Environmental Protection Technology Series
Hydrodynamic Flux Control
for Waste Water Application
of Hyperfiltration Systems
.SB
\
Ul
o
Office of Research and Monitoring
U.S. Environmental Protection Agency
Washington, D.C. 20460
-------
-------
-------
EPA Review Notice
This report has been reviewed by the Environmental
Protection Agency and approved for publication.
Approval does not signify that the contents neces-
sarily reflect the views and policies of the
Environmental Protection Agency, nor does mention
of trade names or commercial products constitute
endorsement or recommendation for use.
11
-------
ABSTRACT
The effect of hydrodynamics on flux decline of cellulose acetate
hyperfiltration membranes was studied using primary effluent from the
Oak Ridge East Sewage Plant as feed. The system contained multiple,
annular-geometry housings which could be operated simultaneously at
different velocities with the same feed. Initially, each tube in an
annular housing was divided into two sections which could be sampled
individually. In the latter part of the program, 3-section tubes were
used so that membranes with three different flux capabilities could be
exposed simultaneously to three different velocities. The range of
fluxes studied was from 2 to 150 gal./fts«day and axial velocities
ranged from 3 to 30 ft/sec. System pressure was 600 psig.
The results demonstrate that there is a threshold velocity above which
flux decline is markedly smaller than at lower velocities. Visual
inspection of the membranes after a test indicates that operation above
the threshold velocity markedly reduces accumulation of solids during
the first 200-300 hours of a test. In scouting studies, addition of a
flocculant appeared to markedly reduce the threshold velocity.
During the initial 200-300 hours of those tests with primary sewage
effluent as feed and in which the axial velocity was above the threshold
velocity, the flux decline parameter, b, [b = — (A log flux)/(A log time)]
was directly proportional to the six-tenths power of the flux and
inversely proportional to the square root of the velocity.
This report was submitted in fulfillment of Project Number 17020 FEV,
Contract lU-12-896, Interagency Agreement AEC Uo-196-70 under sponsorship
of the Office of Research and Monitoring, Environmental Protection
Agency by the Oak Ridge National Laboratory, Oak Ridge, Tennessee 37830
ill
-------
CONTENTS
Section Page
I CONCLUSIONS 1
II RECOMMENDATIONS 3
IH INTRODUCTION 5
IV EQUIPMENT 13
V FEED COMPOSITION AND CHEMICAL ANALYSIS 15
VI RESULTS 17
Reproducibility of Results 25
Effect of Loop Depressurization on Flux 30
Intermittent Application of Chlorine 33
Modification of Membrane Characteristics by Additives .... 40
Additives for Sewage Floe-Size Control . . 44
Composite Results 48
Flux Decline 48
Observed Rejections 50
VII CORRELATION OF RESULTS 65
Threshold Velocity 65
Flux Decline 69
VIII ACKNOWLEDGMENT 77
IX REFERENCES 79
v
-------
FIGURES
?aee
1 Hyperfiltration Test Section with Transparent Pressure
Jacket ............................ 14
2 Variation of Constituents in Primary Sewage Effluent, Oak
Ridge East Sewage Plant ................... 16
3 Effect of Flux and Axial Velocity on Flux Decline Using
Primary Sewage as Feed (Cellulose Acetate Membrane, 600
psig, Run 26) ........................ 26
^ Effect of Flux and Axial Velocity on Flux Decline Using
Primary Sewage as Feed (Cellulose Acetate Membrane, 600
28)- • 29
5 Flux Recovery After Shutdowns as a Function of Axial Velocity
and Initial Membrane Flux (Primary Sewage Feed, 600 psig,
Cellulose Acetate Membrane, Run 33) ............. 32
6 Effect of Flux and Axial Velocity on Flux Decline Using
Primary Sewage as Feed (Cellulose Acetate Membrane, 600
psig, Run 26) ........................ 34
7 Membranes from Run 26 After 768 Hours Exposure to Primary
Sewage Feed . . ...... • ................. 36
8 Membrane with Initial Flux of 50 gfd Exposed to Primary
Sewage Effluent Feed for 768 Hours on Axial Velocity of
10 ft/sec. (Run 26, 600 psig, smallest scale division =
n. ) .......................... 37
9 Comparison of Effect of Adding CaCl(OCl) Before Severe
Fouling Occurs on All Membranes (Solid Points) with
Addition After Severe Fouling Occurs (Open Points).
(Primary Sewage Feed, 600 psig, Cellulose Acetate Membranes
Solid Points - Run 28, Open Points - Run 26) ......... 39
10 Effect of Axial Velocity and Initial Flux (jj) on Flux
Decline in the First 20 Hours After Exposure to Primary
Sewage Feed. (Cellulose Acetate Membrane, 600 psig
Pressure) .......................... 49
11 Effect of Axial Velocity and Time on Flux. (Cellulose
Acetate, Aerojet 6$, 600 psig, Primary Sewage Feed) ..... 51
12 Membranes After 750 Hours Exposure to Primary Sewage Feed
in Run 2h .......................... 52
13 Effect of Axial Velocity and Time on Flux. (Cellulose
Acetate, Eastman HT-00, 600 psig, Primary Sewage Feed) .... 53
vi
-------
FIGURES
?aee
1^ Observed Rejection with Aerojet 6% Cellulose Acetate
Membrane with Primary Sewage Feed .............. 54
15 Observed Rejection with Aerojet 6% Cellulose Acetate
Membrane with Primary Sewage Feed. (Note Scale Difference
on Phosphate Rejection) ................... 55
16 Observed Rejection with High Flux, Low-Salt Rejection
Eastman HT-00 Cellulose Acetate Membrane with Primary
Sewage Feed ......................... 56
17 Observed Rejection with High Flux, Low-Salt Rejection
Cellulose Acetate Membrane with Primary Sewage Feed ..... 57
18 Product Composition in Hyperfiltration of Primary Sewage
Effluent with a Nominal 9^ Rejection Aerojet 6% Cellulose
Acetate Membranes with Initial Flux of ko to 50
gal. /ft2 .day ......................... 60
19 Product Water Composition in Hyperfiltration of Primary
Sewage Effluent with a Nominal Zero-Percent Rejection
Eastman HT-00 Cellulose Acetate Membrane with Initial Flux
of 130 to 150 gal. /ft2 .day .................. 61
20 Product Turbidity from Nominal 9^ Rejection Aerojet 6%
Cellulose Acetate Membrane with Primary Sewage as Feed .... 62
21 Product Turbidity from High Flux, Low-Salt Rejection
Cellulose Acetate Membrane with Primary Sewage as Feed .... 63
22 Effect of Particle Diameter and Flux on Limiting Deposition
Velocity for Small Particles (v = 1 cm/min Corresponds to
35^ gal. /ft2 .day) ...................... 67
23 Threshold Velocity Above Which Catastrophic Flux Decline
is Prevented with Polluted Surface Water Feeds ........ 68
2U Comparison of Threshold Velocity from Results After 20 Hours
Circulation of Primary Sewage (Round Points) with Results
After 200 Hours ....................... 70
25 Effect of Adding ~50 mg/4 FeCl3 to Primary Sewage Feed
Immediately Before Introduction to Hyperfiltration Loop ... 71
26 Effect of Axial Velocity and Initial Flux on Flux Decline
Parameter, b, for Initial Portion of Runs with Primary
Sewage Feed. (600 psig, Cellulose Acetate Membranes) .... 72
vii
-------
FIGURES
27 Summary of Flux Decline Results for Velocities Above
Threshold with Primary Sewage as Feed
28 Effect of Initial Flux on Flux Decline Parameter, b, for
Runs with Primary Sewage Feed and Velocity Above Critical
(600 psig, Cellulose Acetate Membranes) 75
viii
-------
TABLES
Ho. Page
1 Rejection of Various Constituents from Municipal Sewage
Plant Effluents by Cellulose Acetate ............. 6
2 Test with Secondary Sewage Effluent in Plate and Frame
Module ............................ 7
3 Average Value of Flux for One Year of Operation Assuming
a Constant Value of Flux Decline Parameter Throughout
the Year ........................... 8
^ Flux Decline Slopes Obtained Under Identical Conditions
with Different Types of Municipal Wastewater [After
Feuerstein and Bursztynsky (ll)] .............. 9
5 Summary of Cellulose Acetate Runs with City Water and
Primary Sewage Feed ..................... 18
6 Summary of Results at ^KX) and 700 Hours for Long Term Runs
with Cellulose Acetate and Primary Sewage Feed (First 200
Hours of Runs Summarized in Previous Table) ......... 24
7 Reference Flux and Rejection (Run 2.6, System Pressure
600 psig) .......................... 27
8 Effect of Loop Shutdown on Flux (Run 33, 600 psig) ...... 31
9 Effect of Calcium Hypochlorite Wash on Flux (Run 26, 600
psig, CaCl(CCl) Added After 62k Hours Total Run Time) .... 35
10 Comparison of Flux and Rejection in Run 29 Using FeCla
with Flux and Rejection in Run 28 when no Additives
Were Used (Primary Effluent Feed, 600 psig Pressure) ..... 46
IX
-------
SECTION I
CONCLUSIONS
In studies of the effect of hydrodynamics on flux decline in hyperfil-
tration of primary sewage effluent with cellulose acetate membranes,
convincing evidence was obtained demonstrating that there is a threshold
velocity above which flux decline is markedly smaller than at lower
velocities. Based on inspection of membranes at the conclusion of tests,
it is apparent that operation at velocities above the threshold prevents
accumulation of visible amounts of solids on the membrane in the first
200-300 hours of a test. Thus the threshold velocity is that velocity
above which deposition of particulates on membrane surface is minimized.
The results of these studies indicate that for primary sewage with no
additives, the threshold velocity, uc, (in ft/sec) is related to the
initial flux, Jj, (in gal./ft2»day) by
u = (1.3 ± 0.6) J 1/a .
C. J_
Surprisingly, results obtained at ORNL using untreated Tennessee River water
feed and results obtained using Pomona, California secondary effluent as
feed are in remarkably good agreement with present threshold velocity
results using primary sewage effluent as feed. This suggests that floccu-
lation characteristics of polluted surface waters may be similar. If this
proves to be so, then the present results should be applicable to a wide
variety of feeds.
In scouting studies in which ~50 mg/4 FeCla was added to the primary
sewage, there was evidence that the value of the coefficient in the
threshold velocity equation was reduced from 1.3 to ~0.5. With an initial
flux of 50 gal./ft2"day the threshold velocity was reduced from ~9«5 to ~U
ft/sec. Additional studies are required to substantiate this. If proven
true it could have important consequences for the application of hyperfil-
tration to treatment of municipal sewage effluents.
During the initial 200-300 hours of those tests with primary effluent feed
and in which the axial velocity was above the threshold velocity for
particulate fouling, the effect of flux, v, and axial velocity, u, on the
flux decline parameter, b, (b = — A log flux/A log time) was found to be
0.6
b cc I .
u0-5
The value of b may also be a function of concentration of foulants although
additional tests are required for quantitative evaluation of the relation.
For times longer than 200-300 hours the situation becomes more complicated.
Apparently stabilized flux declines can suddenly become markedly worse,
even at velocities above the threshold. Based on visual inspection of the
membranes after long term tests and on the relative times' at which the
-------
increased rate of flux decline began with membranes having different
initial fluxes exposed to different axial velocities, it was suspected
that this increased rate of flux decline was associated with some sort
of nucleation phenomenon, In particular, that solids build-up occurred
preferentially at nucleation sites (of unknown origin) leading ultimately
to a covering of the membrane by a layer of solids and marked flux
decline.
Based on this premise, calcium hypochlorite (~100 mg/4) was added for ~1
hour directly to the primary sewage feed in one test after 620 hours
operation. In that test a markedly increased rate of flux decline had
begun after 300-^00 hours even with membranes operated above the threshold
velocity. Almost immediately there was a pronounced flux recovery with
those membranes operated above threshold; the recovery was to almost the
value based on extrapolation of the flux results for the first 200-300
hours of operation. In a subsequent test, the calcium hypochlorite
additions were made for 1 to 2 hours at approximately 250-hour intervals
in a test lasting 9^5 hours. Such treatment prevented the change in the
rate of flux decline which began after 300-^00 hours in the previous test.
Based on observed rejection of the membranes during this test, there was
no damage to the membrane during this 985-hour test. However, longer
testing times are required to determine whether intermittent treatment
with calcium hypochlorite coupled with operation above the threshold
velocity is a satisfactory procedure for preventing particulate fouling
of hyperfiltration membranes used in treatment of primary sewage effluent.
-------
SECTION II
RECOMMENDATIONS
The present studies clearly demonstrated that deposition of particulates
on membrane surface can be minimized by operation above the threshold
velocity. However, even when the axial velocity was above the threshold
velocity, there continued to be a flux decline. In all probability this
flux decline was due to dissolved materials which interacted with the
membrane surface in an unknown way.
The primary objective of future studies should be the determination of
the'nature of this interaction, evaluation of which feed constituents
cause the most severe problems and development of procedures for
ameliorating the problem.
Secondary objectives should be the development of economic procedures
for increasing the floe size, thereby reducing the threshold velocity
for particulate fouling, as well as further studies of the nucleation
phenomena which occurs after 200-300 hours of operation.
-------
SECTION III
INTRODUCTION
In early studies (l) with a plate and frame module exploring the possible
application of hyperfiltration (reverse osmosis) to the treatment of fil-
tered municipal secondary sewage effluent, it was demonstrated that besides
removing dissolved salts, the cellulose acetate membranes effectively
removed surface active materials, and organic solutes (Table l). These
exploratory studies also indicated that time-dependent flux decline was
a major problem when treating municipal sewage effluent by hyperfiltration.
Tests were usually of 5 to 6 hours duration. It was found that flux
decline was most severe with high initial flux (30 to 60 gal. /ft2»day)
membranes. After a few hours operation, flux from these membranes
approached values observed with much lower initial flux (10 to 20 gal./
ft2•day) membranes. Even with these lower flux membranes, decreases in
flux ranged from ^5 to 6Q% of the initial values in the 5- to 6-hour test
period.
In subsequent studies with a spiral wound and tubular modules (2,3) the
high organic and inorganic solute rejection obtained in the early tests
were again repeated (Table l). Flux decline was again a problem, partic-
ularly with high flux membranes. Among steps taken to ameliorate the
problem were carbon pretreatment, chlorination, and a daily air-water
cleaning. With a membrane having an initial flux of 5 gal./ft2 »day, a
flux decline of ~I5% occurred in the first 250 hours; little further
decline was noted until after 1100 hours of operation. However even with
this pretreatment and daily flushing, a membrane with an' initial flux of
17 gal./ft2«day showed severe flux decline; the flux from this membrane
fell below the flux from the 5 gal./ft2»day membrane after about 800 hours
of operation. The test was continued for a total run time of 5000 hours.
At that time the rejection of the low flux membrane had deteriorated only
by the amount expected due to hydrolysis of the cellulose acetate membrane
(3). Thus in this extended run the treated-secondary effluent appeared to
have no particular deleterious effect on the membrane.
In exploratory studies with membranes dynamically formed from poly-anionic
electrolytes and from sewage constituents, initial fluxes were frequently
greater than 100 gal./ft3»day at 1000 psig pressure; by using high circu-
lation velocities (>30 ft/sec) flux values were frequently over 50 gal./
ft2iday after UOO hours of operation (^,5).
Based on the promising results of these initial studies, several groups
made more comprehensive investigations, a substantial portion of which
were devoted to attempts to develop procedures for ameliorating the flux
decline problem.
Smith, Masse and Miele (7) made pilot plant studies to determine the effect
of feed water quality on membrane performance and to examine methods to
control membrane fouling. Three different configurations were used in
their studies: flat plate, spiral wound and tubular modules. Typical
rejections observed in these studies are given in Table 1.
-------
Table 1. Rejection of Various Constituents from Municipal Sewage Plant Effluents
by Cellulose Acetate
Source
Pretreatment
Flux, gfd
Pressure, psi
Rejections:
(a) BOD
(b) Organic Carbon
(c) Inorganic Carbon
(d) ABS
(e) TDS
(f) ci-
te) NOg
(h) P04 =
(i) Ca++ + Ug++
U) NH4 +
AWTR-lU
(1965)
Secondary
Filtered
-10
750
98
—
—
98
97
69
—
>96
98
86
Merten
and
Bray
(1966)
Secondary
Carbon
~10
^00
—
90
—
>97
96
95
~8o
>99
>99
>85
Hauk,
et al.
(1969)
Primary
None
33
1000
86
—
—
93
97
—
50
>99
—
—
Smith,
Masse,
Miele
(1970)
Secondary
Carbon
10-25
UOO-800
—
90-95
—
—
93-95
—
60-70
95-99
—
80-90
Nusbaum,
Sleigh,
Kremen
(1970)
Secondary
Carbon
13-17
600
—
-90
—
—
9^
9^
5^
>99
96
85
Envirogenics
(1970)
Secondary
None
50-60
600
—
80-100
—
—
89-95
60-85
31-75
95-99
—
80-100
Conn
(1971)
Raw
Filter Aid
6-8
500
—
—
—
85-87
90-93
—
—
96-98
—
—
-------
With the spiral wound module, a daily 15-minute air-tap water flush with
the system in a depressurized condition was successful in controlling
excessive- pressure drops across the brine side of the spiral wound mo'dules.
