WATER POLLUTION CONTROL RESEARCH SERIES
17040—05/70
STUDY AND EXPERIMENTS
IN WASTE WATER RECLAMATION
BY REVERSE OSMOSIS
U.S. DEPARTMENT OF THE INTERIOR • FEDERAL WATER QUALITY ADMINISTRATION
-------
WA1 R POLLUTICI CONTROL RESEARCH SERIES
The Water Pollution Control Research Reports describe the results
and progress in the control and abat nent of pollution of our
Nation’s waters. They provide a central source of information on
the research, development, and dmnonstration ‘activities of the
Federal Water Quality Administration, Department of the Interior,
through in-house research and grants and contracts with Federal,
State, and local agencies, research institutions, and industrial
organizations.
Water Pollution Control Research Reports will be distributed to
requesters as supplies permit. Requests should be sent to the
Planning and Resources Office, Office of Research and Development,
Federal Water Quality Administration, Department of the Interior,
Washington, D. C. 20242.
-------
STUDY AND EXPERIMENTS IN
WASTE WATER RECLAMATION BY REVERSE OSMOSIS
by
I. Nusbaum
J. H. Sleigh, Jr.
S. S. Kremen
Gulf General Atomic Incorporated
San Diego California
for the
FEDERAL WATER QUALITY ADMINISTRATION
DEPARTMENT OF THE INTERIOR
Contract
May, 1970
For sale by the Superintendent of Documents, U.S. Government Printing Office
Washington, D.C. 20402 - Price $1.25
-------
FWQA Review Notice
This report has been reviewed by the Federal Water
Quality Administration and approved for publication.
Approval does not signify that the contents neces-
sarily reflect the views and policies of the Federal
Water Quality Administration, nor does mention of
trade names or commercial products constitute
endorsement or recommendation for use.
-------
ABSTRACT
The basic objective of this program was to apply current reverse
osmosis technology to the treatment and demineralization of secondary
effluents. The development of commercial modules and systems has made
it possible to determine operational problems that might be encountered
with secondary and activated—carbon—processed secondary effluents and
to seek solutions for these problems. Two parallel reverse osmosis
systems were operated with 50 sq ft spiral—wound reverse osmosis
modules. A comparison was made of results obtained with activated
sludge effluent versus activated—carbon—treated activated sludge
effluent, standard flux membrane modules versus high—flux membrane
modules, and large (45—mu) brine spacer modules versus small (22—mu)
brine spacer modules.
Severe fouling problems were encountered almost immediately upon
starting operation with both feeds. Previously developed cleaning
procedures proved generally ineffective. A technique utilizing an
enzyme—based detergent was developed, which restored performance.
Rejection of organic carbon, nutrient, and salt with high—flux membranes
was very good, and the product water quality was not greatly different
than that of the product water produced with standard-flux, high—
selectivity membranes. At the same time, the quantity of product
water produced over an extended period such as three years can be
expected to be substantially more.
Except for a greater pressure drop across modules made with small
brine spacers, there was no appreciable difference in performance between
modules made with large or small brine spacers.
ii
-------
CONTENTS
ABSTRACT
ACKNOWLEDGMENTS
INTRODUCTION
EXPERIMENTAL FIELD
TEST EQUIPMENT
PRETREATMENT
FIELD EXPERIMENTS
TEST RESULTS .
Phase I . .
Phase II .
Phase III .
Phase IV
Phase V
MEMBRANE/MODULE CLEANING
MODULE CLEANING
CHLORINATION
SUMMARY AND CONCLUSIONS.
APPENDIX
REFERENCES
vii
OPERATION
TESTS
viii
1
4
8
14
16
19
19
23
29
39
46
55
61
67
75
81
115
111
-------
INTRODUCTION
This is the final report by Gulf General Atomic Incorporated on
Federal Water Pollution Control Administration (FWPCA), U. S. Department
of the Interior, Contract 14—12—181, “Waste Water Reclamation by Reverse
Osmosis tt , covering the one—year period from July 1, 1968, through
June 30, 1969.
Reverse osmosis appears to have great potential for removing water
of potable or near—potable quality from secondary waste—water effluents
and industrial wastes. Pollution control, by itself, will generally not
require water of the quality that can be produced by this method, but the
fact that the product water may be suitable for direct reuse, or reuse
with minor additional treatment, could be important in making this process
economically feasible. The high membrane packing densities available in
the spiral—wound module concept and the economies attainable with factory
fabrication of modules should make this design a desirable approach.
Gulf General Atomic undertook a program to support and extend the
pilot—plant work on water renovation that had been initiated at the
Los Angeles County Sanitation Districts Pomona facility. The specific
aims of the program were (1) to evaluate the capability of the state—of—
the—art, desalination—oriented, reverse osmosis technology to treat
waste water, (2) to achieve modest extensions of the present technology,
as required, to accommodate waste—water treatment and to achieve a rapid
payoff with respect to demonstrated capability, and (3) to gather the
data necessary to allow a subsequent rapid introduction of the technology
into the fields of waste—water treatment and reclamation.
1
-------
It was recognized that a complete investigation of the possibilities
offered by reverse osmosis would require a broad program, including the
study of completely new membrane materials and configurations, and it
was not proposed to venture into these fields at this time. The effort
was to be directed at adapting the best of present saline conversion
technology to the waste—water field rather than to develop an entirely
new technology. This approach, with the expenditure of some additional
effort, offered the possibility of accelerating the usefulness of a new,
advanced concept of waste-water treatment.
The studies previously conducted at Pomona have served to define
some of the limitations of the direct application of the desalination
module to waste—water treatment. The earlier experiments had a low
level of funding support and proceeded quite slowly since progress was
further complicated by mechanical difficulties of the early systems.
Improvements In the modules and solution of mechanical problems made It
possible for the County Sanitation Districts of Los Angeles County to
conduct an extended reverse osmosis experiment on activated—carbon—treated
secondary effluent. This, in turn, made it possible to consider the
proposed study.
Essentially, the program consisted of the following: (1) design,
construction, and installation of additional field—test facilities at
Pomona, (2) a comprehensive series of short—term (500 to 700 hr) ex-
periments in the new test facilities, and (3) direct and laboratory
support of the short—term field experiments and of the longer—term
testing that the County Sanitation Districts would be conducting in
the 5000—gpd reverse osmosis pilot plant at Pomona. The program was
directed toward an assessment of the state of the art of reverse osmosis
technology and of operational problems that might be encountered. A
computer program available at Gulf General Atomic could be used to assist in
the analysis and interpretation of test results. Laboratory and test
facilities were available at Gulf General Atomic for detailed evaluations
2
-------
of module and membrane parameters and for experiments in support of the
field evaluation. Any new components or techniques developed from the
laboratory support program were to be considered for incorporation under
field conditions in the proposed new field—test program.
It was not expected that this program, because of the relatively short
runs and because a number of variations in modules and operating conditions
were included, would do more than set up the parameters for an extended
program to provide reliable data upon which to base design, construction,
and operation of full—scale reverse osmosis water renovation systems.
The economics of large plants can be proven only by large—scale tests.
Although much information along these lines could be gained from the
operation of brackish water plants, the special problems associated
with processing waste water require that a waste—water program should
ultimately include the design and construction of full—scale plant
modules, in the range of 0.1 to 0.5 mgd, of a plant having a capacity
of 1 to 10 mgd.
3
-------
EXPERIMENTAL FIELD OPERATION
The site selected for this study of the application of reverse
osmosis to waste water reclamation or tertiary treatment was the Pomona,
California, Water Reclamation Plant of the County Sanitation Districts
of Los Angeles County. This plant has become the location for several
important studies in advanced waste treatment by the Sanitation Districts
and the Federal Water Pollution Control Administration. The facility
at Pomona has long played a significant role in water reclamation
studies.
The plant itself is basically a conventional activated—sludge plant
into which has been built some flexibility that permits introduction of
variations in operation that may make the effluent more suitable for
subsequent processing. It is located adjacent to the major trunk sewer
that conducts the wastes of the surrounding community to the central Los
Angeles County system.
Raw sewage is pumped from the trunk sewer and treated at the activated
sludge water reclamation plant, the effluent from which is used in a number
of pilot plant studies. This permits the treatment facilities to operate
under virtually constant hydraulic loading and with minimum variations in
organic and nutrient loadings compared with conventional systems. This in
turn enables the Pomona plant to operate under relatively constant and prob-
ably optimum loading. Primary and waste secondary sludge are returned to
the trunk sewer and, in case of deterioration in treatment for any reason,
operation can easily be controlled to assure a rapid return to satisfactory
conditions. The feeds for the reverse osmosis units consisted of secondary
clarifier effluent from the activated—sludge facility and the same effluent
after it had been processed through a granular—activated—carbon adsorption
system. Typical characteristics of the feeds are shown in Tables 1, 2,
and 3.
4
-------
TABLE 1
MINERAL ANALYSIS OF REVERSE OSMOSIS STREAMS FROM
ACTIVATED-CARBON-TREATED SECONDARY EFFLUENT FEED
*
Feed
Brine
Product
Mineral Content, mg/i
Calcium (Ca) 68 196 0.8
Magnesium (Mg) 12 34 1.5
Sodium (Na) 128 328 14
Potassium (K) 11 35 1.3
Carbonate (C0 3 ) 0 0 0
Bicarbonate (HCO 3 ) 41 67 22
Sulfate (SO 4 ) 85 245 0
Chloride (Cl) 260 720 15
Nitrate (NO 3 ) 1.7 4.2 0.8
Orthophosphate (P0 4 ) 22 56 0.56
Total phosphate (P0 4 ) 27 76 0.6
Fluoride (F) 0.4 5.7 0.3
Boron (B) 0.5 0.7 0.4
Silica (Si0 2 ) 33 85 3.0
Iron (Fe) 0.02 0.07 0
Manganese (Mn) 0 0 0
Hardness (CaCO 3 ) 220 630 8
Total alkalinity (CaCO 3 ) 34 35 18
Total dissolved solids
At 105°C 756 2152 24
At 180°C 672 1756 16
Electrical conductivity,
micromhos/cm at 25°C
1140
2860
104
pH
5.8
6.2
5.7
Turbidity, Jackson units
0.55
0.83
0.27
*
Acidified feed.
5
-------
TABLE 2
SUMMARY OF CHEMICAL ANALYSES OF FEED (F), PRODUCT (F), AND BRINE (B) AT POMONA
(Unit I, Phase III; operated on activated—carbon—treated secondary effluent)
Date
Total COD (mg/i)
F P B
Dissolved COD (mg/i)
N0 3 -N (mg/i)
F P B
N11 3 N (mg/i)
F P B
P a 4 (mg/i)
TDS (mg/i)
F
P
B
F
P
B
F
P (a) B
1/21/69
10.1
0.0
20.8
8.7
———
16.3
5.2
0.6
6.6
10.6
2.3
18.5
18.5
0.2
36.5
578
63
1047
1/23/69
9.8
1.0
26.3
9.4
———
25,7
2.1
1.0
4.6
13.7
2.9
33.6
29.0
0.12
85.0
635
39
979
1/28/69
13.0
0.8
28.0
11.0
———
26.4
9.4
4.2
17.0
2.2
0.5
5.4
29.0
0.12
63.0
674
96
1605
1/30/69
11.7
0.0
27.9
9.5
———
24,6
13.3
4.5
25.5
1.2
0.4
6.0
26.5
0.23
69.0
594
18
888
2/4/69
—
———
0.8
———
1.1
35.5
2.0
53.0
620
79
956
2/6/69
16.7
0.0
26.0
12.9
———
22.0
2.5
0.3
0.6
19.2
2.8
24.0
27.0
0.33
49.0
492
40
562
2/11/69
15.4
1.0
23.9
13.8
———
23.3
1.8
1.2
1.9
20.2
3.0
28.4
34.0
0.4
54.0
616
104
976
2/13/69
———
0.7
0.2
0.3
17.8
1.2
29.6
27.0
0.1
43.5
675
52
1051
2/20/69
14.4
1.0
21.2
13.2
———
18.6
19.3
6.5
26.2
1.8
0.1
2.4
24.0
0.12
48.0
599
44
1003
2/25/69
———
———
21.8
7,1
34.8
———
———
———
502
28
680
2/27/69
5.5
0.0
18.5
6.0
———
12.0
———
———
———
———
———
722
106
1500
NOTE: Samples were taken and ana1y ed by LACSD personnel at their Pomona facility.
(8)Product water COD samples run for total COD only.
-------
TABLE 3
SUMMARY OF CHEMICAL ANALYSES OF FEED (F), PRODUCT (P), AND BRINE (B) AT POMONA
(Unit II , Phase III; operated on secondary effluent)
Date
Total COD (mg/i)
Dissolved COD (mg/i)
N0 3 —N (mg/i)
NH 3 —N (mg/i)
P0 4 (mg/i)
TDS (ing/l)
F
P
B
F
B
F
p
B
F
P
B
F
P
B
‘F
P
B
1/21/69
38.5
1.8
79.7
21.6
———
51.2
8.1
1.9
15.2
9.3
1.4
20.1
17.0
0.7
40.0
603
———
1333
1/23/69
37.6
0.4
84.7
25.1
———
59.4
6.7
2.0
13.0
11.0
1.4
25.9
31.0
0.18
76.0
525
69
1570
1/28/69
40.4
5.8
69.2
28.2
———
53.6
9.4
5.7
16.9
1.4
0.5
3.8
29.0
0.6
54.3
784
99
1248
1/30/69
42.0
0.4
72.3
23.0
———
45.2
16.5
5.3
23.0
1.7
0.2
3.1
27.3
0.15
48.5
321
74
668
2/4/69
———
———
———
———
36.2
1.4
49.2
549
72
974
2/6/69
39.5
1.8
52.7
25.6
———
33.4
1.2
0.5
1.4
16.4
2.0
21.1
22.0
0.03
29.5
452
38
687
2/11/69
40.5
2.0
53.1
27.1
———
32.3
3.9
1.6
4.6
18.2
2.8
24.1
36.0
0.4
48.0
646
60
864
2/13/69
———
2.0
1.3
2.7
17.4
13.8
21.4
26.5
0.1
35.5
703
33
910
2/25/69
———
18.7
10.8
17.5
0.4
0.8
1.2
24.0
0.2
35.0
530
63
659
2/27/69
53.0
0.5
61.5
30.4
———
37.9
22.3
8.1
25.8
———
———
713
160
608
NOTE: Samples were taken and analyzed by LACSD personnel at their Pomona facility.
(a)P OdUC with COD samples ru for total COD only.
