Вы находитесь на странице: 1из 18

THEORETICAL AND EXPERIMENTAL STUDIES ON

ULTRAFILTRATION (UF) PROCESS FOR MILK


CONCENTRATION
A. H. BAHNASAWY* and M. E. SHENANA**
* Agricultural Engineering Department
e-mail: bahnasawyadel@hotmail.com
** Food Science Department,
oshtohor, Faculty o! Agriculture,
"enha #ni$ersity, Egypt
".%. "%& '()(*
ABSTRACT
Ultrafiltration (UF) is an important process in the food industry, particularly
for dairy applications such as concentration of milk. The major problem in membrane
separation process is decline in flux over time of operation. This flux decline is
attributed to the fouling of membrane. embrane fouling is affected by three major
factors, namely, the membrane material properties, the feed characteristics and the
operating pressure. T!o models of UF process (resistance and diffusion) !ere
developed to predict the permeate and retentate fluxes decline during concentration of
milk as a function of time under different transmembrane pressures. "xperiments
!ere carried out to study the effect of different transmembrane pressures on the
permeate and retentate fluxes. #lso, milk concentration !as determined !ith time at
different pressures. Flux recovery during cleaning cycle !as also recorded. The
experimental results !ere used to validate the model results. The experimental results
!ere compared !ith the predicted data and they !ere in a good agreement !ith each
other.
Keywords: Ultrafiltration, permeate, retentate, milk, pressure, model, flux decline,
fouling, flux recovery.
INTRODUCTION
embrane separation processes for li$uid systems are conventionally classified in
terms of the si%e ranges of materials separated ( microfiltration, &' (m)'.& (m*
1
ultrafiltration, '.& (m)+ nm* nanofiltration, + nm)'.+ nm* reverse osmosis '.+ nm).
embrane ultrafiltration is a pressure driven process !hich is capable of separating
macrosolutes or colloidal particles from a solvent or smaller solutes. These particles
may be inorganic (metal oxides), polymeric (lattices) or biological (proteins). The
ultrafiltration process has become particularly important for concentrating
proteinaceous solutions. "xamples of commercial membrane processes involving the
filtration of protein solutions in the presence of electrolytes include the concentration
of !hey proteins in the dairy industry, protein recovery from blood plasma and
protein concentration in do!nstream processing. Ultrafiltration performance is
limited, ho!ever, due to the build up of the solutes at the membrane surface. This is
the so,called concentration polari%ation effect (Bowen nd !"##"$s% &''())
R" e* #) (&''() reported that the major problem in membrane separation
process is decline in flux over time of operation. This flux decline is attributed to the
fouling of membrane. embrane fouling is affected by three major factors, namely,
the membrane material properties, the feed characteristics and the operating
parameters. embrane fouling is of t!o types, namely, reversible and irreversible
fouling. -eversible fouling is mainly caused by a phenomenon, kno!n as
concentration polari%ation. .oncentration polari%ation is the accumulation of solute
particles over the membrane surface. This phenomenon is predominantly a function of
membrane channel hydrodynamics. /n case of reversible fouling, the membrane
permeability is recovered significantly after a proper !ashing protocol.
There have been some theoretical approaches to predict the ultrafiltration
performance of colloidal solutions (e.g., milk). These are based on some models such
as mass transfer model (film theory), gel,polari%ation model, osmotic pressure model,
boundary layer,adsorption model, 0ro!nian diffusion model, shear,induced diffusion
model, inertial lift model and surface transport model (A"$r nd F"e#d% +,,&) and
(B-er e* #)% +,./)) /n addition to the complexity of mathematical e$uations
involved, each of these models has a number of limitations1 (i) they demand some
experimental data for determining the input parameters. 2erhaps this is al!ays
possible in practice, but the e$uipment re$uired are especially sensitive instruments,
!hich might not be readily available. (ii) 3one of the methods can describe the full
flux)time behavior of process* they often predict the steady or pseudo,steady,state
flux. (iii) "ach one has been sho!n to be valid for certain feeds under special
conditions. 4ence, modeling methods based on direct analysis of experimental data
appear to be good alternative to the models based on phenomenological hypotheses.
/t seems that the most important problem of using UF e$uipment is the flux decline
!hich is mainly affected the operational conditions especially the operational
pressure. /n the present study, it is aimed to develop a model to study the effect of the
operational pressure on the flux decline. Filter medium resistance during both
concentration and cleaning processes at different pressures !ith time !ill be
estimated. Total soluble solids !ill be predicted and compared !ith the experimental
values. #lso, flux recovery during cleaning in place (./2) !ill be predicted !ith time
at different operational pressures.
2
MODEL DE0ELOPMENT
/n this study t!o different models !ere proposed. These models are1 first is the
resistance model to study the medium resistance and their effect on the flux decline,
second is the diffusion model to study the T55 change and flux decline.
Res"s*n1e $ode#:
/n UF of milk concentration, permeate flux declines !ith time due to
membrane fouling, !hich is very complicated phenomenon caused by many chemical
and physical properties interactions. /t is believed that membrane fouling is a
dynamic process starting !ith pore blocking follo!ed by continuous cake formation
on the membrane surface. 2ore blocking is a fast process observed at the beginning of
UF for a clean membrane due to its high initial permeate flux. #s UF process goes
on, accumulation and deposition of particles on the membrane surface begin and gel
layer is formed. The layer resistance becomes dominant after the initial stage. # little
reduction in permeate flux is observed in experimental results, and assuming very fast
pore blocking, it could be supposed that the layer resistance is dominant resistance.
/gnoring the pore blocking resistance, the !hole resistance can be considered as
membrane resistance, deposited solute resistance and boundary layer resistance
(Mrs2## e* #)% +,,3)) Therefore, for UF of milk concentration using 6arcy7s la!,
the follo!ing e$uation can be !ritten1
) (
&
b d m m
+ + +
,
dt
d-
A
.
+ +