However, these flushes had little effect on the rate of flux decline. Two
other methods were tested, a citric acid flush and a flush with anionic
detergent solution. Neither of these procedures were successful in
restoring membrane flux.
With the tubular configuration two different runs were made with carbon
treated secondary effluent. The initial flux in each run was 20 to 25
gal. /ft2 »day. At an axial velocity of 2.0 ft/sec, flux decline was 11$
after 300 hours while at an axial velocity of 0.^4- ft/sec flux decline was
during the first 300 hours.
With the flat plate configuration, five different waste streams were
treated (l) lime clarified secondary effluent after dual media filtration
and carbon adsorption, (2) lime clarified secondary effluent after dual
media filtration, (3) lime clarified secondary effluent, (U) secondary
effluent, and (5) primary effluent. Perhaps the most revealing result
comes from comparison of the runs with Feeds 1 and 2 which had initial
fluxes of 12 to 15 gal. /ft2 »day. Feeds for both runs (l and 2) were
passed through a nominal 5-micron cartridge filter just before the module.
Pertinent feed analyses are given in Table 2.
Table 2. Tests with Secondary Sewage Effluent in Plate
and Frame Module
Run d TOG, Turbidity A log flux
ee
No. mg4 JTU ~ A log time
1 lime clarified, dual
media filtration,
2
carbon treated
lime clarified, dual
media filtration
1.0
10.6
0.5
1.3
0.05
O.llt
The greatly increased flux decline observed in the run with feed No. 2
(which had 10 times the total organic carbon of feed No. l) was interpreted
as supporting a positive role for soluble and colloidal organic carbon in
causing flux decline.
In the test with feed No. 3 (lime-clarified secondary effluent) the value
of the flux decline parameter, b, (b = — A log flux/A log time) was main-
tained at 0.03 by flushing every other day for one hour with an enzyme-
detergent solution. With feed No. h (secondary effluent) a one-hour flush
every day maintained a value of b = O.Oh. In both cases the initial flux
was ~8 gal./ft2 •day. A solution with greater than 35 mg/4 enzyme and
3500 mg/jfl detergent was required for an effective flush.
-------
With feed No. 5 (primary effluent) a daily flush with enzyme + detergent
was only able to maintain a value of b = 0.19 starting with an initial
flux of ~7 gal. /ft2 «day.
The significance of the flux decline parameter, b, (cited above and used
in subsequent portions of the report) can be appreciated by calculating
the average flux for a one-year period of operation assuming that the flux
decline parameter is constant throughout the year. Results of such a
calculation for a range of b-values are given in Table 3.
Table 3. Average Value of Flux for One Year of Operation
Assuming a Constant Value of Flux Decline Parameter
Throughout the Year
_ _ A log flux Average Flux as
A log time % of Initial Flux
0.3
0.2
0.1
0.06
0.0*1
0.02
0.01
0.005
2k.k
38.3
61.7
7^.7
82.3
90.7
95.2
97.6
Nusbaum, Sleigh and Kremen (8) used spiral wound modules' exclusively with
carbon-treat.ed secondary effluent rechlorinated to give approximately 2
mg/ji chlorine residual. Using daily air and water flushes and weekly
flushes with a solution of 10,000 mg/4 of an enzyme containing laundry
presoak product, flux decline parameters of 0.0*1-7 to O.o6l were obtained
in 1000-hour tests with modules having an initial flux of 13 gal. /ft2«day
and flux decline parameters of 0.055 to 0.077 were obtained in 1000-hour
tests with modules having an initial flux of 17 gal. /ft2 «day when the COD
of the feed was 3 to 9 mg/jfc. When the COD increased to 9 to 15 ing/A, the
flux decline slope of the higher flux modules was 0.1*1-, indicating a
direct relation between the rate of membrane fouling and the increase in
COD of the feed.
Other studies showed that neither smaller brine side spacer thickness nor
vertical module operation significantly affected module performance.
Beckman et al. (9) made further studies with spiral wound modules. Feed
was from an oxidation pond after secondary treatment. Particularly notable
were a series of three 200-hour tests at 750, 600, and UOO psig resulting
in initial fluxes of 26, 22, and 1*1- gal./ft8 .day. Apparently no flushing
procedures were used in these tests. Severest flux decline was observed
with the highest flux membrane; after 90 hours the production rate from
this membrane crossed below the rate for the lowest initial flux membrane.
The intermediate flux membrane crossed below the lowest flux membrane after
~35 hours of operation.
-------
Another important result of this study (9) was the discovery that sodium
perborate and ethyle^tediaminetetraacetic acid as well as enzyme deter-
gent flushes restored the flux to 80 to 85$, of the initial value.
Recently, high flux cellulose acetate membranes have been developed (10).
Three different formulations were developed: (l) a cellulose diacetate of
moderately-low acetyl content (2) a cellulose triacetate-diacetate blend
and (3) a crosslinked cellulose acetate methacrylate. All produced fluxes
greater than 60 gal. /ft2«day and rejected at least 60$ of a sodium chloride
and 95$ of a sodium sulfate solution when tested at 600 psig with 1000 mg/jfc
feed solutions. The performance of the membranes was evaluated using sec-
ondary effluent containing 5 ml per gallon of 5$ sodium hypochlorite; pH
was adjusted to 5.0 ± 0.25 with dilute sulfuric acid. Rejections with this
feed are given in Table 1.
Flux decline was measured in 5- to 17-day tests with a 3-in. flat-plate
test cell constantly recirculating the feed from a 5-gallon reservoir.
Starting from an initial flux of ~50 gal./ft2»day, the flux declined to
less than 20 gal./ft2»day in ~100 hours. With these high flux membranes,
flux declines using carbon-treated secondary effluent was modest after
30 hours (from 65 to 55 gal./ft2«day) but after ~100 hours flux was only
~20 gal. /ft2 «day. Periodic flushing with an enzyme detergent largely
prevented the flux declines in 100-hour tests.
In a general discussion of design considerations for treatment of solids-
laden wastewaters by reverse osmosis, Feuerstein and Bursztynsky (ll)
present an interesting comparison of the effect of feed water quality on
flux decline under comparable, controlled test conditions. The only
pretreatment given to several grades of municipal wastewater was pH
adjustment to ~5.0. Flux decline slopes, determined by replotting a
graphical comparison of flux declines, are given in Table k. The membranes
had initial fluxes of 15 to 30 gal./ft2«day.
Table U. Flux Decline Slopes Obtained under Identical
Conditions with Different Types of Municipal Wastewater
[after Feuerstein and Bursztynsky (ll)]
o m, -u A 1°S flux
Sewage Type b = — '. ., , .—
•^ A log time
1. Carbon-treated secondary 0.1^+
2. Secondary 0.35
3. Primary 0.56
U. Raw ~0.9
A novel scheme has been developed for preventing fouling of tubular modules
when treating raw sewage which had been filtered through a 100-mesh
screen (12). Before introducing sewage into the reverse osmosis system
-------
the membrane is coated with diatomaceous earth, powdered activated carbon
and surface active agent. This precoat mixture is applied under pressure;
after seven hours 10 minutes of operation the pressure is dropped, and
the accumulated solids are flushed from the system .for 15 minutes. Then
a cleaning solution of ammonium citrate and enzyme detergent at pH 5-5 is
recycled for 15 minutes, the membrane is recoated and raw sewage introduced
to begin the cycle again. No information was given about membrane life-
times achievable.
Results to date have shown that considerable pretreatment beyond the
secondary stage is required to prevent severe flux decline in treatment
of municipal sewage effluent by hyperfiltration using commercial modules
at flow rates recommended for brackish waters. Even with carbon treatment
after the secondary stage, the rate of flux decline of membranes with
different initial fluxes, increases as initial flux increases. The evidence
is that soluble or colloidal organics are an important factor in flux
decline with these highly treated sewage effluents.
Relatively long term runs have been achieved with carbon treated secondary
effluent as feed by using frequent flushea with solutions of enzyme deter-
gent, sodium perborate or ethylenediaminetetraacetic acid.
Extensive studies of cellulose acetate membranes operating on untreated
river water feeds containing 10-100 mg/j? suspended solids have shown that
flux decline can be arrested by operation at high tangential velocities (13).
For example, after an initial flux decline during the first day's operation
(from 36 to 26 gpd/ft2), flux decline was arrested for the next nine days
by the high axial velocity (2U ft/sec) used and after the first 10 days
operation, the flux was still greater than 23 gpd/ft3. When the axial
velocity was dropped to 1.6 ft/sec, there was catastrophic flux decline
during the next two days from 23 to k gpd/ft2. Returning the velocity to
2k ft/sec for one day gave no significant flux recovery.
With some feeds velocities of 2^ ft/sec may not be required to arrest flux
decline. In one experiment velocity was reduced from 2U to 9.8 ft/sec
after two days operation and from 9.8 "to ^.9 ft/sec after six days opera-
tion. After two days operation at k.9 ft/sec (eight days after the start
of the run) flux was still greater than 23 gpd/ft2 and there were no
marked changes in the rate of flux decline as the velocity was successively
changed. When the velocity was decreased to 1.6 ft/sec, there was a rather
abrupt decline in flux from 23 to 13 gpd/fts followed by a slower decline
to 11 gpd/ft2 during the next three days.
The results of flux decline studies frequently produce straight lines when
plotted as log flux, J(t), vs log time, t (at constant pressure and temper-
ature) i. e.,
J(t) = JT (£-)•* , (l)
where Jj = reference flux at reference time, Tj. Thus, small values of
the exponent "b" correspond to small values of flux decline with time.
10
-------
These studies showed that mean values of b (when using Tennessee River
water as feed) increased as the axial velocity decreased. At a velocity
of 2k ft/sec, values of b were small (0.02 to 0.03) and were consistently
in a narrow range regardless of feed composition. However, at smaller
velocities, values of b were not only much larger (a mean of 0.09 to 1.6
ft/sec) but the range was much larger (from 0.07 to 1.0 at 1.6 ft/sec).
In view of the poor reproducibility of results at the lower velocities,
it appears that in this range flux is very sensitive to composition of
the feed. At sufficiently high velocities, however, it seems that -flux
decline can be held to low values and is insensitive to feed composition
for a particular wastewater stream.
Because of the indications that proper hydrodynamic conditions could
markedly reduce the rate of flux decline with polluted surface water
feeds, a program was initiated at the Oak Ridge National Laboratory to
determine the effectiveness of various hydrodynamic and related techniques
for control of fouling and flux decline of cellulose acetate membranes to
improve the proposed hyperfiltration (reverse osmosis) method for treatment
of municipal sewage effluents. Effluents from the primary and secondary
stages of the Oak Ridge Sewage Disposal Plant were to be used. Since this
sewage is of unusually low salinity, it was to be spiked with salts to
facilitate analysis for salt rejection. Since it was anticipated that,
under proper hydrodynamic conditions, it would be no more difficult to
handle primary sewage effluents than secondary sewage effluents, emphasis
in this program was to be on primary and treated primary effluents.
It was anticipated that, in the absence of turbulence promoters, high
tangential velocities would be required to minimize flux decline and
fouling. It was proposed to evaluate the effectiveness of this high
tangential velocity technique using commercial cellulose acetate mem-
branes of various intrinsic fluxes in test sections designed locally to
allow good hydrodynamic control and ease of assembly. In one of these
test sections the membrane is mounted on the outside of a cylindrical
porous tube and this "finger" is inserted into a transparent pressure
jacket which allows visual observation of the membrane during operation.
The principal parameters to be studied in the high tangential velocity
systems were circulation velocity, time, and the effect of permeation rate
on flux decline or fouling rate at velocities ranging from 5 to 20 ft/sec.
11
-------
SECTION IV
EQUIPMENT
Mobile Hyperfiltration Loop. All tests were carried out in a test loop
located at the Oak Ridge East Treatment Plant. The loop was substantially
the same as that developed and used for an earlier FWQA program on appli-
cation of dynamic membranes to treatment of municipal sewage effluents by
hyperfiltration (5). The loop is mounted in a highway trailer having
internal dimensions of 6 1/2 x 7 1/2 x 28 1/2 ft. Feed is circulated by
a canned rotor, oil-cooled pump rated at 120 gpm with 29^ ft head for
fluid at 68°F. Pressurization and feed makeup are provided by a Mil-Roy
Triplex pump rated at h gpm at 2000 psig. Pressure is controlled by an
air-controlled 1-in. letdown valve. Volume of the loop and two test
sections is about 5 gallons.
The test loop has a concentric pipe he at-exchanger to remove pump heat, two
1-in. test sections, and a test section by-pass to permit operation of the
pump at near rated output even with low flows through the test sections.
Flows through the test sections are controlled by ball valves located in
each section and are measured by Venturis which have a range of 6 to 60
gpm.
Auxiliary equipment consists of a cooling oil system for the circulating
pump, product tank, feed pump supply tanks, transfer pump between product
and feed tanks, and an air compressor. Chemical addition to the loop is
through either an air operated piston pump, used for intermittent additions,
or a small diaphragm, pump, used for continuous addition and supplied by
a 50-liter tank. A skid mounted feed booster pump transfers primary
effluent from downstream of the primary settling tanks to the feed pump
supply tanks. Secondary effluent, supplied through a second skid-mounted
pump, is used for the heat exchanger cooling water. Fresh water can also
be supplied to the heat exchanger through a tank connected to the coolant
pump suction line.
Test Sections. Two types of test sections were used. One similar to that
shown in Figure 1 consists of a 5/8-in. test finger that has two 6-in. long
porous sections fabricated from sintered stainless steel frit enclosed in
a transparent jacket. The product sides of the test section are separated
and have their own discharge lines. A cellulose acetate membrane is wrapped
around each porous section and is sealed in place with pressure-sensitive,
water-proof tape. The feed stream flows through the annulus between the
test section and the jacket. The product stream flows through the membrane
and is discharged at atmospheric pressure to the product tank.
The other type of test section is of similar design except it contains
three porous sections each of which is sealed from the others. Three
product lines are required. The pressure jacket is fabricated from
stainless steel. Since these pressure jackets are longer than the ones
used for the two-section fingers, stainless steel was used as a safety
measure to protect against failure in case of an accidental pressure surge
in the loop.
-------
TRANSPARENT JACKETS
POROUS MEMBRANE SUPPORTS
Figure 1. Hyperfiltration Test Section with Transparent Pressure Jacket.
-------
SECTION V
FEED COMPOSITION AND CHEMICAL ANALYSIS
The composition of primary sewage effluent from a municipal sewage treat-
ment plant varies during the day and also appears to have some seasonal
trends. The Cl~ ion concentration and the turbidity (JTU) were monitored
almost daily when the loop was operating by sampling the feed storage
tank. A more complete analysis was made at less frequent periods.
Figure 2 shows the data obtained during the period July 1, 1970 to
August 1, 1971. Run numbers for tests in the present program are shown
at the top of the figure with lines indicating the dates the runs were
started and terminated. The temperature data are for the circulating
fluid in the loop but reflect changes in sewage temperature since secondary
effluent is used as the loop coolant.
No storm sewers are connected to the treatment plant. However, ground
water does leak into the lines so that rains do affect the quantity and
composition of primary effluent.
During the 13-month period that the primary sewage effluent was monitored,
the turbidity varied between 5 and 66 JTU with slightly higher values
during warm weather and following hard rains. Light rain tended to reduce
turbidity. Inorganic carbon concentration was relatively constant varying
between 32 and 60 mg/A, with lower values occurring during cold weather.
The amount of organic carbon varied considerably more ranging from 15 to
Il8 mg/jfc. Phosphate concentration varied from 5 to 67 mg/jj. Normally,
samples were obtained in mid-morning when phosphate concentrations were
20 to 50 mg/4 so that rejection values for membranes would'be more
accurate. 03++ and Mg"1"1" concentrations remained fairly constant — 0.00112
to-O.OOr/rMegVl — showing little seasonal trend. The Cl~ ioni concentration
was always less than 1.0 meq>/l.
Chloride analysis was performed amperometrically with a Buchler-Cotlove
chloridometer. Turbidity was determined with a Hach 2100 turbidimeter.
Ca++ plus Mg"1"1" concentrations were measured together by EDTA titration
using Eriochrome Black T indicator and potassium cyanide as an inhibitor
for metal ions. Total phosphate was determined by a colorimetric method
using aminonapthosulfonic acid during the first part of the program. Then
stannous chloride was used as the reducing agent during the latter part of
the program because of its greater sensitivity. Total and inorganic carbon
were determined with a Beckman 915 analyzer. Organic carbon was obtained
as the difference between these determinations.
15
-------
ORNL-DWG 71-9901
RUN NO.