-------
TEST EQUIPMENT
The reverse osmosis equipment used in the study was essentially a
production model; this nominally rated 10,000 gpd unit is shown pic-
torially in Pig. 1 and schematically in Fig. 2. Additional instrumentation
and piping were installed in order to permit gathering as much information
as possible. Two 10,000 gpd units (Units I and II) were provided, each
identically equipped and piped so that secondary effluent or carbon—column
effluent could be supplied to either one unit or to both units at the
same time. Each system was equipped with six 4—in.—diameter Schedule 40
pressure vessel (tubes); each vessel could hold up to three of the current
3—ft—long spiral—wound modules, each containing a nominal 50 sq ft of
membrane area. Pressure and flow were provided by three submersible
multistage centrifugal pumps connected in series. The pumps were housed
within 4—in. Schedule 40 pressure vessels and mounted on the same frame
as the reverse osmosis module vessels. The pumps provided a 9 gpm feed
at 800 psi. Provision was made to obtain product and brine samples and
flow measurements from each pressure vessel, as well as total samples
for the whole unit. Pressure differentials could be measured across
each tube. Gas feed chiorinators and chemical proportional feeders for
feeding acid and pretreatment additives were provided. A pH controller
was used on each unit to control the acid used for pH adjustment of the
feed. The loops were also piped so that it was possible to flush the
modules easily with an air—water mixture.
Initially (Phase I), the pressure vessels were arranged in a parallel—
series combination, i.e., the feed flowed through three tubes in parallel,
to two tubes, to one tube (3+2+1). This configuration was used with the
high—selectivity, standard—flux membrane modules, which had an initial (1 hr)
water permeation coefficient A of about 1.5 x 10 g/sq cm_sec atm.* (See
Appendix.)
*Water permeation coefficient A of 1.0 x 10 g/sq cm—sec—atm
8.64 gal./sq ft—day at 600 psi net pressure.
8
-------
PRODUCT SAMP
MANIFOLD
PRESSURE VESSELS (6)
FINAL BRINE
FEED MANIFOLD
LC59771
PUMP VESSELS (3)
PRODUCT
BYPASS
BRINE
FEED
Fig. 1. 1O,000—gpd ROGA reverse osmosis system
-------
DEG F
C
FEED
IN.
SAMPLE
PRODUCT
BRINE
BYPASS
PUMPS P-i , P—2, P-3
COMBINED PUMP OUTPUT = 5 GPM AT 800 PSI
COMBINED PUMP POWER 4.5 HP, 230 V, 3 4, GO CYCLE
SUBMERGIBLE
4—IN. SCM 40 PIPE VESSELS
VICTAULIC END CLOSURES
PRESSURE VESSELS V-i THROUGH V-6
4—IN. SCH 40 PIPE
COATED INSIDE
THREE H MODULES
VICTAULIC END CLOSURE
F Ig. 2. Process and instrumentation diagram
-------
During Phase II, the reverse osmosis units were loaded with high—flux
modules [ A (1 hr) = 2.25 x lO g/sq cm—sec—atm]. Only four of the module
pressure vessels were used, and the array was changed to 2+l l since the
desired nominal flow could be produced with only twelve modules rather than
the eighteen required by the 3÷2 1 array. The tubes are placed in a parallel
series arrangement in order to maintain the fluid flow and turbulence through
the modules great enough to minimize concentration polarization at the
membrane. (See Appendix, p. 109.)
Following the third field experiment, Unit I was modified so that a
portion of the brine or concentrate could be recycled through the unit
(see Fig. 3). Studies at the laboratory of Gulf General Atomic had shown
that a brine flow of 3 to 4 gpm through the modules was required in order
to minimize concentration polarization and maintain flow conditions through
the modules which would produce favorable performance. With flows of less
than 3 gpm through the modules, the net effect is to impair product—water
quality and to increase the probability of membrane fouling and pre-
cipitation of marginally soluble salts. This is further aggravated with
high—flux—membrane modules, which require 4 gpm brine flow. In order to
maintain 3 gpm through the 10,000 gpd units, as originally equipped at
Pomona, the water recovery rate for the initial phases of the study was
limited to 67% or less. This is short of the project goal of 80 to 90%
water recovery. By recycling a portion of the final brine to the feed
to maintain brine—channel turbulence, it was possible to attain water
recoveries approaching 90%. It should be recognized that recycling of
the brine increases the dissolved solids concentration of the feed, thus
raising the osmotic pressure slightly and changing the dissolved solids
content of the product water slightly. In other words, once recycling
is introduced into a small reverse osmosis pilot plant of the type
utilized here, the results do not correspond exactly with what can be
expected overall from a large reverse osmosis plant; however, the experi-
mental results do simulate what can be expected from the downstream
one—half or one—third of a large plant. (See Appendix, pp. 83—86.) At
the end of the fourth field experiment, Unit II was also modified by
adding a brine recycle pump.
11
-------
BRINE
1/2
FEED
DRAGON
NEW LINE (TYP)
1/2 IN. x 0.035 IN.
WALL TUBE TYP
1/2 IN.
VENTURI
EXISTING LINE (TYP)
APCO 1/4 IN.
0-10 GPM
40 W.C.
1/2 IN.
3/4 IN.
TO MODULE VESSELS
Fig. 3. Schematic diagram of brine recycle system
-------
The operating pressure selected for the field experiment was 600 psi.
This pressure is the normal operating pressure for which Gulf General
Atomic reverse osmosis loops operating on brackish water have been rated.
The equipment, as designed, can be operated at a pressure as high as 1400 psi,
however. Based on previous operational experience, 600 psi represents
the best compromise of system efficiency and membrane properties for
extended life and operation of the system on water containing less than
10,000 mg/i TDS (total dissolved solids). Flux decline, due to compaction
of the cellulose acetate membrane substructure at 600 psi, has been shown
to be low enough so that economical water fluxes can be maintained for
periods in excess of three years.
At the beginning of the fifth field experiment, the pressure vessels
of Unit II were changed from a horizontal to a vertical orientation. This
variation was introduced because from examination of modules after cleaning
studies, there was reason to believe that some advantage both in cleaning
fouled modules and in controlling fouling might be realized from placing
the modules in a vertical operating position. Sediment dropped out rapidly
from modules removed from the loop and placed in a vertical position in
water. Postmortemed modules that had been subjected to air—water flushing
in a horizontal position showed bands of dirty and clean membrane which
were related to position. The dirty bands were at the bottom of the
module and the clean bands at the top; separation of the air and water
was occurring, which limited the cleaning effectiveness. In addition,
although the anti—telescoping devices also act to center the modules,
there is some evidence to show that when a seal failure occurs, it does
so at the top of the seal when the module is placed horizontally.
13
-------
PRETREATMENT
The two feeds available for study at the Pomona Water Reclamation
Plant, as noted already, are the effluent from a standard municipal
activated—sludge waste—treatment plant and the same effluent after it
had been processed through the granular—activated—carbon—column pilot
plant at Pomona. The carbon columns produced an essentially well—filtered
feed for the reverse osmosis study. During earlier studies at Pomona,
markedly reduced fouling of the modules and membranes was observed when
the feed had been treated by granular activated carbon. This had originally
been attributed primarily to filtration, but data obtained during the
course of this study show a correlation between the chemical oxygen
demand (COD) of the carbon-treated feed and membrane fouling.
Two small, pressurized sand filters (Crystaleen Model C—6) with a
surface area of 3.1 sq ft, each, piped in parallel, were included for
removing gross particulate matter from the activated—sludge-plant effluent.
They were used for only a short interval since it was felt that the purpose
of the project was to investigate processing secondary effluent without
pretreatment. The use of gross screening filters requires further consideration,
however, because small particles resembling minute grease balls are to
be found in even the most well—treated secondary effluents and these particles
may cause plugging of the upstream face of the modules.
The effluent from the activated—sludge plant was delivered chlorinated
with about 2 mg/i combined residual. Carbon—treated effluent had no chlorine
residual since it was rapidly dechlorinated by the activated carbon.
Accordingly, the carbon—column effluent was rechlorinated before introduction
into the reverse osmosis unit to give approximately 2 mg/i combined residual.
Chlorination inhibited the growth of shines within the modules as evidenced
by the postmortem examination of both chlorinated and unchlorinated modules.
14
-------
The pH of the feed was adjusted to 5.0 ± 0.5 with hydrochloric acid. There
are several reasons for adjusting the pH of the feed, the most important of
which are prevention of precipitation of calcium carbonate in the modules
and prolongation of membrane life. At pH 5.G, virtually all carbonate al-
kalinity has been removed, and the hydrolysis rate of the cellulose acetate
ester from which the membrane is formed is at a minimum. The rate of hy-
drolysis increases exponentially above and below this range. Above pH 8
and below pH 2, the rate of hydrolysis of the cellulose acetate membrane
is such that replacement of the module would probably be necessary in less
than one year of operation, resulting in excessive costs (see Fig. A—b).
There is an additional requirement for pH adjustment to 6 or less
when treated waste water is processed, particularly where the original water
source is quite hard. The concentration of orthophosphate, P0 4 , ion in
secondary effluent can be expected to be of the order of 30 to 50 mg/i.
CaHP0 4 2H 2 0 solubility is about 200 mg/i at 25°C. Even with only 67%
water recovery, the concentration of calcium phosphate may exceed saturation
in the boundary layer. Precipitates of calcium phosphate, CaHPO 4 , and
calcium phosphate double salts have been found in the modules when pH
control has been lost. Calcium phosphate solubility increases rapidly
below pH 6 when the equilibrium shifts toward monocalcium phosphate,
Ca(H 2 P0 4 ) 2 , so that pH adjustment is an effective scale—control technique.
It was originally proposed that a threshold inhibitor (Cyanamer P—35)
would be added to the feed to inhibit the precipitation of calcium
sulfate and eliminate this source of interference to the project. After
operating the equipment for a short period with the Cyanamer P—35, it was
decided to eliminate its use since sulfate and calcium concentrations in
the feed were low enough that saturation with calcium sulfate would
not be reached at the water recoveries anticipated in this study. In
Table 1 an analysis of a typical feed and brine at Pomona shows that it
would require a more than lOX concentration of the feed to approximate
saturation of calcium sulfate. Laboratory studies had already shown that
neither sodium hexametaphosphate nor Cyanamer P—35 were effective in in-
hibiting the precipitation of dibasic calcium phosphate.
15
-------
FIELD EXPERIMENTS
The field experiments outlined in Table 4 represented the consensus
of Gulf General Atomic and the County Sanitation Districts of Los Angeles
County as the most meaningful steps that could be taken to develop informa-
tion on the capability of reverse osmosis systems to treat waste waters
of the types encountered at Pomona. The list of experiments was modified
to that shown in Table 5 during the course of the work on the basis of
the results of previous tests and by mutual agreement between the FWPCA,
Gulf General Atomic, and the Sanitation Districts.
On the basis of earlier testing at Pomona, a daily air—water flush
of each test unit was incorporated as a standard operating procedure.
This consisted of a 2—mm tap water flush at line pressure (60 to 75 psi)
followed by a 3—mm flush of tap water with compressed air (6& to 90 psi)
added. This procedure was repeated three times in succession. The purpose
of this flush was to minimize the buildup of any material within the
module and thereby minimize fouling problems.
Early in the second phase of the test program, a cleaning technique
was discovered that has had a significant impact on reverse osmosis
performance. This was the demonstrated ability to clean the system
using a solution of a commercial enzyme preparation. Initially, the
units were flushed (closed loop) with a solution of 10,000 mg/l of BIZ*
for 50 mm at ambient temperature. After this flushing, during which
time a considerable amount of particulate matter was removed from the
system, each loop showed approximately 20% increase in product water
flow. The development and results of the cleaning procedure are covered
in a later section of this report.
*BIZ is a Proctor and Gamble enzyme—containing laundry presoak product.
Mention of commercial products does not imply endorsement by the Federal
Water Pollution Control Administration.
16
-------
TABLE 4
PROPOSED EXPERIMENTS
Experiment
Test
Unit
Feed
Treatment(s)
Flushing
Mejnbrane 1
Spacers
Array
Comments
I
I
Carbon—
column
effluent
Continuous
Cl 2 at 2 mg/i
Air/water
15 mm/day
A
Large
Series
To evaluate relative system per—
formance on carbon effluent and
secondary effluent, using large
spacers and highly selective mem—
branes. Although each unit will
be operated at high recovery (>80%)
sampling ports will be provided to
allow an evaluation of performance
at lower recovery as well.
II
Secondary
effluent
Continuous
Cl 2 at 2 mg/i
Air/water
15 mm/day
A
Large
Series
II
I
Carbon—
column
effluent
Continuous
Cl 2 at 2 mg/i
Air/water
15 mm/day
A
Large
Series
To evaluate relative system per—
formance operating on carbon ef flu—
ent, using high— and intermediate—
selectivity membranes. As before,
the units will be operated at high
recovery, with provision for moni—
toring performance at low and
intermediate recovery.
II
Carbon—
column
ef fluent
Continuous
Cl 2 at 2 mg/i
Air/water
15 mm/day
C
Large
Series
ill
I
Carbon—
column
effluent
Continuous
Cl 2 at 2 mg/i
Air/water
15 mm/day
A
Small
Series
To evaluate relative system per—
formance on carbon effluent and
secondary effluent, using small
spacers and high—selectivity mem-
branes. Again, operation will be
at high recovery, with provision
for monitoring lower—recovery
performance.
II
Secondary
effluent
Continuous
Cl 2 at 2 mg/i
Air/water
15 mm/day
A
Small
Series
I V
V
I
Secondary
effluent
Continuous
Cl 2 at 2 mg/I
Air/water
15 mm/day
A
Large
Series
To evaluate relative system per—
formance operation on secondary
effluent when using high— and inter-
mediate selectivity membrane.
Again, high—recovery operations are
planned, with provision for monitor-
ing lower—recovery performance.
Feed, treatment, membrane types,
operating parameters, etc., to be
selected on the basis of results
of Experiments I through IV.
II
Secondary
effluent
Continuous
Cl 2 at 2 mg/l
Air/water
15 mm/day
—
C
——
Large
——
Series
——
I-I
values are combined residuals.
(b)A — high—selectivity membrane; C — intermediate—selectivity membrane.
-------
TABLE 5
FIELD E) PERIMENTS CONDUCTED
Experiment
(Phase No.)
Test
Unit
Feed
Treatment
Flushing
(b)
Membrane
Spacers
Module
Array
I
I
Carbon—
column
effluent
Continuous
Cl 2 at 2 tag/i
Air/water
15 mitt/day
A
Large
Series
II
Sand—filtered
secondary
effluent
Continuous
Cl 2 at 2 mg/i;
10 mg/l
Cyanamer P-35
Air/water
15 mitt/day
A
Large
Series
II
I
Carbon—
column
effluent
Continuous
Cl 2 at 2 mg/i
Air/water
15 mm/day;
BIZ cleaning
once a week
A
Large
Series
II
Carbon—
columm
effluent
Continuous
Cl 2 at 2 mg/l
Air/water
15 mitt/day;
BIZ cleaning
once a week
C
Large
Series
III
I
Carbon—
column
effluent
Continuous
Cl 2 at 2 mg/i
Air/water
15 mm/day;
BIZ cleaning
once a week
C
Large
Series
II
Secondary
effluent
Continuous
Cl 2 at 2 mg/i
Air/water
15 mm/day;
B1Z cleaning
twice a week
C
Large
Series
IV
I
Secondary
effluent
Continuous
Cl 2 at 2 mg/I
Air/water
15 mm/day;
BIZ cleaning
twice a week
C
Large
Series
II
Carbon—
column
effluent
Continuous
Cl 2 at 2 mgll
Air/water
15 aim/day;
BIZ cleaning
once a week
C
Small
Series
V
I
Carbon—
column
effluent
Continuous
Cl 2 at 2 mg/i
Air/water
15 mitt/day;
BIZ cleaning
once a week
C
Large
Series
II
Carbon—
column
effluent
Continuous
Cl 2 at 2 mg/i
Air/water
15 mm/day;
BIZ cleaning
once a week
C
Large
Series (c)
(&)Chlorjfle values are combined residuals.