(&)
"$uation & could be re!ritten as1
) (
&
p m m
+ +
,
dt
d-
A +

(8)
9here, -
p
is polari%ed solute resistance, using conventional filtration theory, the
follo!ing e$uation can be derived1

,
_

m
p
p p
A
-/
+
(:)
.ombining e$uations 8 and : and integrating gives the follo!ing filtration e$uation1
, A
+
-
, A
/
-
t
m
m
m
p

8
8
(;)
3
5olution flo! -etentate
.
g
2ermeate
C
4
embrane
0oundary layer <el layer
Fig. &. schematic of formation of external concentration boundary layer and gel layer
over the membrane surface.
The 5perry e$uation has been modified to account for changing resistance !ith
increasing time. De # 56 nd Bo7#*on (+,.8) assumed #lpha to be constant and
modified the 5perry e$uation such that the filtration rate is a po!er function of filtrate
volume.
1
1
]
1

,
_

m
n
+
A
-
c
, A d-
dt

(+)
/ntegration of e$uation + gives1
1
]
1

+
+

+
- +
n
A - c
,A
t
m
n
&
) = (
&

(>)
0ayindirli et al., &?@?, tested e$uation + and discovered that a common n in e$uation
> could not be found to describe all data at different body feed concentration.
.onse$uently, as alternative e$uation !as proposed !hich combines the specific cake
resistance and solids concentration into a parameter, k.
[ ]
A 0-
m
e +
, A d-
dt
=


(A)
/ntegration of e$uation A gives1
4
[ ]
A 0-
m
e +
,0
t
=


(@)
Taking a natural logarithm of e$uation @1
-
A
0
,0
+
t
m
+
1
]
1


ln ) ln(
(?)
D"997s"on Mode#:
/n cross,flo!, fluid flo!s over membrane at a fast rate. # laminar boundary layer of
thickness (B) exists at the membrane surface. /f . is the solid concentration in the
boundary layer and . is the concentration in the li$uid bulk, a mass balance of solids
entering the boundary layer !ith the solvent and those leaving the boundary layer by
diffusion !ill be (To#edo% +,,,) 1

