United States
Environmental Protection
Agency
Risk Reduction
Engineering Laboratory
Cincinnati, OH 45268
Research and Development
w EPA Project Summary
EPA/600/SR-92/023 April 1992
Reduction of Disinfection
By-Product Precursors by
Nanofiltration
J.S. Taylor, C.R.Reiss, P.S. Jones, K.E. Morris, T.L. Lyn, O.K. Smith,
L.A. Mulford, and S. J. Duranceau
This document summarizes a project
that investigated the cost and perfor-
mance of employing a membrane pro-
cess to remove disinfection by-product
(DBF) precursors at highly organic
groundwater and surface water sites.
The groundwater investigation was fol-
lowed by the surface water investiga-
tion. The main phases of the project at
the groundwater and surface water sites
were site selection, membrane selec-
tion, pretreatment studies and pilot
plant operation for 1 yr at each site.
Eleven different membranes were in-
vestigated at the groundwater site,
Daytona Beach, FL, using a 1000 gpd
membrane pilot plant to produce flow,
pressure, and water quality samples.
Nine membranes with molecular weight
cutoff below 300 produced Jess than
0.10 mg/L DBF formation potential (FP)
as Cl- in the permeate. Only mem-
branes classified as ultrafilters did not
achieve DBPFP reduction of more than
90%. The pretreatment study at the
groundwater site clearly demonstrated
that only scaling control and
prefiltration were required to control
nanofilter fouling. A three stage 50,000
gpd pilot plant was operated for 8650
of 8704 hr available for production, ex-
periencing less than 1% downtime. Av-
erage DBPFP in the permeate stream
was 20 ng/L as C|- and was more than
96% rejection of DBP precursors in the
raw water. The average rate of mass
transfer coefficient (MTC) decline dur-
ing operation at the groundwater site
was taken as 2.2E-7/d2, which did not
include a period when new wells were
placed on line before adequate flush-
ing was achieved. DBPFP in the per-
meate was independent of variation in
pressure and recovery. The surface
water site nanofiltration (NF) project
was conducted at Melbourne, FL. Nine
different membranes were tested for
long-term operation using the same
technique as used at the groundwater
site. Seven of the tested membranes
with molecular weight cutoffs of 300
removed 95% of the DBPFPs to less
than 100 )ig/L as Ch Membranes se-
lected for long-term operation were the
DuPont A15s* (A15s) and the Desal DS5
(DS5) on the basis of precursor rejec-
tion and high productivity. Pretreatment
studies demonstrated that conventional
pretreatment alone or with sand filtra-
tion was not adequate to remove
foulants at the surface water site. Alum
coagulation-settling and rapid sand fil-
tration (ACSSF), microfiltration (MF),
and granular activated carbon (GAC)
filtration were found to significantly re-
duce fouling in the pretreatment stud-
ies and were used in the pilot plant
studies. Six different NF systems were
studied at the surface water site con-
sisting of three different pretreatments
and two different membranes. These
systems were operated for 16,770.8
out of 17,944.5 hr with less than 7%
downtime. The AC-DS5 system was
found to have the lowest rate of MTC
decline, 0.00010/d2. GAC pretreatment
was the least effective means of reduc-
ing MTC decline. Alum coagulation
* Mention of trade names or commercial products does
not constitute endorsement 'or recommendation lor
Printed on Recycled Paper
-------
(AC), and the DS5 nanofilter were the
most effective pretreatment process
and membrane for minimizing MTC de-
cline. The DBPFP averaged 22 |ig/L as
Ch in the permeate stream and repre-
sented more than 98% reduction of
the raw water DBPFP. Cost estimates
for a 10 mgd groundwater NF plant
were $18,424,250 to construct and
$0.58/Kgal for operation and mainte-
nance costs. Cost estimates for the 10
mgd surface water NF plants using
alum, MF, and GAC pretreatment were
$26,402,250, $30,537,250, and
$24,571,750 to construct and $1.41/Kgal,
$0.96/Kgaf, and $1.04/Kgal for opera-
tion and maintenance costs, respec-
tively.
This Project Summary was developed
by EPA'a Risk Reduction Engineering
Laboratory, Cincinnati, OH, to announce
key findings of the research project
that Is fully documented in a separate
report of the same title (see Project
Report ordering information at back).
Analytical and Field Procedures
Standard analytical procedures as ref-
erenced in the 17" Edition of Standard
Methods (SM) and EPA were utilized for
this project with the exception of DBP,
which were measured by interim EPA
methods 551 for the base neutral and 552
for the acid extractable DBPs. The
trihalomethanes (THMs) were measured
by the approved EPA method 501 and
were also measured as a base neutral in
EPA method 551. Both 501 and 551
THMs are contained in the final report.