CT ION
(meq/l)
1.0
0.5
0
:o
wf" °-0015
0.0010
60
TOTAL „..
PHOSPHATE ION w
20
0
10O
80
ORGANIC
CARBON 60
mg/ Si 40
20
INORGANIC 6o
CARBON
40
JTU
CIRCULATING 30
TEMPERATURE
(°C) 20
60
40 H-b
20
0
5 15 25 5 15 25 5 15 25 5 15 25 5 15 25 5 15 25 5 15 25 5 15 25 5 15 25 5 15 25 5 15 25 5 15 25 5 15 25
JULY AUG SEP OCT NOV DEC JAN FEB MAR APR MAY JUNE JULY
1970 1971
Figure 2. Variation of Constituents in Primary Sewage Effluent, Oak Ridge East Sewage Plant.
-------
SECTION VI
RESULTS
Three types of membrane with markedly different characteristics were used
in this study: (l) an unheat-treated membrane (Eastman HT-00) with a
nominal initial flux of 150 to 200 gal. /ft2»day at 600 psig and nominal
zero rejection, (2) a membrane (Aerojet 6%} with nominal initial flux of
^ to 50 gal./ft2'day at 600 psig and chloride rejection of ~9^0, and (3)
a membrane (Eastman) heat-treated at ~90°C to give a nominal initial flux
of 2 to 10 gal./ft2«day and 97 to 99% chloride rejection. With our experi-
mental equipment, three sections of a membrane were always exposed to the
same feed at three different velocities. In the first 10 tests^two of the
three membrane types were compared simultaneously; in the remainder of the
tests all three membrane types were compared simultaneously at three
different velocities, i.e., three membranes each exposed to three different
velocities or 9 tests per run.
Before starting a run the test loop was cleaned by circulating two batches
of 1 M NaOH for one hour each; then the loop was rinsed with city water
until the pH of the rinse water was 7- At the start of every run, mem-
branes were exposed for 2 to l8 hours to a feed of city water spiked with
NaCl to obtain baseline performance for the membrane. The flux one-half
hour after startup on city water was taken as the initial flux, Ja.
The runs made during this study are summarized in Tables 5 and 6. Table
5 gives the total run time in hours, the velocity at which the membranes
were exposed, the duration of the baseline city water portion of the test
and the flux and observed rejection for chloride measured one-half hour
after startup with city water. The bulk of the table gives flux and
chloride rejection at 20, 50, 100, 150, and 200 hours after switching to
primary sewage feed. Samples were taken at 20 and 200 hours for analysis
for P04 , Ca++ plus Mg++, organic carbon, and inorganic carbon. These
latter results will be summarized in a subsequent part of the report. Since
unscheduled shutdowns frequently may affect flux and rejection, the time of
occurrence and duration of the shutdowns is given in the last two columns
of Table 5. A few tests were extended to times longer than 200 hours,
results for the extended time portion of the tests are given in Table 6.
Flux decline results are frequently presented as log flux vs log time or
as log flux vs arithmetic time. The first form implies the possibility
of a power law dependence according to Equation (l) while the second
implies the possibility of an exponential dependence:
Since there is no theory for fouling which supports either Equation (l) or
Equation (2), it is immaterial which form is used in presenting experimental
results. More important, great caution should be used in extrapolating
results for flux decline for any appreciable time beyond that for which
data are available using either Equation (l) or Equation (2). As will be
illustrated with the present results, apparent linear relations may be
followed on one or the other of the two types of plot for hundreds of
17
-------
Table 5 . Summary of Cellulose Acetate Runs with City Water and Primary Sewage Feed
00
Run No. Total
/StartingN Run
k Date ; T^e,
13 177
(8-4-70)
14 156
(8-14-70)
17a.a 190
(9-17-70)
18 167
(9-25-70)
19 235
(10-5-70)
16 Vel.
/ 15.3
B f 9.9
( 5.2
, 15.3
A \ 9.9
* 5.2
/ 20.0
B { 13.6
1 7.2
I 20.0
A { 13.6
1 7.2
/ 30.0
B { 7.6
1 4.0
/ 30.0
A f 7.6
'• 4.0
/ 30.0
B { 12.0
1 6.4
, 30.0
A { 12.0
I 6.4
, 15.3
B { 9.9
1 5.2
I 15.3
A { 9.9
I 5.2
Water Run
Run 1/2 Hr.
THr?' #* Y8
2 56.1 82.6
74. 3 66. 2
68.5 72.8
199.8 11.2
171.3 8.7
180.6 0
2 52.2 86.3
56.0 8o.O
61.9 68.4
495.5 2.0
180.7 12.5
166. 5 18. 2
6 1/2 61.9* 75.3*
336. l 16. 2
213.1 24.2
211.1 10.8
434.9 7.0
220.2 9.6
1 3/4 53. 45 89.
-------
Table 5 . (Continued)
Run No. Total
(StMtiagN Bun Membtie Vel.
V Date / Tin*, ?PS
Hr
20s 31+2
(10-19-70)
21 288
(n-n-70)
22 W+
(11-25-70)
23 1*79
(12-17-70)
21+9 752
(1-7-71)
B
A
C
A
C
A
C
A
B
A
/ 12.0
1 9.0
1 U.75
, 12.0
{ 9.o
* 1*.75
/ 15.3
1 9.9
' 5.2
, 15.3
I 9.9
' 5.2
(12.0
5.7
• 3.0
, 12.0
{ 5.7
1 3.0
. 12.0
( 5.1
* 3.0
, 12.0
{ 5'7
1 3.0
/ 15.0
\ 9.9
1 5.2
, 15.0
\ 9.9
1 5.2
Water Run
Run 1/2
Ti*e, gfd
18 50.6
60.6
50.7
163.3
181.7
181>.U
3 1/2 2.1
0.7
0.7
129.U
11*1.0
125.2
2 1/2 7.6
16.6
20.1*
156.6
133.0
126.U
3 30.0
13.6
11.6
122.1*
115.6
123.2
2 1/2 1*1.7
1*7.1*
1*0.1*
ll*6.5
lUl.9
122.0
Time After Change to Primary Sewage
Hr.
Robs
%
92.7
77.0
89.0
2.0
7.0
"t.O
22.0
38.5
1*8.0
27.0
20.3
21.5
76.0
26.6
16.7
21.1*
1.0
5."*
18.8
17.0
1*1.3
31.0
13.0
5.6
92.9
78.9
91.5
18.7
10.3
9.8
20 Hr.
gfd
36.8
31*. 1
2U.3
35.0
53.0
15.1+
0.8
0.7
0.6
71.5
1*0.5
13.0
1+.3
3.1*
3.0
51.7
9.6
l*.9
5.0
3.1*
l*.l
1*0.0
11.6
5.7
3l*. 2
35.5
6.1*
So.l*
38.5
8.30
Robs
%
95.2
93.3
85.0
6.0
20.0
25.5
90.0
90.8
97.0
1*3.0
1*2.0
1*1*. 5
97.5
70.0
91.6
50.8
26.5
30.0
93.1*
96.0
95.1*
61.2
1+1.6
31.9
56.0
38.0
86.3
39.5
38.0
1*1+.5
50
gfd
36.0
33.3
10.0
38.0
1*7.1*
7.5
0.8
0.7
0.6
66.5
37.8
6.5
1+.3
3.3
2.9
1+7.5
l*.8
2.9
1+. 7
3.3
3.6
38.0
6.2
3.6
32.6
33.0
1+. 3
77.2
36.0
3.8
Hr.
Robs
%
95.5
9>*. 6
83.1
10.0
20.5
29.0
92.5
92.7
98.5
Itlt.o
1*2.2
39-5
99.8
70.0
91*.!
51.0
25.2
28.5
95.1+
97.1+
96.lt
6o.lt
36.7
29.5
57.2
1*2.5
82.5
39.5
36.1
1*1*. 0
100 Hr.
gfd
35.3
32.3
5.1
1+0.2
38.8
l*.l*
0.7
0.7
0.6
62.3
35.7
l+.O
U.1
3.2
2.8
1+1+.8
2.9
1.9
1+.5
3.3
3.2
36.7
3.3
2.2
31.5
31.2
2.9
75.6
3"+. 3
2.6
Robs
%
95.8
95.1+
75.5
10.0
28.1+
31.8
91*.!
93.7
99.1
1+1*. 1
1+2.1+
36.3
99.9
70.0
95.6
51.1
21+.0
28.0
96.5
98.1
97.0
29.5
33.2
27.0
58.1
1*6.0
78.7
39.5
35.0
1+3.5
150
gfd
33.3
29.0
3.1*
38.8
27.1*
3.2
0.7
0.6
0.6
61.7
31*. 8
2.9
I+.3
3.7
2.8
1*1.8
2.1+
1.8
1+.1+
3.1
2.8
3>+.l
2.3
1.8
30.8
30.1
2.3
69.7
30.2
2.1
Hr.
Robs
%
95.9
95.8
69.1*
9.0
2l*. 7
27.9
9l+. 8
9l*. 3
99.3
¥+.2
1*2.5
31*. 5
99.5
56.0
95.7
53.0
23.5
27.5
97.0
97.8
97.3
28.7
31.0
26.9
59.0
1*7.5
75.5
39.5
31*. 2
1*3.3
200 Hr.
gfd
32.1
27.7
2.6
39.3
21.7
2.5
0.7
0.6
0.6
60.0
31*. 0
2.3
1*.2
3.2
2.5
33.1+
1.9
1.5
!*.!+
2.8
2.2
30.0
1.8
1.5
30.0
29.5
1.9
65.0
27.7
1.8
Robs
%
96.0
9!*. 2
61*. 0
7.-6
2l*.8
27.0
95.3
91*. 7
99.0
1*1*. 3
1*2.6
33.0
99.6
52.0
97.9
57.0
23.0
27.0
97.3
97.5
95.6
28.0
29.5
2l*. 5
59.1*
1+8.5
70.0
39.5
31*. o
1*3.0
Shutdowns
Total Run Duration
Time, Hr. Hr.
21+1+ lit 1/3
30l+ 1/1+ 11 1/2
315 3/1+ 11 1/2
131 13
299 lU
631 3/1+
-------
Table 5 . (Continued)
to
O
_ „ Total
Run No. „ ,
/StartingN Memb.16 ^
\ Date / jjj. '
25
(2-18-71)
26
(3-11-71)
283
(4-19-71)
gglO
(6-8-71)
352 , 17.6
C < 10.0
1 3.1
/ 17.6
B < 10.0
1 3.1
, 17.6
A { 10.0
1 3.1
768 / 17.6
C < 10.0
3.1
i 17.6
B < 10.0
1 3.1
/ 17.6
A < 10.0
1 3.1
985 / 17.6
C \ 10.0
1 3.1
, 17.6
B { 10.0
1 3.1
, 17.6
A I 10.0
I 3.1
72 17.6
C i 10.0
' 3.1
Water Run
Run 1/2
THr! gfd
2 1/2 4.4
5.7
11.7
81.3
41*. 6
43.0
81.7
117.2
101.0
2 1/2 5.4
3.6
6.3
55.9
49.3
39.2
143.9
115.9
113.7
2 1/1* 2.8
8.1
10.8
40.7
48.6
59.3
177.7
197.1
154.8
2 1/1* 4.4
3.6
15.3
Time After Change to Primary Sewage
Hr.
Robs
%
80.5
60.0
77.4
48.6
84.5
77.6
52.5
25.9
21.5
81.5
88.0
70.5
73.9
88.9
90.4
27.0
23.9
19.1
92.6
33.9
49.5
93.0
83.2
69.8
16.5
6.3
12.1
51.0
92.6
22.8
20
gfd
2.9
2.9
3.1
36.0
35.0
4.4
64.0
45.0
4.3
4.8
3.3
3.4
34.8
5.0
93.6
50.8
5.9
2.1*
3.0
3.2
31.2
31.6
5.4
96.5
39.0
6.2
3.1
2.9
3.2
Hr.
Rqbs
%
99.3
99.0
98.7
95.6
95.3
67.2
72.1
50.2
4l.o
98.1
98.0
97.6
94.4
94.1*
70.5
45.8
4o.O
36.0
99.5
96.7
95.3
95.5
94.8
83.3
30.9
29.1
37.0
99.2
98.2
98.2
50
gfd
2.8
2.9
2.7
34.2
33.5
2.5
60.6
42.3
2.6
4.6
3.2
3.2
31.8
33.0
3.1
87.3
48.5
5.5
2.4
2.9
3.2
30.8
30.2
2.8
89.0
39.9
3.5
3.1
2.7
3.4
Hr.
Rqbs
99.9
99.7
99.1
96.5
95.6
65.8
74.2
50.4
4o.2
98.4
99.3
98.1
94.7
94.7
67.5
50.0
4l.o
36.0
99.9
98.5
99.4
95.4
94.8
74.0
32.2
27.0
35.7
99.5
98.3
99.2
100
gfd
2.7
2.8
2.5
30.7
29.4
1.8
55.5
30.2
1.8
4.6
3.1
3.1
29.0
29.8
2.2
78.0
34.8
2.6
2.4
2.8
3.0
28.3
29.9
2.0
75.0
39.9
2.2
Hr.
Rqbs
99.8
99.6
98.7
96.3
95.7
64.5
75.7
49.7
4o.l
98.7
99.7
96.4
94.9
95.0
66.2
53.0
42.0
36.0
99.8
99.2
99.5
95.4
94.8
71.0
4o.O
28.2
36.5
150
gfd
2.7
2.8
2.6
29.1
26.5
1.6
51.8
25.9
1.6
4.5
3.1
2.8
27.2
25.9
1.7
72.0
26.8
2.2
2.3
2.7
2.8
27.6
29.0
1.6
60.0
37.9
1.8
Hr.
Rqbs
%
99.7
99.5
98.5
95.8
95.2
63.6
76.6
49.4
4o.O
98.8
99.8
98.3
95.0
95.1
63.0
56.0
42.3
36.0
99.8
99.3
99.5
95.6
95.2
68.0
46.0
26.5
36.4
200
gfd
2.8
2.9
2.4
28.0
21*. 8
1.5
49.3
19.0
1.4
4.5
3.1
2.5
26.0
2.2
1.5
69.3
22.5
1.9
2.4
2.7
2.7
26.7
28.3
1.5
59.2
35.0
1.7
Hr.
Rqbs
%
99.6
99.4
98.2
95.4
94.9
62.9
77.1
49.2
4o.O
99.0
99.9
98.2
95.5
95.2
60.5
61.2
45.0
36.2
99.9
99.2
99.5
96.0
95.9
68.5
49.2
29.0
36.4
Shutdowns
Total Run Duration
Time, Hr. Hr.
303 1/3 7 1/2
208 5/6 7 1/2
-------
Table 5 . (Continued)
Run No. Total
/StartingN ?™ Memb.16 Vel'
( Date )*£>, *•
29
(Continued)
3011 192
(6-23-71)
3l13 58
(7-13-71)
321* 51
(7-21-71)
/ 17.6
B < 10.0
1 3.1
, 17.6
A J 10.0
I 3.1
, 17.6
C ^ 10.0
' 3.1
. 17.6
B J 10.0
* 3.1
. 17.6
A ) 10.0
< 3.1
i 2k. 0
C | 12.0
7.1*
, 2l*.0
B < 12.0
V 7.1*
2l*.0
A I 12.0
* 7.1*
i 19.1
C { 6.7
1 2.9
< 19.1
B \ 6.7
2.9
Water Run
Run 1/2
T^e, gfd
75.8
56.6
76.8
192.6
20l*. 3
167.5
2 1/6 U.l
6.6
9.1
63.0
53.8
86.6
70.2
68.9
89.9
11 1/2 13.1
5.8
1*1*. 9
53.0
113.0
180.7
63.5
89.0
81*. 7
3 1/2 3.9
2.6
3.6
50.5
55.2
57.2
Time After Change to Primary Sewage
Hr.
Rqbs
%
67.9
92.1
60.9
16.7
10.7
15.9
61.9
22.7
23.1*
78.5
87.0
56.9
39.3
31*. 6
21.7
12.1*
21.6
3.3
93.1
1*3.3
26.6
1*2.5
22.2
25.1*
68.3
95.9
69.2
93.5
89.8
87.8
20
gfd
38.0
26.7
10.9
71.5
28.8
11.3
2.1
1.3
1.3
28.8
28.7
6.8
35.6
33.0
7.1
0.5
1.3
3.8
1.1
It. 5
5.8
0.9
5.3
It. 6
2.2
2.0
1.8
33.0
22.5
8.6
Hr.
