EPA/6UU/A-94/UI4
THE REDUCTION OF MEMBRANE FOULING BY
PHYSICAL, CHEMICAL AND BIOLOGICAL PRETREATMENT
Todd L. So
CTI Enviror
Louisville, K:
Dr. R. Scott Summers
University of Cincinnati
Civil and Environmental Engineering Department
Cincinnati, Ohio 45221
Thomas F. Speth
Environmental Protection Agency
Risk Reduction Engineering Laboratory
Cincinnati, Ohio 45268
BACKGROUND
Membrane processes have become more attractive for potable water
production in recent years due to the increased stringency of drinking water
regulations. Membrane processes have excellent separation capabilities and
show promise for meeting many of the existing and anticipated drinking water
standards. The Surface Water Treatment Rule (SWTR) and the anticipated
Groundwater Disinfection Rule have led to the investigation of ultrafiltration
(UF) and microfiltration (MF) for turbidity and microbial removal (1). The
anticipated new disinfection by-product (DBP) regulations have increased
interest in nanofiltration (NF) and ultrafiltration (UF) membranes for DBP
precursor removal.
The main impediment to the development of membrane processes in drinking
water treatment is the reduction in membrane efficiency caused by membrane
fouling. Fouling is the build-up of material in the pores and on the surface
of the membrane which restricts flow through the membrane. By reducing the
flux through a membrane, fouling makes the process less economically
attractive. Flux loss due to fouling may be operationally defined as either
reversible or irreversible. Reversible fouling, also termed colmatage, is
recoverable by backflushing or chemically cleaning the membrane.
Irreversible fouling is operationally defined as flux loss which is not
recoverable by backflushing or chemically cleaning the membrane.
The two ways a membrane can be fouled are by adsorption and cake layer
(gel layer) formation. Materials may adsorb on the surface or within the
pores of the membrane. Hanemaaijer et al. (2) conducted adsorption
159
-------
experiments on solid, non-porous polysulfone beads and on polysulfone
membranes, which had the same surface area as the beads. They found that
the membranes adsorbed 100 to 400 times more protein than the beads. The
greatly enhanced adsorption was attributed to adsorption within the
membrane pores, indicating that pore adsorption plays a significant role in
membrane fouling. Adsorption of materials onto pore walls reduces the area
of the pore, increases resistance to flow and reduces flux through the
membrane. Fouling due to adsorption is thought to be largely irreversible
(3) and presents a major impediment to the success of membrane processes.
Adsorption is difficult to control and chemical cleaning may be required to
remove a portion of the adsorbed materials from the membrane.
Cake or gel layers are formed on the surface of the membrane when material
is rejected by the membrane and accumulates on the membrane surface. This
phenomenon is closely associated with concentration polarization. Cake
layers are often associated with hydraulically reversible fouling, because the
cake layer is often easily removed by hydraulically cleaning the membrane (3).
Cake layer formation can often be controlled by adjusting the operating
parameters which control the hydraulic conditions within the membrane
module (4).
The hydrophobicity of a membrane is important in determining the fouling
potential of the membrane-water system. Researchers have shown that
hydrophobic membranes foul more readily than hydrophilic membranes.
Laine (4) used a stirred batch cell reactor to filter Lake Decatur water with
two hydrophobic acrylic membranes and two hydrophilic ccllulosic
membranes. The hydrophobic membranes were fouled much more severely
than the hydrophilic membranes. Bonner and O'Mclia (5), using a 30,000
MWCO polysulfone (hydrophobic) membrane and a 30,000 MWCO cellulose
(hydrophilic) membrane, found that bovine serum albumin (BSA), natural
organic matter (NOM), dextran (a poylsaccharide) and triton-X (a non-ionic
surfactant) all caused much greater fouling of the hydrophobic membrane
than the hydrophilic membrane. The greater fouling tendency of hydrophobic
membranes is probably due to a hydrophobic interaction between the
membrane and hydrophobic components of the feed water.
Membrane feed pretreatment is a method that can be used to reduce
membrane fouling and improve filtration efficiency. Coagulation,
biotreatment, microfiltration and activated carbon filtration are some of the
pretreatment techniques that have been employed in an attempt to reduce
membrane fouling and enhance organics removal.
Coagulation is used in water treatment to remove suspended solids and
organic colloids. Coagulation selectively removes high molecular weight,
hydrophobic and acidic compounds. UF can be used to efficiently remove the
floe from a coagulated water in a coagulation/UF system. In this system, UF
replaces the settling and media filtration processes used in conventional
coagulation treatment. Coagulation improves the organics removal over direct
UF filtration and also reduces membrane fouling by removing organic
material and forming a more porous cake on the membrane.
160
Lain. <4) used water aluminum
coagulate Lake Decatur water prioi_ waJ. increased by at least a
that after coagulation the final membra f r tw0 hydrophobic
factor of two and that TOC removal improved by 18-40% to two y ^
(10W0° a0d re^o0rmatIdroFtCc
membranes, Lainestates h ^ ^ (he resuU is an improved
imparts a porous character to coagulated Seine River water
permeate flux. Lahoussine-Turcaud et a . ( ) ,n8reduce fouline of a 1000
with a polyaluminum coaplant in an at e P Although the rate of flux
MWCO hollow nber polysu'fone mem^ fou,ing at the
decline was reduced, the ex e untreated Seine River water,
end of the filtration run was t e sam^ Jn Seine Rjver water that is
This suggests that there is a fouing P archers used pyrolysis-gas
JESSE" %SL"rs m - -
conclusions.