THMs discussed in this document were
measured using EPA method 551. Com-
parison of permeate THMs measured by
EPA methods 501 and 551 found statisti-
cally equivalent measurements for total
THMs, dichlorobromomethane and
dibromochloromethane, but different for
chloroform and bromoform. All methods
and instruments were standardized and
calibrated before use.
Field procedures included the installa-
tion and operation of a membrane pilot
plant at the Daytona Beach, FL, ground-
water site, and the installation and opera-
tion of three membrane pilot plants at the
Melbourne, FL, surface water she. Field
samples were transported to University of
Central Florida (UCF) in ice filled coolers
on the day of collection. Laboratory analy-
sis of field samples was completed within
the specified SM or EPA holding times
which never exceeded 2 wk.
Groundwater Study
The groundwater NF study for DBPFP
reduction was divided into phases for
membrane selection, pretreatment, and
long-term productivity. The groundwater
site selected for the year of operation was
located at the Daytona Beach Brennen
water treatment plant on the east coast of
Florida. This site was selected because
of the high DBPFP, utility support, and
proximity to UCF.
Groundwater Membrane
Selection
Eleven membranes were tested at
Daytona Beach for DBP control and pro-
ductivity with the use of a small-scale
membrane unit. Nine of the 11 mem-
branes tested met a 0.1 mg/L as Cl~ DBP
concentration limit, and from these, the
A15s was selected on the basis of pro-
ductivity. The A15s had an MTC of 0.0094
day1.
Groundwater Pretreatment
Studies
A pretreatment study was performed
using the bench-scale membrane system
and one A15s membrane. The raw water
was adjusted to pH 6.2 and continuously
passed through a 5m filter for 571 hr at 30
to 65 psi and 20% to 25% recovery. The
stability of the MTC demonstrated that no
pretreatment other than acidification and
filtration was required. Consequently only
conventional membrane pretreatment was
required for pilot plant operation at the
groundwater site.
Groundwater Operation
The plant personnel recorded pressure
and flow daily and university personnel
conducted trailer repairs, chemical
makeup, and sample collections. The flow
diagram of the groundwater site NF pilot
plant is shown in Figure 1. Each pressure
vessel contained three 4-in. x 40-in. mem-
brane elements in series. During the year
of operation, 8938.4 hr were available for
operation. The pilot plant operated 8650
hr with 54 hr of unavoidable downtime
and produced water more than 99% of
the time.
A change in either pressure or recovery
defined operational periods that are des-
ignated by number and vertical line in
Figure 2, the system water MTC versus
time. System recovery was varied from
70% to 90%, and feed pressure varied
from 110 to 170 psi. In this figure three
natural operational time periods were es-
tablished by the MTC slope. The rate of
MTC decline during the first period, from 0
to 4750 hr, was 6.53E-6/d2. This decline
rate was representative of normal mem-
brane fouling and deterioration. From 4750
to 5750 hr, the MTC declined at a rate of
3.96E-4/d2. During this time, new unde-
veloped wells were put on-line and caused
increased colloidal fouling. The mem-
branes were cleaned with a high phos-
phate cleaner at hr 5750, and the MTC
decline rate was 4.08E-6/da for the re-
maining operation to 8650 hr. The MTC
did not change as operating conditions
changed but did decline steadily during
operation. The overall MTC decline dur-
ing the year of operation was 24%, 15%
was attributed to normal membrane dete-
rioration and 9% was attribute to irrevers-
ible well field fouling.
The raw, feed, and permeate DBPFP
for the surface water and groundwater
sites are reported in Table 1. Average
DBPFP concentrations in the groundwa-
ter feed and system permeate were 503
and 20 ng/L as Ch, respectively. THMs
accounted for 72% and 75% of the feed
and permeate DBPFP. The other major
DBPs found in the feed water were
haloacetic acids (HA), averaging 105 u.g/L
as Ch, and chloral hydrate, averaging 32
u.g/L as Ch. The same DBPs, THMs, HA,
and chloral hydrate found in the feed wa-
ter were dominant in the system perme-
ate. THM, FP, HA, and chloral hydrate
averaged, 15, 3, and 2 ug/L as Ch in the
system permeate. Chloroform was the
dominant species in the feed and system
permeate, comprising 67% of the feed
and 40% of the permeate DBPFP. The
two major HA in the feed were di- and
trichloroacetic acid, comprising 9% and
12% of the DBPFP. Mono-, di-, and tri-
chloroacetic acid were the major HA in
the permeate accounting for 5% each of
the DBPFP. Over the year, 96% of the
feed stream DBPFP was rejected and the
permeate stream DBPFP never exceeded
0.1 mg/L as C|-
The total organic halogen formation po-
tential (TOXFP) and non-purgable dis-
solved organic carbon (NPDOC), and se-
lected inorganic water quality parameters
for the surface and groundwater sites are
reported in Table 2. On average, 96% of
the TOXFP was rejected, corresponding
to the rejection of DBPFP. NPDOC aver-
age rejection was 98%. Changing operat-
ing conditions of feed pressure and sys-
tem recovery did not appear to affect the
rejection of either TOXFP, NPDOC, or
DBPFP. This result would indicate that
the large organic molecules that comprise
these groups are rejected by sieving rather
than by diffusion. Calcium hardness aver-
aged 264 mg/L as CaCO3 in the feed and
22 mg/L as CaCO3 system permeate. TDS
averaged 368 mg/L in the feed and 48
mg/L in the system permeate.