Robs
91*. 7
91.9
78.2
29.0
30! o
97. !t
93.1
91.1*
97.8
96.5
69.2
71.8
77.2
29.8
87.0
91*. 6
56.0
1*8.5
55.5
1*8.0
1+6.1
26.2
23.1*
91.9
82.0
90.9
92.9
88.6
82.5
50 Hr.
gfd
36.8
26.9
U.l*
70.0
2U. 7
U.3
2.1
1.3
1.3
26.8
26.5
3.9
30.8
27.7
l*.9
0.31
1.U
3.6
0.7
3.2
U.5
0.5
2.U
3.9
2. 21
1.9
1.8
23.7
lU.2
u.u
Rqbs
%
91.6
73.7
27.8
28.3
31.0
99.0
96.6
95.9
95.3
97.8
55.6
82.7
77.5
26.0
eu.o1
95.5
U6.5
U2. 2
53.1*
Uo.o
Ul.O
30.8
18. U
90. o1
87.0
85.0
9^.7
93.6
7U.O
100
gfd
2.1
l.U
l.U
25.9
17.8
2.8
25.8
19.1*
2.8
Hr.
*?'
99.5
96.3
97.0
98.2
96.5
51.9
85.1
77.1
2U.3
150
gfd
2.1
l.U
1.3
23.7
11.5
2.1
21.0
12.8
2.0
Hr.
Hobs
99.5
96.6
98.0
98.3
95. U
U8.1
86.9
70.0
22.8
200
gfd
2. 11
l.U
1.3
21.3
8.U
1.7
18.0
9.5
1.6
Shutdowns
R9bs Time, Hr. Hr.
7°
96. o1
96.2
97.0
98.8
9U.3
U7.3
88.7
63.8
22.1
10 7 1/2
-------
Table 5 . (Continued)
K:
N3
Run No. Total
/StartingN »» Memb.ie
V Date / Time>
Hr
32i4
(Continued)
33 1*09
(7-29071)
3^16 1^
(8-23-71)
3516 25
(8-31-71)
A
C
B
A
C
B
A
C
B
A
I
I
/
{
I
/
{
I
/
I
|
\
/
{
I
Vel.
fps
19.1
6.7
2.9
21.8
8.0
3.3
21.8
8.0
3.3
21.8
8.0
3.3
9.0
5.0
3.0
9.0
5.0
3.0
9.0
5.0
3.0
9.0
6.0
3.7
9.0
6.0
3.7
9.0
6.0
3.7
Water Run
Kun 1/2
Time,
Hr. 6"
73.7
77A
80.8
2 2.1
2.5
2.0
1*9.8
51.2
1+7.7
65.5
61*. 8
61.1*
3 1/2 7.5
8.2
7.5
51*. 6
51*. 6
71.6
61*. 8
71.6
75.0
2 1.9
3.3
9.5
52.6
68.6
56.6
76.6
79.3
92.6
Time After Change to Primary Sewage
Hr.
Robs
%
1*3.5
32.3
3l*.l
99.5
91.6
92.8
9l+. 1*
93.1*
92.6
51.5
1*5.9
50.7
32.5
17.3
21.0
9l+. 5
10.2
68.2
1*9.9
38.2
36.9
99.9
99.3
23.1*
99.1
66.1+
98.6
30.0
36.0
19.6
20
gfd
1*3.1
30.3
8.7
1.7
2.2
1.5
35.8
29.1
8.5
1*6.9
31.1*
9.0
1.7
1.6
1.9
21.8
15.0
12.7
25.9
15.0
13.1
1.1*
1.7
1.8
23.8
19.0
16.1
25.2
20.3
16.2
Hr.
Robs
%
1*9.8
1*8.6
1*1*. 5
97.8
98.1+
92.1*
95.7
9l*. 5
85.2
63.2
66.5
58.8
97. >*
97.2
91*. o
86.3
88.1*
88.5
60.0
3!*. 2
30.8
88.5
58.6
67.0
93.7
85.7
79.8
51.9
1*5.3
3.5
50 Hr.
gfd
27.5
26.3
1*.7
1.7
2.2
1.7
32.0
25.7
!*.7
1*1.7
22.9
5.3
l.l*1
1.5
1.7
9.6
8.3
5.7
12.3
7.5
6.3
Robs
*
65.0
58.0
59.0
99.3
99.2
86.8
96.1
9!*. 2
80.0
68.5
67.2
51*. o
98. I*1
97.9
97.6
89.9
86.6
85.1*
67.1
51.2
W.5
100 Hr. 150 Hr.
gfd «<*« gfd *£>•
1.6 99.9 1.6 99.9
2.2 97.8 2.2 97.9
1.6 93.8 l.l* 98.5
30.8 96.3 29.2 96.3
21.0 93.9 I1*.1* 93.0
7.8 87.0 3.6 80.3
1*0,7 70.9 38.2 71.8
21.0 67.1* 13.9 70.0
7.8 62.9 3.8 57.7
200 Hr.
,, Robs
gfd %
1.6 99.9
2.1 99.0
1.5 98.9
28. 2 96. 8
7.1* 95.2
3.8 73.0
37.6 7!*. 8
11.0 70.8
2.6 57.0
Shutdowns
Total Run Duration
Time, Hr. Hr.
86 9 1/3
159 1/2 1 min.
288 ll* 1/6
-------
Footnotes for Table 5. Summary of Cellulose Acetate Runs
with City Water and Primary Sewage Feed
1. Extrapolated data.
2. Run 15 terminated prematurely because system pressure controller
failed.
3. Run 16 terminated prematurely because of defective membranes.
h. One-and-one-half hour sample.
5. One-hour sample.
6. Seal on test section loosened at Il8 hours.
7. 101 mg/jj polyvinylmethyl ether (PVME) circulated during water run.
8. Membranes soaked for three days in FVME before being installed.
9. Membranes soaked 30 days in polyacrylic acid (100 mg/j£) while in
test section before testing.
10. Ferric chloride added during sewage circulation (66 mg/j& Fe).
11. Cat-Floe added during sewage circulation (20 mg/jfc).
12. Eastman membranes from a different batch of material than used in
previous tests.
13. Alum [commercial A12 (S04)3«nH20] added during sewage circulation
(50 mg/4 Al).
lU. Aqua Nuchar A added during sewage circulation (100 mg/jfc).
15. Ferri Floe [commercial Fe2(S04)3»nH20] added during sewage circula-
tion (~50 mg/J& Fe).
16. Membrane type, A: Eastman HT-00, B? Aerojet 60;
C: Eastman heat-treated @ 90°C.
17. All runs were made at a pressure of 600 psig.
23
-------
Table 6 .
Summary of Results at ^00 and 700 Hours for Long Term Runs with
Cellulose Acetate and Primary Sewage Feed (First 200
Hours of Runs Summarized in Previous Table)
Run No.
22
23
21+
26
28
33
Membrane
Eastman
Heat Treated 90°C
Eastman
Unheat-Treated
Eastman
Heat Treated 90°C
Eastman
Unheat-Treated
Aero j et
Eastman
Eastman
Heat Treated 90 °C
Aerojet
Eastman
Unheat-Treated
Eastman
Heat Treated 90°C
Aerojet
Eastman
Unheat-Treated
Eastman
Heat Treated 90°C
Aerojet
Eastman
Unheat-Treated
Velocity
ft/ sec
12.0
5.7
3.0
12.0
5.7
3.0
30.0
13.6
11.6
30.0
13.6
11.6
15.0
9.9
5.2
15.0
9.9
5.2
17-6
10.0
3.1
17.6
10.0
3.1
17.6
10.0
3.1
17.6
10.0
3.1
17.6
10.0
3.1
10.0
3.1
21.8
8.0
3.3
21.8
8.0
3.3
21.8
8.0
3.3
Time After Change
kOO Hours
gfd
3.89
1.79
1.07
19-7
1.1
0.9
1+.08
2.17
1.20
23.1
1.32
1.05
2k. 3
20.6
2.lk
39-2
8.9
1.1+3
k.6o
3.29
1.21
20.9
16.3
1.22
1+5.0
12.7
1.22
2.1*0
2.80
2.07
2k. 5
27.8
1.03
29-9
1.12
1.61+
2.21
i.ks
23.9
5.11
1.1+2
30.1
5.11
1.50
R, %
99.9
^1.5
96.8
66.5
22.3
29.6
97.6
97.1
92.7
68.1
30.0
25.5
95.9
95.0
73.5
1+9.0
36.6
39.5
99.8
99.9
97.5
96.7
95.6
56.1+
73.5
50.8
36.5
99.9
99.8
99-2
96.1
95.1
65.3
32.1
31+.0
99.9
99.8
99. ^
96.6
88.9
73.0
81.8
6l+. 1+
55.0
to Primary Sewage
700 Hours
gfd
19-1
18.8
2.37
25.1+
21.1
1.18
1+.59
3.37
0.7!+
17.1+
12.9
0.72
33.0
7.65
0.81
2.U6
2.88
1.28
21.7
23.8
0.89
21.2
1.00
E, %
97. ^
96.1
81+. i
5^.5
29.5
56.7
99.7
99-9
98.0
96.5
95.3
56.0
6l+. o
56.8
39-0
99.9
99.8
98.9
96.7
96.1+
61.0
1+0.0
3^.0
24
-------
hours before marked deviations become apparent. For this reason we prefer
as a definition of the flux decline parameter, b:
^ _ _ A log flux , ^
A log time * ^ '
With this definition, b may be a function of time and there is no implica-
tion of what the functional relation between "b" and "time" might be.
The bulk of our results will be presented as composite plots of data from
many runs. This approach was chosen because the inherent day-to-day and
week-to-week variation of sewage composition introduces an unknown varia-
bility in the results. Although the number of tests made was insufficient
to represent a good statistical sample, enough data were obtained to give
an idea of the range of results which might be expected under different
operating conditions.
Before presenting the composite results, several individual runs will be
discussed because they illustrate a particular point or because special
treatment was given to the membrane or to the primary effluent feed.
Reproducibility of Results
Run 26. Figure 3 shows typical results of a single run, using primary
sewage as feed, in which three different membrane types were exposed
simultaneously to three different axial velocities. This test is identi-
fied as Bun 2.6 in Tables 5 and 6. ~The three membranes used were Aerojet-
General 6%, Eastman HT-00 as received (unheat-treated), and the same type
membrane heat-treated in water for five minutes at 90°C. Axial velocities
in the test sections were 17.6, 10, and 3.1 ft/sec. Loop pressure was -
maintained at 600 psig. Temperature of the circulating fluid ranged
between 21. 2 and 28°C showing an increasing trend with time resulting
in part from the increasing temperature of the cooling water to the heat
exchanger. One hour after the start of sewage circulation a high pressure
excursion shut the loop down. The cause of the excursion was not deter-
mined. The run was restarted in about one minute. The shutdown, which
caused the system pressure to drop quickly to 0 psig, did not appear to
affect the operating characteristics of the membrane at this initial period
of circulation.
City water spiked with NaCl to ~0.01 M was circulated for the first 2 1/2
hours of the test to determine the initial flux and rejection for each
membrane before the feed was changed to primary sewage effluent. Reference
fluxes, Jj, and rejections are shown in Table 7.
Initial fluxes for the heat-treated Eastman membranes ranged from 3.6 to
6.3 gfd with rejections (71 "to 88$,) that varied inversely with flux. Wo
velocity dependence is apparent. Flux for the Aerojet membranes ranged
from 39 to 56 gfd increasing as the axial velocity increased. Rejections,
which varied from 7^ to 90$, decreased as the flux increased. For the
unheat-treated Eastman membranes the flux (llU-lMt gfd) and the rejection
(19-27$) increased as the axial velocity increased.
25
-------
ORNL DWG. 71-4701
2 5 10 2 5 102 2 5 103! 2 5 10 2 5 102 2 5 1031 2 5 10 2 5 102 2 5 103
TIME ON PRIMARY SEWAGE (hi)
Figure 3. Effect of Flux and Axial Velocity on Flux Decline Using Primary Sewage as Feed.
(Cellulose Acetate Membrane, 600 psig, Run 26)
-------
Table 7. Reference Flux and Rejection (Run 2.6, System
Pressure, 600 psig)
Membrane
Eastman HT-00
Heat-treated 90°C
Aero jet -General 6%
Eastman HT-00
Velocity
ft/sec
17.6
10
3.1
17.6
10
3.1
17.6
10
3.1
Jl
gfd
5.U
3.6
6.3
56
1*9
39
nM
116
nU
Robs
%
82
88
71
7^
89
90
27
2k
19
Flux and rejection as a function of axial velocity and time after the
introduction of primary sewage effluent are shown in the left portion of
Figure 3 for the heat-treated Eastman membranes, in the middle portion
of Figure 3 for the Aerojet membranes, and in the right portion of Figure
3 for the unheat-treated Eastman membranes. Total run time for this test
was 523 1/2 hours.
During the run there were temperature fluctuations as large as 2.5°C
between successive data points which resulted in flux variations of as
much as 5$ from the changes in viscosity. The data shown in Figure 3
are the values actually measured. Flux decline parameters, b,
(b = — A log flux/A log time) in this report were determined from plots
where the flux data had been normalized to 25°C by assuming that flux
is inversely proportional to viscosity. The irregularities seen in the
figures, especially between 200 and 350 hours, were almost removed in
the normalized plots.
For the heat-treated Eastman membranes operated at 17.6 and 10 ft/sec
velocities the flux decline was mild for the first 14-30 hours (b = 0.030
and 0.017, respectively) with fluxes going from U. 9^ to h. 30 gfd and
from 3.^0 to 3.20 gfd, respectively, between 1.3 and ^30 hours of sewage
circulation. The rate of decline then increased (b = 0.79 and 0.52) so
that the flux was 3.55 and 2.82 gfd after 621 hours of operation. Rejec-
tions for these membranes increased with time from 97-9^$ to greater than
99$. The membrane operated at 3.1 ft/sec axial velocity also showed a
slight drop in flux (b = O.OUo) from 3.8 to 3.06 gfd during the initial
1^1 hours of the test. Then a steep decline (b = 1.11) was observed
during the remainder of the run with the flux being only 0. 733 gfd after
621 hours. The rejection increased during the initial period from 89 to
98$ and then decreased slightly to 97$ as the fouling layer increased in
thickness.
27
-------
For the Aerojet membranes flux behaviors for membranes at 17.6 and 10
ft/sec were similar to each other during the first 52 hours with flux
decreasing from 37 to 32 gfd and from 38 to 33 gfd, respectively, (b =
0.043). The rate of decline then increased slightly (b = 0.17 and 0.33>
respectively) until at about 350 hours when severe flux declines started
(b = 1.5 and 1.6). After 621 hours fluxes were only 10.1 and 7.6 gfd,
respectively. Rejections increased slightly during the test (from 94 to
95-970/0). The membrane at 3.1 ft/sec had a severe flux decline dropping
from 18.8 to 0.77 gf
-------
ORNL DWG. 71-4116R
JO
VJ
u = 17.6 ft/sec]
100 200 300
TIME (hr)
0 100 200 300 0 100 200 300
TIME (hr) TIME (hr)
Figure ^4. Effect of Flux and Axial Velocity on Flux Decline Using Primary Sewage as Feed.
(Cellulose Acetate Membrane, 600 psig, A >O>V; Run 25; A,,Q,V> Run 26;
l^'l7' Run
-------
For clarity, the 27 different sets of results from Runs 25, 26, and 28
are separated into nine individual graphs in Figure k. Graphs in each
column are for the same velocity but different membranes (easily identi-
fiable by the megnitude of the flux), while graphs in each row are for
the same membrane but different velocities, i.e., the top row in the
figure is for the high flux unheat-treated membrane at velocities of 3)
10, and l6 ft/sec.
The results shown in Figure k are notable for several reasons. First the
rather remarkable agreement of the three different tests. If the fluxes
had been normalized to the same initial flux the agreement would have
been even better in several cases. The greatest divergence among indivi-
dual tests in the three runs occurred with the lowest initial flux mem-
branes exposed to a velocity of 3 ft/sec and the highest initial flux
membrane exposed to a velocity of 10 ft/sec. As will be shown later
the conditions for these two tests place them very close to the "threshold
velocity" for prevention of particulate fouling. In other words the
deviations within these two particular combinations of initial flux and
axial velocity are probably not due to variations in the primary effluent
but to the proximity to the threshold velocity.
Second it is clear that the rate of flux decline is a function of both
initial flux and the axial velocity. For the high flux membrane, within
50 hours of operation at 3 ft/sec the flux had decreased from over 150
gal./ft2«day. However flux from the same type membrane exposed to an
axial velocity of 17. 6 ft/sec was still over 90 gal. /ft2 »day after 50 hours
and at the end of 300 hours the flux was ~6o gal./ft2»day. This will be
discussed further in a subsequent section of the report.
Effect of Loop Depressurization on Flux Recovery
Run 33. The primary objective of this test was to obtain additional
information on the effect of axial velocity and initial flux on the rate
of flux decline. However there were three unscheduled loop shutdowns
-during the test and as a result the most informative aspect of these
tests was the flux recovery after the shutdown for the different operating
conditions.
Membranes used were Eastman HT-00 unheat-treated, similar material heat-
treated at 90°C for 5 minutes, and Aerojet-General 6% membrane. Axial
velocities were 3.3> 8, and 21.8 ft/sec. 'System pressure was 600 psig.
Temperature of the circulating fluid ranged from 31.3 to 33.3°C.
City water spiked with Nad to ~0.01 M was circulated for the first 2
hours to determine the initial flux and rejection of each membrane.