Bamina.i.n of ,he use of biotreatrnen, to -educe membrane fouling ha, been
limited. Taylor et al. (8) reportani udte,F„ , b,o,0gicall(.
County, Florida demonstrated that on-s g ^ rovided adequate
active sand bed, of hig y brane The improvement in flux
pretreatment to maintain e to a combinatj0n of biological activity,
due to ground storage is probab y biodegradation of aquatic
adsorption and filtration. Jhe Mudof t «J>^8 that ,0w
organic matter Meyer et al. W J«™«handwe™
molecular weight organics ( ) mw MOOO-lO 000) or high MW
removed to a greater e*tent presses
(>10.000) fractions. ™ese ft^ "toride-based material,)
- m»^.< « ¦»«—
in irreversible membrane fouling.
Microfiltration <0.1-10 »m> is capable of removingparti.ulat«
colloids and is now being inveistiga^ ^ ^ cffecI of panicle size on
nanofiltration. Wiesner et al. ( ) 8 narticles lareer than 3
hollow fiber UF membrane fouling 1.0 »
had minimal effect on flux and t P Dretreatment in the 0.1 to 1.0 jxm
caused the greatest flux re uc 10 . memhrane fouling by removing the
range has the potential to reduce UF m«nbrane louiing^ Tay,or(ii)
particulates which hav
-------
Granular activated carbon (GAC) and powdered activated carbon (PAC) have
been used in conjunction with membrane processes to reduce membrane
fouling and to provide enhanced organics removal. Cruver and Nusbaum (12)
used GAC to pretreat primary and secondary wastewater effluent prior to RO
filtration and found that GAC filtration significantly improved membrane flux.
Reiss and Taylor (11) used GAC to treat a highly organic surface water prior
to NF filtration. They demonstrated that GAC prefiltration allowed the long-
term operation of an NF membrane, while conventional pretreatment (acid
addition, cartridge filtration) was inadequate to maintain membrane flux.
Laine (4) and Laine et al. (13) demonstrated that PAC pretreatment of Lake
Decatur water prior to UF marginally improved the membrane flux while
improving DOC and THMFP removal significantly. Anselme and Charles
(14) treated a groundwater and a surface water with a PAC/UF combination.
Their research indicates that PAC addition has little effect on membrane flux,
but that DOC and THMFP removal are greatly improved over UF treatment
alone. These studies indicate that activated carbon pretreatment has the
potential to reduce membrane fouling and enhance organics removal.
Modeling flux decline with a series resistance model can aid in the
understanding of fouling phenomena. The resistance model is often used to
describe flux decline:
J =
R,
where J is the permeate flux (permeate volume/time'membrane surface
area), aP is the trans-membrane pressure and R, is the total resistance to flux
(pressure'time'membrane surface area/permeate volume). In the series
resistance model the total resistance, R„ is composed of a membrane
resistance term (Rm), a cake layer resistance term (Rc), a chemically reversible
fouling resistance term (R„) and an irreversible fouling resistance term (R,).
Laine (4) and Laine et al. (13, 15) used the series resistance model to
elucidate how coagulation and PAC pretreatment of a UF membrane feed
affected the resistances of the cake layer and the irreversible fouling layer.
This approach proved to be an effective method to facilitate the
understanding of the effect that various pretreatment schemes had on the
extent of membrane fouling.
The kinetics of the fouling process can be elucidated by modeling the flux
decline curve; normalized flux (J/J0) versus time. Wiesner et al. (16)
successfully used second order kinetics to describe the flux decline of ceramic
MF membranes and to compare the effect that various pretreatment schemes
had on the rate of membrane fouling. This type of modeling can be a useful
tool in understanding how pretreatment schemes affect the rate of membrane
fouling.
The cumulative normalized water production is an indication of the ability of
a membrane to sustain water production over a prescribed period of time.
Integrating the normalized flux (J/J0) over a specified time period yields the
162
™mulr irssiT.
Turcaud et al. (6) found, for a vaneiy fn-eversible fouling for each
UF, that the exten' °fbu;etVhea^Si rates of flux decline and the cumulative
pretreatment was similar, but t cipnificantlv different for each
normalized normalized water production is a
STSltSparing .lie ability of membrane, .0 susta.n ware,
production.
npiFmVES
The objective of GAC fixation, alum
processes to reduce membrane fou g. ofiitration ^ investigated as
coagulation, biofiltration and / ydrophobic UF membrane
pretreatment schemes to re p— -
turbidity was also investigated.
METHODS A NO MATERIALS
Water Source
The natural water used in this sUidyttaClifton Camp J of tl
is aerated at 2 locations in the "™er (Whatman 934-AH) within 2 hours aft
and filtered with a 1.5 „m Z 37.8 NtO to 0.52 NTU. Tl
collection. This filtration redu . Dractjce 1-10 filtration is usually usi
1J filtration was performed because n pracnce 1 1U m jculateS. T
•nte molecular
filtering the water with 3 r_.a,- camDles The data indicate that 8:
^ -h - m»]or,,y of ,he 000,hc w
10,000 fraction.