Surface Water Study
The surface water investigation was
similar to the groundwater NF investiga-
tion in that a membrane selection, pre-
treatment, and long-term productivity stud-
-------
Stage 1
Stages
Stage 3
Raw Water
Concentrate
t*d Valve
«P W/p/i Pressure Pump
Numbers indicate gage and/or sample locations
Figure 1. Groundwater membrane pilot plant flow diagram.
Permeate
a
0.015
0.012
§0.009 -
O.OOS -
0.003
7000 2000 3000 4000 5000 6000 7000 SOOO 9000
Hours of Operation
Figure 2. Groundwater system (MTC) for the nanofiltration pilot plant.
les were conducted. After several poten-
tial surface water sites were considered
for pilot plant operation, Melbourne, FL,
was selected because of the very high FP
of the water, city support,'and the prox-
imity to the university.
Surface Water Site Membrane
Selection
A total of 10 different spiral-wound thin-
film composite membranes were investi-
gated for pilot plant operation by monitor-
ing flow and pressure and by analyzing
water quality samples collected during the
operation of each of the membranes in a
single element system. Seven of the
tested membranes reportedly have mo-
lecular weight cutoffs less than 300
and removed more than 95% of the
DBPFP from 1522 u.g/L to less than 100
(ig/L as Ch The productivity of the mem-
branes as measured by the water MTC
ranged from 0.0041/d to 0.0296/d averag-
ing 0.0094/d. Two membranes, an A15s
and a DS5, were selected on the basis of
DBPFP precursor rejection and high pro-
ductivity.
Surface Water Site
Pretreatment Studies
Four different pretreatment techniques
in addition to conventional pretreatment
for a NF membrane process were investi-
gated on a short-term basis. Each of the
four techniques included conventional pre-
treatment, which is typically defined as
prefiltration through a filter with pore di-
ameters of 5 to 20m and acid or antiscalent
addition to keep a salt from precipitating
on the feed stream side of the membranes.
The pretreatment techniques were con-
ventional pretreatment, sand filtration,
ACSSF, MF, and GAC filtration. The ef-
fectiveness of the pretreatment techniques
was judged by the resulting MTC decline
over time of operation. Limiting the MTC
decline to a maximum of 15% over an
assumed 2-wk operation period results in
an MTC decline rate 0.0001/d2 for a typi-
cal nanofilter.
The conventional pretreatment system,
5m prefiltration and scaling control, re-
sulted in less than 2 hr of operation be-
fore the prefilter had to be replaced. Sand
filtration in addition to conventional pre-
treatment resulted in less than 8 hr of
operation before the prefilter had to.be
replaced. These experiments clearly dem-
onstrated that conventional pretreatment
and sand filtration in addition to conven-
tional pretreatment using a disbursant was
not adequate for fouling control at the
surface water site.