(b)A — high—selectivity membrane; C — intermediate—selectivity membrane.
(C) jj tubes in vertical orientation.
18
-------
TEST RESULTS
The results of the five sets of field tests are described below,
along with the objectives and conclusions.
PHASE I
This experiment was designed to compare system performance of reverse
osmosis units operating with carbon-column effluent and with sand—filtered
secondary effluents as feeds. The modules were fabricated with large
(polypropylene Vexar) brine—side spacers to minimize system pressure
drop and to further minimize potential fouling problems, and with
high—selectivity, standard—flux membranes. A gross evaluation of the
capability of reverse osmosis modules fabricated with highly selective
membrane to treat both carbon—column effluent and secondary effluent
was to be made from this experiment.
Unit I operated on clarified secondary effluent that had been passed
through a carbon adsorption column. Unit II operated on sand—f iltêred
secondary effluent. The pH on both units was adjusted to 5.5 ± 0.5 with
hydrochloric acid. In addition, 10 mg/i of Cyanamer P—35 was added to
minimize any possibility of calcium sulfate precipitation. It was decided
that the- addition of Cyanamer P-35 would not be necessary in future work
as long as the recovery rates of 75 to 80% were not exceeded.
Because of the varying pressures employed during the first 300 to 400
hr of operation, it is difficult to interpret compaction and fouling data.
Since the majority of membrane compaction takes place during the first few
hundred hours of operation, a more meaningful flux decline slope can be
drawn from the data available after 300 hr. In addition, the pressure was
constant at 600 psi from that period on, and steady—state operation was
achieved. This test, as well as all the succeeding ones, was operated
at 600 psi.
19
-------
Figures 4 and 5 are log—log plots of the water permeation coefficient
A versus time for Units I and II. A line is plotted that shows a flux
decline slope of —0.06. This is a reference slope. Data obtained from
laboratory studies on compaction and from field installations on flux
decline have shown that a flux decline of —0.06 is typical of what can
be expected from modified cellulose acetate membranes subjected to feed
waters free of undissolved solids at 600 psi and 25°C (see Appendix).
As shown by the curves, the flux decline slope for Unit I (carbon—
column effluent) is equal to —0.16. Unit II exhibits a slope of —0.425
over the same period. Since this time period (400 hr to 800 hr) represents
steady—state operation for the two systems and —0.06 is typical of the
compaction slope, it can be seen that fouling causes the log—log slope
of the water flux decline curve to be approximately three times greater
than normal when carbon—column effluent is the feed and approximately
seven times greater than normal when sand—filtered secondary effluent
is the feed.
Water flux declines of this magnitude cannot be tolerated since at
the end of one year a slope of —0.16 would represent a 50% loss in water
flux and a slope of —0.425 would represent an 85% loss in water flux. A
module with a flux decline slope of —0.06 would show a decline in water
flux of 23% in the same time. The undissolved materials must either be
removed upstream of the reverse osmosis unit or development of a procedure
that would prevent fouling or enable rapid, economical cleaning was
necessary.
Both units showed excellent rejections throughout this phase of the
program. Unit I consistently gave 95% TDS rejection as measured by
conductivity rejection, while Unit II was 93% or better. This lower
rejection by Unit II can be explained by boundary—layer phenomena
resulting from the higher degree of fouling.
The results of this phase did indicate, however, that it was feasible
to operate reverse osmosis modules on both carbon—treated and secondary
effluents and that further testing was warranted.
20
-------
1 .2 -i
w
z
ir
0
— REFERENCE FLUX DECLINE
SLOPE = -0.06
1.0 __________________
— 0.8 ACTUAL FLUX DECLINE
w
0
C-)
z
2 0.6
I -.
CARBON-COLUMN EFFLUENT
STANDARD-FLUX MODULES
w
0)4 I I I E
100 200 +O0 600 800. 1OO0
TIME. (HR)
Fig. 4. Water permeation coefficient A versus time; Unit I, Phase I
-------
__ 1.2
z
I-
U,
C 4
SAND-FILTERED SECONDARY EFFLUENT
LI STANDARD-FLUX MODULES
><
• 1.0
I-
z
U i
0 8 REFERENCE FLUX DECLINE
SLOPE = -0.06
O.6 V
SLOPE = -0. 25
ACTUAL FLUX DECLINE
&
U i
I-
I I I I. I
100 200 L OO 600 800 1000
TIME (HR)
Fig. 5. Water permeation coefficient A versus tune; Unit II, Phase I
-------
PHASE II
This experiment was designed to compare the performance of modules
fabricated with high— and intermediate—selectivity membranes when they
were operated on carbon—column effluent. The formulation and casting of
high— and intermediate—selectivity modified cellulose acetate membranes
is identical; differences in performance in terms of water flux and
salt rejection are produced by varying the annealing temperature (see
Appendix). Since, in general, waste waters are not highly saline, there
would appear to be an advantage to using a higher—flux, less—selective
membrane to increase product water output and reduce treatment costs.
Earlier work had indicated that there might be a plugging problem with
intermediate—selectivity membranes, with a resulting rapid flux decline.
The purpose of this experiment was to examine the relative merits of the
high— and intermediate—selectivity membranes and to evaluate possible
plugging problems. Carbon—column effluent was selected as the feed because,
of the two feeds being evaluated, it seemed to offer the highest probability
of successful operation.
Unit I was loaded with eighteen standard—flux modules (A 1.5) in
a 3÷2÷1 array. Unit II was loaded with twelve high-flux modules (A 2.0)
in a 2 l l array. Both systems were operated at 600 psi on carbon—column
effluent (pH 5,5 ± 0.5, adjusted with HC1). During working days, each
system was flushed with air and water at low pressures. In addition,
each unit was cleaned once a week with a solution of 10,000 utg/l of BIZ
for 50 mm at ambient temperature.
Some pump problems were encountered during this run, especially on
Unit I. These problems were identified as being electrical in nature,
and modifications were made on the three pumps to solve these problems.
In addition, inadequate feed pressure was also a problem. This was
attributed to the chlorination system; plumbing changes that were made
solved this problem. As a result of this problem, the systems were not
chlorinated from approximately Hour 150 to the completion of the run.
23
-------
Figures 6 and 7 represent the rejection performances for Units I and
II, respectively. Unit I (standard—flux modules) dropped from an initial
TDS rejection as measured by conductivity of greater than 96% to around
94.5% after 825 hr of operation. Unit II (high—flux modules) dropped
from 94% to 91.5% after 1000 hr of operation. Decline in rejection of
this magnitude over these periods of time cannot be explained by hydrolysis
at pH 5.5. There are several possibilities that might explain this decline.
The first is the fact that this was the period when BIZ cleaning was first
utilized. The enzyme detergent (BIZ), at the concentrations used of
10,000 mg/i, exhibits a pH of 10. At this pH, the hydrolysis rate for
the membrane is high and could feasibly explain the lower rejections.
Also, in addition to the weekly BIZ cleaning, these modules were not
chlorinated from Hour 150 to the completion of the run. There were also
problems associated with conductivity equipment used which necessitated
changing the meter and cell. The occurrence of these problems coincided
rather closely with the noticeable decline in rejection, as shown in
Figs. 6 and 7. When these variables are considered, it is difficult
to project a meaningful explanation for the relatively rapid decline in
rejection. In any event, the product quality was quite acceptable.
Figures 8 and 9 are log—log plots of the water permeation coefficient
A versus time for Units I and II, respectively. On each figure, there
are two straight lines; one of these is a conservative representation
of flux decline, and the other, an optimistic representation.
The slope of each straight line is noted on each curve. The broken
line curves clearly show a cyclic function. The peaks on both curves
represent data that was taken approximately 20 hr after the weekly BIZ
cleaning. The minimum points are representative of highly fouled values.
By using the weekly BIZ cleaning procedure, reasonable flux decline
slopes have been maintained. Even the conservative slopes are reasonable
for clarified secondary effluent. Flux decline slopes on the order of
—0.16 were noted for comparable membrane on this type of feed during the
first phase of this program.
24
-------
100 I
-
80-
f
— 1 2
60 -
0
I—
( )
CARBON-COLUMN EFFLUENT
STANDARD-FLUX MODULES
N) 4O - 1. CHANGED FROM
CHLORINATED FEED
TO NON-CHLORINATED
FEED.
2. CHANGED CONDUCTIVITY
20 INSTRUMENT.
0
0 200 LfOO 600 800 1000
TIME (HR)
Fig. 6. Percent TDS rejection versus time; Pomona Unit I, Phase II
-------
00 I V
80 H
H
60
z
0
I—
L)
CARBON-COLUMN EFFLUENT
HIGH-FLUX MODULES
L O-
I . CHANGED FROM
CHLORINATED FEED
- TO NON-CHLORINATED
FEED.
2. CHANGED CONDUCTIVITY
20 - INSTRUMENT.
0 1 1 1 _________
0 200 L+oo 600 800 1000
TIME (HR)
Fig. 7. Percent TDS rejection versus time; Pomona Unit II, Phase II
-------
6.0 I I
140
L)
w
(j
c.,1I
2.0
1 SLOPE = -0.0147
1.0 -
CARBON-COLUMN EFFLUENT
STANDARD-FLUX MODULES
I—
0. 1 1 __1 _ _ _ 1 1 1 1 I I I I I I I I I I I I I I I I I I I
1 10 100 1000
TIME (HR)
Fig. 8. Water permeation coefficient A versus time; Unit I, Phase II
-------
6.0
I I I I I I
I
Li
In
c . 1I
L)
U.’
0
‘C
I-
z
LU
LA.
Li
0
00 z
0
Li
I
LU
0 .
LU
I .-
4.O
2.0
1.0
0. 1
TIME (HR)
II
100
S
LOPE = -0.055
SLOPE = -0.0
I I I I I I
C RB0N-COLUMN EFFLUENT
HIGH-FLUX MODULES
10
1000
Fig. 9. Water permeation coefficient A versus time; Unit II, Phase II
-------
As anticipated, the higher—flux modules have a slightly greater flux
decline slope (—0.055 to —0.077) than the standard—flux modules (—0.046
to —0.061). The differences are not abnormal and are in fact reasonably
close to the values obtained in controlled laboratory experiments. Figure
10 is a log—log plot of the expected ranges of A versus time for the high—
flux modules. If the extrapolation is valid (and past experience has shown
this to be the case for pure compaction, i.e., no fouling) and the data
are representative of compaction with little or no fouling, then an eval-
uation can be made of the merits of high—flux versus standard—flux systems.
The poorest situation would be to take the optimistic flux decline slope
for standard—flux membrane and the conservative slope for high—flux modules.
If this is done, it can be seen (Fig. 10) that the two curves cross at
approximately 10,000 hr (about 1.2 yr). Based on these considerations,
it appears that a significant increase in water flux may be obtained
over a 25,000—hr (3—yr) operating period using high—flux modules with a
slight decrease in product water quality. However, additional work
comparing the results for the high— and standard—flux membranes over
longer periods of time is required before a judgment can be made.
PHASE III
This phase, by mutual agreement of participating agencies on technical
grounds, deviated slightly from the proposed program. In view of what had
been learned in prior phases, it was agreed that an evaluation of high—flux
modules on secondary effluent and on carbon—column effluent was in the
best interest of the research program. After minor plumbing changes
and upgrading of some components, both units were loaded with twelve
high—flux modules. The respective feedwaters were adjusted to a pH of
5.5 ± 0.5, and each unit received a daily air—water flushing.
Unit I (carbon—column effluent) was loaded with high—flux modules
having an initial A of 2.6, while Unit II (secondary effluent) was loaded
with high—flux modules having an initial A of 2.5. Both systems were
29
-------
z
C-)
-
C’
C-) U
z V)
C
Li .)
0
- t%
0
I II
l T
10
1.0
0.1
-0.055
-0.077
EXTRAPOLAT ION
FIELD DATAI
-0. 0L 3 7
-0.061
1 10 100
HIGH FLUX
0
TIME (HR)
STANDARD FLUX
CARBON EFFLUENT
1 000
10,000
100,000
Fig. 10. Water permeation coefficient A versus time; extrapolated compaction data based on Phase II
-------
operated at 600 psi in a 2 l l array and with an exit brine flow rate of
3 gpm. Unit I received a once—weekly BIZ cleaning, while Unit II was
cleaned with 10,000 mg/i of BIZ twice a week. Recoveries varied between
40 and 70% during the length of the run.
TDS rejection versus time is plotted in Fig. ii for Unit I and in
Fig. 12 for Unit II. As indicated, the rejection remained essentially
constant at 92%. Figure 13 is a log—log plot of the water permeation
coefficient A versus time for Unit I. The flux decline slope is —0.14.
This is approximately twice the value found for this type of system (high—
flux modules, carbon—column effluent) during Phase II. It appears that
the carbon—column effluent quality may have an effect on the rate of fouling.
Phase II was started immediately after the activated—carbon column had
been regenerated. At that time, effluent from the carbon columns contained
an approximate chemical oxygen demand (COD) of 3 mg/i.
Phase III was started after the carbon columns had been in operation
for some time. The carbon—column effluent gradually declines in quality
over a 2—1/2 to 3 month period, from COD = 3 to COD = 15 mg/i (see Table 2).
During Phase II the unit operated on an effluent with a COD of between
3 and 9 mg/i. During Phase III, the COD was between 9 and 15 mg/l.
The flux decline slopes were —0.07 and —0.14 for Phases II and III,
respectively. A definite correlation could be seen from the feed qualities
and flux decline slopes during these two phases; this Indicated that
cleaning schedules should be adjusted to changing feed conditions.
Unit II was plagued with frequent unscheduled shutdowns due to loss
of feedwater. During these shutdowns, the modules experienced severe
pressure shocks. In addition, the diameters of several of the modules
were found to be slightly small. As a result of these problems, brine
seal failures were observed in this unit. Several of the modules were
removed, the diameters built up, and new seals applied. When the modules
were reinstalled, the rejection improved (see Fig. 12, Hour 300). More
31
-------
‘::
z 60 -
0
I .-)
w
w
1+0
( )
t )
CARBON-COLUMN EFFLUENT
20 - HIGH-FLUX MODULES
0 I 1
0 200 400 600 800 1000 1200
TIME (HR)
Fig. 11. Percent TDS rejection versus time; Pomona Unit I, Phase III
-------
100 I UItTT
SECONDARY EFFLUENT
20 HIGH-FLUX MODULES
0 1 I I I
0 200 L 00 600 800 1000 1200
TIME (HR)
Fig. 12. Percent TDS rejection versus time; Pomona Unit II, Phase III
-------
T I T
BARS DENOTE
BIZ CLEANING
. I
I I I I I J
BRINE RECIRC.PUMP
INSTALLED AT 9140 HR
600 800 1000
TIME (HR)
(-)
w
e.4
L)
LA
0
z
L U
L
L)
0
z
w
a-
w
10
8
6
14
2
100
CARBON-COLUMN EFFLUENT
HIGH-FLUX MODULES
I _____ I I I I I
10,000
200
Fig. 13. Water permeation coefficient A versus time; Unit I, Phase III with extension
-------
shutdowns were experienced during the remainder of the run. Three
modules were removed at the termination of the experiment; each had a
brine seal failure, even though the diameters were within specification.