Ultrafiltration Apparatus
, Am;™„ niaflo PM10) was used exclusively in
A 10,000 MWCO disc membrane (^.co^Da ^ it ,
fouling tests. In preliminary membranes (PM series) fouled m
discovered that the hydrophobic Membranes (YM series). '
more rapidly than the hydrop 11 because a membrane which foi
hydrophobic PM10 *^"treatment schemes on memb,
extensively was desired, so that membtil„e used in this research had
flux could be readily discerned. The disc mem
mm diameter which yields a surface area of 41.8 cm .
163
-------
A 5 liter stainless steel pressure vessel (Alloy Products Corp.) containing the test
water was used to feed into the 400 mL stirred UF cell (Amicon Model 8400) which
contained the membrane. The UF cell housed a magnetic stir bar which was used
to limit concentration polarization. The permeate flux was measured by timing the
collection of permeate volume in graduated cylinders.
Fouling Test Procedure
Fouling tests were conducted on raw BWP water and BWP water pretreated with
microfiltration, GAC filtration, alum coagulation, biofiltration and
ozonation/biofiltration. Fouling tests were also conducted on laboratory clean (LC)
water. LC water is produced by treating Cincinnati tap water with RO membrane
filtration, cation/anion exchange and GAC filtration. The LC water was used to
rinse and backflush membranes in this research. Membrane fouling was measured
as the reduction of permeate flux as a function of permeate volume at constant
pressure. A pressure of 55 psi was utilized for the fouling tests (17).
Two fouling test procedures were followed for most of the waters in this study: a 1-
cycle test and a 4-cycle test. A new membrane was used for each fouling test. The
1-cycle fouling test was conducted by first determining the initial "clean water" flux,
J„, of the membrane by filtering LC water at 55 psi. The flux after 5 minutes of
applied pressure was used as the initial membrane flux because 5 minutes is
sufficient time for the membrane to compact and the flux to stabilize (17). The test
water was then fed to the membrane at 55 psi. The permeate flux was measured
during the course of the fouling test by timing the collection of permeate in
graduated cylinders until 1.0 liter of permeate had been collected. The run was then
terminated, the membrane was removed from the cell and the surface of the
membrane was rinsed with LC water. The membrane was then placed in the cell
with the skin-side down and 100 mL of LC water was filtered through the membrane
at 15 psi. This procedure is termed backflushing. The membranes were backflushed
in order to remove the cake layer which had deposited on the membrane. The
membrane was then placed skin-side up in the cell and a LC water flux test was
performed to determine the amount of flux recovered by backflushing. The
membrane was then removed from the cell and soaked in 200 mL of 0.1 N NaOH
for 30 minutes to remove base soluble materials which had adsorbed onto the
membrane (17). After the base cleaning, the membrane was rinsed with LC water
and placed skin-side up in the UF cell. An LC water flux test was then conducted
on the membrane to determine the amount of flux recoverable by base cleaning.
Figure 1 shows a typical flux curve for a 1-cycle fouling test. The normalized flux,
J/Jj, is plotted as a function of total permeate volume. The normalized flux is the
flux divided by the initial "clean water" flux of the membrane. Normalization of the
data in this manner allows comparison of fouling tests conducted on membranes with
different initial fluxes.
The 4-cycle fouling tests were conducted by repeating the 1-cycle fouling test
procedure four times on the same membrane, but only including the base cleaning
step in the third cycle of the test. Figure 2 shows a typical flux curve for a 4-cycle
fouling test.
164
Static adsorption tests
Static adsorption tests were ^
fouling due to adsorption. Static adsorptio waters were
raw BOT «». alum coasted waterand bofilte0,de,
f,leered will, a 0.22 wn membrane After pr or CO '^™csorfa£ Lc wate, flu,
o— «>"d -3-8 32
hours of adsorption.
Pretreatment Schemes
Microfiltration
£ rSiore'SlL' warn (VY30-090-00). Tfce 1- and 4-cycle fonl.ng tests
were conducted on this microfiltered water.
Coagulation ,
The raw BWP water was coagulatcd with by slow
50 mg/L. The water was rapid mixed a /j*.• , ts were conducted on the
mixing a, 30 rpm '<>'^VJ^in s^nln and on .he coagulated wa.er
coagulated water with floe still in suspension
supernatant after 1 hour of settling.
GAC
A *-ss; iWirf5
empty bed contact time (hbCi; /Fiino Calnon Carbon Corp.) was
57'5 04)'m
volumetric flow rate was 3 mL/min.
The water passed through the GAC 254 nm. Tli<
The breakthrough of the carbon was Vuv absorbance at 254 nm i
breakthrough curve is shown in Figure , (Saq The breakthrougl
expressed in terms of the spectra a rp ^ ^ tha( passe(1 throug!
curve was broken down into four sectio . ^ ^ through 5th liters collecte
the column were combined into Section . wefe combjned in,
were combined into Sfed|.on :2 wate^;'h^ J into Section 4 water. Hi
sssi
,he seciion 4 wa,er ,""e"n
ware, treated by carbon in a near exhausted state.