The remaining pretreatment systems
(ACSSF, MF, and GAC) were more effec-
tive. ACSSF pretreatment resulted in an
MTC decline rate of 0.00035/d2 during
runtimes varying from 40 to 70 hr. A
Memcor cross-flow microfilter (CFMF) was
-------
T«W« 1. Organic Water Quality Summary (ng/L as Cl1 for Groundwater and Surface Water Site Nanofiltration Pilot Plant Site
Raw Water
Source
Ground Surface
NPDOC (mg/L)
TOXFP
DBPFP
Bromochtoroacetonitrite
Dichtoroacetonitrilo
DibromoDcotonitrite
Trichloroacetonitrite
Carbon tetrachtorida
Trichtoroethyfane
Totrachloroethylene
1,1,1-TrichIoroethane
1,2-Dibromoe thane
1 ^-Dibromo-3-chloropropane
Chtoroform
Bromodichlorormethane
Chlorodibromomethane
Bromoform
1,1-Oichloropropanone
1,1,1-Trichforopropanone
Chtoropfcrin
Chloral hydrate
Monochtoroacetic acid
Monobromoacetic acid
Dtchloroocetic acid
Dibranoacetic acid
Trichtoroacelic acid
2-ChIorophenol
2,4-Dichlorophenol
2,4,6-Tnchtorophenol
9
1120
503
0
2
0
0
0
0
0
0
0
0
335
27
1
0
0
1
0
32
2
0
43
0
60
0
0
0
AC
29 10
6232 1248
2446 478
4
27
1
1
1
0
0
0
0
0
865
80
5
3
2
43
1
187
9
0
336
2
878
0
0
0
3
4
0
0
0
0
0
0
1
0
176
65
20
3
0
13
3
40
4
0
60
3
80
0
0
0
Surface
Feed
MF
27
4925
1939
3
35
6
1
0
0
0
6
0
0
735
84
5
3
40
0
146
8
0
267
2
630
0
0
0
GAC
23
3467
1342
3
3
0
0
0
0
0
0
0
0
502
68
5
26
62
6
0
182
3
452
0
0
0
Ground
PERM
A158
0.2
43
20
0
0
0
0
0
0
0
0
8
5
2
0
2
1
0
1
0
1
0
0
PERM
AC-
A15s
0.1
56
18
0
0
0
0
0
0
0
3
4
4
1
1
0
0
1
2
1
0
0
PERM PERM
CGMG- GAC-
DS5 DS5
0.2
47
30
0
0
1
0
3
8
7
3
0
0-
2
1
0
0
0.1
94
14
0'
0
Q
0
0
0
1
3
5
'3
0
0
o
0
0
A
1
•1
1
Q
0
0
Surface
PERM
AC-DS5
0.4
74
35
0
o
o
0
0
0
4
7
10
6
0
0
0
1
2
o
2
1
i
o
0
0
PERM
DFMF-
DS5
1.6
56
26
o
0
o
0
o
o
o
0
1
0
1
4
7
5
0
0
0
1
4
o
•(
2
o
0
0
0
PERM
GAC-
A15s
BDL
62
16
0
0
0
0
o
o
0
0
0
0
1
2
5
3
0
0
0
0
0
1
2
1
•(
0
0
0
PERM
GAC-
A15S
1.3
45
18
0
0
0
0
0
0
0
0
0
0
2
3
5
2
0
0
0
1
1
1
1
2
0
0
0
0
used to pretreat the raw water before NF.
Operational times up to 365 hr using only
the CFMF were achieved before CFMF
fouling terminated the experiment. CFMF-
NF series operation achieved run times of
110 to 150 hr before a scheduled termina-
tion was implemented. The NF MTC de-
cline rate was 0.0006/d2 to 0.0008/d2 in
these studies. The CFMF flux was varied
from 70 to 161 gal/ftVday (gsfd). CFMF
lluxes less than 100 gsfd resulted in longer
times o! operation. GAC pretreatment
resulted in NF MTC declines of 0.0010/d2
or higher. The longest time of operation
(approximately 40 to 60 hr) was achieved
at the lowest GAC surface loading rates
(SLR). GAC SLRs were varied from 1.54
to 0.72 gal/ftVmin .
The surface water site pretreatment
studies demonstrated that some form of
pretreatment would significantly decrease
the rate of NF MTC decline and cleaning
frequencies. AH pretreatment processes
had reduced NF fouling relative to con-
ventional pretreatment. CFMF was the
most effective process for reducing the
rate of MTC decline, but the results of the
pretreatment study showed ACSSF and
GAC as effective processes for reduction
of NF fouling. The pilot plant was modi-
fied from a three-stage system to two sepa-
rate two-stage systems. The CFMF was
used in series with the single-stage pilot
plant, and the modified pilot plant was
used to investigate GAC and alum coagu-
lation pretreatment. Consequently all pre-
treatment processes were evaluated in
long-term operation at the surface water
she.
Surface Water Site Operation
Six separate systems were evaluated
on a long-term basis at the surface water
site that involved three different pretreat-
ment processes and two different
nanofilters. The pretreatment systems —
ACSSF, GAC, and both crossflow and
direct flow microfiltration (CFMF, DFMF)
— were used in advance of the DS5 and
the A15s nanofilters. The systems are
abbreviated as AC-A15S, CFMF-DS5,
GAC-DS5, AC-DS5, DFMF-DS5, and
GAC-A15S. A single-stage NF system
was used to pilot the CFMF-DS5 and the
DFMF-DS5 systems. The remaining NF
systems were operated using a two-stage
pilot plant. The AC-A15S, GAC-DS5, and
CFMF-DS5 systems were operated simul-
taneously in the initial phase of the project.