Figure 14 is a log—log plot of A versus time for Unit II with a flux
decline slope of —0.24. This is a severe decline in water flux, even
though the system was cleaned twice each week with 10,000 mg/i of BIZ.
The results of this experiment were almost certainly perturbed by the
seal failures. Efficient cleanings could not be conducted if the brine
seals were bypassing brine, and fouling during operation would be more
rapid and severe.
Investigations of modules removed after Phase II and of several
removed during and after Phase III indicated some form of attack on the
rubber and possibly on the Silastic RTV adhesive. The adhesive deteriora-
tion looked more pronounced in the system running on secondary effluent.
The brine seals are made of Buna—N rubber, and oxidative attack on the
rubber by the chlorine in the effluent was a strong possibility. A
survey of available compounds indicated that ethylene—propylene is not
subject to this chlorine attack, and a sample lot of the copolymer brine
seals was procured. All the molecules from the two systems were removed,
checked for dimensions, diameters built up with tape, and new ethylene—
propylene brine seals installed using vinyl tape. Figures 15 and 16 show
the appearance of several of the brine seals upon removal. One module
was removed 120 hr after the new seals were installed, and the brine
seal was visibly unaffected at that time.
It was planned to correct the brine seal situation before Phase IV
was started, and this was accomplished.
A short—term run, designated Phase III extension, was conducted to
evaluate seal performance. This brought to light further seal problems,
which were successfully corrected.
35
-------
SECONDARY EFFLUENT
HIGH-FLUX MODULES
I I I I T—T U
I—
w
c -c
c, .4
2:
L)
IJ
0
I-
z
w
C’
3.0
2.0
1.0
0.1
100
I I I
TIME (HR)
1000
10,000
Fig. 14. Water permeation coefficient A versus time; Unit II, Phase III
-------
Fig. 15. Photographs of displaced brine seals (Phase III)
37
-------
Fig. 16. Additional views of displaced brine seals (Phase III)
38
K75558
-------
Brine recirculation was also investigated during this short—term
run but no meaningful conclusion was drawn at that time.
PHASE IV
Previous experiments (Phases I, II, and III) were conducted on
modules using standard polypropylene Vexar brine spacers to minimize overall
system pressure drop and possible fouling problems. By using modules
with small brine spacers, however, it would be possible to increase the
active membrane area per unit volume of the system and thus increase
the product water output. Experiment IV was designed to evaluate system
performance using modules with small spacers.
Unit I was loaded with twelve high—flux modules made with standard
brine spacers and operated on secondary effluent feed. Unit II was also
loaded with twelve high—flux modules. These, however, were fabricated
with special Chicopee polypropylene brine spacers. The special brine
spacers have less open area and thickness, with resulting higher (but
acceptable) system pressure drops; they were studied to provide further
enlightenment with respect to the relationship between brine channel
geometry, turbulence, and fouling. Unit II was operated on carbon—column
effluent.
All of the modules had the more resistant brine seals and carefully
checked diameters. Unit I operated for 1550 hr; at the end of this
time, there was no evidence of brine seal attack or failure.
Both systems were plumbed in a 2÷l l array and initially operated
at the same system conditions as those of the previous run, i.e., 600 psi
with 3 gprn brine flow and pH of 5.5 ± 0.5 adjusted with HC1. Unit I
(secondary effluent) was cleaned twice a week with 10,000 mg/l of BIZ,
while Unit II (carbon—column effluent) was cleaned once a week. Each
system was flushed daily with tap water and air. Recoveries range
from 55 to 75% in both units during this run.
39
-------
Rejection versus time is plotted in Fig. 17 for Unit I and in Fig. 18
for Unit II. As shown, the rejection remained essentially constant at
95 ± 1% for the first 1000 hr. Figure 19 is a log—log plot of the water
permeation coefficient A versus time for Unit I. Since the unit was cleaned
with 10,000 mg/i of BIZ twice a week, it was not possible to obtain a
single flux decline curve. An envelope that contains most of the points
for the first 1000 hr is shown with a slope of —0.090. This is by far
the best sustained performance obtained on secondary effluent. The effect
of the cleaning is evident on this figure as well as on Fig. 20, a linear
plot of product flow (in gpm) versus time. Brine recirculation to increase
brine flow to the level recommended for the modules was initiated at 625
hr, and system pressure was increased to 625 psi to offset the increased
L P. This enabled the recovery to be increased from 60 to 75% at 760 hr
while brine flow was maintained. The unit operated at 625 psi and
4.5 gpm brine flow (3.0 being recirculated). The modules were flushed
individually with tap water at 623 and at 980 hr. One of the pumps
failed at 855 hr and was replaced with a spare.
Unit I operated for 1550 hr during Phase IV. The length of this run
was dictated by several reasons. First, there was time available before
the Phase V modules were available, and there were also some additional
parameters of BIZ cleaning, i.e., concentration and temperature, that
warranted investigation. The cleaning procedures were altered beginning
at about 1000 hr and continued during the final 500 hr of operation,
during which time the unit was cleaned seven times.
Note in Fig. 17 that during the final few hundred hours, when high
temperature (— l15°F) and higher concentrations of BIZ (pH 10, 20,000 mg/i)
were used, there was some indication that rejection was decreasing.
Figure 21 shows the water permeation coefficient A versus time for
Unit II, as well as the effects of the altered BIZ cleaning procedures.
There is a marked increase in water flow; in fact, the flow is as good
as was observed in the first several hundred hours.
40
-------
z
0
(-)
w
Li
H
100
90
80
70
60
I I I I I I
SECONDARY EFFLUENT
HIGH-FLUX MODULES
STANDARD BRINE SPACERS
0 200 400 600 800 1000 1200 1400 1600 1800
TIME (HR)
2000
Fig. 17. Percent TDS rejection versus time; Pomona Unit I, Phase IV
-------
I I I I I
80-
CARBON-COLUMN EFFLUENT
HIGH-FLUX MODULES
70 - SPECIAL BRINE SPACERS
60 I I I I - ___________
0 200 LeoO 600 800 OO0
TIME (HR)
Fig. 18. Percent TDS rejection versus time; Pomona Unit II, Phase IV
-------
1 —
I 1111111 I i1 kli
TIME (HR)
100
I I I I I liii
1000
BARS DENOTE
BTZ CLEANING
SLOPE = -0.090
SECONDARY EFFLUENT
HIGH-FLUX MODULES
STANDARD BRINE SPACERS
I I I __ 1 I I I
10,000
(
z
L)
LA
0
I-
I J
C-)
L )
w
0
C-)
z
0
I—
w
w
a-
U i
I-
3.0
2.0
1 .0
0.1
10
Fig. 19. Water permeation coefficient A versus time; Unit I, Phase IV
-------
I
0
0
-J
U-
I .-
0
0
TIME (HR)
BARS DENOTE
BIZ CLEANING
SECONDARY EFFLUENT
HIGH-FLUX MODULES
STANDARD BRINE SPACERS
/
BRINE RECIRCULATION
STARTED AT 625 HR
8
6
14
2
0
0 200 14oo 600 800
1 000
Fig. 20. Product water flow versus time; Unit I, Phase IV
-------
100
TIME (HR)
I I I I II
—BARS DENOTE
BIZ CLEAN!NG
SLOPE — -0.078
CARBON-COLUMN EFFLUENT
HIGH—FLUX MODULES
SPECIAL BRINE SPACERS
J_____J I I III I I
1000 10000
-— I
z
I-
w
c..J
z
L)
LA
0
I-
w
(-)
3.0
2.0
1.0
0.1
10
I I I III
Fig. 21. Water permeation coefficient A versus time; Unit II, Phase IV
-------
Thus, indications are that there is still much to be gained by further
investigations of BIZ cleaning parameters and the mechanisms of fouling.
Unit II had operated for 632 hr when the run was terminated. High
product conductivity values were observed at 490 hr for one of the first
two parallel tubes. When the modules were removed, the outer tape wrap
of the lead module in the bad tube was found to be split for about one half
its length. This failure could have been caused by a too rapid pressuriza-
tion of the loop. Also, the lead module in the other initial tube was
badly deformed as a result of slippage of the brine seal adapter ring.
These rings were used to adjust the module diameter since the modules
were fabricated with the small Chicopee brine spacer. There was no
evidence of brine seal failure in either loop during this run.
Figure 18 shows TDS rejection versus time for Unit II, which averaged
about 93% until module failure. Figure 21, the log—log plot of A versus
time for Unit II, indicated a flux decline slope of —0.078. During the
period of this run, the carbon—column effluent contained between 10 and
15 mg/i COD. The effects of the once-a--week BIZ cleaning is shown in
Fig. 22.
In an attempt to further identify the fouling agents, samples were
scraped from a module that was removed from Unit II, which operated on
carbon—treated effluent; these samples were submitted to two independent
laboratories for analysis. Results of these analyses are listed in
Tables 6 and 7.
PHASE V
Phase V of the contract was selected to demonstrate a “best effort”
based on feed, treatment, membrane type, spacers, and other operating
parameters. In addition, there was some evidence that warranted a
systematic study of the value of vertical—oriented pressure vessels
to minimize fouling and/or enable more effective cleaning.
46
-------
0 200
L OO
TIME (HR)
600
CARBON-COLUMN EFFLUENT
HIGH-FLUX MODULES
SPECIAL BRINE SPACERS
800
BARS DENOTE
BIZ CLEANING
0
0
-J
I-
L)
0
0
—4
0
8
6
14
2
0
1000
Fig. 22. Product water flow versus time; Unit II, Phase IV
-------
TABLE 6
ELEMENTAL ANALYS IS
(In 7.)
SLIME FROM POMONA REVERSE OSMOSIS UNIT II (5/21/69)
Carbon 45.33
Hydrogen 6.61
Nitrogen 6.04
Oxygen 34.27
Sulfur Not detected (<0.1)
Phosphorus 2.31
Ash 11.02
EMISSION SPECTROGRAPHIC ANALYSIS OF ASH
(% of Element in Ash)
Sodium . . 7.1 Copper . . . 0.42
Silicon . . 5.0 Nickel . . . 0.09
Calcium . . 12 Molybdenum . 0.09
Manganese . 0.31 Tin 0.49
Iron . . . 1.4 Silver . . . 0.002
Magnesium . 2.9 Zinc . . . . 0.20
chromium . 0.58 Titanium . . 0.19
Aluminum . 1.5 Zirconium . . 0.10
Cobalt . . 0.002 Potassium . . 1.4
Boron . . . 0.06 Strontium . . 0.06
Lead . . . 0.21 Gallium . . 0.02
48
-------
TABLE 7
ANALYSIS OF SLIME FROM POMONA REVERSE OSMOSIS UNIT II (5/21/69)
Gross Composition, %
Moisture
Total Solids
Solids Composition, %
Crude protein
Crudefat
Crude fiber + carbohydrate
Ash
Subculture and Growth
Growth in thioglycollate medium
Growth on Sabourauds medium
Growth on wort agar
Growth on Difco Micro Assay agar
Growth on potato dextrose agar
32.3
1.8
57.3 (by difference)
8.7
Very heavy, 24 hr
Heavy, 72hr
None
Extensive; favors
aerobic conditions
Extensive
97.0
3.0
Cultures did not sporulate or change color of media.
49
-------
Thus, both systems were plumbed in a 2-’-l-’-l array, with Unit I
horizontal as before and Unit II having vertical pressure tubes. Both
systems were to be operated as nearly identically as possible. Both
units were loaded with high—flux modules with standard Vexar brine channel
spacing and operated at 600 psi, with 4.5 gpm brine flow (3 gpm being
recirculated) and pH of 5.5 ± 0.5 adjusted with Rd. Both units operated
initially on carbon—column effluent with weekly BIZ cleanings and daily
air—water flushing.
Figures 23 and 24 are graphs of TDS rejection versus time for Units
I and II, which averaged about 93% for both units.
Figures 25 and 26 are log—log plots of A versus time for both units.
The rather severe flux declines experienced in the first 100 hr were
caused by lack of acid addition and by effluent of extremely poor quality
due to the excessive suspended solids. Both units were flushed with
acidified tap water (pH 3 to 6) for several days before operation was
continued.
Water recoveries as high as 87% were attained in Phase V with no
apparent operating problems.
The flux decline slopes indicate that there is no advantage in
vertical orientation over horizontal orientation, but this may warrant
another investigation under a larger range of testing conditions.
Further testing, utilizing these units and modules, was undertaken
by the FWPCA and the County Sanitation Districts of Los Angeles County.
50
-------
100 I I I I
I I
I I
90
PLANT SHUTDOWN
OF ABOUT 1 WEEK
z __________ ____ I I
o 1 . - -. 2 3
— I I
-‘ 80- I I -
IaJ I
w I I
HIGH-FLUX MODULES I
(I ,
I CARBON-COLUMN EFFLUENT I
I.- I I
2 SAND-FILTERED SECONDARY I I
70 - EFFLUENT (NO CHLORINE) I
3 UNFILTERED SECONDARY I
EFFLUENT (CHLORINATED) I
6c I I I I I I I I
0 200 400 600 800 1000
TIME (HR)
Fig. 23. Percent TDS rejection versus time; Pomona Unit I, Phase V
-------
i::
PLANT SHUTDOWN I PLANT SHUTDOWN
OF ABOUT 1 WEEK OF ABOUT 1 WEEK
z
0
‘ -‘ 80-
w
U i Li
I HIGH-FLUX MODULES
70 1 . CARBON-COLUMN EFFLUENT
1 —2-— 2. SAND-FILTERED SECONDARY
EFFLUENT (NO CHLORINE)
60 1 I I I I I
0 200 400 600 800 1000
TIME (HR)
Fig. 24. Percent TDS rejection versus time; Pomona Unit II, Phase V
-------
z
I —
LU
0
I .-
z
LU
L)
U i LU
L&) 0
L)
0
I-
LU
LU
0 ..
LU
3.0
2.0
1.0
0.1
10
11 .—BARS DENOTE
BIZ CLEANING
100 1000
TIME (HR)
10,000
Fig. 25. Water permeation coefficient A versus time; Unit I, Phase V
-------
__________ — i till i i i i i itit
I — ______________________ ______________________________
_______________ 1 1 2—.’-
(-) I
w 2.0 . •
. .—BARS DENOTE
‘ BIZ CLEANING
c..J . . . .
(_) . I •II 4I.