165
-------
Biofiltration
The biofilter was comprised ofa^OO mL^eparatorv ft,rec'rculating aerobic biofilter.
acclimated sand. The.water was ciraXrf,L u containin8350mLofbio-
downflow mode. The water was pumped from 6 fil,cr 31 15 mL/min «
penstaJt.c pump into a 20 liter glass s,Xer, J -T™ of the biofilter with a
[ e Stora8e r«ervoir dropped through 12 inche^'h ^ Wa,er bdn8 PumPed >nto
¦•quid surface. The free-fall of the water .hrnnth .i headsPace before reaching the
was designed to oxygenate the water for use bvLmh" OXygen:containin8 headspace
water m the reservoir was pulled into the biofiherhi h °rgan'sms in ,he f'"er. The
separatory funnel by the peristaltic pump The b.o^, VaCUUm Created inside th«=
temperature room at 20-C The hfnfiit biofilter was maintained in a constant
™-
-------
LC water fouling tests
A 1-cycle fouling test was conducted on LC water. The LC water had a DOC of 0.19
mg/L Figure 6 is the fouling curve for the LC water. The curve exhibits a steep
initial drop in flux in the first 50 mL of permeate followed by a steady decline in flux
until the end of the test. The membrane lost 16% of the initial flux by the end of
the fouling test. There was apparently a material which penetrated the RO
membrane, cation/anion exchanger and G AC filter which was responsible for fouling
the membrane. Organic materials composed of low MW polysaccharides have been
implicated in penetrating ire:::-.? it processes and irreversibly fouling membranes
(6,7). Compaction of the membranes during the fouling test could also be responsible
for the lost flux. However, the membrane manufacturer (17, 20) stated that for the
conditions used in our initial flux determination (5 minutes filtration at 55 psi), the
membranes should have been sufficiently compacted such that flux loss due to
compaction should not occur during the fouling tests.
Raw BWP water fouling tests
The results of the 1- and 4-cycle fouling tests of raw BWP water appear in Figures
7 and 8, respectively. The flux decline for the 1-cycle fouling test was rapid in the
initial stages, but gradually leveled off. The flux dropped to 11% of the initial flux
after 1 liter of permeate had been collected. A continuous, brown film layer was
observed on the membrane after filtration, which was easily removed by surface
rinsing. Backflushing recovered 36% of the initial flux, and base cleaning recovered
an additional 21%. Each cycle of the 4-cycle fouling test exhibited the same flux
decline pattern as the 1-cycle test, reaching 11 % of the initial flux after 1.0 liter of
permeate. A removable continuous, brown film layer was observed on the membrane
after each cycle. The flux recoverable by backflushing dropped in each of the first
three cycles of the 4-cycle test from 52% to 44% to 41%, indicating that
hydrodynamically irreversible fouling increased with each cycle. Base cleaning after
the 3rd cycle recovered 18% of the initial flux. The 4th cycle of the 4-cycle test also
followed the same fouling pattern as the 1-cycle test.
Microfiltered water fouling tests
The 1-cycle and 4-cycle fouling tests which were conducted on the microfiltered BWP
water were similar to the fouling curves for the raw BWP water. Figure 9 shows the
4-cycle fouling test for the microfiltered water. The solid line with the asterisks
represents the raw BWP water 4-cycle fouling test. These results indicate that
microfiltration did not remove the materials from the raw BWP water which were
responsible for membrane fouling. From these tests we can conclude that
particulates larger than 0.22 Mm were not responsible for significant membrane
fouling.
GAC pretreated water fouling tests
The raw BWP water was pretreated with GAC and the breakthrough curve was
monitored with UV absorbance until a constant effluent concentration was
approached as shown in Figure 3. The breakthrough curve was broken into four
sections and the water within each section was composited. The DOC of Section 1
168
water was 1.59 mg/L and that organic matter which has passed
water contains only non- an w' * 4 water has a DOC sjmilar to that of the
through the GAC system. The Sect Q -c matter is not present. 1-
influent and thus only the AC Section 1 and Section 4 waters and
cycle fouling tests were conductedw,^ ^ ^ ^ ^ astensks represents the
the results are shown in Figure . Sectjon i and 4 waters exhibited a rapid
raw BWP water 1-cycle fouling test. Both Sectional ^ ^ similar
initial flux loss, followed by a more gra permeate for both GAC
to the raw BWP water. The extent of^ the flux recovered by
waters was also very similar to the ™ ^ brQwn film ,ayer was observed on
backflushing and base cleaning. was easily removed by surface rinsing,
both membranes after filtration. re,Donsible for the membrane fouling, then
If strongly adsorbing comPou"ds * ,he fouling. The flux curves indicate that
the GAC pretreatmentshouldreduc ^ ^ GAC (Section 1), removed the
neither nearly exhausted GAC: (Set«o ) mcmbrane fouling. These results were
components of BWP water re p There apparently was a component of
unexpected, especially for the fresh GA . foul membrane.