For the heat-treated Eastman membranes initial fluxes were 2.0 to 2.5 gfd
with rejections of 92~99^ with the highest rejection associated with the
highest velocity. For the Aerojet membranes fluxes were kQ to 51 gfd with
rejections of 92 to 914-0. The unheat-treated Eastman membranes had fluxes
of 6l. U to 65.5 gfd with rejections of U6 to
30,
-------
Flux and rejection as a function of time are shown in Figure 5; separate
plots are given for each membrane at each different velocity. As in
previous figures the upper row of plots are for the Eastman unheat-treated
membranes; the middle row is for Aerojet 6% membranes and the bottom row
of plots is for the Eastman membrane heat-treated to 90°C. Fluxes before
and after shutdown are given in Table 8.
Table 8. Effect of Loop Shutdown on Flux (Run 33, 600 psig)
Shutdown No.
Run Time, hr
Duration
9
Axial
Membrane Velocity
ft/sec
Eastman
Heat-treated
Aerojet 6%
Eastman
Unheat - 1 r e at e d
21.8
8.0
3.3
21.8
8.0
3.3
21.8
8.0
3.3
32 hr
before
1.7
2.2
1.7
32
26
k.6
k2
25
5.1
1
8k
1/2 hr
2
158
1 min
Flux, gal. /ft2 <
0.8 hr
after
1.6
2.2
1.7
31
23
11
k2
27
11
1 hr 6 hr
before after
1.6
2.1
1.^
29
lU
3.3
38
Ik
3.5
1.6
2.2
1.5
30
ih
3.. 3
39
13
3.5
•day
3
286
I1-!- hr
9. 5 hr 1 hr
before after
1.6
2.2
1.5
26
7
1.8
39
7.1
1.9
1.6
2.3
l.k
26
12
2.0
36
11
2.1
The first shutdown (of 9-5 hours duration) occurred after 8^ hours of
sewage circulation when the city power failed during a thunderstorm. When
the loop was restarted, there was no significant change in flux for six
of the membranes, a small increase for one membrane and appreciable
increases in flux for two membranes. The greatest increase was observed
with the two higher flux membranes exposed to the lowest velocity, well
below what we now know is the threshold velocity for these membranes.
However, at the time of the shutdown, flux from both membranes had
declined from an initial value of ~50 gal. /ft2»day to ~4 gal./ft2»day
and since the flux after startup was only ~10 gal./ft2«day the recovery
is largely of academic interest.
The second outage of the test, caused by a momentary power failure during
an electrical storm, occurred after 158 hours of sewage circulation and
lasted only about 1 minute. This shutdown had little or no effect on the
flux.
The third shutdown, also caused by a power failure during a storm,
occurred after 286 hours of sewage circulation and lasted ik hours. This
shutdown had no effect on the lowest flux membranes at all three velocities
or on the two higher flux membranes at the highest velocity, 21.8 ft/sec.
31
-------
ORNL DWG. 71-13057
LO
100
SHUTDOWNS 9 1/3 hr 1 min
0
100
200
TIME (hr)
14 hr
300
9 1/3 hr 1 min
400 0
100
200
TIME (hr)
14 hr
300
9 1/3 hr 1 min
400
100
200
TIME (hr)
14 hr
300
400
Figure 5. Flux Recovery After Shutdowns as a Function of Axial Velocity and Initial Membrane Flux.
(Primary Sewage Feed, 600 psig, Cellulose Acetate Membrane, Run 33)
-------
The greatest effect was observed with the two higher flux membranes at
the intermediate velocity, 8.0 ft/sec. With both membranes the flux had
declined from ~6o gal. /ft2»day to ~7 gal. /ft2 »day before the shutdown.
After the shutdown the flux increased to ~11 gal. /ft2 «day, again an
amount which is largely of academic interest.
From the results of this and other tests we believe that the principal
effect of a depressurization is to loosen heavy layers of particulate
fouling sufficiently that at least a portion of them can be flushed from
the system. Because the Aerojet and Eastman unheat-treated membranes at
3.3 ft/sec had such a low flux after the first shutdown, the axial velocity
was sufficiently near "threshold" that little additional particulate
fouling occurred. Hence subsequent shutdowns had little effect on their
flux.
Intermittent Application of Chlorine
Run 26 (Continued). In Run 26 membranes with initial fluxes of ~120,
~50, and ~5 gal. /fta «day were exposed to primary sewage feed at axial
velocities of 3.1, 10, and 17.6 ft/sec. During the first 300 hours the
flux decline of the two higher initial flux membranes at the higher axial
velocities appeared to stabilize at a rather modest value. However, from
~300 hours to 600 hours a severe flux decline occurred with these membranes
(see Figure 3).
After 622 hours of sewage circulation, HTH, a commercial chemical compound
containing 70% calcium hypochlorite, was mixed with the primary sewage feed
stream to give a concentration of about 100 mg/4 calcium hypochlorite and
was used as the feed for about one hour. This was done to determine
whether part of the deposit that had formed on the loop surfaces could be
removed. The treatment was repeated four hours later. The immediate
effects on the flux are summarized in Table 9 which shows flux data taken
^5 minutes before the start of the treatment and those taken one hour
after it was completed. Longer-term results are shown in Figure 6 which
is a plot of flux as a function of total run time with the time of the
calcium hypochlorite addition indicated by a dashed line and the notation
"add HTH."
The effectiveness of the treatment of the flux appeared to be velocity
dependent. Little or no change occurred for membranes operated at 3 ft/sec
velocity. For membranes at 10 ft/ sec velocity the flux increased by 28
to 75$. For membranes at 17.6 ft/sec the flux increased 36 to 317%.
Rejections decreased only a few per cent except for the unheat-treated
Eastman membrane operated at 17.6 ft/sec which dropped from 86.^4 to 65.7$.
This membrane had the greatest flux increase, from 8.39 to 35.0 gfd.
The run was continued for about lUo hours to observe the effect the
calcium hypochlorite treatment had on the flux decline. During this
period the flux decline was mild for all nine membranes as can be seen
in Figure 6. For the Aerojet and the unheat-treated Eastman membranes
at the two higher velocities the severe flux declines observed during
33
-------
OKNL DWG. 71-4702
n
T3
in2
2
10
5
2
1
u = 3 ft/sec
ADD HTH1
N^4ijta^^
2
0
2
1
c
V —
^v
Kv,
^=ac
=
ooo-i
-too
ADD
ao-c
HT
H —
}
\
-\
H —
J 0
|u = 17.6 ft/sec |
0 200 400 600
TIME (hr)
800 0
200 400 600 800 0
TIME (hr)
200 400 600
TIME (hr)
800
Figure 6. Effect of Flux and Axial Velocity on Flux Decline Using Primary Sewage as Feed.
(Cellulose Acetate Membrane, 600 psig, Run 26)
-------
Table 9. Effect of Calcium Hypochlorite Wash on Flux (Run 2.6,
600 psig, CaCl(OCl) Added After 62k Hours Total Run Time)
Flux, gfd
Velocity
ft/sec
3.1
10
17-6
Initial
39
6
116
k
56
^4-5 Minutes
Before Wash
0.85
0.77
0.73
k.6
7.6
2.8
8.U
10.1
One Hour
After Wash
0.85
0.76
0.73
7.0
13.3
3.3
35.0
18.3
lUd Hours
After Wash
0.76
0.71
0.72
6.5
12.1
3.3
29.9
16.6
the 200 hours preceding the treatment did not occur but rather the
declines were similar to those which occurred between 50 and UOO hours
of circulation. The fluxes after the calcium hypochlorite wash had
recovered back to the extrapolate^ flux from the 50 to ^00 hour portion
of the run.
Figure 7 shows the membranes after they were removed from the test sections
at the conclusion of the run. Membranes from the low velocity test section,
shown on the left side of the figure, were coated with a heavy, soft sludge.
The quantity of deposited material decreased as the axial velocity increased.
The unheat-treated Eastman membrane at 17.6 ft/sec velocity had essentially
no solids on the surface. Note that the membrane with an initial flux of
~50 gal./ft2»day exposed at an axial velocity of 10 ft/sec was only par-
tially covered with a continuous deposit; the uncoated part of this mem-
brane, however, was speckled with small deposits of sewage particulates'
elongated in the direction of flow. An enlarged photograph of this mem-
brane is shown in Figure 8. The appearance of this membrane suggests that
some sort of nucleation phenomena may play a role in the adherence of
particulate solids to cellulose acetate membranes.
Run 28. The objective of this test was to determine whether addition of
calcium hypochlorite to the primary feed at approximately 250-hour intervals
would prevent the severe flux decline which was observed after ~UOO hours
of operation in Run 26.
Membranes used in this test were Eastman HT-00 unheat-treated (jj = initial
flux ~150 gfd), the same type membrane heat-treated at 90°C (jj ~ 5 gfd),
and Aerojet-General 6$ (jj ~ 50 gfd). System pressure was maintained at
600 psig. Temperature of the circulating fluid ranged between 25.5 and
31.8°C. Axial velocities in the test sections were 3.1, 10, and 17.6 ft/sec.
Total run time was 983 hours; two shutdowns occurred one after 209 hours
with a duration of 7 1/2 hours, the other after 890 hours with a duration
of lU 2/3 hours.
35
-------
PHOTO 78674
u (ft/sec)
120
50
10
17.6
Figure ?. Membranes from Run 26 After ?68 Hours Exposure to Primary Sewage Feed.
-------
PHOTO 79324
OJ
Figure 8. Membrane with Initial Flux of 50 gfd Exposed to Primary Sewage Effluent
Feed for 768 Hours at an Axial Velocity of 10 ft/sec. * ""
(Run 26, 600 psig, smallest scale division = l/l6-in.)
-------
City water spiked with NaCl to ~0.01 M was circulated for the first
2 1/U hours of the run to determine the initial flux and rejection for
each membrane before the feed was changed to primary sewage effluent.
Flux and rejection data obtained 30 minutes after the system was started
are given in Table 5.
Flux as a function of time is shown in Figure 9 as solid points. For
comparison purposes the results from Run 26 are also shown on this figure
as open points.
The loop shutdown which occurred after 209 hours was thought to have
resulted from a momentary city power outage during a thunderstorm. The
system was restarted 7 1/2 hours later. Data taken 1/2 hour after the
system was restarted showed modest flux increases for all three unheat-
treated Eastman membranes and for the Aerojet membrane at the lowest
velocity. A day later these fluxes had decreased to about the same values
as observed before shutdown. The shutdown did not appear to have much
effect on membrane rejection.
After 231 hours of sewage circulation, the loop was washed with calcium
hypochlorite for about one hour by adding calcium hypochlorite directly
to the primary sewage effluent feed to give a concentration of nearly
500 mg/jJ calcium hypochlorite. The treatment appeared to have little or
no immediate effect on the fluxes or rejections'. This was anticipated
since the wash was made before the time that sharp flux declines were
observed during Run 26. Flux from all membranes appeared to be stable
during the next 200 hours.
The second calcium hypochlorite wash, made after 5^2 hours, had no
immediate effect on the flux of the heat-treated membranes. The Aerojet
membrane fluxes increased slightly — from 26 to 28 gfd and from 23 to 25
gfd from membranes at 10 and 17.6 ft/sec, respectively. Larger increases
occurred for the unheat-treated Eastman membranes at these two velocities
with fluxes increasing from 25 to 32 gfd and from 39 to 52 gfd. The sharp
flux declines noted in Run 26 for these four membranes during this period
were not observed during this run.
The third wash was made after 758 hours in an attempt to restart two feed
pump heads that had stopped working. The calcium hypochlorite concentra-
tion in the system was the same as for the other two washes but the rate
of fresh feed addition (and letdown) was'reduced to 1/3 of the prior rate.
Treatment time remained the same — one hour. There was no effect on the
pumps and flux data taken 70 hours later indicated little or no increase
in fluxes.
For the heat-treated Eastman membranes operated at 17.6 and 10 ft/sec very
small flux declines occurred between one hour after the start of sewage
effluent circulation and 9^5 hours. Fluxes decreased from 2.8 to 2.6 gfd
and from 3.1 to 3.0 gfd, respectively. For the membrane at 3.1 ft/sec the
flux decreased slowly from 3.7 to 2.7 gfd at 308 hours. The rate of
decline then increased so that the flux was only 0.89 gfd at the end of
the test. The break in the time-flux curve occurred after a longer time
38
-------
ORNL DWG. 71-13058
VD
2
102
5
lu = 3 ft/sec I
t
-------
period than in Rim 26. Rejections increased slightly with time for all
of the membranes going from 93-96$ to
Calcium hypochlorite treatment had no effect on the intermediate and
high flux membrane exposed at an axial velocity of 3 ft/sec. However,
at axial velocities of 10 and 17.6 ft/sec there was a dramatic effect
of the intermittent chlorine addition on the flux from both the inter-
mittent and high initial flux membranes. After the initial flux decline
in the first 30-^0 hours flux decline was stabilized at a very small value
for the remainder of the run by the chlorine additions at 231, 5^2, and
758 hours. With the Aerojet membranes, chloride rejections were better
than 95$ for the entire 950 hours indicating that, at least for this
period of time, the brief exposures to chlorine did not damage the
rejection capability of the membranes.
Although the amount of chlorine, the length of time between flushes and
the duration of flushes was not optimized in these studies, it is clear
that unless the axial velocity is above the threshold velocity for parti-
culate fouling the use of an intermittent chlorine flush is totally
ineffective in restoring flux. With proper hydrodynamics, intermittent
chlorine flushes, at least in this one test, prevented an expected
catastrophic flux decline in the period from UOO to 950 hours run time.
Modification of Membrane Characteristics by Additives
Several exploratory studies have indicated that modification of membrane
characteristics (i.e., charge on the membrane) may offer protection
against fouling since most polymers in sewage are probably anionic in
nature.
Kraus (5) reported qualitative observations supporting this hypothesis in
studies of the application of dynamic membranes to treatment of sewage.
Specifically it was noted in those studies that although rejections of
cation-exchange membranes were decreased on exposure to sewage, the
decrease was not as serious as that observed on exposure of anion-exchange
membranes to sewage.
The Environgenics Company (10) developed procedures for attaching enzymes
to cellulose acetate membranes. Although enzyme activity was markedly
reduced in the process, the activity remaining was significantly stabilized
by attachment of the enzyme to the membrane. No tests were reported of
the characteristics of this membrane in the presence of sewage. However,
the cellulose acetate hydrogen succinate (CAHS) used in preparing the
enzyme membrane was tested in contact with secondary effluent. The first
six days of the test indicated no change in flux while rejection increased
from 80 to 87$. From 6 to 8 days the flux decreased from 15.5 to 13.5 gfd
and then to 6 gfd after l6 days. The complete absence of flux decline in
the first 6 days was taken to be an indication that the negatively charged
CAHS may have had an inherent fouling resistance.
Keilin (lU) reported promising results in improving salt rejection with
polyvinylmethyl ether (PVME). Interestingly, the additive was reported
40
-------
to increase rejection with only a slight reduction of the flux. However,
the effects of the PVME decayed over a period of about one day after the
ether was flushed from the system. Also, it was reported (15) that PVME
will precipitate from dilute solutions at a temperature above about 33°C.
A secondary objective of the present program was to determine the effect
on flux and rejection, if any, of a pretreatment of the membrane with
additives. Two additives were tested: polyvinylmethylether and poly-
acrylicacid; the first with the objective of increasing the rejection of
the unheat-treated high flux membrane, the second with the objective of
developing a charge on the membrane to minimize adherence of foulants.
The results of each test will be described individually.
Run 19. Axial velocities for-this test were 5.2, 9.9, and 15.3 ft/sec,
feed was primary sewage effluent. Temperature of the circulating fluid
varied between 30 and 32.5°C and the system pressure was maintained at
600 psig.
During the first 19 hours a 0.01 M solution of NaCl containing 100 mg/4.
PVME was circulated for pretreatment with the letdown and product streams
being returned to the feed. During the pretreatment the flux from the
Aerojet membranes decreased (at all velocities) from 37-^9 to 26-28 gfd
and rejection increased from 77-79 to 95-97$. The flux and rejection for
the Eastman membrane, operated at an axial velocity of 15.3 ft/sec,
remained constant at about 120 gfd and 1%, respectively. At 9.9 and 5.2
ft/sec velocities, the rejection increased from a 31-^5$ range to 78-81$
range while the fluxes decreased from 80-87 to 23-35 gfd. Most of the
change in the rejections occurred during the first two hours of circula-
tion. Although the increased rejection is of interest, the fact that the
flux declined markedly at the same time means that the result is of little
practical interest.
Because flux through cellulose acetate membranes varies with applied
pressure and viscosity, measured fluxes for this test were corrected for
small variations in temperature and pressure by the relation
where J = normalized flux,
s
J = measured flux,
P = system pressure,
(j. = viscosity of water at system temperature, and
|j.30 • = viscosity of water at 30°C.