—S
I
LA 1.0 •
0
— I
‘C
I- I
z
U i I
U, U
HIGH-FLUX MODULES
Ui
C
L) - 1 CARBON-COLUMN EFFLUENT
z
o 2 SAND-FILTERED SECONDARY
EFFLUENT (NO CHLORINE)
Ui
Ui
0 I ji
I ii
Ui
‘—0.1 1 I I j liii I iii iii iitl I I I I 1111
10 100 1000 10,000
TIME (HR)
Fig. 26. Water permeation coefficient A versus time; Unit II, Phase V
-------
MEMBRANE/MODULE CLEANING TESTS
Almost immediately after the initiation of operation of the two new
10,000 gpd reverse osmosis pilot plants at the Pomona facility, it became
apparent that the available techniques for controlling membrane fouling
were not adequate.
Preliminary procedures for screening various compounds and techniques
in the laboratory were established. Earlier consideration had been given
to three basic approaches. The first involves the removal from the feed
of substances to which fouling could be attributed; the second involves
addition to the feed of additives that would inhibit fouling; and the
third is the use of cleaning techniques and additives to remove from the
membrane rejection surface and brine channels the materials that seriously
affect plant performance and efficiency. This discussion is primarily
concerned with the third approach, although there is included within this
report some information on the first and second approaches obtained during
the field operations.
The criteria for any approach to cleaning required that the technique
or materials should be economical, generally available, easily controlled,
and not detrimental to the reverse osmosis modules or system. One
classification of materials which came to the attention of the project
was the new, commercial, enzyme—containing laundry presoaks, and enzyme—
containing laundry detergents. The use of enzymes as an aid to cleaning
reverse osmosis modules was under review since gross analyses performed
on scrapings from modules used at Pomona and at Santee in earlier studies
had shown that these scraped materials were principally organic. Dean
et al . (Ref. 1) and Busch and Stumm (Ref. 2) have described the colloidal
and suspended material in secondary effluents as mostly biological debris
55
-------
consisting of proteins, lipids, and polysaccharides. It thus appeared
possible that an enzyme product might be found that would loosen the
bond of the deposited material on the membrane and hasten its removal.
The first tests were made with BIZ (Proctor and Gamble), an enzyme
laundrey presoak. A sample of membrane from a “dirty” Pomona module was
immersed in a solution containing 10,000 mg/i of BIZ at 35°C for 30 mm.
The solution had a pH of 9.9. The membrane cleaned up rapidly, compared
with no change in a control sample in water.
A fouled Pomona module (3009A, No. 423) was placed in a laboratory
test loop at 600 psi and checked on a sodium chloride solution. The
following performance was recorded:
Rejection 92.42%
A (water permeability coefficient) . .0.92 x l0 g/sq cm—sec—atm
B (salt permeability coefficient). . .2.95 x 10 cm/sec
The module was then given a static soak in a BIZ solution (pH 9.9,
2000 mg/i) for 30 mm at 40°C. The module was retested at 600 psi,
with the following results:
Rejection 92.16%
A 1.016 x 10 g/sq cm—sec—atm
B 3.4 x 10 cm/sec
There was an approximately 10% increase in water flux. A postmortem
showed that the module was cleaner than a similar uncleaned module from
the same pressure vessel. To determine the effect of the BIZ solution
on salt rejection, samples of cellulose acetate membrane were immersed
in 10,000 mg/l BIZ solutions. One solution had a pH of 9.9, and one was
adjusted to pH 7.5 with 11C1. The membrane was tested several times over
a 48-hr period. The sample at pH 7.5 was unaffected; the sample at
56
-------
pH 9.9 showed a more than tenfold increase in B. As expected, the
membrane could be seriously damaged by hydrolysis on exposure to a pH
of about 10 in relatively short periods. However, the potential of being
able to clean the membrane with the laundry presoak was important enough
to extend the tests to the field operation and to expand the laboratory
investigation to determination of the parameters that might be controlled
to improve cleaning while limiting membrane damage. The field units were
flushed in a recycle mode (closed loop) with a BIZ solution of 10,000 mg/i
for 50 mm at ambient temperatures. Each pilot plant showed a 20% increase
in product water flow following the cleaning, with no significant change
in TDS rejection.
A survey was made of the available compounds on the market which might
be of value in module cleaning. These were tested on “dirty” membrane
samples from a used module from Phase I of the field study. A proteolytic
enzyme, Rhozyme PF, and non—ionic detergent were included. The results,
based on visual observation as noted, are shown in Table 8. Any attempt
to test these samples in a high—pressure reverse osmosis cell would have
caused scouring of the rejection surface and affected the results.
Since BIZ appeared to be clearly superior to the other products
tested at that time, additional tests were run on BIZ in an effort to
optimize the cleaning conditions for reverse osmosis modules. The results
are detailed below:
(a) pH 10.2 (normal BIZ solution pH) . Samples of membrane from fouled
Pomona modules were subjected to concentrations of 10,000, 20,000, and
40,000 mg/i at 23° (ambient) and 40°C for 10, 20, and 40 mm. At 23°C,
the best visual cleaning results were obtained at 20,000 mg/i in 40 mm.
The results at 40,000 mg/i showed no improvement over 20,000 mg/i. At
40°C, excellent cleaning was found at 10,000 mg/i in 10 mm.
(b) pH 7 (BIz solution adjusted with HC1) . Tests listed in (a), above,
were repeated at pH 7. The increase in temperature did not have as great
57
-------
TABLE 8
INITIAL ENZYME DETERGENT SURVEY(a)
The concentration level for all tests was 10,000 mg/i.
The soak time was 30 mm.
(b)
Product Sample
pH
at 23°C
Soak Test
At 23°C
At 55°C
At 23°C with
pH 7.5, UC1
Axion (1)
(Colgate Palmolive)
10.4
Fair
Good
Poor
BIZ (1)
(Proctor and Gamble)
10.20
Good
Excellent
Fair
Family Tree (2)
(Calusa Chemical Co.)
10.15
Poor
Good
Poor
Gain (2)
(Proctor and Camble)
10.25
Fair
Good
Very poor
Tide XK (2)
(Proctor and Gamble)
9.92
Poor
Good
Very poor
Drive (2)
(Lever Bros.)
10.0
Fair
Good
Poor
Rhozyme PF
(Robin and Haas Enzyme
with 0.1% Triton X—lOO)
7.5
Poor
Good
Poor
(a)Defjfljtjon of visual appraisal terms: Excellent = definite slime
removal
Good = definite loosening
of slime
Fair = small amount of
slime is loosened
Poor = very little slime
loosening
Very poor = very little to no
slime loosening
(b)S ples are categorized as follows: (1) = laundry presoak, and
(2) = enzyme detergent. Use of trade name does not constitute an endorse-
ment or recommendation by the Federal Government of the item or product
mentioned.
58
-------
an effect on cleaning as in (a). At pH 7, higher BIZ concentrations were
required. The best cleaning results at pH 7 were not as good as could be
obtained at pH 10 and required a 50,000 mg/i concentration. Satisfactory
cleaning can be reached with the 50,000 mg/i at pH 7 and 40°C in 40 mm.
Solutions containing 60,000 mg/i were difficult to prepare.
(c) pH 5 (BIz solution adjusted with Hci) . Very poor cleaning resulted
at pH 5 with all concentrations, time, and temperatures. An unknown pre-
cipitate is formed in the low pH solution.
Membrane was then subjected to 20,000 mg/i BIZ solutions at pH values
of 7, 8, 9, and 10.3. Hydrochloric acid was used to adjust the pH. The
initial membrane parameters were sodium chloride rejection 97.3% with
an A of 1.9 x 10 g/sq cm—sec--atm.
At pH 7 no significant changes were observed after two days and after
four days. Slight changes were observed at pH 8. After two days at pH 9,
rejection had dropped to 92.4% and an A of 1.9 x l0 . Membrane exposed
at pH 10.3 gave 60% salt rejection and A of 1.8 x l0 . These data are
consistent with previous information obtained on the effect of pH on
cellulose acetate membranes. The rapid membrane deterioration observed in
the static tests, however, has not been observed when the enzyme prepara-
tion was used at its normal solution pH in the weekly field cleaning
procedures. Figure 27 shows the results of the 28—day static exposure
but is somewhat misleading since during the period of the experiment the
pH changed, probably due to acetic acid from membrane hydrolysis and
atmospheric carbon dioxide absorption.
The pH values of the solutions at 4 and at 28 days are shown below:
Original pH 4 days 28 days
7 7 6.4
8 7.7 7.3
9 8.9 8.0
10.3 9.2 8.3
59
-------
100
80
z 60
0
I-
IjJ
20
0
800
TtME (HR)
Fig. 27. Percent TDSre ection versus time; membrane life test in BIZ solutions at various pli
levels
4 25 00 200 300 L 00 500 600 700
-------
MODULE CLEANING
A portable individual module cleaning system (Fig. 28) was built
at the laboratory and taken to Pomona. Modules were removed from the
10,000 gpd reverse osmosis ioops and each was cleaned in a 10,000 mg/i
solution of BIZ. Modules from Unit II operating on secondary effluent
were cleaned by recirculating the BIZ solution for 30 mm at 29°C and were
then flushed with tap water for 10 mm. Modules from Unit I operating
on carbon—column—treated secondary effluent were cleaned for 20 mm at
a temperature of 19°C and then flushed with tap water for 10 mm. The
cleaning took place during Phase III of the field operations.
Unit II modules had an average pressure drop of 6.2 psi before cleaning
and 3.9 psi after cleaning under the cleaning—loop flow conditions of about
3 gpm and low pressure. After cleaning, the flow rate through the brine
channels was observed to increase. Before the modules were cleaned,
Unit II (secondary effluent) was producing only 0.75 gpm of product water.
After cleaning, the product flow had increased to 3.3 gpm. After 100 hr
of operation, the product water flow rate had dropped to 2.3 gpm. One
of the modules, the first one in tube 1, Unit II (secondary effluent),
Phase III, was removed from Unit II at about 100 hr after the individual
module cleaning and was postmortemed to find out why a flux level approaching
the original value had not been restored and why the pressure drop was
higher than that for Unit I. Solid inorganic deposits were present in
substantial amounts. Samples were collected and submitted for analysis
with the results shown below. The composition of the solid material
was based on X-ray diffraction data.
% Ash (600°C) 80.9
% Calcium in ash 29.8
% Magnesium in ash 0.026
% Sulfate in ash 48.4
% Phosphate in ash 15.4
Crystalline material Ca(SO 4 ) (HP0 4 )y2H 2 0
where y/x = 4.
61
-------
SYSTEM PRESSURE
PRODUCT WATER
PVC
VESSEL
Fig. 28. Schematic diagram of portable cleaning loop
16-LITER
SUMP
62
-------
Deposition of this material could have been caused in several ways
basically associated with poor flow conditions within the brine channels,
such as can be caused by seal failure that permits water to bypass the
modules or loss of pH control.
The modules from Unit I which treated carbon—column effluent had ex-
perienced much less fouling and had an average initial pressure drop of
1.26 psi before cleaning, which decreased to 0.9 psi per module after
cleaning. The product water flow from Unit I increased from 4.4 gpm to
4.8 gpm after the modules were cleaned. The module cleaning procedure
had not been entirely effective, inasmuch as the initial performance,
on January 16, 1969, had been 8 gpm. This decline could not be attributed
solely to compaction but obviously included fouling.
Several modules from Phase III of the field study which had operated
on carbon—column effluent were returned to the laboratory for cleaning.
Individual modules were placed in the portable cleaning loop noted above,
and a number of different solutions were tried at room temperature. The
following results are based on visual observations of the disassembled
modules.
(a) A solution containing 10,000 mg/i of BIZ was recirculated for 30
mm at 22°C; the pH of the solution was 10.2. The flushing solution turned
dark. Much of the membrane remained covered with a loose coating of slime.
(b) A commercially available heavy duty detergent (tt409 ) solution*
prepared by diluting 200 ml in 12 liters of water was circulated for 30 mm
at 22°C. The water remained clear. Postmortem showed loosening of the
slime, but no removal.
(c) A solution of hydrogen peroxide was prepared by diluting 500 ml
of 30% H 2 0 2 in 12 liters of water and flushing for 30 mm at 22°C. The
*Mention of commercial products does not imply endorsement by the
Federal Water Pollution Control Administration.
63
-------
water became dark in color, and effervescence due to oxygen appeared in
the sump. There did not appear to be any slime removal from the membrane.
(d) An attempt was made to use osmosis to clean the membranes by
filling the brine channels with a 37 sodium chloride solution. The
product water side of the modules was filled with distilled water by
connecting the product water tube to a carboy of distilled water. In
effect then, a flow of clean water was created back through the membrane
since there was an osmotic pressure difference of about 300 psi. The
brine remained clear, and 10,000 mg/i of BIZ was added after 15 mm.
The test ran for 30 mm at 22°C. On examination, compared with previous
subjective examination, the cleaning was considered to be fair to good
except near the product water tube, where limited recirculation occurs.
(e) A 10,000 mg/i solution of Rohm and Haas Bate li3OD (pancreatic
enzyme) was tried for 30 mm at 22°C. The slime was loosened but not
removed.
Two modules from Phase III that had operated on an activated—carbon- -
processed feed were cleaned with a BIZ solution in the laboratory to
determine the effects of elevated temperatures and BIZ concentrations.
Both modules had shown a significant reduction in water flux over the
initial quality control performance. The modules were cleaned with a
20,000 mg/i BIZ solution at 45°C for 60 mm. The solution was adjusted
to pH 5 with HC1 to minimize any possible change in membrane characteristics.
There was little change in the module performance after cleaning. The
modules were opened, and the membrane examined. The membrane appeared
very clean; in fact, it was characterized visually as the best that had
been observed. However, after samples of the membrane were cleaned ultra-
sonically, there was a 50% increase in water flux and the A values in-
creased from 1.24 x l0 to 1.74 x l0 .
A module made with a thinner brine—side spacer (Chicopee woven
fabric) from Phase IV of the field study was tested and examined after
64
-------
650 hr of operation on carbon—column effluent. The module had been
subjected to a regular cleaning program while in operation and visually
was in very good condition and clean. Ultrasonic cleaning of the
membrane samples gave about a 10% increase in the membrane wat3r flux.
The ultrasonic cleaning procedure for cleaning membrane samples restores
the surface to a “new” condition and is valuable for preparing controls.
There is no damage to the rejection surface by ultrasonic cleaning. It
has also been demonstrated that the cleaning procedures are not entirely
adequate. It appears that the slimes or membrane coatings can be loosened
by treatment with various cleaning solutions, but that some sort of turbu-
lent action within the brine channels is necessary for wholesale removal
of fouling. Increasing the flow of the cleaning solution may be of
significant help in attaining clean membrane.
A series of preliminary tests were run with enzyme samples obtained
from several manufacturers. Slime—coated membrane from used modules were
exposed to enzyme solutions with and without a non—ionic detergent,
Triton X_100.* The results were poor. None of these materials produced
results as good as the best compound found thus far, BIZ. Data sheets
obtained from the manufacturer, Proctor and Gamble, indicate that BIZ
contains, in addition to the enzyme, a detergent, a builder (probably a
polyphosphate), and sodium perborate. Evidently, this is a synergistic
combination as far as membrane cleaning is concerned.
Only a relatively small amount of effort has been expended as yet,
however, in developing satisfactory cleaning techniques, compounds, and
systems for reverse osmosis loops, particularly in waste water treatment
and renovation. The work performed in this contract has been useful in
demonstrating that a certain group of materials may have value. Additional
screening of enzymes and combinations of enzymes and other compounds is
necessary to develop a “cleaning” compound that will produce the maximum
cleaning in the minimum time and with minimum damage to the modules.