,
_


,
_

1
/
D /
t
-
(&')
6 is the mass diffusivity of solids. /ntegrating !ith respect to x, designating the
transmembrane flux
t
-

CD, using the boundary conditions, .C . at xC B1


) ( ln 1
D
.
/
/

,
_

(&&)
#t the surface, xC', and .C.
s

D
.
/
/
s

,
_

ln
(&8)
5ince flux must be maximi%ed in any filtration process, reduced polari%ation
concentration can be achieved at maximum flux if high li$uid velocities can be
maintained on the membrane surface to decrease the boundary layer thickness, B.
"$uation can be arranged to 1

,
_

/
/ D
.
s
ln

(&:)
The ratio of 6= B may be represented by a mass transfer coefficient for the solids, k
s
.
Thus 1
5
This e$uation sho!s that a semilogarithmic plot of the bulk concentration, .,
against the transmembrane flux under conditions !here .
s
is constant is linear, !ith a
negative slope of &=k
s
. Thus, transmembrane flux decreases !ith increasing bulk
solids concentration, and the rate of flux decrease is inversely proportional to the mass
transfer coefficient for solids transport bet!een the surface and the fluid bulk.
ass transfer coefficient can be determined using 6ittus,0oelter e$uation for
turbulent flo! as follo!s1
:: . ' @ . '
-e '8: . ' Sc Sh
(&;)
,
(&+)

h
d
-e
, (&>)

D
Sc
(&A)
:: . ' :: . '
:: . '
@ . '
@ . ' @ . '
'8: . '

D
d
d
D
0
h
h
s

(&@)
;A . ' 8 . '
;A . ' @ . ' >A . '
'8: . '


h
d
D

d
h
is the hydraulic radius C ;(cross,sectional area)=!etted perimeter, m
ass diffusivity (6) !as estimated by Por*er (+,(,) using the follo!ing e$uation1
r
2 0
>

6
0

(&?)
To make the UF operation more clear and understandable to those !ho run or
redesign such e$uipment, a process calculation using "xcel as a program tool is used.
Using "xcel as a calculation tool, it is easy to calculate any UF process kno!ing only
a fe! basic entering data.
6
D
d 0
Sh
h s