For the Aerojet membranes only modest flux declines were observed during
the 215-hour run at velocities of 9.9 and 15.3 ft/sec (from 26 to 22 gfd
and from 26 to 21 gfd, respectively). Rejections were greater than 87$,.
There appears to be a slight offset in the flux decline curve at ~35 hours
41
-------
particularly for the results from the membrane operating at 9. 9 ft/sec.
At a velocity of 5.2 ft/sec a severe flux decline occurred with this
membrane, from 27 to k. 5 gfd, while the rejection remained fairly con-
stant at 97%.
For the Eastman membrane at a velocity of 9.9 ft/sec, flux increased
from 36 to ^3 gfd during the first 30 hours on sewage and then decreased
to 2k gfd by the end of the run. Rejection decreased from 83 to 41% "
during the first 30 hours and to 3k% by the end of the test. At a
velocity of 15.3 ft/sec, flux decreased from 128 to 80 gfd during the
215-hour run with little or no discontinuity at 30 hours after addition
of primary sewage (in contrast with the bahavior at 9-9 ft/sec). Rejec-
tions were less than 20% for the entire run. At an axial velocity of
5.2 ft/sec there was a severe flux decline during the run, from 2k to
2.6 gfd and rejection decreased from 90 to
It is interesting to note that of the six different combinations of
initial flux and axial velocity, only the results from the membrane with
an initial flux of ~80 gfd and axial velocity of 9. 9 ft/sec resembles
the brackish water results of Keilin (1*0. That is, the flux increase
from 31 to k5 gfd and rejection decrease from 83 to ki% in the first 30
hours apparently corresponds to Keilin' s observation that the effect of
the PVME disappeared ~1 day after the PVME was flushed from the system.
Apparently with primary sewage as feed there is an optimum velocity
(not necessarily 9-9 ft/sec) required to obtain the benefit of the PA/ME
membrane treatment. In none of the tests was there a negligible effect
of PVME on flux.
Run 20. All membranes used in this test were soaked for 65 hours in water
containing 100 mg/^-PVME prior to installation on the test unit. The tem-
perature during pretreatment was maintained below 30°C to prevent precipi-
tation of the PVME. This pretreatment procedure was used to permit the
ether to diffuse into the membrane and perhaps prolong the effect of the
PVME.
Axial velocities were U. 75, 9> and 12 ft/sec; the temperature of the
circulating fluid varied between 28.5 and 31. 5°C, and system pressure was
600 psig. City water spiked with NaCl (~0.01 M) was circulated for the
first 18 hours. During this time fluxes for the Aerojet membranes
remained essentially constant at ko to 60 gfd except for the last measure-
ment at the 9 ft/sec velocity which showed an increase to 182 gfd. The
rejection for this membrane dropped from 77 to 16% indicating that a
defect had developed in the membrane. Rejection for the membrane at 12
ft/sec was constant at 92% while at a velocity of ^.75 ft/sec it decreased
from 89 to 58%. For the Eastman membranes flux was about l8o gfd at
velocities of 9 and 12 ft/sec while it decreased from 18*4- to 86 gfd at
k. 75 ft/sec. Rejections of the Eastman membrane was less than 15% at all
axial velocities.
After changing to primary sewage mild flux declines were observed for the
Aerojet membranes at 9 and 12 ft/ sec during the next 215 hours (37 to 27
42.
-------
gfd at 9 ft/sec and 39 to 33 gfd at 12 ft/sec). At U.75 ft/sec, flux
dropped from 36 to 2.6 gfd during the 215-hour test. At the two higher
velocities rejections ranged from 92 to 97$. At the low velocity it
decreased from 92 to
For the Eastman membrane a severe flux decline occurred at k. 75 ft/sec
(from 36 to 2 gfd in 215 hours). At an axial velocity of 9.9 ft/sec there
was a mild decline during the first 79 hours on sewage followed by a
sharper decline to the 215-hour point (from 8l to kk gfd and finally to
20 gfd). The flux data from the membrane at 12 ft/sec velocity were
scattered with higher fluxes measured in the mornings than in the after-
noons. Turbidity of the feed did not show this type of variation. During
the 215-hour period the flux varied between 29 and k8 gfd with an increasing
trend. Rejections remained less than 30$ during this period for all Eastman
membranes.
There were three flow interruptions after 215 hours of sewage circulation
and before the conclusion of the test at 322 hours. In general fluxes
were higher after a startup than prior to the stoppage.
Comparison of the results from the run in which the FVME was circulated
past the membrane for l8 hours with results from the run in which the
membrane was soaked for 65 hours in a FVME solution suggests that circu-
lation of the pretreating solution is necessary for the development of
appreciable rejection by the unheat-treated membrane. At 10 ft/sec
circulation velocity, pretreatment of the unheat-treated membrane with
the circulating FVME solution for 18 hours developed a rejection of
for the nominally zero rejection membrane. (However the flux decreased
from 80 to 23 gfd in contrast to Keilin's observation of negligible effect
of FVME on flux.) This rejection decayed to ~Ul$ in the first 30 hours
after FVME treatment and the flux increased slightly from 36 to ^3 gfd
during this same period. Soaking the unheat-treated Eastman membrane in
FVME had little effect on the rejection for any of the velocities tested.
Run 2k. The object of this run was to investigate the effect on flux and
rejection of long-term soaking of membranes in polya'crylicacid (PAA) a
negatively charged polymer and.to determine the change of flux with time
for circulation periods in excess of the normal 200-hour test. Membranes
used for this test were Aerojet-General 6$ and Eastman HT-00. Membranes
were secured to the test fingers and placed in a solution containing 100
ppm FAA for 31 days.
Axial velocities for this test were 5.2, 9.9, and 15 ft/sec. Temperature
of the circulating fluid ranged from 22.5 to 27°C. System pressure was
maintained at 600 psig. However, some pressure transients, < 5 seconds
duration, occurred during the run. System pressure would drop 200 psi,
then rise 250 psi, and finally return to normal. The transients were
caused by erratic operation of check valves in the feed pump.
City water spiked with NaCl (~0.01 M) was circulated for the first 2 1/2
hours. Reference fluxes for the Aerojet membranes ranged from ^7 to ko
gfd with rejections of 79 to 93$. The membrane having the highest flux
43
-------
had the lowest rejection. For the Eastman membranes initial fluxes varied
from 122 to 1^7 gfd with rejections of 10 to 19$. Flux and rejection
increased with increasing axial velocity. The PAA appeared to have little
or no effect on initial flux or rejection for either membrane.
Primary sewage effluent was circulated for a total of 752 hours. Two
flow interruptions occurred during this period. A lU-hour shutdown
occurred after 296.5 hours of circulation when the fuses in the air
compressor motor circuit opened. Then a second shutdown (^5 minutes
duration) was caused by one of the pressure excursions tripping the low
pressure cutoff switch.
For the Aerojet membrane at 15 ft/sec velocity the flux decline was small
(b = 0.055) for the initial period with the flux decreasing from ho to 28
gfd between 1 and 283 hours, just prior to the first loop shutdown. The
membrane at 9.9 ft/sec velocity also had a mild flux decline (b = 0.083)
during the first 187 hours (Ul.5 to 30 gfd) and then the flux decreased
to 25 gfd before the shutdown. Rejection increased slightly with time
for both membranes (95 to 96$, and 92 to 95$, respectively). For the
membrane at 5.2 ft/sec the flux dropped sharply (b = 0.565) from Uo to
1.7 gfd before the shutdown. As the flux decreased the rejection dropped
from 9^- to J0% presumably due to 'increased concentration polarization as
the thickness of the foulant layer increased with time.
Eastman membranes at 15 and 9.9 ft/sec velocities showed discontinuities
in the slope of the flux-time curves at about 187 hours. At 15 ft/sec
flux decreased from 88 to 70 gfd during this period (b = 0.060) and then
to 57 before shutdown. Rejection remained nearly constant at about
At 9-9 ft/sec the rate was larger (b = 0.13) with flux dropping from
to 28 gfd in the first 187 hours and then to 21 gfd before shutdown.
Rejection decreased with time from k-2 to 3^$. A severe flux decline
(b = 0.565 occurred for the membrane at 5.2 ft/sec from 31 to 1.7 gfd
just before shutdown). Rejection decreased from 50 to 38$ during this
period.
Based on the combined results of other runs without additives the PAA soak
did not appear to have an effect on either the flux or rejection for
either membrane during this 287-hour time period. Since the membrane
had been soaked in PAA for 30 days prior to the test, any beneficial
effect of the PAA was expected to disappear ~30 days after exposure to
circulating feed as the PAA diffused out of the membrane.
Additives for Sewage Floe-Size Control
The results of many tests made with primary sewage effluent from the Oak
Ridge East Sewage Plant showed that for each type of cellulose acetate
membrane tested there was a critical or threshold axial velocity below
which severe flux declines occurred (see Correlation Section, this report).
The threshold velocity is a function of the initial flux, Jj (jj = flux
30 minutes after pressurization with city water feed). The threshold
velocity may also be a function of the floe size in the circulating fluid.
44
-------
Therefore, to study the effect of floe size on threshold velocity and on
the flux decline parameter, b, flocculating agents must be added to the
feed stream.
Jar tests were made using several additives to observe the type of floe
formed and to determine the required concentration. Based on the results
of the jar tests five runs were attempted with different additives:
ferric chloride, Cat Floe (a commercial organic cationic polymer used
in sewage treatment), alum, Aqua Wuchar A and Ferri Floe [commercial
Fe2(S04)3J. Because of experimental difficulties with the addition
system, addition of flocculant was often intermittent and tests were
often of short duration. Results of the different tests will be briefly
reviewed in the following section.
Run 29. The objective of this run was to determine the effect on flux
and rejection of the addition of 70 mg/j£ Fe as ferric chloride (FeCl3)
to a primary effluent feed. Membranes used were Eastman HT-00 unheat-
treated, similar material heat-treated at 90°C, and Aerojet 6%. Axial
velocities in the test sections were 3-lj 10, and 17.6 ft/sec. System
pressure was maintained at 600 psig. Temperature of the circulating
fluid varied from 31.5 to 33.5°C.
During the initial part of the test while using city water feed only NaCl
was added to the loop. When the feed was switched to primary sewage
effluent, Nad and FeCl3 were added. This was done by pumping a 2 M NaCl
and 0. 2 M FeCl3 solution into the circulating stream at a rate of about
one liter per hour. This concentrated solution (pH 1. 2) had a high
corrosion rate for the type of stainless steel used in the addition
system. A leak developed in the tubing 12 hours after the start of the
addition. The addition system was isolated and repaired without depres-
surizing the loop. After operating for about 35 hours with no additives,
FeCl3 solution was again added for about 12 hours before another leak
developed. The run was then terminated after a total of 71 2/3 hours
of circulation.
Since the flocculant was added for only the first 12 hours and then for
another 12 hours later in the run, the flux and rejection data are useful
mainly as indicators of the possible effect of floe size during a short
period of operation.
Table 11 shows flux and rejection data for each membrane taken after
5 3/U hours of sewage circulation and data from the previous run (Run 28,
which contained no flocculating agent) taken after a similar time period.
The most notable effect occurred with the two higher flux membranes
exposed to the lowest axial velocity, where in each case the flux with
FeCl3 additive was ~3 times greater after 5 3 A hours exposure to sewage
than in a similar run in which there was no FeCl3 addition.
45
-------
Table 10.. Comparison of Flux and Rejection in Run 29 Using FeCl3
with Flux-and Rejection in Run 28 when no Additives Were Used
(Primary Effluent Feed, 600 psig Pressure)
„___. _n Run 29, Flocculant
Membrane
Eastman HT-00
Heat-treated 90°C
Aerojet 6%
Eastman HT-00
Unheat-treated
J1.A.J.CIJ.
Velocity
ft/sec
3.1
10
17.6
3.1
10
17.6
3.1
10
17.6
Flux,
JI
15
3.6
77
57
76
168
204
193
gfd
5 3/4
Hr on
Sewage
3.4
3.3
3.2
31
32
30
62
87
Robs*
95
98
99
84
92
95
28
20
29
Run 28, No Flocculant
Flux,
JI
10
8.1
2.8
59
49
4i
155
197
178
gfd
5 /34
Hr on
Sewage
3.6
3.1
2.6
10
48
35
12
43
108
R?bS'
96
98
99
88
95
95
48
33
31
As will be discussed in detail in the "Correlation of Results" section,
for primary sewage effluent with no additives the threshold velocity, uc,
(in ft/sec) is related to the initial flux, Jj, (in gal./ft2 »day) by
uc = (1.3 ±0.6)(J];)1/2 .
The preliminary results given in Table 11 indicate that, by adding a
suitable flocculant, it may be possible to reduce the numerical coefficient
in the equation above from 1.3 to ~0.5, i.e., for an initial flux of 50
gal. /ft2«day, the threshold velocity may be reduced from ~9-5 ft/sec to
~4 ft/sec.
Run 30. The objective of this test was to determine the effect of Cat Floe
(an organic cationic polymer) on the flux decline using primary effluent as
feed. Although jar-tests indicated that the rate of flocculation with this
additive was slower than with FeCl3, the pH of the concentrated additive
solution was 7.5 compared with 1.2 for the FeCla. The higher pH was
expected to ease corrosion problems with the addition system. Fluxes at
various times during the 192-hour test are given in Table 4. Comparison
of fluxes from this test with results from earlier tests with no additive
indicated that the Cat Floe had little or no effect on the rate of flux
decline. In all probability this is a reflection of the much slower rate
of flocculation observed in the jar scale tests.
The test sections were removed at the end of this test before the loop was
flushed. The inside of the pressure jacket was coated with thick, dark
46
-------
deposit, much heavier than was found after 900 hours of operating with
no flocculant. The membranes at the lowest velocity were heavily coated
with a soft sludge. Membranes at the higher velocities had hard film
over parts of the surfaces with nucleation spots for heavier deposits
evident. The loop piping also appeared to have a heavier coating of
sludge than is normally seen.
Run 31. The objective of this run was to determine the effectiveness of
adding 50 mg/4 Al in the form of commercial alum [A12(S04)3 containing
17 wt % A1203] to the primary effluent feed. Membranes used were Eastman
HI-00 unheat-treated (new batch), similar material heat-treated at 90°C
for 5 minutes, and Aerojet 6%. Axial velocities in the test sections
were 7. ^, 12, and 2h ft/sec. System pressure was maintained at 600 psig.
Circulating fluid temperature increased from 32.5 at the beginning to
39-5°C at the end of the 58-hour test. This is a much larger increase
than normal and indicates solids deposition on the sewage side of the
heat exchanger.
Continuous alum, additions were started when the feed was switched from
city water to primary sewage. Flux and rejections for this test are
given in Table 5.
Severe flux declines were observed with both the high flux membranes at
all velocities in this test. When the test sections were examined at
the end of the run a heavy deposit was observed on all surfaces. The
finger that operated at 7. ^ ft/sec had the heaviest deposit as is normally
observed. The other fingers were also completely coated but the deposit
appeared more dense, a probable explanation for the lower fluxes for
these membranes. Spectrographic analysis of the deposit indicated that
it was primarily aluminum oxide.
Run 32. The object of this run was to test Aqua Nuchar A (100 mg/j£) as
an additive to the primary effluent feed. Membranes used were Eastman
HT-00 unheat-treated, similar material heat-treated at 90°C, and Aerojet
6%. The velocities chosen for this run were 2.9, 6.7, and 19.1 ft/sec.
System pressure was maintained at 600 psig. Temperature of the circu-
lating fluid varied from 32.5 to 35°C.
The Aqua Nuchar A addition was started with the start of sewage circu-
lation. It was added to the 2 M Nad solution that is used for the 01"
ion addition. The Aqua Nuchar formed heavy floes in the addition solution
which were dense enough to prevent reading the rotameter used to monitor
this flow. Flow through the 1/U-in. tubing was not fast enough to prevent
the floes from settling. Within the first hour, flow through the addition
system was blocked. The lines were cleaned out without changing the
pressure in the circulating system. Flow was restarted but continued for
only a short period of time. Further attempts to unplug the lines and
maintain flow were unsuccessful. The run was terminated after Vf hours
of circulation.
47
-------
Runs 3^ and 35. The objective of these tests was to determine the
effect of ~50 mg/j£ Ferri Floe [commercial Fe2(S04)3] on the rate of
flux decline using primary sewage effluent as feed. It was hoped that
switching from the chloride to the sulfate would alleviate the severe
corrosion problems experienced with the addition system in Run 29.
Although there were no piping or tubing failures in these tests, both
were terminated by failure of the chemical addition pump diaphragm.
Examination of the diaphragm after each test showed that they both
failed by a pitting type attack.
Flux and rejections for these two very brief runs are summarized in
Table 5. During each test there were equipment failures or addition
lines plugged every night so that Ferri Floe addition was not continuous.