*Mention of commercial products does not imply endorsement by the
Federal Water Pollution Control Administration.
65
-------
Additional work is required on the “direct” osmosis approach, including
combining this approach with cleaning compounds. More data are required
on the effect of increasing brine—channel flow and turbulence as an aid in
cleaning. Pressure differential measurements across the modules or
pressure tubes have had only limited usefulness as an indicator of
fouling since a module with coated membrane may have suffered a major
flux decline with virtually no change in pressure drop.
66
-------
CHLORINATION
Disinfection of reverse osmosis systems and of the feed and product
water appears to be desirable for a number of reasons. The presence of
biological forms such as bacteria, algae, and molds in the raw water feed
may cause fouling of the membrane and brine channels. Some evidence has
been published (Ref. 3) that demonstrates membrane degradation by bacteria
or other microbiological organisms on membranes that were contaminated
before or during service.
The disinfectant, if possible, should be low in cost, generally
available, effective in low concentrations, easily handled and fed, and
the solution concentration readily determinable at low concentrations.
Chlorine in its various active forms fits this picture of a suitable
disinfectant for reverse osmosis systems. It is the most commonly and
widely used chemical for biological control of potable, waste, and
industrial waters, so much so that chlorine may be almost impossible
to avoid in feed waters and from potable or waste water sources.
An important consideration is that the disinfectant must not damage
or degrade either the cellulose acetate membrane currently in use for
reverse osmosis or other module or system components. Periodic references
to free chlorine in feed waters have been ambivalent about the effect of
chlorine on cellulose acetate membranes. Previous experiments (Ref. 4)
have shown that continuous exposure of the membrane to relatively high
concentrations of free chlorine (10 to 50 mg/i) in San Diego tap water
had measurable effects on cellulose acetate membranes in comparatively
short times of 12 to 15 days. Definite deterioration had been initiated
in the membranes with 50 mg/l chlorine exposure. Salt rejection had de-
creased and the water flux had increased. At the 10 mg/l continuous
67
-------
exposure level in 15 days, the salt rejection measurements showed no change,
however, and the membrane coefficient had shown a slight increase which
did not appear significant.
Little or nothing has been known about the effect of prolonged exposure
to concentrations of about 2 mg/l or less of free chlorine. Prior experience
at the Pomona facility was obtained on the 5000 gpd spiral—wound system
operating on activated—carbon—treated secondary effluent at approximately
2 mg/i combined chlorine residual, and on secondary effluent at the same
residual processed through single modules at low pressures. At Pomona,
2 mg/i combined chlorine residual appeared to exercise satisfactory biolog-
ical control and to have little effect on the membrane in exposures exceeding
six months.
The laboratory experiment on the effects of chlorination on the
membrane was an effort to obtain quantitative long—term exposure data
since the field experiments were limited to 30 days.
The laboratory experiment was conducted using City of San Diego tap
water as a basic supply. A set of controls was established by using
membrane samples that were immersed in flowing tap water that was de-
chlorinated in an activated—carbon column. Two complete sets of membrane
samples, including controls, were run. One set consisted of high—selectivity,
standard—rejection membrane, and the other set consisted of high—flux
membrane. The selectivity and flux of the semipermeable cellulose acetate
membrane are controlled by the heat treatment to which the membrane is
subjected; the higher the temperature, the greater the rejection and the
lower the water flux. Enough membrane was used in the experiment to
permit at least four samples each of the chlorinated and control membranes
to be taken for testing every week. The membrane samples of each type
were taken from the same roll or production lot. Since the pH of San
Diego tap water may be as high as 8.0 or higher, which could easily affect
the membrane by hydrolysis of the cellulose acetate, the pH of the water
was adjusted to below 7 using sulfuric acid, and was between pH 5.5 and 7
68
-------
during the test. An Advance Chlorinator operating on gaseous chlorine was
used to control the concentration at 2 mg/i in the test stream to which
the chlorinated samples were exposed. Chlorine concentrations were
tested frequently and at least daily to ensure that the concentration
was maintained. The ortho—tolidine test was used for control and was
checked against the starch—iodide titration at the start. The laboratory
tests were based on free chlorine exposures. Samples of the chlorinated
membrane and the controls were removed weekly and tested at 800 psi with
a 10,000 mg/i sodium chloride solution.
Figures 29 and 30 show the solute (NaC1) rejection versus time for
high—flux and high—selectivity membranes, respectively, exposed to approxi-
mately 2 mg/i chlorine over a 29—week period. Figures 31 and 32 are plots
of B, the salt permeability coefficient, for the same membranes during the
same period. It is apparent that there have been some detrimental changes
in the chlorinated membranes. There has been a small but significant
change in the water permeation coefficient A and a large change in the
salt permeation coefficient B of the exposed membranes. This is reflected
in a decrease in salt rejection for the chlorine-exposed samples. A
substantial decline in salt rejection occurred after the eighth week of
exposure. The laboratory facility suffered a power and water failure
during the eighth week, and a remote possibility exists that the samples
could have been exposed to elevated chlorine levels for many hours.
However, the Advance Chlorinator is a vacuum—feed unit, and in the event
of a water failure should shut off completely. The high—selectivity membrane
did not change between the eighth and twenty—ninth week following the
initial loss in performance. The high—flux membrane did show some
continuing change, but at a much slower rate than the change that occurred
between the eighth and tenth weeks. This tends to lend credence to the
possibility that excessive exposure occurred. Nothing that happened affected
the control samples. The total chlorine exposure at 2 to 2.5 mg/i for
the eight—week period was of the same order as the earlier 10 mg/l exposure
for 15 days. During the 10 and 50 mg/i chlorination tests reported earlier
(Ref. 4) the membrane changes were gradual, as expected, and not abrupt.
Addtional chlorine exposures will have to be made under controlled laboratory
69
-------
100
___ .-. S — S S—•-- _ •.--•-’-• - -*-__.____.
CHLORINE-FREE FEED -
80
CHLORINMED FEED
(2 MG/L FREE RESIDUAL)
6O -
L)
w
-)
4o
Q I
I .-
HIGH-FLUX MEMBRANE
20
0 I - ___________________________ I
0 10 20 30
TIME (WK)
Fig. 29. Percent TDS rejection versus time; high—flux membrane operating on chlorinated feed (2 mg/i free
chlorine residual) and on chlorine—free feed
-------
100 _____
________________ s—v--.—- - CHLORINE-FREE FEED
80 CHLORINATED FEED
(2 MG/L FREE RESIDUAL)
60
w
HIGH-SELECTIVITY MEMBRANE
20
0 - ____________________ _j _________
0 10 20 30 40
TIME (WK)
Fig. 30. Percent TDS rejection versus time; high—selectivity membrane operating on chlorinated feed (2
mg/i free chlorine residual) and on chlorine—free feed
-------
100 I I I F I I 11 1 I I I I I I I I I
,
• •_/.
/
/
0 /
• •/ S
/
/
0
w
U,
S.- / S •
z
/
/
LA /
0
— / CHLORINATED FEED
/ (2 MG/L FREE CHLORINE
/ RESIDUAL)
/
/
w /
9/
L .
L i / a
0 - ,/ S
0 / 0 S.
• CHLORINE-FREE FEED
7’ - ————a— -
Li S S
Z • •
Li.I / •
0
I— 0.
-J
HIGH-FLUX MEMBRANE
III 1111111111111111111111111
10 20 30
TIME (WK)
Fig. 31. Salt permeation coefficient versus time; high—flux membrane
operating on chlorinated feed (2 mg/i free chlorine residual) and
on. chlorine—free feed
72
-------
100
I I I I I I I I I I I I 1 I I I I 1 I I I I I I I I I
C -,
‘7
0 •i
C-)
0
ir • —,
0
. ,-
><
I ,’
/
I— -, I
‘ CHLORiNATED FEED
— ‘ (2 MG/L FREE CHLORINE
10—
RESIDUAL)
Li
w
0
0”
— 1
I— “0 I
w
x -,
c 4
_ _ _ • CH LO R I NE - FREE FEE D •
—
I— II
S I
HIGH-SELECTIVITY MEMBRANE
I I I I I I I I I I I t I I I I I I I I I J_ I
0 10 20 30
TIME (wK)
Fig. 32. Salt permeation coefficient versus time; high—selectivity n einbrane
operating on chlorinated feed (2 mg/i free chlorine residual) and
on chlorine—free feed
73
-------
conditions, and tests using chioramines as well as free chlorine should
be included since the ammonia concentration in secondary effluents such
as at Pomona are high enough to ensure that all available chlorine will
be present as chioramines. Weekly tests of intermittent chlorination
at the 10 mg/i level were not made.
74
-------
SUMMARY AND CONCLUSIONS
A principal objective of the program was to find and develop the
parameters and conditions that would best permit operation of reverse
osmosis systems on waste waters receiving secondary treatment. To a con-
siderable extent, the relatively short-duration tests made under this
contract have accomplished the goals outlined in the proposal. The short-
term tests that were conducted cannot substitute for prolonged runs of
one to three years, which are necessary to show that reverse osmosis systems
can produce a high—quality product water economically. The techniques
that have been developed within the scope of this contract for maintaining
membrane water flux have shown great promise but certainly are at an early
stage of development. For example, the enzyme—based cleaning compound
that made it possible to get sustained runs with feedwaters which can
cause serious module fouling should be improved, if possible, to produce
more complete and rapid cleaning with little or no effect on the membrane
properties.
The first phase of the project was designed to compare and to attempt
to achieve continuous operation of the two reverse osmosis pilot plants
utilizing 50 sq ft modules made with high—selectivity membrane. One loop
was fed with sand—filtered secondary effluent and the other with secondary
effluent treated with granular activated carbon. (Prior to this study,
sustained operation at 600 psi had been attained only with the carbon—
column effluent. The air—water flushing method had been moderately successful
in keeping the unit running.) Phase I demonstrated that the plants could
not be operated continuously either on secondary effluent or carbon-column
effluent without a great improvement in pretreatment or cleaning techniques
to maintain water flux.
75
-------
Phase II of the project was originally established to compare the
performance of high— and standard—water—flux membranes in the 50 sq ft
modules operating under the best possible feedwater conditions. Operational
data indicated this was the effluent from the carbon columns. At this point,
the enzyme-based cleaning compound that had demonstrated definite cleaning
potential in the laboratory was introduced into the field study. The
flux decline slope of both the high— and standard—flux modules was
maintained close to that which might be expected from compaction without
fouling by use of the enzyme compound. Phase II demonstrated that it
would be economical to operate with high—flux membranes in the modules
since it was possible to maintain a water flux approaching reference
values and that the product water from the high—flux membranes was of
high quality.
The standard—flux system operating on carbon—column effluent reduced
the total chemical oxygen demand (COD) from 6.3 to 0.25 mg/l (average
values), the phosphate as P0 4 from 25.3 to <0.1 mg/l, and the total
dissolved solids (TDS) from 651 to 40 mg/l (93.8% removal). The high—
flux system operating on carbon—column effluent reduced the total COD
of the feed from 7 to 0.8 mg/i (average values), the phosphate as P0 4
from 24.8 to <0.1 mg/i (>99% removal), and the TDS from 555 to 35 mg/l
(93.7% rejection). The rejections noted are not based on a feed—brine
average concentration but are based on the ratio of feedwater concentra-
tion to product—water concentration. Normally, the feed—brine concentra-
tion or a weighted mean is used in calculating membrane parameters and
system operational effectiveness. To illustrate actual results obtained
or desired in water renovation, the product—water values are compared
directly with the feed concentrations. Of particular interest in organic
carbon rejection is that the level of organic carbon (as shown by the
COD) is very much lower in the product water from a reverse osmosis loop
(<2 mg/i COD) than the effluent from activated—carbon treatment alone
(6 to 12 mg/i COD). The constituents of the COD which are not removed
by activated carbon are not necessarily the same as those that are not
76
-------
rejected by reverse osmosis. For example, phenols, in general, will be
transported through a reverse osmosis membrane, whereas they may be
readily adsorbed by carbon.
In Phase III, comparisons were made of the results of operating
modules containing high—flux membranes on unfiltered secondary effluent
and on activated—carbon—processed secondary effluent using enzyme—based
cleaning compounds to keep the systems functioning. The average total
COD in the carbon—column feed (12.1 mg/i) was reduced by 96%, to 0.5
mg/i; the reduction in COD using secondary effluent (41.6 mg/i) was 95.7%,
to 1.8 mg/i. Phosphate reduction was greater than 99% and TDS reduction
was greater than 90%. Ammonia rejection on carbon—column feed was 85%
and on secondary effluent, 70%. Nitrate rejection on carbon—column feed
was 66% and on secondary effluent, 58%. Product—water COD values are
so low that their accuracy may be subject to some suspicion. The flux
decline slopes for Phase III were -0.24 for secondary—effluent feed and
-0.14 for the carbon—column—treated feed. These were higher than expected.
It was found that the COD of the feed from the carbon columns in Phase
II was between 3 and 9 mg/i since the carbon had just been regenerated,
whereas during Phase III, the feed had a COD of between 9 and 15 mg/i.
A review of the operating data indicates that there is a direct relationship
between the rate of membrane fouling and the increases in COD of the feed.
The brine seals, which were made of Buna—N rubber, failed in the
loop operating on secondary effluent, which created serious boundary—layer
problems since much of the brine bypassed the brine channels within the
modules. This also caused excessive fouling to occur. New brine seals
made from ethylene—propylene copolymer, which has superior resistance to
oxidation and solvents, were installed, and the module diameter tolerances
were adjusted to make the seals operate more effectively. Brine recircula—
tion was initiated, which permitted higher product water recoveries.
77
-------
Phase IV was concerned with the use of brine spacers thinner in
section than the standard polypropylene Vexar, which is 0.045 in. thick.
The advantages to a thinner brine spacer are that it permits higher membrane
packing densities per unit pressure—vessel volume and offers the possibility
of greater turbulent flow conditions in the channel. The disadvantages
are increased pressure losses and possibly greater module fouling. The
study showed no marked differences in performance between modules made
with large and with small brine—side spacers, except for higher pressure
differentials. The pressure drop across the loop with the standard brine
spacer was 16 to 20 psi; the pressure drop across the thin spacer was 70
to 75 psi. The 3.75—in.—diameter module made with the thin spacer would
contain 25% more membrane area.
Phase V of the field operations was modified to permit the modules
in one loop to be reoriented vertically. In single—module studies, it
had appeared that cleaning was improved and fouling more readily controlled
In a vertical position versus horizontal; however, the short—duration test
showed no major differences in the operation of vertical versus horizontal
pressure vessels. The pressure loss for the vertical system was 90 psi,
due principally to the piping arrangement, almost 5 times that of the
horizontal unit. At the start of Phase V, pH control of the feed was
temporarily lost. Water production dropped very rapidly but was restored
when operation of the acid control was resumed. The flux decline was due
to the precipitation of calcium carbonate and calcium phosphate within
the modules.