Table (&) 12arameters that are used in the models1


Pr$e*ers :#7e 7n"*
9hole milk density
8
&':' kg=m
:
9hole milk Eiscosity
:
'.''@ 2a.s
9ater viscosity
&
'.''& 2a.s
2ermeate viscosity
&
'.'': 2a.s
k0 is 0olt%mann constant
8
&.:@&F&'
G8:
m
8
kg s
G8
H
G&
ilk initial concentration '.&: decimal
the molecular radius
&
(r) &.8 I &'
,?
m
embrane features1
total area '.@; m
8
membrane length &8' cm
3umber of modules :A ,
external diameter &.' cm
internal diameter '.> cm
edium resistance (-m)
For retentate ?.&"J'? &=m
For permeate +.A"J'@ &=m
.leaning solution :"J'A &=m
.ombined cake layer resistance, k
For retentate '.8AA &=m
For permeate '.&&A &=m
.leaning solution 8.@&A &=m
6iffusivity, 6 '.AI&'
,&'
m
8
=s
ass transfer coefficient in the diffusion
model, k
s
'.+;?@@ K=m
8
.min
Temperature :': H
&
-enner and #bd "l,5alam, &??&.,
8
Toledo, &??&,
:
2rokkopek et al., &?A+.
7
MATERIALS AND METHODS
M*er"#s:
Fresh !hole .o!7s and 0uffalo7s milk !as obtained from the herds of the
Faculty of #griculture, oshtohor, 0enha University, "gypt. T55, fat, protein
percentages and p4 !ere &:.', +.:, :.;+, and >.A8 , respectively.
UF e;7"<$en* des1r"<*"on:
The UF unit consists of 8''K stainless steel storage tank, t!o scre! pumps
(feeding and pressure pumps), 8 pressure gages, .arbosep 8 membrane
(?&.5c.:A.+8'>.&.'@ model, '&A': iribel, .odex, France) it has the follo!ing
specifications1 :A tube=module, total area of '.@; m
8
, membrane length is &8' cm, has
an external and internal diameters of &.' and '.> cm, respectively and empty !eight is
@.@ kg, The details of schematic diagram of the experimental set up are presented in
fig. 8.
Bypass
Permeate
Modules
Heat exchanger
Cooling
water
P TH
Retentate
p
Fig. 8. schematic of the experimental setup.
8
CP
FP
Feed
tank
Me*2ods:
-etentate and permeate fluxes !ere measured by dividing the volume by
product of the effective membrane and sampling time. Flo! velocity !as calculated
by dividing the flo! rate of both retentate and permeate by the cross,sectional area of
the UF module. -eadings !ere taken every &+ min.
uring !F e"periments t#e e$e%ts o& t#e operating pressure
on t#e permeate 'u" and retentate %on%entration steady state were
studied( )#e operating transmem*rane pressures were sele%ted as
3+ 4+ 5 and 6 *ars( ,a%# treatment needed 15- . o& milk and was
repli%ated 3 times+ and measurements a/erages were taken(
C#en"n= nd sn"*"6"n= *2e $e$4rne:
embranes !ere cleaned and saniti%ed at different pressures as follo!s1 flushing
!ater for &+ min, alkaline solution (3aL4, '.+M) for 8' min at A'
o
., then flushing
!ater for &+ min, nitric acid ('.:M) for 8' min at +' ., and finally, !ater !ashing for
&+ min. 6uring cleaning, the flux should be monitored continuously until it reaches
the maximum at the time proceeds. The cleaning efficiency !as evaluated by the
ratio of flux during cleaning to the pure !ater flux measured under the same
conditions for each cleaning procedure (B"rd nd Br*#e**% &''&))
RESULTS AND DISCUSSION
F#7> 4e2:"or:
Re*en**e 9#7>)
Figure : sho!s the experimental and predicted retentate fluxes at different
operational pressures (:, ;, + and > bars) !ith time. "xperimentally, it
could be seen that the retentate flux declines rapidly during the first ;'
min, slo!s do!n gradually during the period of ;' to &8' min, after this
period, flux seems to be constant until it reaches the end of concentration
process. -etentate flux increases !ith increasing the operational pressure.
This increment !as clear during the first +' min of process, it seemed to
have little difference by the end of process as affected by pressure. #fter
&+ min, the flux !as as high as 8.&A K=m
8
.s at > bars operational pressure,
!hile it !as as lo! of &.:+ K=m
8
.s at : bars. 0y the end of process (after
&@' min), flux decreased to reach '.&& K=m
8
.s at : bars and '.;& K=m
8
.s at >
bars. -etentate flux !as predicted as a function of time (at the range of &+
to &@' min) at different operational pressures (:, ;, +, and > bars). The
predicted retentate flux !as in a reasonable agreement !ith the
experimental ones. These results are in agreement !ith that has been
reported by Ton= e* #) (+,..)) /n early stage, of milk concentrating using
UF, adsorption fouling is probably the primary mechanism of flux decline
(M**2"sson% +,./)) /n the second stage, flux decline is probably due to
the concentration polari%ation. 0ut the majority of declination !as due to
the adsorption fouling.
0
Figure ; sho!s the relationship bet!een the average experimental and
predicted retentate fluxes (K=m
8
.s), !hich !as linear !ith coefficient of
determination of '.??.