Despite this, the flux from the membrane with an initial flux of ~50 gfd
was at least twice as large as the upper limit defined by the scatter
band on Figure 11. This supports our belief that increasing the floe
size with a suitable additive has promise as a means for markedly reducing
the threshold velocity for particulate fouling. Of course the economic
attractiveness of this possibility depends on whether the reduced pumping
costs due to lower velocity offset the additional chemical costs. Further
studies are required to determine optimum, additive and additive concentra-
tion for maximum reduction in threshold velocity.
Composite Results
The primary objective of this program was to determine the effect of
hydrodynamics on the performance of cellulose acetate membranes. The
principal hydrodynamic factors studied were the axial velocity (3 to 30
ft/sec) and the velocity normal to the membrane surface (i.e., the membrane
flux) which had a range of from ~0.01 cm/min (~3 gfd) to ~0.^ cm/min (150
gfd). In this section of the report, results from all runs made in the
present study are combined into composite figures.
Flux Decline. In prior studies a distinction was often made between an
initial rapid flux decline followed by a substantially smaller rate of
flux decline. The effect of initial flux and axial velocity on the flux
decline in the first 20 hours after exposure of the membrane to primary
sewage feed is illustrated in Figure 10. The three parts of the figure
show results for membranes with initial fluxes, Jj, (jj = flux 1/2 hour
after startup with city water feed) of ~9> ~^6, and ~130 gfd. Because
initial fluxes varied somewhat from run to run, the flux after 20 hours
exposure to primary sewage feed, J2o, was normalized by dividing by the
initial flux. The dashed lines on these figures are drawn to include
the bulk: of the experimental results (Tables 5 and 6) for each set of
conditions; as such they define the variability of results due to natural
variation of the primary sewage feed.
Above a certain velocity range, which was different for each initial flux,
the range of flux declines in the first 20 hours is essentially independent
of the axial velocity. For a membrane with an initial flux of ~9 gfd,
the threshold velocity is h to 5 ft/sec; for an initial flux of ~46 gfd
48
-------
ORNL-DWG 71-5680
1.0
0.5
0.2
0.1
0.05
0.03
f
f
/
0
/
L}
f
t
\/
JT "
^t
•
/
•0-
-9 c
Jd
•
1
n
MH
/I
•
^-•75:
v — v-
fl
day
_ _
A
,
/
r~
t
5 /
/
i
f
i
/
i
«>
4(
j
I
/
j
t
t
7
i^
^
t
\
1
g
PB
I
/
a
z
3 TM » X
E- _ZLU
l/ft2-da\
l
==b
^
«
^•^^
/
••
7^
/
y
/
/
}
(_,
/
r
1
/
7
n u /
fly
'
/
L
i
r
/I
-1
>
'l
y
2
/l
d
4
*
>c
— ^=-
/ \ ¥
^nfA'
' /
7A /
V
i D
) gal/ft
— A.
=^
2-da
r-
y-
10 20 30 2 5 10 20 30 2
AXIAL VELOCITY (ft/sec)
10 20 30
Figure 10. Effect of Axial Velocity and Initial Flux (jj) on Flux Decline in the First 20 Hours
After Exposure to Primary Sewage Feed. (Cellulose Acetate Membrane, 600 psig Pressure)
-------
the threshold velocity is 8 to 10 ft/sec and for an initial flux of —130
gfd the threshold velocity is 15 to 20 ft/sec.
Below the threshold velocity the range of flux declines in the first 20
hours is strongly dependent on axial velocity. Some initial flux declines
in the first 20 hours were so severe that the flux was only k to 5$> of
the initial value.
The extent of the range of flux declines above the threshold velocity was
surprisingly similar for the three different initial fluxes; for an
initial flux of ~9 gfd, 0.6h < (J2o/Jj) < 0.96; for an initial flux of
~U6 gfd, 0.58 (Jso/Jj) < 0.86 and for an initial flux of -130 gfd, 0.61+
< (J2o/Jl) < 0.86. Additional studies are required to determine factors
responsible for the extent of this range and to develop means for insuring
that results nearer the upper limit can be achieved routinely.
Flux after 20, 100, and 200 hours exposure to primary sewage feed is shown
as a function of axial velocity in Figure 11 for the Aerojet 6ff0 permeation
membrane. As in the previous figure there is a threshold velocity above
which the axial velocity had little effect on the range of fluxes observed.
The threshold velocity for the first 200 hours exposure was substantially
independent of exposure time. For this membrane, having an initial flux,
J"l, of ~50 gfd, the threshold velocity ranged between 8 and 10 ft/sec.
For axial velocities above the threshold there was a definite flux decline
during the first 200 hours, from 35 ± 5 gfd at 20 hours to 27 ± 5 gfd at
200 hours. It is notable that this is about the same percentage range of
fluxes as in the previous figure.
Membranes exposed at velocities above the threshold showed little or no
deposit when removed from the loop. For example, Figure 12 is a photograph
of membranes removed from the loop after 750 hours exposure in Run 2U.
Although the membrane with an initial flux of —50 gfd exposed to an axial
velocity of 9- 9 ft/sec was right at the threshold condition, there are
only a few speckles on its surface. Nevertheless the flux from this
membrane had declined from ^7 to 19 gfd during this time. In all proba-
bility this flux decline is due to either soluble or colloidal organic
material and not to gross particulate fouling.
Figure 13 is similar to Figure 11 except that the results were obtained
with a membrane having an initial flux of —130 gfd. With this membrane
the range of threshold velocities was 1^ to 20 ft/sec, again substantially
independent of exposure time for times up to 200 hours. The flux declined
from 90 ± 10 gfd at 20 hours to 60 ± 10 gfd after 200 hours, an average
decline of 33% compared to 23% with membranes having an initial flux of
-50 gfd.
Observed Rejections. In previous sections of this report all rejections
referred to were chloride rejections. Routinely samples were taken at
—20 and —200 hours run time with primary sewage feed for more complete
analysis. Results for chloride, turbidity, organic carbon, Ca++ plus Mg4"1",
inorganic carbon and phosphate are shown in Figures lU and 15 for the
nominal °lk% rejecting membrane and in Figures 16 and 17 for the nominal
50
-------
102
D
-O
CO
D
cn
2
1
20 hr
2
7
10 20 40 2
100 hr
/
fO
o
5 10 20 40 2
AXIAL VELOCITY (ft/sec)
ORNL-DWG 71-5482
200 hr
•
10 20 40
Figure 11. Effect of Axial Velocity and Time on Flux. (Cellulose Acetate,
Aerojet 6%, 600 psig, Primary Sewage Feed)
-------
TIAL FLUX
140 gfd
45 gfd
PHOTO 78527
VELOCITY
(ft/sec)
_—
15
Lr>
N>
9.9
Figure 12. Membranes After 750 Hours Exposure to Primary Sewage Feed.
-------
102
20 hr
D
T3
CM
O
~S"
x
10
5
/
Vc ,
7^71
ES*
X
r
100 hr
/
*
^_
y<
-d—
e-M
/
t
A
10 20 402 5 10 20 40 2
AXIAL VELOCITY (ft/sec)
ORNL-DWG 71-5480
200 hr
i
OT
10 20 40
Figure 13. Effect of Axial Velocity and Time on Flux. (Cellulose Acetate,
Eastman HT-00, 600 psig, Primary Sewage Feed)
-------
OHNL DWG. 71-13557
1.0
0.5
0.2
(O
.O
a? 0.1
i
0.05
0.02
0.01
B
O
D
A
O
e>
e
e
a
V
V
a
B
KUN
2
2
2
2
2
1
1
1
1
1
3
3
3
3
3
3
8
5
5
4
0
9
8
7
4
J
1
2
3
4
6
lr
<.
_
A
i
CHLORIDE
200 hours
L)
1
1 ;
\
A
A
2
•V
fe^_,
H
^7
[
O
V
-t
C
rURBI
200 h
*
-O
-1
n
-
\t
§
GL
DITY-
ours
i D
q
'
.
A A
*
o
]-(£«>-<;
(
»
e
a
ID"
4-
O
I
t
fc
^
_l
J
»0
;
— <
1
^
o
3
3
/
\\
V
\ i
D
^
j
1
L
n— A-
\
ORGANIC CARBON
200 hours
r
>—
A
V7
10 30 2 5 10 30 2
AXIAL VELOC-ITY (ft/sec)
10 20 40
Figure I1*. Observed Rejection with Aerojet 6% Cellulose Acetate
Membrane with Primary Sewage Feed.
54
-------
ORNL DWG. 71-13555
1.U
0.5
0.2
0.1
0.05
0.02
0.01
1.0
0.5
0.2
0.1
0.05
0.02
n m
B
O
n
A
V
O
c
o
e
e
a
A
V
a
B
RUN
28
26
25
24
23
22
21
20
19
18
17
14
13
30
31
32
33
34
36
CaH t Me
20 hour
n
a
— fc
7
Jps(
t
—i
H_
V
1
>-
•!
±
K
jt1
s
^
i
-
1 -
^
%
c
A
A
^=M
^— o
i5l ^
*
c&
* V
ol« a
'
a
>— g
I.
3
^
y
;a++ +
200 h
D
-1
<>
^i
A
B-
^
_^
++ —
IviO
ours
-o n.
1 1
B ^
d Ay
A
Is*
»
e
V
l.U
O "i
n 9
- 0.1
U.Uc
nn i
d
-a
= INC
I
-i
^i
s
3
<
?
V
i
-i
— i
)RG^
7r
^
"j
^^g"
c
1 O
O 9
0.1
O.O£
nm
-
-INORGA
U$
l<
/
r>
4
S'~~
\NIC CAR
-H 20 ho
i O
OA
a
-g^-l
™
ra a^
A
o
BON
— t-
I
e
NIC CARBON
)0 hours
kJ
"A A
p\ /^ i
vj O^
CO
LJ
A
O
3
4
0.1
nrm
U.Ud
0.01
UxXj£
-i ni
n 09
- 0.01
— ___
U.UUO
-I nnni
oise^ta
f-
B
D
T—
-U-
-
0,1t
~o~ ~
c
S--4-
-1
TT<
o
V
I
_J__
— PH
5- 2
"O"~
'b
-A-
— "-O
jNs'PHOSF
41 - - 20 f
4^
^rd A
e a a
1 x t
JHA1
lours
fitfv
•E;
1
)
OSPHATE — 1
00 hours
1
e
01
^^
-«A— A-
*
•>^
?
10 30 35 10
AXIAL VELOCITY (ft/sec)
10 20 40
Figure 15. Observed Rejection with Aerojet 6% Cellulose Acetate Membrane
with Primary Sewage Feed. (Note Scale Difference on Phosphate
Rejection)
55
-------
ORNL DWG. 71-13558,
1.U
0.5
0.2
W
Q? 0.1
1
0.05
0.02
0.01
1.0
0.5
0.2
(A
-Q
Q? 0.1
$
V
RUN
B 28
0 26
D 25
A 24
V 23
> 22
< 21
O 20
€> 19
£
_
a
rbfit
»S&
5fVi
I
HL
2C
C
0
R
no
it
S=
Ss
v
^
€
/
a
I
i
1C
u
\
13
)E
rs
jNfei
1. ,
1V
1 A
\
E
17
V> §)
Fir-?
Yv A
:i-^l
O
a
— 0-
V
V
1
^
5
P*
i
r
a.
D1h
A
T
-7
V
LI
n
0
V
NP-
<
R
2(
C
lUKt
2C
e
c
A
%
-^
R
DC
&
|c=
!
*
1
J2
BIC
) h<
^
e
'
>
3IDITY
hours
C
C
o _r^
A A
S>
|5k A
J?2^
V
A
_J
4
0
.-«
^
5
$
OR
B A n
-1ITV
3urs
C
nV
U
\
S
Kfe
V
n
~c
<
G/
•7T1
z
r
r
^
\^
i-
£
-H
-
[:1
5/
D
o-
d
>
i
^
Ji
?c
i
>
«
.
^
F
5
>
i
<2
|7
A 1
IV
c ^
A ^^
n
e
<
^^
V
*w
CARBO
hours
e
v1
A
a
n
\J
c
(
I
•»
>
0.05
0.02
0.01
10
30 2
10
30 2
10 20 40
AXIAL VELOCITY (ft/sec)
Figure l6. Observed Rejection with High Flux, Low-Salt Rejection,
Eastman HT-00, Cellulose Acetate Membrane with Primary
Sewage Feed.
56
-------
ORNL DWG. 71-13556
1.0
0.5
0.2
!i-
J^
i
' J_CP wo
Sfflef>
7W
v
Ca+++MgH
20 hours
-1--
-J
fe-
>~:
>
n
rJ
T
IN
<
V
Of-
I | I I I
-|
br
t<^
5 A"
D
J-^
^
T
;GAN c
20 h
I I
,_
L
K
tr
V
— A"
x
,
$
D
?
r
O_t VI
eL*Sjv
H "*
A ^
, 1
111
O
i!
V
A
CARBON
ours
1
ti
A
1 /\
|Jr3 TD
P
; V °
D
?
r
^
\
f
j
i <
b
7
v
i
~i
i« -
i
k
»*
V
<
F
f
V
;
V
t9*
D
-
<
^
D —
vv
a
i
3HOS
2C
^k
^ *
8a
^^
<^J / \1
-t-A
<
t
>
^
0.05
0.02
0.01
10 20 2 5 10 20 2
AXIAL VELOCITY (ft/sec)
Figure 17. Observed Rejection with High Flux, Low-Salt Rejection
Cellulose Acetate Membrane with Primary Sewage Feed.
10 20 40
57
-------
zero-percent rejecting membrane. In these figures (l — Robs) is plotted
vs axial velocity for the samples at 20 and 200 hours run time. One minus
the observed rejection was chosen as the ordinate in order to spread out
the high rejection results.
Although there is no clearly defined trend of rejection with velocity,
there seems to be some tendency for higher rejections to be observed at
higher velocities. Sheppard and Thomas (l6) have shown that thick non-
rejecting layers covering a membrane result in marked decrease in rejec-
tion due to increased concentration polarization in the fouling. Because
thick fouling layers occurred at low velocities in the present study, a
more pronounced effect of velocity on rejection was expected.
After 200 hours the median rejection and the range for the. nominal
rejecting Aerojet 6% membrane was:
_ , . . Rejection, %
Material l—L—
Median Range
Chloride 9^ 50-98
Turbidity 97 30-99
Organic Carbon 88 15-98
Ca++ plus Mg++ 95 80-98
Inorganic Carbon 88 kO-^B
Phosphate 99 93-99+
These rejections compare favorably with those reported by others as shown
in Table 1.
Some rejection developed with the nominally zero rejection membrane;
apparently a dynamic membrane formed from sewage constituents as has been
reported in previous studies. After ~200 hours the median rejection and
the range for the nominal zero-percent rejecting Eastman HT-00 membrane
was:
Material Rejection, %
Median Range
Chloride ho 10-75
Turbidity 90 ^5-99
Organic Carbon 75 25-96
Ca++ plus Mg++ 55 25-80
Inorganic Carbon 50 15-88
Phosphate 85 55-99
Although the median phosphate rejection is somewhat higher than expected,
the relatively good rejection of turbidity and organic carbon is not
surprising since much of this material i~ probably filterable and its
removal does not depend on the rejection characteristics of the membrane.
Although rejection is of interest in defining membrane characteristics,
in the final analysis the quality of the product water is the deciding
factor. The composition of the primary sewage feed is shown as a function
of time in Figure 2; run numbers are identified across the top of the
58
-------
figure. Selected product analyses are shown in Figures l8 through 21.
After 200 hours exposure to primary sewage, the product water composi
tions from the nominal $k% rejecting Aerojet 6% membrane were:
Material Concentration, ing/ 1
- Maximum Median
Organic Carbon 30 7
Ca++ plus Mg++ 30 8
Phosphate 8 0. 2
Turbidity (JTU) 10 1
With the nominal zero rejecting Eastman HT-00 membrane comparable values
were
,, , . n Concentration,
Material - — -
,
Maximum Median
Organic Carbon 65 12
Ca++ plus Mg++ 130 60
Phosphate 2h 6
Turbidity (JTU) 10 1.5
59
-------
102
5
2
10
5
ORNL DWG. 71-13059
J 0.5
I 0.2
o 0.1
2 5
o
o
^ORGANIC CARBON
£ L A
_^ DO-
B V?-"ri /
A
c — =
g g B -A-Ch— e-a
MJ' 9
0 a a *
B e
a
- 20 hr
c
:
B
)
3 —
Mq++ - 20
l-azs
-5T
-O-A-fe
O
Q
O
E 5
2
10
2
1
0.5
0.2
0.1
— ORGANIC CARBON - 200 hr-)—
D
«P
A± A±
— a O(^ — ^ A~
u
^CW jr. A
r*. *^~
1
n
O
B
_D_
O
"P04~ - 20 hr:
A
pn++
AC°
i,* -
D i
B
e *o
e <£
i ^Jl^1++ f
3A
) 3
^
;
^ a
O
tr\c\ h^
_uu nr
c
H C
D
)
"P0d- - 200 hr
A
0 10 20 30 0 10 20 30 0
AXIAL VELOCITY (ft/sec)
10
20
30
Figure 18. Product Water Composition in Hyperfil"tration of Primary Sewage Effluent with
a Nominal SM° Rejection Aerojet 6% Cellulose Acetate Membrane with Initial
Flux of ho to 50 gal./ft2*aeiy.
