and upper membranes
20 o Jf_Lower (L/h/m2) a Jf_Upper (L/hlmz) A Pi (bar)
/ 0.2 in a concentration test of silica suspen-
0 -0 sion in Klebosol with a 50 kD PES
10 20 35
%Si 25 membrane.
84 R. Bouzemr et al. ’Desalination I58 (2003) 79-85
Catlontc effluent
100 Membrane POO5F5 kD _,_ ”
its maximum of 30 bar, where it reaches 110 I/hm’. were filtrated using a DESAL DS DL membrane
The COD in permeate was reduced to 2700 mg/l of 150 D equivalent cut-off and a disk equipped
and its conductivity, proportional to ionic content, with vanes and rotating at 1.500 rpm. The variation
to 95 pS/cm at 22 bar, from a feed value of of permeate flux with inlet pressure is represented
500 ps/c111. in Fig. 7 together with data obtained on a similar
Similar tests were performed on an anionic effluent, but with an initial COD of 49,500 mg/l,
effluent ofthe same origin, but with an initial COD using the VSEP equipped with the same mem-
of33,200 mg/l, using a 10 kD PES membrane under brane and operated at resonant frequency. The
the same experimental conditions. The permeate permeate flux provided by the rotating disk
flux increased linearly with inlet pressure up to increases linearly with inlet pressure up to 40 bar
12 bar and reached 300 I/hm’ at the maximum and is higher than that of the VSEP (200 I/hm’
pressure of 20 bar. The permeate COD was for the disk at 40 bar vs. 139 I/hm? for the VSEP
reduced to 5400 mg/l. at the same pressure). Permeate CODS were
similar in both systems (730 mg/l at 20 bar for
3.5. ,7. NmoJiltrution at I50 D the rotating disk vs. 760 for the VSEP at the same
pressure).
Other anionic effluents from the detergent
industry, but with an initial COD of 8 1,600 mg/l,
250
4. Conclusion Acknowledgment
The high performance of rotating disk filtra- The financial support of R. Bouzerar by the
tion devices, which has been widely reported in Valorisation Dept of CNRS is acknowledged.
MF, subsists also in UF and NF. The reason is
that the very high shear rates (up to 150,000 s?
References
for water at 20°C in our module at 1500 rpm, when
averaged over the membrane area, or 220,000 s-’ [I] U. Frenander and A.-S. Jiinsson, Biotech. Bioeng.,
at membrane rim, depending upon speed and disk 52 (1996) 397-403.
geometry) reduce concentration polarization to [2] S.A. Lee, A. Burt, G. Russoti and B. Buckland,
Biotech. Bioeng., 48 (1995) 386400.
much lower levels than in conventional crossflow
[3] C. Harscoat, M.Y. Jaffrin, R. Bouzerar, and J. Courtois,
filtration, especially when disks are equipped with
Biotech. Bioeng., 65 (1999) 50&5 11.
radial vanes that increase fluid velocity for the [4] 0. Al-Akoum, M.Y. Jaffrin, L.H. Ding, P. Paullier and
same disk rotation speed. We estimate from tests C. Vanhoutte, J. Membr. Sci., 197 (2002) 37-52.
conducted at 3000 rpm in the small unit that even [5] R. Bouzerar, M.Y. Jafii-in, A. Lefevre and P. Paullier,
higher permeate fluxes could be obtained by J. Membr. Sci., 165 (2000) 111-123.
increasing the rotation speed in the large device
and using higher vanes.