Per$e*e 9#7>)
Figure + sho!s the experimental and predicted permeate fluxes at different
operational pressures (:, ;, + and > bars) !ith time. /t indicate that the
permeate flux declines rapidly during the first ;' min, slo!s do!n
1-
gradually during the period of ;' to &8' min, after this period, flux seems
to be constant until it reaches the end of concentration process. permeate
flux increases !ith increasing the operational pressure. #t : bars,
operational pressure, the flux decreased from &.A8 K=m
8
.s after &+ min to
'.&;: K=m
8
.s by the end of process (after &@' min), #t : bars, operational
pressure, the flux decreased from &.A8 K=m
8
.s after &+ min to '.&;: K=m
8
.s
by the end of process (after &@' min). #t ; bars, operational pressure, the
flux decreased from 8.8? K=m
8
.s after &+ min to '.&?& K=m
8
.s by the end of
process (after &@' min). #t + bars, operational pressure, the flux decreased
from 8.@A K=m
8
.s after &+ min to '.8:? K=m
8
.s by the end of process (after
&@' min). #t > bars, operational pressure, the flux decreased from :.;;
K=m
8
.s after &+ min to '.8@A K=m
8
.s by the end of process (after &@' min).
The predicted permeate flux !as in a reasonable agreement !ith the
experimental one.
Figure > sho!s the relationship bet!een the experimental and predicted
values, !hich !as linear !ith coefficient of determination of '.??.
11
To*# so#74#e so#"d "n $"#- * d"99eren* o<er*"on# <ress7res:
Figure A sho!s the total soluble solids (T55) as indicator for milk
concentration !ith time at different UF operational pressures
experimentally and theoretically. T5 !as '.&: as an average for the !hole
milk at the beginning of UF process. /t increased gradually until it reaches
around from '.&A to '.8 during the first +' min. #s the process goes on,
T5 increment !as slo!ing do!n until it reaches around '.8: !hich is
suitable for cheese making, and then the T5 of milk increases very slightly
!ith time at different UF operational pressures under study (:, ;, + and >
bars).
12
13
The predicted values !ere in a reasonable agreement !ith the experimental values
during the first A' min of process and it !as in a good agreement after that time until
the end of process. The relationship bet!een the predicted and experimental T55 of
milk !as sho!n in Figure @. This relationship !as linear !ith '.?&: coefficient of
determination, !hich may be attributed to effect of operational pressures in causing
membrane fouling !hich happens $uickly during the first +' min even before !hich
made variations bet!een both the predicted and experimental values of T55 during
the first +' min of process.
F#7> re1o:ery d7r"n= 1#en"n= "n <#1e)
Figure ? sho!s the flux recovery of the cleaning solution !ith time at different UF
operational pressures. /t is sho!n that the solution flux increased rapidly during the
first ;' min of cleaning process, !here it reaches 8+' K=m
8
.h, and then it !as
approximately constant until the end of the cleaning cycle. /n this process, initially,
small cleaning flux increases to a maximum, this occurs due to the removal of surface
deposits happened gradually !ith the cleaning time until it reaches the complete
cleaning, flux restores as a pure !ater. #fter &' min, flux recovery percentages !ere
+.8, &A.8, :&.' and ;'.; M at :, ;, + and > bar operational pressures., !hile, by ;'
min, flux recovery percentages !ere the same (@>.8M) at different pressures under
study. #t both : and ; bars, flux recovery M !ere the same by the end of cleaning
process (?>.++M), !hile it reach &''M flux recovery after ?' and @' min !hen it
!orks at + and > bar operational pressure
F"#*er $ed"7$ res"s*n1e:
F"#*er $ed"7$ res"s*n1e d7r"n= 1on1en*r*"on o9 $"#- 4y UF)
Figure &' sho!s the effect of operational pressure of UF during the concentration of
milk on the predicted medium resistance (-
m
) !ith the process time. /t indicated that
the -
m
increased linearly !ith the time at different operational pressures. -m
increases !ith increasing the pressure. #lso, it is !orthy to notice that at the higher
pressures (+ and > bars), -m seems to have no big difference as affected by those t!o
pressures, !hile there !ere big differences bet!een -
m
values !hen it !orks at : and
; bars. These results are in agreement !ith those obtained by -ai et al., (8''+a).