The use of the enzyme—based cleaning compound (BIZ) on a regular
cleaning schedule has made it possible for the first time to get extended
operation of reverse osmosis systems running on secondary effluents and on
activated—carbon—treated secondary effluents. The most successful cleaning
compound has an optimum pH of approximately 10; at this pH, membrane
degradation can occur rapidly. Additional work is required on formulating
and testing a compound that will work at a less alkaline pH. However,
78
-------
the limited field experience with the alkaline enzyme detergent using
short contact periods of 30 mm has not shown the severe membrane
degradation found in the laboratory soak test. This may be due partly
to the less severe exposure in use under dynamic conditions and to a
protective effect of membrane coatings and slimes. More work is also
required in optimizing methods of flushing to remove the fouling material
from the membranes. The cleaning techniques developed on spiral—wound
modules will be useful for all kinds of reverse osmosis configurations.
One specific advantage to the spiral—wound module for this purpose is the
ability to examine individual modules easily and to test membranes during
extended runs with a minimum amount of disturbance to the test as a whole.
Some of the variations that were introduced require long—term tests
to determine whether any advantages are to be gained or improvements made,
as in vertical versus horizontal module orientation and large versus
small brine spacers. Additional work is required in optimizing the
frequency of cleaning during long-term operation.
The chlorination studies did not provide, at this time, a firm
answer to the effect of low—level chlorination on membrane life and
performance. Additional investigations are to be undertaken on specifically
evaluating the effect on cellulose acetate membranes of low—level continuous
chlorination versus intermittent application of chlorine. A comparison
should also be made of the effect of chloramines as compared with free
chlorine since the prevalent forms of active chlorine in waste-water
chlorination are the chloramines.
79
-------
APPENDIX
NOMENCLATURE
There are several relationships and expressions that have been used
in this report which may require some amplification or clarification.
Since no standard terminology is yet in use for reverse osmosis technology,
the expressions shown are those used at Gulf General Atomic. A comprehensive
review of reverse osmosis is available in Ref. 5, and there is a great
deal of information in the current literature.
The basic behavior of semipermeable cellulose acetate reverse osmosis
membranes can be described by two basic equations.
The product water flow through a semipermeable membrane may be
expressed as:
F = A( P — , (1)
where
F = water flux (g/sq cm—sec),
A = water permeability coefficient (g/sq cm—sec—atm),
= pressure differential applied across the membrane (atm), and
7T = osmotic pressure differential across the membrane (atm).
The water permeability coefficient A has the following components:
DCV
A= , (2)
RT Ax
81
-------
where
D 1 = diffusion coefficient for water in the membrane (sq cm/sec),
C 1 dissolved water concentration in the membrane (glcc),
= partial molar volume of water in the external phase,
R = gas constant,
T = absolute temperature, and
Ax = membrane thickness (cm).
The salt flux through the membrane may be expressed as
F=BAC,
where
F salt flux (g/sq cm—sec)
B salt permeability coefficient (cmlsec), and
= C 1 — C 2 = concentration gradient across the membrane (g/cc).
The salt permeability B has the following components:
D 2 K
(4)
where
= diffusion coefficient for salt in the membrane (sq cm/sec),
K = distribution coefficient ratio of the salt concentration
in the membrane (g/cc) to the salt concentration in the
solution (g/cc), and
= membrane thickness (cm).
The water permeability and salt permeability coefficients are characteristic
of the particular membrane which is used and the processing that it has
received.
82
-------
An examination of Eqs. (1) and (3) shows that the water flux is
dependent upon the applied pressure, whereas the salt flux is not. As
the pressure of the feedwater is increased, the flow of water through the
membrane should increase while the flow of salt remains essentially
constant. It follows that both the quantity and the quality of the purified
product should increase with increased driving pressure.
In Eq. (1), the water flux through the membrane, F , is directly
proportional to and increases with the net driving pressure ( P —
The osmotic pressure of a water solution is proportional to the concentra-
tion of the solute. Since the water flux F is proportional to the net
driving pressure, the amount of water from a given system operating at
a specific applied pressure will depend on the osmotic pressure of the
feedwater. As the feed passes through the reverse osmosis systems, and
water (and, to a very small extent, salt) passes through the membrane,
the concentration of dissolved substances in the feedwater (or brine)
increases. This causes a marked increase in the osmotic pressure differ-
ence t 1T and reduces the available driving pressure, which in turn results
in a lower water flux. The average osmotic pressure to which the membrane
is exposed is a function of the amount or percentage of the water recovered
as product. In a system operating at 50% water recovery, the osmotic
pressure of the brine will reach a value approximately twice that of the
feed. If a system is operated at 75% recovery, the osmotic pressure in
the brine will approach four times that of the feed. From this considera-
tion and from Eq. (3), it can be seen that the product quality will also
be affected by the amount of product water recovered.
These points are well illustrated by Figs. A—i and A—2. Figure A—i shows
the variation in water flux for a reverse osmosis system operating under
fixed pressure conditions as a function of water recovery, with feed
salinity as a parameter. It is seen that the water flux per unit of area
83
-------
1.2
I I I
30
I I I I I
40 50 60
70
80 90
WATER RECOVERY (%)
LU
>-
-J
><
-J
U-
LU
0.9
0.6
0.3
1000 mg/i
3000 mg/i
5000 mg/i
Fig. A—i. Variation in performance with feedwater concentration
-------
1000
800 - &00 PSI
0 )
E
>-
I-
-I
1
I-
U I
L)
0
0
200 -
0 AvL
0 30 40 50 60 70 80
WATER RECOVERY (%)
I I I I I I
TYPE A MEMBRANE
5000 mg/i
3000 mg/i
1000 mg/i
I I
90
Fig. A—2. Effect of feedwater concentration on product—water quality
-------
decreases with increasing feed salinity and with increasing recovery
percentages. Figure A—2 is a similar plot, except that product—water
quality (salinity) is plotted versus recovery. Again, we see that the
quality of the product water decreases with increasing feed salinity
and increasing recovery percentages. It is noteworthy that for recoveries
below about 75%, the reduction in water flux and water quality is fairly
low, particularly on feed of low salinity.
As noted above, the specific water and salt flows depend on the
water and salt permeation coefficients (A and B) of the membrane in question.
It is possible, as will be discussed later, to prepare membranes over a
range of specific water and salt permeation constants. When A = 1.5 x lO
g/sq cm—sec—atm, the water flux rate corresponds to 13 gal./sq ft—day
at 600 psi.
As noted earlier, B is given in units of cm/sec. An alternative method
of expressing B, in more readily measurable terms, is
B = CPa FP
where
C , = salt concentration of product (g/cc),
F = flow of product (cc/sec),
average concentration of feed and concentrate (g/cc), and
a = area (sq cm).
The percent salt rejection (S.R.) is defined by the equation
/ C \
% S.R. = fi — —Jx 100
C/
where
C = concentration (or conductivity) of the product, and
= average concentration (or conductivity) of the feed and brine.
86
-------
It then follows that a membrane exhibiting a water flux ratio of 13
gal.Isq ft—day at 600 psi and 95% salt rejection will have a B value
of 3 x 10 cm/sec.
RELATION BETWEEN MEMBRANE PROPERTIES AND ACETYL CONTENT OF CELLULOSE ACETATE
General—purpose reverse osmosis membrane is cast from solutions
prepared with commercial grades of cellulose acetate, usually described
as the 2.5 acetate, whose corresponding acetyl content is approximately
40%. Since each glucose unit in the original cellulose backbone contains
two secondary and one primary hydroxyl group, the terminology, T? 2 5 acetate”
means that an average of five out of every six hydroxyl groups are esterified
by an acetyl group (see Fig. A—3).
The permselectivity and transport properties of a particular membrane
result in part from what may be simply envisioned as a porosity at the
molecular dimension level. This porosity is attained and controlled by
several factors in combination:
1. The acetyl content of the cellulose acetate used to cast the
film.
2. The additives to the casting solution, such as magnesium perchiorate,
zinc chloride, etc.
3. The choice, and proportions, of casting solution solvents,
generally among acetone, water, and formamide.
4. The casting conditions, including air—drying rate, temperature, etc.
5. The heat—treatment (annealing) conditions.
The most fundamental consideration is the acetyl content of the
polymer. Once it is established, the other factors are employed to
achieve the desired range of properties under reproducible conditions.
Figure A—4 illustrates the changes in water and salt permeability values
over a range of acetyl contents. The semilog plot shows a much greater
slope for salt permeability than for water permeability. The values range
from insignificant salt rejection at 33% acetyl content to very low water
87
-------
I
ci i i
4 cu$
,c=.
I
S
U,
I I
S I C=S
I I
!
c C1 2 0
_ _ y
•
I’
c= s S c=s I
I I
âi , c=S ci , 0=S
,‘ CU, I
CU,
CU,
0=I
I C is
S
, Ii I C=S
CU,
C I ’
‘I
“1
‘c=S S C=S I
I ci, =I CU, C=l
I I
c i , CI,
Fig. A—3. Molecular representations of cellulose, cellulose acetate, and
cellulose triacetate; for clarity, the hydrogen atoms have been
omitted in the ring structure of the anhydroglucose units (from
Eastman Kodak Bulletin on Eastman Cellulose Acetate, 1959)
U
Si
Si
‘I
CUl l
I
88
-------
(-)
LU
U,
C-)
I-
( -)
I-
I .
>-
F-
-J
LU
LU
a-
LU
I —
— -9
10
LU
LI)
C-)
LI,
C 4
a
>-
F-
‘-4
-J
‘-4
LU
LU
I-
io 2
32 34 36 38 40 42 44
ACETYL CONTENT (X)
Fig. A—4. Permeability to water (DiC 1 ) and ,NaCL (D 2 K)• of cellulose acetate
membranes as functions of acetyl content
-8
10
ARROWS DENOTE VALUES FOR
GENERAL-PURPOSE SPIRAL-
WOUND-MEMBRANE MODULES
FOR USE IN INDUSTRIAL
WATER PRODUCTION
89
-------
flux, with very high salt rejection, at 43.2% acetyl content. This
relationship is illustrated in more practical and common terminology and
units in Table A—l. A hypothetical system operating at 75% recovery on
a 700 mg/l feed at 600 psi will produce purified water over the listed
range of water flux and salt rejection values (and corresponding product
salinity).
As a comparison of the two examples for the 39.8% acetyl value with
the 37.6% acetyl value shows, increased flux at a modest rejection penalty
can be obtained more readily by heat treatment than by varying the acetyl
content. The data in Table A—i illustrate the many combinations available
for a very wide range of operating requirements. The 39.8% acetyl content
membrane is a good general—purpose membrane that has been extensively studied
and utilized. Continued intensive developmental work will result in an
ever—expanding availability and performance range of proved membrane.
EFFECT OF HEAT TREATMENT ON MEMBRANE PROPERTIES
For practically all applications involving demineralization, heat
annealing of the cellulose acetate membrane is required. Heat treatment
involves immersion of the membrane in water for periods of up to 30 mm
at closely controlled temperatures.
The heat—treatment temperature profoundly influences the water flux
capability, the salt rejection properties, and another property (compaction),
which is defined and discussed in more detail below. Simply stated, the
higher the heat—treatment temperature, the lower the water flux and the
better the salt rejection. This can be seen from the curves in Fig. A—5,
which show the influence of heat—treatment temperature on water flux rate
and salt rejection capability. Thus, modules containing general—purpose
membranes, heat—treated in the range of 85°C, exhibit (1) an initial water
flux capability of approximately 18 gal./sq ft—day at 800 psi net driving
pressure, and (2) a salt rejection capability (NaCl), at the same pressure,
in the range of 95%. These values are derived from standard—membrane
test cell conditions (1% NaC1 feed, 25°C, 800 psi, “zero” water recovery).
90
-------
TABLE A—i
SPIRAL MEMBRANE MODULE PERFORNANCE(a) VERSUS ACETYL CONTENT
Acetyl Content (%)
33
37.6
39.8
398 (b)
43.2
1120 flux, gal./sq ft—day
Product salinity, mg/i
Rejection (based on avg.
conc. 1120 mg/i), %
49
Comparable
to feed
Not significant
17
225
80
10
45
96
15
67
94
3.7
2
99.8
(a)Feed H 2 0 contains 700 mg/i TDS; brine contains 2500—2800 mg/i TDS;
pressure = 600 psi; recovery 757 . All membrane was heat—treated at 85°C
except as noted.
(b)Htttd at 81°C.
91
-------
Li.
op
‘0 ,
-J
Li.
L i i
I-
0
I-
Lu
I.,
Lu
I—
1
V)
40
30
20
10
0
100
90
80
75
t )
r600TEI
— AT 800 PSI
77
79
81
83
TEMPERATURE (°C)
85
87
89
91
Fig. A—5. Dependence of water flux and salt rejection on membrane heat—treatment temperature
-------
CHANGES IN MEMBRANE WATER FLUX PROPERTIES WITH TIME
As mentioned above, the third important property of cellulose acetate
membrane which is significantly determined by the heat—treatment temperature
is compaction. Even under idealized conditions, i.e., a pure water feed
and no fouling of the membrane surface due to other factors, there is a
decline in water flux with operational time. Flux decline behavior is
illustrated in Fig. A—6, which is a log—log plot of water flux versus
time. A plot of the membrane permeation constant A versus time gives
precisely the same curve. Membrane compaction is considered to be comparable
to creep observed in plastics, metals, etc., under compression—stress
conditions. The fact that compaction exhibits behavior that is amenable
to a straight—line log—log plot means that if the negative slope is of
an acceptable magnitude, a practical reverse osmosis system can be designed
to ensure rated water output performance over the estimated membrane
lifetime. As shown by the figure, the 1—hr water flux at 600 psi on a
typical feed is 13 gal./sq ft—day. The flux decline rate, which is the
slope of the straight line, is shown as —0.05. The term “flux decline”
is used deliberately in preference to “compaction” because, in real—world
systems, the observed flux decline is the sum of the fundamental compaction
effect already defined plus a tolerable amount of membrane surface fouling.
The flux decline slope can be obtained by plotting observed flux values
as a function of time on suitable log—log paper. The best straight line
is then drawn. By taking the values from the graph, the slope (m) may then
be calculated from the equation
log F 2 — log F 1
m = log T 2 — log T 1
If T 2 and T 1 are selected so that they differ by a factor of 10, then the
denominator is equal to one, and the above equation can be simplified to
m = log (F 2 /F 1 )
93
-------
101. i I
80
60
40 600 PSI NET PRESSURE
25°C TEMPERATURE
20
v)
-J
10
< 6
-J
La 4
w
I-
1 1 1
1 10 100 1000 10,000 25,000 100,000
TIME (HR)
Fig. A—6. Module water flux decline versus time
-------
Since F 2 will always be smaller than F 1 , m equals the log of a
fraction less than one; hence, m is a negative number, consistent with
the observed flux decline behavior. Experience has indicated that a
reasonable value of m for a properly controlled system is —0.05, as
illustrated. This means that the water flow rates a 3—yr period will
be as follows:
Time Flow Rate
( hr) ( gal./sq ft—day )
10 13
100 11.7
1,000 10.2
10,000 8.3
25,000 8.0
The design point used by Gulf General Atomic is currently the
10,000—hr value: 8.3 gal./sq ft-day at 600 psi. It decreases very
slightly over the next two years, and incremental increases of pressure
will provide the rated capacity during that period. Initial output will
be substantially in excess of rated capacity, and it is desirable at all
times to run the system at the lowest pressure in the 400 to 600 psi
range that will produce the rated output.