BWP water which was able to penetrate the OAL
Alum coagulated water fouling tests
i rwp water which resulted in a 28 fo
An alum dose of 50 mg/L was used toconductcd wlth the coagulated
reduction in DOC. T^o 1-cycle o g ^ ^ jn suspension ,n one test
water. The coagulated water was u tranitere ^ ] ^ tQ removc |he floc pn0r
and in the other test coagulated waterwas se ^ settled tests. Figure
,o UF. These two tests will be referred to^ the susp^ ^ wa(er ^ sohd
11 shows the 1-cycle fouling test °r p watcr fouling test. Compared
line with the asterisks represents he raw after j ,uer of
'to the suspended water, the & rate of flux decline for
permeate, after backflushing and afer^eCea g ^ ^ af(er ,
Soth waters was substantially ess than th^ raw BWp water. ,nstead.
liter of permeate did not rea p ^ ^ (he fj|tration. This indicates
the flux continued to decline stea y ^ ^ flux yalue after \ llter 0f
that coagulation altered the kinetics of d 31% than for the raw
permeatlwas substantially^greater for material from the BWP
BWP water, 11%. This indicates th P1 n In contrast to the raw BWP
water which was responsible for cmbraSc after filtration. Table 1 shows
water, a film layer was not obiserved on any additional DOC. This data
that ultrafiltration of coagulatedwaite molecuiar weight material which would
indicates that coagulation completed he h,g. mole™ ^ ^mbrane flu* after
have otherwise formed a de"se- " f ,h coagulated waters than that for the raw
SS-S—"did not ,ed"",he eaem
of irreversible fouling.
Biofiltered water fouling tests
T* BWP water was biof ,l.e,«d in ba.ch mode un.il ,h« DOC co„e«n„a,ion reached
169
-------
iPsS^-illssH~=s
No film layer was obs^rv^d Se me^b3"3' ,han th*1 oHhe'rai BVVPwa?'6
of biofiltered BWP water h h membrane after filtration Tahle 1 i, Puwater-
"•« Woflltra.ion LXt,^ir0™ "» TO UF
membrane. The fli.v ->e. . , W organics so thev wer» ,ki 'ndicates
SSS55??S=SSSSpS
third cycle was sTmihr^ ^ Cydes from «% * S t0 it? ba?kflushin8
hydraulically irreversible f p3' lhe raw BWp water This • a-' after lhe
'he raw BWT *?*"" WOrse ^"ST* that the
"re^ed to only 33% after four
Ozonated/biofiltered waler fouling tests
contactorandwLMthtn°f0Hate(!.Wlth a 15 mg 0,/mR DOC h™
DOC reached a^S /° e biofi"er, where ciralar?,, In 3 n°w trough
was then used ?/, " °f 3 3 after 5 dan ?h? COnt'"^d until the
instead of a 4- cycle ? c^c'e ^ou''ng tests. A^3-cycle °^J)ate<'/k,'ofiltered water
fouling test for th* CS' t0 a ''m'ted water suddIv P' ->'eSt was conducted
^rattd/«iK^».1«an/Si;.^s.*¦— Fisrz
Static adsorption tests
appe-
-* - is ss£ rmg -»¦— «^™~rr,tp;rs
170
Modeling of Fouling Phenomena
Series resistance model
The extent of fouling was modeled with the series resistance model. The model can
be expressed as:
J = aP
R,
where aP is the trans-membrane pressure, R, is the total flux resistance and J is the
permeate flux. The total resistance, R,, is composed of a membrane resistance, Rm,
term, a cake layer resistance, Rn term, a chemically reversible fouling resistance, RCP
term and an irreversible fouling resistance, R,, term. These terms are calculated
using the following equations:
R. = &E
Jo
R, = aE - Rm
J.
R„ = 4E - Rm - R,
h
Rc = aE - Rm - Ri - Rn
where J0 is the initial flux of the membrane, Jr is the flux after 1 liter of permeate,
Jb is the flux after backflushing and J, is the flux after base cleaning. This model is
valuable in elucidating the effect that various pretreatment schemes have on
membrane fouling. Table 2 presents values of the resistance terms calculated for
each pretreatment scheme in the 1-cycle fouling tests and Table 3 presents the
percentage of the total resistance contributed by each resistance term.
The data in Tables 2 and 3 show that the total resistance to flux was not affected by
the microfiltration or GAC filtration pretreatment schemes. The tables show that
for the raw, microfiltered and both GAC filtered BWP waters, the cake resistance
term dominated, 70 to 80%, the total resistance to flux. The resistance from the
membrane, chemically reversible fouling and irreversible fouling were small
compared to the cake layer; 3 to 13% each. Alum coagulation reduced the total
resistance by 50% through reduction of the cake layer resistance. The membrane
resistance became more important in the resistance to flux when alum pretreatment
was applied. Biofiltration reduced the total resistance by about 70%, through a 90%
reduction of the cake layer resistance, to a point where the membrane resistance
term represented the greatest resistance to flux. The irreversible fouling resistance
became more important and approached the resistance posed by the cake layer.
Ozonation/biofiltration further reduced the cake layer resistance. The membrane
resistance term was significantly larger than any other resistance term for this
pretreatment scheme and the irreversible fouling resistance was equivalent to the
cake layer resistance.