14
F"#*er $ed"7$ res"s*n1e d7r"n= 1#en"n= "n <#1e)
The model !as able to predict the medium resistance !ith the time of cleaning at
different operational pressures. Figure && sho!s the effect of operational pressure of
UF during cleaning process on the medium resistance (-
m
) !ith the time. /t indicated
that the -
m
increased linearly !ith the time at different operational pressures during
the first ;' min and then decreases. This maybe due to removing the surface deposits
and loose particles tend to block off the membrane !hich in turn increase the
resistance, after flushing this solution !ith these loose particles, -
m
decreases (0ird
and 0artlett, 8''8). The results revealed that -
m
increased !ith increasing the
pressure. -
m
reached the maximum (&.8x&'
&'
&=m) at the highest operational pressure
(> bars) after ;' min. mean!hile, it !as >x&'
&'
&=m at the lo!er pressure (: bars) at
the same time.
CONCLUSION
T!o models of UF process !ere developed to predict the permeate and retentate
fluxes decline during concentration of milk as a function of time taking into account
membrane resistance (resistance model) and concentration polari%ation (diffusion
model). Fouling status during cleaning se$uence cycle under different transmembrane
pressures !as also studied. The model has the capability to predict the follo!ing 1
, -ententate and permeate fluxes at different operational pressures !ith a
reasonable accuracy.
, Total solids in milk at different operational pressures.
, Flux recovery of the cleaning solution !ith time at different UF operational
pressures.
, Filter medium resistance during concentration of milk and during the cleaning
process.
, This model can be used to optimi%e the operating transmembrane pressure and
to identify the trade off bet!een productivity and !orking life of an
ultrafiltration membrane.
15
NOMENCLATURE
# surface area of the membrane (m
8
)
#m membrane area (m
8
)
. solid concentration, kg=m
:
.
N
concentration in the li$uid bulk, kg=m:
.
s
concentration at xC'
6 effective diffusivity, m
8
=s
k
0
0olt%mann constant, m
8
kg s
G8
H
G&
k
s
mass transfer coefficient, m=s
-e -eynolds number, dimensionless
5c 5chmidt number, dimensionless
5h 5her!ood number, dimensionless
t filtration time (h)
E filtration volume (m
:
)
D permeate flux, m
:
=m
8
s
cross,flo! velocity, m=s
2 transmembrane pressure, 2a
r the molecular radius, m
-
b
0oundary layer resistance (m
,&
)
-
d
deposited solute resistance (m
,&
)
-
m
membrane medium resistance (m
,&
)
filtrate viscosity (2a.s)
6ensity (kg=m:)
O specific cake resistance (m,&)
0oundary layer thickness, m
d
h
hydraulic radius, m
T temperature, H
REFERENCES
#imar, 2. and -. Field .&??8. Kimiting flux in membrane separations1 # model based
on the viscosity dependency of the mass transfer coefficient, /hemical
Engineering Science 8( (:) (&??8), pp. +A?)+@>.
0aker, -.D, #.<. Fane, ..D.6. Fell and 0.4. Poo. &?@+. Factors affecting flux in