The inherent compaction properties of the membrane are initially
set, in large measure, by the heat—treatment temperature. The lower the
heat—treatment temperature, the less the compaction resistance of the
membrane. For operation in the range of 600 psi, the inherent compaction
properties of membrane do not vary widely over a heat—treatment range of
80° to 90°C, and it is in this heat—treatment range that maximum resistance
to compaction is attained. For higher operating pressures, however, the
differences in both initial flux value and compaction slope are substantial,
as illustrated in Fig. A—7.
95
-------
V)
C-,
0
C ’ ,
40 60 100
TIME (HR)
Fig. A—7. Change of membrane coefficient A with time at various pressures for modified cellulose acetate
membrane annealed at 80°C
1
0.1 0.2 0.4 0.6 1.0 2.0 4.0 6.0 10 20
-------
For any one combination of membrane heat-treatment temperature and
operating pressure, compaction rate is affected by feedwater temperatures.
At temperatures up to approximately 25°C, the effect is not very significant.
Beyond that, it accelerates, as illustrated in Fig. A—8.
EFFECTS OF FEEDWATER TEMPERATURE AND pH ON MEMBRANE PERFORMANCE
In reverse osmosis operation, feedwater temperature has a significant
effect on membrane performance and must therefore be taken into account
in system design. Essentially, the value of the water permeation constant
is only constant for a given temperature. The value of A will increase with
increased temperature and vice—versa. Figure A—9 illustrates the effect
of temperature on the water flux rate of the membrane. The flux rate at
25°C is taken as unity, and the appropriate multiplier shown on the ordinate
is used to calculate the added or lessened membrane area that is required
to produce the same quantity of water at any other temperature, assuming
equal feed salinity, pressure, and other conditions.
At any given time, the observed salt rejection does not vary substantially
with temperature, because the salt flux varies by approximately the same
magnitude as the water flux in the normal operating temperature range. It
must be recalled, however, that the membrane is an ester and therefore
subject to long—term hydrolysis. Hydrolysis results in a lessening of
salt rejection capability. The rate of hydrolysis is accelerated by
increased temperature and is also a function of feed pH. Slightly acidic
pH values (5 to 6) ensure the lowest hydrolysis rate, as do cooler temperatures.
This relationship is illustrated in Figs. A—lO and A—il.
It is noted that the hydrolysis rate increases sharply above 40°C, below
a pH of 3, and above a pH of 7. This phenomenon is taken into account in
the system design, particularly with reference to the choice of initial
salt rejection capability. The addition of mineral acids to the feed in
small amounts (equal to approximately 50% of alkalinity) performs a double
function by providing the optimum pH to minimize hydrolysis while controlling
and preventing CaCO 3 scale formation in the system.
97
-------
0.12
w 0.10
0 .
0
-J
I ,,
LU 0.08
-J
I - )
LU
o 0.06
-J
Li. 0.04
C D
•0
-I
C D
0.02
c0 10 20 30 40 50
FEEDWATER TEMPERATURE (°C)
Fig. A—8. Effect of feedwater temperature on log—log flux decline slope
I I I I I I I I
I I I I I I I
98
-------
L J
-J
I .-’
I —
—I
Fig. A—9. Dependence of water flux on temperature; for values other than 25°C, use indicated multiplier to
determine the membrane area required for equal water flux
FEEDWATER TEMPERATURE (°C).
-------
pH
Fig. A—b. Effect of pH on rate of hydrolysis
100
1
LU
U,
I
2 3 4 5 6 7 8 9 10
-------
6
I - I
I I I I I
>-5-
I —
-J
V)
0
LU
0 LU
I-I
-I
0
I—
LL1
0 I I I I I
0 1 2 3, 4 5 6 7 8 9 10
TIME (YR)
Fig. A—il. Performance basedon hydrolysis of membrane
pH = 7.0
6.0
pH = 5.0
-------
Based on this information and experience, design and predicted de—
mineralization performance over the membrane lifetime is summarized in
Table A—2. If other pH values prevail, salt concentration in the product
water will increase more rapidly. This is illustrated in Fig. A—il,
which is a plot of the time required for given factor increase in salt
concentration of the product water at a temperature of 23°C, with pH as
a parameter, based on the curves of Fig. A—b.
GENERAL PERFORMANCE
Figure A—12 shows the effect of pressure on the product water flow
for a standard—flux module having a membrane coefficient of 1.5 x 10
g/sq cm—sec—atm. The three curves represent the predicated performance
of a system operating on a solution containing 1000, 5000, and 10,000 mg/i
of sodium chloride. These concentrations are the average in the system
rather than the specific values in the feed or brine solution. The
deviation from linearity at low pressure is due to the fact that at
low pressures osmotic pressure is a significant fraction of the applied
pressure. The deviation from linearity at high pressure reflects the
compaction that can be expected after a period of operation.
Figure A—13 shows the effect of pressure on the sodium chloride
rejection for a standard—flux module. The three curves again represent
1000, 5000, and 10,000 mg/i of sodium chloride, the average concentration
in the module. The decrease in salt rejection with increasing salt
concentration results from the increase in the osmotic pressure of the
feed/brine. If the plot were made using net driving pressure, the
three curves would be superimposed. The improvement in rejection with
increasing pressure results from the fact that the salt flow is es-
sentially independent of pressure, whereas the water flow increases
with pressure.
Three grades of membrane are currently in use. These are usually
designated “high—rejection”, “standard”, and “high—flux”. The high—
102
-------
TABLE A—2
REVERSE OSMOSIS EQUIPMENT PERFORMANCE
Time After Startup (hr)
1
10,000(a)
25,000
Product Volume, gal./day
(at 600 psi and 25°C)
Product Quality, mg/i TDS
(assuming 500 mg/i TDS in feed
of typical ion species and
distribution at 75% recovery)
13
15—30
8.3
25—50
8.0
40—80
(a)Desjgfl basis (Gulf General Atomic Incorporated): standard
flux membrane.
103
-------
20
1000
PRESSURE (PSI)
Fig. A—12. Product flow versus applied pressure
18
>-
C
I - .
I i .
V)
-J
-J
LU
I-
16
14
12
10
8
6
4
2
0
0 200
400 600 800
1200
104
-------
I
I’,’, I I I I I I
98
94
9O—
88
86
82 I I I I
0 200 400 600 800 1000 1200 1400
PRESSURE (PSI)
Fig. A—13. Sodium chloride rejection versus applied pressure
SODIUM CHLORIDE
1,000 mg/i
5,000 mg/i
10,000 mg/i
STANDARD-FLUX MODULE
96 —
84 —
105
-------
rejection membrane is normally used in the home drinking water units.
The high-flux membrane is used in experimental applications. Typical
performance characteristics of modules made with these membranes are
given in Table A—3.
If the salinities indicated in Figs. A—12 and A—13 are taken as
feed salinities, the values represent the condition of zero recovery.
In a unit with a significant water recovery, the concentration of the
brine is proportionally higher than the feed concentration because of
the removal of purified product water. An accurate method for cal-
culating the average salt concentration in the brine channels is to
divide the average salt flow through the brine channel by the average
brine flow; i.e.,
— ( Cf) (Ff) + (Cb) (Fb )
C= Ff+Fb
where C = average brine concentration, g/cc or mg/i,
Cf = feed concentration, g/cc or mg/i,
Cb = brine concentration, g/cc or mg/i,
Ff = feed flow, cc/sec or gpm,
Fb = brine flow, cc/sec or gpm.
The effect of increasing concentration due to recovery on the
product water flow and quality Is shown in Fig. A—l4. This curve is
appropriate for a feed of 1000 mg/i with new standard—flux membrane
operated at 600 psi but does not reflect any boundary layer effect
(see the next section). The relatively small change in product water
flow is due to the fact that the osmotic pressure is relatively small
in all cases. It should be noted that with a membrane giving 95%
rejection, if the recovery is carried to 95%, the product from the final
section of membrane is approximately the same as the feed concentration.
106
-------
TABLE A—3
TYPICAL PERFORMANCE OF MODULES AT 600 psi
WITH A FEED OF 1000 mg/i OR LESS
High
Rejection
Standard
High
Flux
Initial membrane coefficient, A
(g/sq cm—sec—atm x iO )
Initial water flux (gal./sq ft—day)
Sodium chloride rejection (%)
Flux after 1 yr (gal./sq ft—day)
1.2
10.6
97
7.5
1.4
12.7
95
8.4
2.25
19
92
12 (a)
(a)E . d from short—term data.
107
-------
>-
I-
Li
V)
-J
C
-J
U-
U - I
I —
(. n —
0)
E
>-
-J
C
(-)
C
14
12
10
8
6
4
2
20
0
RECOVERY (%)
Fig. A—l4. Typical performance of a new standard—flux module as a function of recovery
-------
This section of membrane is thus not contributing useful product water,
and could be considered the limit of recovery. Normally, however,
system recovery is limited by the solubility of calcium sulfate or
other sparingly soluble salts and very high recoveries will not be
practical. Figures A—i and A—2 show the gross effects of increasing
salinity on the flux and product quality versus recovery.
BOUNDARY LAYER
The previous illustrations of typical performance are for systems
where the flow in the brine channel is sufficient to keep the concen-
tration polarization to a minimum. In many systems, this may not be
the case. When water permeates through the membrane, nearly all of
the salt is left behind in the brine channel. In any dynamic hydraulic
system, the fluid adjacent to the wall of the vessel is moving relatively
slowly. Even though the main body of the stream is turbulent, a thin
film adjacent to the wall (membrane) is laminar. This thin film is
called the boundary layer. The salt left behind by the product water
must diffuse through this boundary layer.
The concentration polarization is defined as the ratio of the salt
concentration at the membrane surface to the salt concentration in the
bulk stream. The concentration polarization has been found to fit the
equation
C
= exp [ K(F /Fb)]
C p
where Cm = concentration at the membrane surface, g/cc or mg/i,
C = average bulk concentration, g/cc or mg/i,
= concentration polarization,
K = a proportionality constant,
F = product flow, cc/sec or gpm,
Fb = average brine flow, cc/sec or gpm.
109
-------
The proportionality constant K is inversely proportional to the
diffusion coefficient of the salt and inversely proportional to the
length of the brine channel. Typical values of the concentration
polarization for a 3—ft module as a function of the product—to—brine
ratio for a sodium chloride solution and for a magnesium sulfate
solution are shown in Fig. A—15.
There have been numerous requests for a curve showing concentra-
tion polarization as a function of brine flow. Constructing such a
curve is somewhat impractical in that there is a separate line for each
pressure and each type of membrane. The ratio of the product flow to
the average brine flow is approximately the same as the recovery ratio.
Thus, a 10% recovery per module gives a concentration polarization
factor for sodium chloride of approximately 1.1. For a standard
50—sq—ft module operating at 600 psi, this corresponds to a brine
flow of about 4.5 gpm.
The effect of concentration polarization on the performance of a
unit is shown by a comparison of Fig. A—14 with Fig. A-l6. The plot
in Fig. A—16 shows the initial performance of a 10K unit operating
at 600 psi with a feed of 1000 ppm NaC1. The decrease in product flow
with increasing recovery is still very small. The decrease in product
quality, the result of an increase in salt concentration in the feed
with increased recovery, is considerably larger when the boundary layer
is considered.
At recoveries below 75%, the effect of concentration polarization
on the product quality and quantity where the feed is a naturally oc-
curring water with a total dissolved solids content of less than 5000 mg/i
is of little importance when the ratio of the brine flow to the product
flow ?8:l.
110
-------
S
C )
0
C - )
a
C
w
w
x
C )
0
C -)
0
I. - ’
‘ -2
C
N .J
‘-4
C
-J
0
0
0
‘-4
I —
I—
w
C .)
0
0
1
0 0 ,6
PRODUCT FLOW/AVERAGE BRINE FLOW
Fig. A—15. Concentration polarization versus product—flow—to—brine—flow
ratio for magnesium sulfate and sodium chloride in a 3—ft module
4
0 ,1 0.2 0 ,3 0.4 0.5
ill
-------
14
12
10
8
6
4
>-
U-
V)
-J
I- ><
I-
-J
LL
I—
r
E
>-
I-
-J
=
C-
I-
C -)
=
C
2
0 20 40 60 80
RECOVERY (%)
20
100
Fig. A—16. Typical performance of a new 10,000 gpd unit as a function of recovery
-------
The most common problem resulting from concentration polarization
is the increasing tendency for precipitation of sparingly soluble salts
and the deposition of particulate matter on the membrane surface. As
a general rule, the sparingly soluble salts have a smaller diffusion
coefficient than does sodium chloride. Although the diffusion coef-
ficient of calcium sulfate is not readily available, it can be expected
to be similar to that of magnesium sulfate. The concentration polari-
zation effect for magnesium sulfate is somewhat more than twice that
of sodium chloride at any given conditions of flow.
To illustrate the effect on precipitation, assume a system with
10% recovery in the final module. The concentration polarization for
magnesium sulfate is approximately 1.25. Assuming that calcium sulfate
behaves the same way, the brine at the membrane surface is approximately
25% more concentrated than the average brine in this module. This
local concentration effect markedly lowers the system recovery which
may be obtained without adding precipitation inhibitors.
113
-------
ACKNOWLEDGMENTS
Gulf General Atomic acknowledges the invaluable assistance that was
given by the County Sanitation Districts of Los Angeles County in providing
the study site and facilities, day—to—day operation and maintenance of
equipment, data acquisition, and waste water analyses. In particular, the
assistance and advice of Charles Carry, Robert Miele, and James Gratteau of
the County Sanitation Districts are gratefully acknowledged.
The authors wish to thank J. E. Beckman and C. E. Foreman of Gulf
General Atomic for their contributions in this work.
115
-------
REFERENCES
1. Dean, R. B., S. Claesson, N. Gellerstedt, and N. Boman, “An Electron
Microscope Study of Colloids in Waste Water,” Environ. Sd. Technol . 1,
147 (February 1967).
2. Busch, P. L., and W. Stumm, “Chemical Interactions in the Aggregation
of Bacteria — Bioflocculation in Waste Treatment,” Environ. Sd.
Technol . 2, 49 (January 1968).
3. Cantor, P. A., D. J. Mechalas, 0. S. Schaeffler, and P. H. Allen III,
“Biological Degradation of Cellulose Acetate Reverse Osmosis Membranes,”
Aerojet—General Rept. 3504, January 1968.
4. Vos, K. D., I. Nusbaum, A. P. Hatcher, and F. 0. Burns, Jr., “Storage,
Disinfection, and Life of Cellulose Acetate Reverse Osmosis Membranes,”
Desalination 5, 157 (1968).
5. Merten, U. (Ed.), Desalination by Reverse Osmosis , MIT Press, Cambridge,
1966.
116
* U. & GOVERNMKWr PRIN’IlNG OFFXCE: 1970 0 - 405-434
------- |