The AC-DS5, GAC-A15s, and DFMF-
DS5 were operated simultaneously in the
latter phase of the project.
Run Time Analysis
The six NF systems were available for
operation a total of 20.141.3 hr and expe-
rienced a total downtime of 3373.5 hr,
which was subdivided into avoidable and
unavoidable downtime. Avoidable down-
time was classified as time generated due
to research preparation. Unavoidable
downtime, lost because of normal produc-
tion activities, totaled 1173.7 hr. The un-
avoidable downtime, which varied from
5% to 10%, averaged 7% of the produc-
tion time taken as the sum of the runtime
and the unavoidable downtime. Two of
the major categories of unavoidable down-
time were power interruption due to main
plant shutdown and filter backwash, which
accounted for 23% and 10% of the un-
avoidable downtime and would be avoided
in an actual membrane plant environment.
The surface water site NF systems con-
-------
Table 2. Inorganic Water Quality Summary for Groundwater and Surface Water Site Nanofiltration Pilot Plant Sites
Raw Water
Parameter
Color
TDS
Sodium
Total
Hardness
Calcium
Hardness
Chloride
Sulfates
Alkalinity
pH
Iron
Turbidity
Heterotopic
Plate Count
Source
Units
Ground Surface
(CPU) -
(mgn.)
(mg/L)
(mg/L as CaCOJ
(mg/L as CaCO3
(mgO.)
(mgfl.)
(mgfl.)
6.3
(mgO.)
(NTU)
(CFU/mL)
31
368
20
289
264
29
0.3
295
7.7
374
0.67
63
224
379
46
138
99
93
0
95
7.3
304
5.33
3174
AC
12
458
47
196
148
98
90
65
7.8
71
0.42
1427
Surface
Feed
MF
167
364
45
141
101
94
0
111
7.5
154
0.16
978
GAC
149
358
39
135
101
87
4
82
4.4
165
0.93
2030
Ground
PERM
A15s
1
48
7
25
22
5
2
22
3.7
32
0.14
410
PERM
AC-
A15s
1
38
15
26
23
34
6
4
6.5
11
0.11
409
PERM
CFMF-
DS5
1
228
32
56
53
85
1
88
3.2
3
0.12
351
PERM
GAC-
DS5
0
169
29
49
37
68
29
25
2.9
4
0.02
173
Surface
PERM
AC-DS5
1
162
36
27
18
77
16
0
4.8
35
0.00
146
PERM
DFMF-
DS5
1
230
37
-65
45
84
0
29
6.4
25
0.04
1129
PERM
GAC-
A15s
0
149
28
43
26
51
0
24
5.6
120
0.20
1123
PERM
GAC-
AlSs
1
119
20
26
18
40
0
8
2
0.04
tooo
sistently produced water on average more
than 93% of the time.
System Water Quality
The organic solute average concentra-
tions for the most frequently observed
DBPFPs from different systems were ana-
lyzed statistically. The analysis indicated
that systems involving DFMF pretreatment
and the AC-DS5 system produced statisti-
cally different average permeate concen-
trations. These occurrences were most
common to the THM and HA species;
when the occurrences were compared with
the average DBPFPs of each system, the
AC-DS5 system and systems involving
DFMF pretreatment were found to be the
least effective for DBP precursor removal.
Percent NPDOC removal through only
the nanofilter for the corresponding sys-
tem percent recovery and pressure gradi-
ent, as shown in Figure 3 for the CFMF-
DS5 system, is independent of the oper-
ating conditions or percent recovery. This
same trend was found for all systems. In
a diffusion controlled process, the solute
percent removal would increase as pres-
sure increase and recovery decreased.
That trend may be slightly evident for the
AC-DS5 and GAC-DS5 systems, but the
opposite or, more likely, no trend was
indicated for the remaining systems. Siev-
ing is indicated as the major mechanism
of NPDOC removal because the percent
of NPDOC removal is largely unaffected
by variations in pressure and recovery.