171
-------
• ^Uli6 ul Iliu dec|inc
'£E 5r-r "•«
^r^r£rf-dBw-r
high lvalues, >08, j" lh/n'°f'he model to the data to p1"'"8 "" P0W"
kine,i<5 "' "*« wateL * ' "* ^ well -eprese'Ttl^
The flux decline results nf th» . ,
'CogSC^£fioia y": a"d OZOnati°"/^ofii?auo1t were^e™"' a'Um
kinetic, of ,hesc ,„,ee ^ 1 .he logarithm ^ ^-e IS ,h,
hiofiltered''water ^fth^-iTand*?C' 'he '"t"yclc foul'ng tesis of the
^ « "OttreXr,^ Z Z"rTr
Water production
ss ijecS'r °'how w<"«
effect that fouling has onT'0" W"h rCSpect to initial flux allow?"' '° the initl'al
fluxes. The n.mMi water Production of memhf comparison of the
a membrane by integrating1'"11 Wa',Cr produc,ion (NWP) ^b^T ,°f 'nitial
tegrating the normalized flux with risPlt \o'*£?£$£ '«
NWP = ; (J/Jo) dt
sipsSiifSsE
as the integration tim "h r 30 m,nu,es of filtratioa Thirt^3-" produc,lon for
minutes to complete "^ aUSe ,he shortes« 'ouhng tc™7™""'" ** chosen
ea water production. The lower rate
172
of flux decline and less extensive membrane fouling exhibited by these waters yielded
higher water productions. These three pretreatment processes were successful in
increasing water production because of their ability to reduce the resistance of the
cake layer formed on the membrane.
The water production decreased in each of the first three cycles of the multi-cycle
tests. This trend corresponds to the increased fouling trend displayed by the multi-
cycle fouling tests. Base cleaning the membranes increased the water production
slightly, in correspondence with the improved membrane flux after base cleaning.
Comparing the total water production for the first three cycles of the multi-cycle tests
shows that the improved water production is sustained over the three cycles for the
biotreated and ozonated/biotreated waters. Table 6 shows the increase in water
production achieved by each pretreatment scheme. Microfiltration and GAC
treatment did not increase water production. Coagulation, biotreatment and
ozonation/biotreatment increased water production by 68%, 60% and 100%,
respectively, in the 1-cycle fouling tests.
SUMMARY AND CONCLUSIONS
BWP water was pretreated with microfiltration (0.22 um), GAC filtration, alum
coagulation, biofiltration and ozonation/biofiltration. The raw and pretreated waters
were fed to ultrafiltration membranes with a 10,000 MWCO. The UF membranes
provided excellent turbidity removal for all waters tested with the permeate turbidity
at or below the detection limit of 0.02 NTU. However, the membranes provided low
DOC removal, 10 to 20%, for nonpretreated water and nearly no removal for
pretreated water. They were not selective for the removal of TOX forming materials.
This research indicates that this UF membrane can be used to produce potable water
which meets the SWTR requirements for turbidity removal, 0.5 NTU, but will not
significantly aid in the control of DBPs.
The raw BWP water caused significant fouling of the membranes, yielding an 89%
reduction in flux. A major portion of the initial flux, « 40%, could be recovered by
backflushing the membrane. Base cleaning provided additional flux recovery of =
20%. The irreversible fouling was significant, resulting in a 30% non-recoverable flux
reduction. Microfiltration and GAC adsorption of the BWP water were unsuccessful
in reducing membrane fouling and yielded fouling curves very similar to the raw
BWP water. Apparently these pretreatments were unable to remove the components
of BWP water responsible for membrane fouling. This was surprising for the GAC
because GAC filtration provided excellent DOC removal. Alum coagulation of raw
BWP water yielded a significant improvement, 90%, in membrane flux. Alum
coagulation was thought to reduce the reversible membrane fouling by complexing
the high MW organics which would otherwise form a dense cake layer on the
membrane. Irreversible fouling, however, was not decreased by alum treatment.
Biofiltration and ozonation/biofiltration significantly decreased reversible membrane
fouling. These processes oxidized the high MW organics and enabled them to
permeate the membrane. Breaking down the high MW organics greatly reduced the
resistance of the cake layer on the membrane. Biofiltration and
ozonation/biofiltration did not decrease irreversible membrane fouling.
173
-------
Series resistance modeling of the raw, microfiltered and GAC filtered BWP waters
showed that the resistance of the cake layer dominated the resistance to permeate
flux. Alum coagulation, biofiltration and ozonation/biofiltration all reduced
membrane fouling by significantly reducing the resistance of the cake layer formed
on the membrane (reversible fouling), but had little impact on the chemically
reversible and irreversible fouling.
Kinetic modeling of the flu* decline curves showed that raw, microfiltered and GAC
filtered BWP waters were modeled best with a power function. These flu* curves all
exhibited a sharp initial decline in flux followed by a plateau region of relatively
constant flux. The power function constants were similar for all three waters. Alum
coagulated, biofiltered and ozonated/biofiltered waters were modeled best with a
logarithmic function. These fouling curves all exhibited a moderate initial decline
in flux followed by a steady flux decline region. The logarithmic function constants
for all three waters were similar. Water production rates for the raw, microfiltered
and GAC filtered BWP waters were approximately the same. Alum coagulation,
biofiltration, and ozonation/biofiltration all yielded significant improvements in water
production, increasing water production by 50-100%. This increase is attributed to
the attenuated fouling kinetics and the reduced extent of reversible fouling.
In conclusion, microfiltered and GAC filtered BWP water had the same membrane
fouling characteristics as raw BWP water. Alum coagulation, biofiltration and
ozonation/biofiltration all yielded significant improvements in membrane flux. These
pretreatments are successful because they are able to reduce cake layer resistance
and attenuate fouling kinetics. The last three pretreatments lead to less extensive
cake layer formation, a slower rate of flux decline and higher water productions than
raw BWP water. However, none of these pretreatment processes were able to
remove the components of BWP water responsible for irreversible membrane fouling.