crossflo! filtration, Desalination /3 (&?@+), pp. @&)?:.
0ayindirli, K., L%ilgen, ., 5. Ungan. &?@?. odeling of apple juice filtration. D.
Food 5ci. +;(;)1&'':,&''>.
0ird, .-. and . 0artlett. 8''8. easuring and modeling flux recovery during the
chemical cleaning of F membranes for the processing of !hey protein
concentrate. .ournal o! Food Engineering /3, pp. &;:)&+8.
0o!en 9., and . 9illiams. 8''A. Quantitative predictive modelling of
ultrafiltration processes1 .olloidal science approaches. #dvances in .olloid
and /nterface 5cience &:;)&:+ (8''A) :)&;
6e la <ar%a, F. and -. 0oulton. &?@;. The modeling of !ine filtration. #m. D. "nol.
Eitic. :+, &@?,&?+.
16
arshall, #.6., 2. #. unro, and <. Tragardh. &??:. The effect of protein fouling in
F and UF on permeate flux, protein retention and selectivity, 6esalination,
?&1&'@.
atthiasson, ". &?@+. Fouling in membrane filtration. /n Fouling and cleaning in
food processing. 6. Kund, ". 2lett and .. 5andu, Univ. 9isconson, adison
"xtension 6uplicating, adison, 9/. 2age ;8?.
2orter, . .. &?A?. embrane filtration. /n1 4andbook of separation Techni$ues for
.hemical "ngineers, 2.#. 5ch!eit%er, ed. ac<ra!,4ill 0ook .o., 3e!
Pork.
2rokopek, 6., Eoss, ". and Thomaso!, D. &?A+. olk. Rtg. 9elt der ilch, 8?, ?:?.
-ai, 2., <... ajumdar, 5. 6as<upta and 5. 6e. 8''+a. Quantification of flux
decline of depectini%ed mosambi (/itrus sinensis (K.) Lsbeck) juice using
unstirred batch ultrafiltration, .ournal o! Food ,rocess Engineering &. (8''+),
pp. :+?):AA.
-ai, 2., <... ajumdar, 5. 6as<upta and 5. 6e. 8''A. "ffect of various pretreatment
methods on permeate flux and $uality during ultrafiltration of mosambi juice,
.ournal o! Food Engineering (. (8''A), pp. +>&)+>@.
-enner, ". and . 4. #bd "l,5alam. &??&. #pplication of ultrafiltration in the dairy
industry. "l 5evier #pplied 5cience. Kondon and 3e! Pork.
Toledo, -. T. &??&. Fundamentals of food process engineering. Ean 3ostrand
-einhold, 3P.
Tong, 2.5., 0arbano, 6.5., and . #. -udan. &?@@. .haracteri%ation of proteinaceous
membrane foultants and flux decline during the early stages of !hole milk
ultrafiltration. D. 6airy 5ci., A&1>';,>&8.
17
! (UF)
"# $% &
?@ABCDE FGAH )G STUVW XYZ[ .\
S]^_`ab_ ScXVdb_ efg Shijk_ lmn^ efg
UdVo Sp[Uq ) rdstYo S^_`ab_ S]nu
1 v .w ) S]om]nxb_ SyzUZ[ ) {m| ) rdst[ 13736
IEJKLM NOPQLM
}~~[ _i~~b_ ]V~~sb_ z gmb_ ]Wrsb_ S]Vx l_Xsc_ Ux] e_ }[ ]nZb_ a]ur rsph
Sn~~b_ \_m~~Yb_ eu_rs~~b S]sT a]ursb_ ]Wrsb_ kXp[ UT_ gmb_ ]Wrsb_ uUt[ e_
]~~t r~~yo r~~s s~~b_ Uhr~~fnb S]~tjk_ i~ S~[Ux[ }~[ Xhah UY[ ]Wrsb_ S]tj_ n^
bi~~b. ]~~tsb_ UdY_ }[ sb_ UZWrYb_
= . = -~= ~ ~-~ ~=`
=` ~ = ~` --- - .~~ ~ 1Permeate2
~ 13etentate2 --` - - = .~ ~
=~ ) 3 4 5 & 6 - ( ~ --- - . . -~~ -=
- _-- -- +- - -= . ~ ~~ -`~~ ~ --` .~
= .~ . ~ ~~ ` = .~ -`~~ ~ =
~-~ ~=` ~ = .~ . - _-- -= -~ .~- ~=
_~ _-- -- +- = .~ -~+ - (
'
= -`~~ ~ ~~~ ~ --`
.~ = .~ (
18

Вам также может понравиться