As NPDOC is representative of DBPFP
precursors, the majority of DBPFP reduc-
tion would be realized by sieving and would
be independent of process optimization at
high flux and low recovery. A similar
relationship for general DBPFP and
TOXFP independence from pressure and
recovery was also observed. Conse-
quently these data indicate that a very
high DBPFP reduction may be realized by
NF but that optimization of a given mem-
brane for DBPFP reduction is not feasible
by varying pressure and recovery. The
data in Table 1 show that the permeate
DBPFP varied from 14 to 35 ug/L as Gl-
and averaged 22 ng/L as C|- for all sys-
tems. The two lowest permeate DBPFP
average concentrations were realized by
the GAC-DS5 and the GAC-A15s two-
stage systems. The highest DBPFP aver-
age concentration was realized by the AC-
DS5 system. These results indicated that
a THM concentration below 50 jig/L, but
not below 25 ug/L, as the species could
be consistently maintained in a field appli-
cation. The ratio of DBPFP, THMFP, and
HAFP to TOXFP was statistically equiva-
lent in the raw, feed, and permeate
streams for each system at all sites. The
ratio of brominated species was greater
in the permeate than in the feed or raw
streams at the surface water site for all
systems. The concentrations of chloro-
dibromomethane, bromoform, and
dibromohaloacetic acid was in general
greater or equal in the permeate stream
as opposed to the raw and feed streams.
The THMFP of the AC-A15s, GAC-DS5,
and GAC-A15s systems was 11 or 12 ug/
L as Cr and was the lowest for the six
systems. The AC-DS5 system recorded
the highest THMFP, 27 u.g/L as Ch The
HA concentration in .the permeate from
the six systems ranged from 2 to 7 ug/L
as Ch with dibromoacetic acid being the
most prevalent.
The inorganic species was dependent
on NF pressure and recovery, which is
indicative of a diffusion controlled solute
mass transfer process. Consequently in-
organic permeate concentration varied by
changing pressure and recovery but or-
ganic permeate concentration did not in
these studies. All NF systems effectively
reduced color to 0 or 1 cpu on average.
The calcium hardness ranged from 18 to
53 mg/L as CaCO3. All systems realized
moderate sodium removal. The AC-A15S
system achieved the lowest sodium con-
centration in the permeate stream, 15 mg/
L. The microfilter was the most effective
pretreatment process for removal of tur-
bidity and heterotrophic plate count re-
duction.
System Productivity
The NF pilot plants were operated in
time intervals or periods specified by dif-
ferent combinations of flux and recovery.
Four operating conditions were used to
evaluate the performance of the 2-stage
pilot plant using GAC or alum coagulation
pretreatment: 10 gsfd/45%, 10 gsfd/65%,
15 gsfd/45%, and 15 gsfd/65%. Six dif-
ferent operating conditions were used to
evaluate the performance of the single-
stage pilot plant, which specified nanofilter
flux, nanofilter recovery, and microfilter flux:
-------
140
1 2O
I
JB *0°
§
°- 80
60
100
98
98
100
100
98
J_
10
20
30
Recovery (%)
40
SO
Figure 3. Percent NPDOC rejection for the CFMF-DS5 surface water site nanofiltration system.
0.03
rooo
4000
2000 3000
Hours of Operation
Figure 4. Mass transfer coefficients for the various pretreatment systems for the surface water
nanofiltration site.
10gsfd/15% and 15 gsfd/24% each in com-
bination with 50 gsfd, 75 gsfd, and 100
gsid. The GAC-A15s system was also
operated as a single-stage system so that
a lower GAG SLR could be evaluated.
The MTC declined with time within peri-
ods of operation and is shown in Rgure4
for all the surface water site NF systems .
The abrupt changes occurred when the
nanofilters were cleaned.
The productivity of each of the six sur-
face water site NF systems was analyzed
by determining the change of the MTC
(1) with respect to time of operation by
linear regression and (2) with respect to
time of operation, surficial membrane ve-
locity, membrane flux, feed stream NPDOC
concentration, and microfilter flux by mul-
tivariate linear regression. The multivari-
ate linear regression equation is shown
below:
MTC = A + B(FR) + CT + DF+
EO + G(MF)
where: MTC = mass transfer
coefficient, 1/d
A = constant
B,C,D,E,G = linear regression
coefficients
FR = (F/2)(1/R + (1-R/R)), gsfd
T = time of operation, hr
O = NPDOC, mg/L
F = nanofilter flux, gsfd
MF = microfilter flux, gsfd
R = decimal fraction NF
recovery
The results of this analysis are shown
in Table 3. Only the time and MF coeffi-
cients have a consistent sign, negative.
This would indicate a changing relation-
ship between fouling and the remaining
variables identified in the regression equa-
tion. The lack of a consistent sign on the
FR, F, and NPDOC coefficients for the
model will not support any readily appar-
ent overall interpretation of horizontal flux,
vertical flux, and NPDOC on water pro-
ductivity. The system productivity did al-
ways decrease with time and MF flux.