It is possible that low molecular weight polysaccharides are responsible for
penetrating the treatment processes and irreversibly fouling the membranes (6,7).
The order of success in reducing cake layer formation (reversible fouling) was:
ozonation/biofiltration > biofiltration > alum coagulation > GAC filtration «
microfiltration.
REFERENCES
1. Laine, J.-M., Jacangelo, J.G., Patania, N.L, Booe, W. and Mallevialle, J., 1991.
"Evaluation of Ultrafiltration Membrane Fouling and Parameters for its Control", In
Proceedings. AWWA Membrane Technology Conference, March 10-13, Orlando, FL
2. Hanemaaijer, J.H., Robbertsen, T., van den'Boomgard, T. and Gunnink, J.W.,
1989. "Fouling of Ultrafiltration Membranes: The Role of Protein Adsorption and
Salt Precipitation", J. of Mem. Sci.. 40, 200-217.
3. Wiesner, M.R., Clark, M.M. and Mallevialle, J., 1989. "Membrane Filtration of
Coagulated Suspensions", J. of Environ. Eng.. ASCE. 115(1), 20-40.
4. Laine, J.-M., 1989. "Optimization of Organic Removal in Ultrafiltration of a
Natural Water", Masters Thesis. University of Illinois at Urbana-Champaign.
174
_ . . cym-i:.. pr 1091 "Some Aspects of the t-ouung 01
asl'Memb™£* NaturL Organic Matter nj Wa,„ T,=r. U
AWWA Membrane Technology Conference, March 10-13. Orlando.
i Lahoussine-Turcaud. V, Wiesner. M.R..
¦Coagulation Pretreatment for Ultrafiltration of a Surface Water. LA
7 Mallevialle. J.. Anselme. C and Marsigny, O, 1987. "Effects of Humic Substances
on Membran; Processes", In ItofifidmfiS, ACS Congress, Denver, CO.
8 Taylor JS Mulford. LA., Duranceau, SJ. and BarreuW_M,l989. Cost and
Performance of a Membrane Pilot Plant". LAVfflA, 81(11), 52-60.
9 Meyer J L.. Edwards. R.T. and Risley. R.. 1987. "Bacterial Growth o„ Dissolved
Organic Carbon From a Blackwater River", Microbial Ecology. 13,
„ , 1 A Rntt TI 1983 "Microbial Heterotrophic Utilization of
Dissolved "organic Matter in a Piedmont Stream", Fmhwata PiolQgV- 13, 363-377.
~ jTo.,ir.r i q 1001 "Membrane Pretreatment of a Surface Water ,
AdWTwt^mb™ T"*no,og, Con,.. March ,H3. Orlando. FL
12. Cruver, J.E. and Nusbaum. I.. 1974. -Application of Reverse Osmosis to
Wastewater Treatment", J. WPCF. *6(2), 301-311.
13. Laine. ,,M, Hags,rem. ,.P.. Clark. M.M and M.'^'IO '989. "Eftec* 0,
Ultrafiltration Membrane Composition , J, AWWA, 81(11), 61 6 •
„ p 1001 "The Use of Powdered Activated Carbon For
JlJe Removal of^pecif.c Pollutants in Ultr^a^
AWWA Membrane Technology Conference, March
\ AWWA. 82(12), 82-87.
S32SS33SSSH
13, Orlando, FL.
17. Amicon, Inc., }°°<) 1 Products Catalog-
18 Mogren, E.M., 1990. "Measurement of Biodegradable Dissolved Organic Carbc
in Drinking Water", M^rs Thesis, University of Cincinnati.
19. Dejmek. P. and Nilsson. J.L., 1989. 'Fta-b^d Measures of Adsorption
Ultrafiltration Membranes", J, gf Mem. Sa, 40> 18y 1V •
20. Vorhees, D., 1991. Amicon Technical Service, Personal Communication.
175
-------
0.9
0.8
1 07
-5
J 0.6
u.
S 0.5
N
| 0.4
o
2 0.3
0.2
0.1
0
0 0,2 0.4 , 0.6 0.8 1 1.2 1.4 1J
Total Permeate Volume (L)
Figure 1 Typical 1-Cycle Fouling Test
1
0,9
0.8
§ 07
J 0.6
u.
S 0.5
M
| 0.4
Z 0,3
0.2
0.1
0
0.0 0.5 1.0 1.5 2.0 2.5 3.0 3.5 4.0 4.
Total Permeate Volume (L)
Figure 2 Typical 4-Cyde Fouling Test
JO = 924 Uhr*m2
Irreversible
fouling
Flux recovered -
by base cleaning
Chemically
reversible
colmatage
Flux recovered
by backflushing
Hydrodynamically
reversible
colmatage
Flux recovered
by backflushing"
Flux recovered x
by base cleaning
176
Average
DOC SAC
(mg/L) (1/m)
Section 1
Section 2
Section 3
Section 4
Influent
1.59
3.07
3.73
4.15
5.00
2.00
5.31
6,94
8.25
11.0
o
O 2.5
1
0.5
0
5000 10000 15000 20000 25000
Throughput, Bed Volumes
Organic Mailer Breakthrough in GAC System
If
Non-ozonated
+¦ ~ Ozonated
-
Recirculating
batch biofilter
1 1 1 1 ¦ i • 7
\ 2 3 4 5 6 7 8 9 10 11
Time (days)
Figure 4 Kinetics of Biodegradation
17?