Both linear models were descriptive of the
MTC with a high level of confidence, but a
significant amount of the variation between
the actual and predicted MTCs cannot be
accounted for by the models.
Initially, during simultaneous operation,
the CFMF-DS5 system had the least MTC
decline with time, 3.99E-6, which was 36%
less than the AC-A15S system and 255%
less than the GAC-DS5 system. All
NPDOC coefficients were negative during
the initial operation as expected. The
CFMF-DS5 system MTC was less effected
by NPDOC than were the other systems.
The CFMF-DS5 NPDOC coefficient was
-------
-1.2E-4, which was approximately one-
third of the GAC-DS5 and one-sixth of the
AC-A15S NPDOC coefficients during this
time period. Interestingly, the highest
NPDOC concentrations were received by
the CFMF-DS5 system, 25 mg/L, as com-
pared with 20 mg/L for the GAC-DS5 sys-
tem and '8 mg/L for the AC-15s system.
The rate of MTC decline with respect to
NF flux was unexpectedly positive varying
from 1.46E-4to2.57E-4.
During the simultaneous operation in
the latter phase of the project, the AC-
DS5 system was found to have the least
rate of MTC decline with time as shown
by the multivariate regression coefficient
for time, -5.26E-7, which was more than
one order of magnitude less than that of
the DFMF-DS (-7.77E-6) or that of the
GAC-A15S (-8.42E-6). The AC-DS5 sys-
tem had the least rate of MTC decline
with respect to time for both the linear and
multivariate regression. The regression
coefficients for FR, flux, and NPDOC var-
ied positively and negatively during the
latter simultaneous operation period; this
does not indicate any apparent relation-
ship to MTC and was unexpected.
The multivariate MTC analysis indicated
that any surface water system using GAC
pretreatment had the most rapid rate of
MTC declines, which was approximately
an order of magnitude more than other
systems; the DS5 membrane was less
fouled during operation; and the alum co-
agulation pretreatment produced the least
rate of MTC decline. The GA.C-A15s sys-
Tablo 3. Surface Water Site Solvent Mass Transfer Coefficent (MTC) Multiple Regression
System Constant Coefficients R2 F d(MTC)
AC-Al5sOld*
AC-Al5sNew
CFMF-DS5
GAC-DS5
AC-DS5
DFMF-DS5
GAC-A15S
GAC-A15s"
0.013
0.021
0.013
0.025
0.016
0.014
-0.009
0.002
FR
105
-4.27
-16.6
6.55
25.7
45.1
-1.43
95.7
-35.6
Time
105
-0.70
-0.54
-.040
-1.02
-0.05
-0.78
-0.84
-3.93
NF
Flux
•10*
2.02
2.45
1.46
2.57
-4.27
-1.12
5.97
9.96
NPDOC
10*
-2.41
-7.47
-1.20
-3.39
1.68
-1.77
1.80
8.47
MF
105
0.450
0.452
-3.38 0.317
0.648
0.475
-7.12 0.668
0.610
0.662
100
67
46
146
58
96
76
31
dt
10*
-2.53
-2.10
-1.89
-2.25
-1.00
-1.88
-6.02
-17.4
' Used at Groundwater Site
' Single Stage
Table 4. Summary Cost Table for Various Water Treatment Processes for a 10 MGD Membrane
Nanofiltration Plant
Type of plant
Groundwater
Capital:
O&M
Surface water
Capital:
O&M:
Surface water
Capital:
O&M
Surf act water
Capital:
O&M
$
$/Kgal
$/yr
$/Kgal
$
$Kgal
$/yr
$/Kgal
$
$/Kgal
$/Vr
$/Kgal
$
$/Kgal
$/yr
$/Kgal
Advanced
Pretreatment
None
0
0
0
0
Alum
5,494,000
0.18
2,835,000
0.78
GAC
3,663,500
0.12
760,400
0.20
Microfiltration
9&80.000
0.32
938,000
0.27
Membrane Total Cost $/Kgal
Nanofiltration
18,424,250
0.59
2,109,000
0.58
Nanofiltration
20,908,250
0.67
2,304,000
0.63
Nanofiltration
20,908,250
0.67
3,054,000
0.84
Nanofiltration
20,557,250
0.66
2,514,000
0.69
1.17
2.26
1.83
1.94
tern was operated as a single-stage sys-
tem to determine if the rate of MTC de-
cline could be lessened by a reduced GAC
SLR; however, the rate of MTC decline
increased as the GAC SLR decreased.