-------
i Burnet Woods Pond (BWP) Water
1
-jj.5 ^m membrane { Pretreatment
i
! j_0-22^m_membrane J - Microfiltered
! 50mg/L alum j . Coagulated
¦ ^ 5 min EBCT of GAC ; ¦ GAC Adsorbed
! [ 10 day batch biofilter ¦ ¦ * Biotreated
' f" 1.S mg 0,/mg DOC » bio j " Qj/Biotreated
Figure 5 Pretreatmeni Schemes
1.1
1
0.9
o 0.8
*0.7
X
if 0.6
T3
J 0.5
m
§ 0.4
o
Z 0.3
0.2
0.1
0
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1 1.1 1.2
Total Permeate Volume (L)
Figure 6 LC Water 1-Cycle Fouling Test
1 70
0.9-
J0 = 624 L/hr*m2
DOC = 5.0 mg/L
o 0.8
ll- 0.6-
47
E 0.4-
0.3-
0.2-)
0.1-
Total Permeate Volume (L)
Figure 7 Raw BWP Water 1-Cycle Fouling Test
JO = 990 L7hr*m2
DOC = 5.0 mg/L
0.9-
0.8-
l 0.7-
u. 0.6-
S 0.5-
44
44
E 0.4
0.2
0.1
4.5
3.5
2.5
0.5
Total Permeate Volume (L)
Figure 8 Raw BWP Water 4-Cyde Fouling Test
179
-------
1
0.9
o 0.8
H
3
0.7
0.6
¦o
® A C
jS 0.5
0.4
o
2 0.3
0.2-
0.1-
J0 = 828 Uhi^m2
DOC = 4.88 mg/L
0+
0
Figure 9
1.1
1
0.9
0 0.8
"I 0.7
X
£ 0.6-
1 0.5
m
§ 0.4-
o
2 0.3-
0.2
0.1
0
0.5
1
1-5 2 2.5 3 3.5
Total Permeate Volume (L)
Microfiitered Water 4-Cycle Fouling Test
4.S
¦ section 1 q section 4
JO = 1030 L/hr*m2 JO = 876 Lihr*m2
DOC = 1.59 mg/L DOC = 4.15 mg/L
62 1
50,1
46' r
0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1 1.1 1
Total Permeate Volume (L)
Figure 10 GAC Filtered Water 1-Cycle Fouling Tests
180
JO = 714 Uhr*rrr
DOC = 3.6 mg/L
d 0 6
1 0.5
0,3
0.2-
Total Permeate Volume (L)
Figure 11 Alum Coagulated Water 1-Cycle Fouling Test
JO = 1010 LVhr'm2
DOC = 4.16 mg/L
X
if 0.6-
1 0.5
¥ 46
(0
E
o
z
0.1
3.5
15 2 2.5 3
Total Permeate Volume (L)
0.5
Figure 12 Biofiltered Water 4-Cycle Fouling Test
181
-------
1.1
JO = 966 L/hr*m2
DOC = 3.34 mg/L
TJ
8 0.5
E 0.4
0.5
1
T"
1.5
2 2.5 3
Total Permeate Volume (L)
Figure 13 Ozonated/BioGltercd Water 3-Cycle Fouling Test
3.5
1.0
0 9 -
Power function
Y = o • X"
0.8 - )
0.6 -0
0.J -
9-S-©
Best-fit
a = 0 644
b = -0.351
= 0 997
u o o o o o
"T
0 20 40 60 80 100
Operation Time (mm)
Figure 14 Kinetic Modeling of Raw BWP Water 1-Cycle Fouling Test
182
0 9 H
0.8 -
Natural log function
Y — o — b»ln(X)
0.6 -
0.5 i
0.4
0.3 H
0.854
-0.156
0.999
0.0-
Operation T.me (mm)
183
-------
Sample
BWP Water
Microfiltered
Coagulated
GAC Adsorbed
JOXFP rtipri/ij
Influent
Effluent
Influent
Effluent
5 0
4 1
945
744
4.9
4.1
929
758
3 6
3 5
514
section 1
16
4 2
section 4
15
3.7
166
746
Biotreated
42
4 2
668
Oi/Biotreated
3 3
3 3
469
Table J Organic Matter Removal
- Flux Resistance* -
699
* (atm«hffi)2/n)3)
Table 2 Seri« Re.btee M
-------
Sample
l-cycle
4-cvcle TW
3-cycle
Total
leil
lil-L
2nd I,
taLL
4HlL
Raw
9.02
8.43
7.65
7.54
7.90
23.6
Microfiltration 8.37
9.37
8.40
7.99
8.72
25.8
GAC filtration
section 1
section 4
8.80
9.08
Alum coagulation
suspended 13.6
settled 15.2
Biotreated
14.4
16.8
13.6
122
12 5
42 5
Ozonated/
biotreated
17.9
17.1
13.7
11.8
42.6
Table 5 Cumulative Normalized Water Production after 30 Minutes of Filtration
Sample
Microfiltcred
GAC Adsorbed
Coagulated
Biotreated
Oj/Biotreated
I-CycleTe^ 3-CvcIeTe
------- |