These results demonstrated that the DS5
nanofilter was less fouling than the A15s
nanofilter; that decreasing GAC SLR did
not reduce nanofilter fouling; and that the
indicated rate of nanofitter MTC decline
by pretreatment was alum coagulation <
MF < GAC. The linear regression of MTC
with time indicated that GAC pretreatment
was the least effective treatment for the
reduction of fouling; that the AC-DS5 sys-
tem experienced the least rate of MTC
decline (0.000100/d2); that the DS5 mem-
brane had a slightly better rate of MTC
decline; and that alum coagulation was a
better pretreatment process for fouling re-
duction.
Cost Estimates
Cost estimates (Table 4) for the con-
struction, operation, and maintenance of
representative 10 mgd NF plants at
groundwater and surface water sites were
developed for an NF system that con-
sisted of prefiltration, acid addition, NF,
packed tower aeration, disinfection, stabi-
lization, and storage. Additional pretreat-
ment costs were incorporated in the sur-
face water estimates: ACSSF, MF, and
GAC filtration. All capital costs were am-
ortized for 20 yr at 10%, and all unit costs
($/Kgal) are based on 10 mgd of finished
product water.
The groundwater cost estimate for the
NF plant was based on 85% recovery
and 15 gsfd flux. The total cost estimate
for constructing, operating, and maintain-
ing an NF plant at the groundwater site
was $1.17/Kgal. Only conventional pre-
treatment (acid addition and prefiltration)
is required. Construction cost was esti-
mated at $18,424,250 ($0.59/Kgal); the
O&M cost at $2,109,000/yr ($0.58/Kgal).
An NF system constructed and oper-
ated at Melbourne, utilizing a surface wa-
ter source, would have a higher cost due
to required pretreatment. The surface
water NF plants were based on a 10 gsfd
flux and 75% recovery. The cost esti-
mate to build and operate a 10 mgd GAC-
NF system was $24,571,750 or $1.83/
Kgal. The total cost estimate to build and
operate a 10 mgd AC-NF system was
$26,402,250 or $2.26/Kgal.. The total cost
estimate to build and operate a 10 mgd
MF-NF system was $30,537,250 or $1.94/
Kgal.
Summary
NF pilot plants processing highly or-
ganic raw waters were operated at a
•&U.S. GOVERNMENT PRINTING OFFICE: »»3 - 75047I/80N8
-------
Daytona Beach, FL, groundwater site for
8650 hr and at a Melbourne, FL, surface
water site for 16,770.8 hr. Both plants
consistently removed 98% of the DBPFP,
whfch averaged 20 and 22 u.g/L as Cl~,
respectively, in the finished water. The
rate of MTC decline averaged approxi-
mately 2E-7/d2 and 2E-4/6?, respectively,
and the approximate cleaning frequency
at the groundwater and surface water sites
was 6 mo and 1 wk, respectively. The
order of the pretreatment processes based
on the least rate of NF MTC decline was
alum coagulation, MF, and GAG filtration.
The removal of DBP precursors by NF
was independent of pressure and recov-
ery variations at both sites, which indi-
cates DBP rejection by NF can be pre-
dicted by a sieving as opposed to a diffu-
sion model. The cost estimate for the
groundwater plant was $0.59/Kgal for con-
struction and $0.58/Kgal for operation. The
least costly surface water plant used GAC
pretreatment and was estimated to be
$0.79/Kgal for construction and $1.04/Kgal
for operation and maintenance.
The full report was submitted in fulfill-
ment of CR 815288 by the University of
Central Florida under the sponsorship of
the U.S. Environmental Protection Agency.
J.S. Taylor, C.R.Reiss, P.S. Jones, K.E. Morris, T.L. Lyn, O.K. Smith, LA. Mulford,
and S. J. Duranceau are with the University of Central Florida, Orlando, FL 32816.
J. Keith Carswell was the EPA Project Officer (see below).
The complete report, entitled "Reduction of Disinfection By-Product Precursors by
Nanofiltratlon," (Order No. PB92-149 269/AS; Cost: $73.00, subject to change)
wilt be available only from: '.
National Technical Information Service ;
5285 Port Royal Road :
Springfield, VA 22161
Telephone: 703-487-4650
For further Information, Jeff Adams can be contacted at:
Risk Reduction Engineering Laboratory
U.S. Environmental Protection Agency
Cincinnati, OH 45268
United States
Environmental Protection
Agency
Center for Envirpnmental Research
Information ;
Cincinnati, OH 45268
BULK RATE
POSTAGE & FEES PAID
EPA
PERMIT NO. G-35
Official Business
Penalty for Private Use $300
EPA/600/SR-92/023
------- |