-------
ORNL DWG. 71-13062
102
5
p
10
5
P
*
^"AC
A
c
/.
Cc
B
VCa++ + M
0
r-°-e
f— *A=q
j r^
i « a
n
i «
+;+^g++-
f-a
> »e
* ["> A
\7 (>
n++ 9O h
.
TI
B
2OOhr
1
-i—
j—
^
f
n
^O/
2 <3 <
A
B «
a a
— A-A
B
A «
H"
PO,
r-r
—8
f^J y\
B
-------
ORNL DWG. 71-13560
Q
00
CC
=)
o
Z)
Q
O
o:
Q_
10
0.5
0.2
0.1
a
Q
®
3
€
O
A
D
O
E
A
/\
V
Cl
K
KUN
13
14
17
18
19
20
24
25
26
28
30
31
33
34
36
20 h
2<
17
c*
ours
K
/^
^>
>
C
^
A
ii
»1
6
h
c
17
c
z
'
1*
«o
L
€
A D
in AFI
B ^KJ
•T ^
J O
HT
-10
^
W7
V
^
«
oa
£
5
200 he
-w
n
>urs
V
/•<
J
f
^
1
t/
1^
k
€
C
V
^
10 20 40 1 2
AXIAL VELOCITY (ft/sec)
10
20
40
Figure 20.
Product Turbidity from Nominal 9^% Rejection Aerojet
Membrane with Primary Sewage as Feed.
Cellulose Acetate
-------
ORNL DWG. 71-13559
10
Co
Q
m
cc
Z)
O
D
O
O
or
o_
0.5
0.2
0.1
A
O
€
O
<1
L>
V
A
n
c;
±
n
B
nuiM
1^
i -j
Iyi
4
j -7
i f
1«
1 O
19
^0
21
2Z
23
24
25
26
? 8
3/^v
O
31
3-2
O
34
36
f-
L
A
9
L
\
^
j
L
n
>
7 «
ML
n
LJ
t
Z
>
^
1
^
**
^
rm
§
2(
J
I-
L
J
D
v
-O
a
u
•4^7
-TV
/. n
]
-------
SECTION VII
CORRELATION OF RESULTS
Threshold Velocity. The results shown in Figures 9, 10, and 12 clearly
demonstrate that there is a threshold velocity above which the flux
decline is markedly smaller than at lower velocities. Results of fouling
studies for the Office of Saline Water using untreated river water as
feed are also strongly indicative of a threshold velocity.
With support of the National Science Foundation, an analysis was developed
for the effect of flux on the minimum velocity required to prevent partic-
ulate solids from depositing or bouncing along the bottom of a horizontal
tube. Because the limit deposit velocity given by that analysis is
related to the threshold velocity for particulate fouling, the results
will be briefly summarized.
The analysis was based on earlier studies of the limiting deposit (or
minimum transport) velocity of flocculated particles in horizontal
impermeable pipes (l8, 19, 21). In those studies it was shown that
Bernoulli forces accompanying the velocity gradient in the vicinity of
the wall, acceleration due to the gradient of velocity fluctuations near
the wall and possibly Magnus forces, all were the correct order of magni-
tude to play a role in preventing deposition of small particles on
horizontal surfaces. Dimensional considerations suggest that, for
particles with diameter smaller than the viscous sublayer and which
follow Stokes ' law, the friction velocity for the minimum transport
condition, (u*.)^, may be related to properties of the particulates by
where Um is the terminal settling velocity of a particulate (or floe) of
diameter D^ and v is the kinematic viscosity. K and k are to be determined
from theory or experiment.
If Bernoulli forces arising from the velocity gradient were assumed to be
the principal lift force, then it was shown (l8) that the Bernoulli forces
equaled gravitational forces on the particle when
(5)
Friction velocity may be related to the mean velocity, u, in a system by
u = % ^/2 , (6)
where f = Fanning friction factor.
Based on a review of existing data on lift forces the most probable value
for CT was between 0.5 and 1. Equations (U) and (5) have identical form
and the exponent k has a value of 3 based on the results of the analysis.
65
-------
Comparison of Equation (5) with minimum transport results obtained with
both flocculated and deflocculated suspension (l8,2l) indicates an
entirely satisfactory fit provided the left coefficient is assigned a
value of 1, the upper limit of the most probable values. It seems likely
that the lift coefficient in this flow regime is somewhat less than one
(l8). Apparently, although Bernoulli forces may play a major role in
supplying particle lift in the vicinity of a wall, particle acceleration
and Magnus forces also play a role and their effect is primarily reflected
in the value of the coefficient in Equations (U) and (5).
When there is flow through the wall as in hyperfiltration, it is a simple
matter to extend the analysis to include the additional drag force on the
particle due to the product water flux, v, at the membrane surface (l?).
The result in the same form as Equations (^) and (5) is
Flocculated particles in polluted waters are often of the order of 1-micron
diameter. For a 1-micron particle, the terminal settling velocity is of
the order 0.002 cm/min. Hence from Equation (l~) , any fluxes greater than
0. 002 cm/min (~1 gf d) may be expected to require increased axial velocity
to exceed the minimum transport condition when the particles are of order
1 micron in diameter.
Equation (7) indicates that in the limit of zero wall flux or large
particle diameter (>20 microns), the velocity required to prevent particle
deposition is inversely proportional to the one-fourth root of the particle
diameter, while, in the limit of very small particle diameter (D < 0. 5
micron), the velocity required to prevent particle deposition is^propor-
tional to the one-fourth power of the flux and inversely proportional to
the three-quarters power of the particle diameter. Calculated values of
the limit deposit velocity are shown schematically on Figure 22 as a
function of particle diameter with the flux as a parameter. For purposes
of illustration a floe density of 1.1 was assumed. Also shown on this
figure is the expected variation of floe diameter with rate of shear based
on results of earlier studies (20). Of course if the particles were not
flocculated the diameter would be independent of the rate of shear.
Assuming that the particles in primary sewage effluent are flocculated
and that the minimum transport or limit deposit velocity must be exceeded
to prevent particulate fouling (and hence catastrophic flux decline) then
from the results shown in Figure U the threshold velocity should be pro-
portional to the square root of the flux through the membrane.
Threshold velocities determined from the 200 hour results with primary
sewage feed in the present study (Figures 11 and 13) are shown in Figure
23 as a function of initial flux. Also shown on this figure are results
obtained using untreated river water as feed (22), and an estimate of
threshold velocity for Pomona, California secondary sewage, based on data
obtained with a Havens tubular unit (7) and with a spiral wound module (3).
66
-------
ORNL-DWG 71-5683
10
5
2
1
5
2
10'1
L 5
2
10-2
12 5 10 2 5 102 2
AXIAL VELOCITY (ft/sec)
Figure 22. Effect of Particle Diameter and Flux on Limiting Deposition
Velocity for Small Particles (v = 1 cm/min Corresponds to
gal./ft2 .day).
CT
LU
h-
UJ
Q
LJ
_l
O
h-
cr
ATYPICAL FLOC DIAMETER
VERSUS VELOCITY
v (cm/min)
67
-------
oo
102
ORNL DWG. 71-4114R
o
>- £
H
O
O
LJ 10
Q
C/)
UJ
(T
X
O OR
D OR
A HAN
o GU
(EG
^
NL PRIMARY SEWAGE
NL RIVER WATER
y/ENS POMONA SECONDARY
LF GA POMONA SECONDARY
IUIVALENT EMPTY CHANNEL VELOCI1
^
t-
T
ff
-e:
?
4?
^
rT ^^^
S"
[1
I'
^i
I
1
{
V
V
^
^fT
^
>•
FY)
^
s
*
f^r~
f^
-THEORETICAL SLOPE
(~1/2)
10 20 50 100
INITIAL FLUX (gal/ft2 day)
200
500 1000
Figure 23. Threshold Velocity Above Which Catastrophic Flux Decline is Prevented with
Polluted Surface Water Feeds.
-------
In the latter case the threshold friction velocity was estimated and this
was converted to an equivalent velocity for tubes using Equation (6).
Surprisingly, all of the results are in rather good agreement with each
other, possibly indicating similar flocculation characteristics of polluted
surface waters. All results are consistent with the expectation tnat the
threshold velocity is proportional to the square root of the initial flux.
The results shown in Figure 23 may be represented by
u = (1.3 ± 0.6)(jT)1/s , (8)
C -L
where u = threshold velocity in ft/sec, and
J = initial flux in gal./ft2 »day.
Because the "initial" flux decline at the first of a run has often been
considered as due to some other phenomenon than fouling, the results for
the present "initial flux declines" (Figure 10) are shown in Figure 2k
as threshold velocity vs initial flux. Clearly the short time results
are in good agreement with long time results. This suggests that at
least the catastrophic flux declines observed in the first few hours of
a test with sewage effluent, are due to particulate fouling because of
poor hydrodynamics.
Both Equation (7) and Figure 22 indicate that the threshold velocity could
be reduced by increasing the particle size in the feed. From Equation (j},
the threshold velocity is inversely proportional to the 3A power of the
particle diameter. Hence an increase of floe size that might be of little
interest in terms of increased settling rate could result in significant
reduction in threshold velocity.
As described in the 'section on "Additives for Sewage Floe Size Control,"
several different additives were tested to see whether they might be
effective in increasing floe size under the high shear conditions in the
loop. Although a series of experimental difficulties with the flocculants
addition system prevented acquisition of conclusive results, some short
time results with FeCl3 as the additive indicate that this might be a
promising technique. As estimate of the threshold velocity for primary
sewage feed in the presence of ~50 mg/j2, FeCl3 is shown in Figure 25 as
a function of initial flux. For comparison the range for untreated
primary feed is shown as the cross-hatched area. Based on these scouting
studies it appears that a suitable flocculant might reduce the threshold
velocity to ~0. ^ the value without flocculant.
Flux Decline. Although the flux decline parameter, b, (b = — A log flux/
A log time) is of little use for extrapolation to obtain long term fluxes
from short term results, it is useful as a parameter in correlating short
term results. The flux decline parameter determined for times from 1 to
100 hours after primary sewage addition, is shown as a function of axial
velocity in Figure 26. Three different initial fluxes are represented
and the threshold velocity for each flux is indicated by the cross-hatched
region.
69
-------
ORNL DWG. 71-13566
ou
-o 20
10
h-
0
o
j
^ R
> o
Q
_J
I
C/)
UJ p
cr ^
i
i-
i
X'^
x^
^
^^
^x^*
X
^»c
V
V
h
.'
L^
^
^^^
"
^
X*
X"*^
^1
.x**
-THRESHOLD VELOCIT
LONG TIME RESULTS
jjp
y
Y
X
•
, i
~l
X
X
X
^
1 ^^
xL^^i
b^H
5 10 2 5
INITIAL FLUX (gal/ft2 day)
10'
Figure 2k. Comparison of Threshold Velocity from Results After 20 Hours Circulation
of Primary Sewage (Round Points) with Results After 200 Hours.
-------
o
0)
c/)
o
o
_l
LU
O
X
CO
Ld
CC
I
3
2
10
ORNL DWG. 71-13565
UNTREATED
PRIMARY FEED
FeCI3 TREATED PRIMARY FEED
5 10 2 5 10'
INITIAL FLUX (gal/ft2 day)
Figure 25. Effect of Adding ~50 mg/A FeCl3 to Primary Sewage Feed Immediately Before
Introduction to Hyperfiltration Loop.
-------
= 3 to 15 gfd
^_
^>
' n R
U.O
CU
o
_l
^ c\ o
<] \J.f-
X
~» n 1
, U-1
j
i ,
u_
,_. nnR
(3 U.UD
_j
r^ n no
1 U.Uc
n
-° nm
"*>
>v
V
-^^
xf
1
4
^,
'
.
?V
IL
Bs
i>
Up
fj^.
'//
//
^
I
//
//
^/
/.
//$
//,
^
P
r
\
/\
^a
2
t
(
A
t
\
)
^ A
" J
= ~46 gfd
ORNL-DWG 71-5481
Jl = -130 gfd
10 30 2 5 10 20 40 2
AXIAL VELOCITY (ft/sec)
75
91
5 10 20 40
Figure 26. Effect of Axial Velocity and Initial Flux on Flux Decline Parameter, b, for
Initial Portion of Runs with Primary Sewage Feed. (600 psig, Cellulose
Acetate Membranes)
D
C
M—
O
-------
For the lowest initial flux, flux decline parameters were always less
than 0.1, they were always less than 0.06 above the threshold velocity
and b continued to decrease with increase in velocity. The median value
of the flux decline parameter for velocities above the threshold was 0.025
with a range from 0.01*1 to 0.06.
For intermediate values of the initial flux, there was a rather pronounced
discontinuity in the value of the flux decline parameter at the threshold
velocity. Below the threshold velocity values of b were from 0.3 to 0.6
while above the threshold velocity the value of b had a median value of
0.065 with a range from 0.035 to 0.15.
Similar results were obtained with the highest initial flux membrane except
that the median value of b for velocities above threshold was 0.09 with a
range from 0.0^5 to O.lU.
Figure 27 is a composite plot showing values of the flux as a function of
time for velocities above the threshold and times up to 700 hours. For
the two higher flux membranes the points represent median values and the
bars give the range of results. Because of the uncontrolled variation in
initial flux with the low flux membrane, two representative runs with
similar initial fluxes are shown. The flux decline parameter clearly
increases with initial flux. Also the value of log flux is not a linear
function of log time, with the nonlinearity increasing as the initial flux
increases. In all probability this is due to increased rate of fouling by
organic s at the higher fluxes.
Figure 28 shows the effect of initial flux on the value of the flux decline
parameter for velocities above threshold and times from 20 to 200 hours
(a period when all data showed a linear log flux-log time relation). As
in the previous figure the values for the two higher fluxes are indicated
by median points with bars giving the range. For the lowest flux membrane,
results of individual tests are given. The slope of the line drawn on the
figure is 0. 6 although the data are not inconsistent with values from 1/2
to 2/3.
From results studied with untreated river water it appears that the flux
decline parameter is proportional to the reciprocal of the square root of
the axial velocity. Although insufficient results were obtained above the
threshold velocity to substantiate the velocity dependence for primary
sewage feed, limited results for individual runs shown on Figure 25 are
not inconsistent with a u 2 dependence.
Tentatively it appears that both initial flux and axial velocity are
equally important in determining flux decline and that a threshold velocity
must be exceeded to prevent a catastrophic flux decline. When the velocity
is above the threshold value for particulate fouling, the effect of flux
and axial velocity on the flux decline parameter for times less than 200
hours is given by
0.6
UO.B
73
-------
ORNL DWG. 71-13572
10'
-o
OJ
-5 10
b = 0.019
10 2 5 1CT 2 5
TIME AFTER SWITCHING TO PRIMARY
SEWAGE EFFLUENT FEED (hr)
Figure 27. Summary of Flux Decline Results for Velocities Above
Threshold with Primary Sewage as Feed.
74
-------
ORNL-DWG 71-5682
Ul
0.3
LU 0.2
CD
O
_J
<
X
CD
O
11
.a
0.1
0.05
0.02
0.01
5 10 2 5
INITIAL FLUX (gal/ft2 day)
10'
Figure 28. Effect of Initial Flux on Flux Decline Parameter, t>, for Runs with Primary
Sewage Feed and Velocity Above Critical (600 psig, Cellulose Acetate Membranes)
-------
For times longer than 200 hours, the present results suggest that fouling
by dissolved and colloidal organics became important. Additional studies
are required to confirm this.
76
-------
SECTION VIII
ACKNOWLEDGEMENTS
The studies performed herein were performed by the Oak Ridge National
Laboratory at Oak Ridge Tennessee. The work was carried out by the
Water Research Program, James S. Johnson, Director, R. B. Gallaher,
J. R. Love, and David G. Thomas. We have received helpful assistance
from K. A. Kraus, who was Program Director during early portions of
this study, as well as from W. R. Mixon, J. D. Sheppard, A. J. Shor,
P. H. Hayes, M. J. Adair, and D. C. Michelson.
We have received extremely helpful cooperation from the City of Oak
Ridge, particularly from Thomas C. Stephens, Supervisor of the
Municipal Sewage Plants and 0. K. Richman, Director of Public Works.
The support of the project by the Office of Research and Monitoring,
Environmental Protection Agency and the help provided by Mr. John M.
Smith, the Project Officer, is acknowledged with sincere thanks.
77
-------
SECTION IX
REFERENCES
1. Summary Report: The Advanced Waste Treatment Research Program,
January 1962 - June 196^, U. S. Department of Health, Education,
and Welfare, Public Health Service, AWTR-lU, April 1965.
2. Summary Report: Advanced Waste Treatment, July 196^ - July 1967 >
U. S. Department of the Interior, Federal Water Pollution Control
Administration, AWTR-19, 1968.
3. Merten, Ulrich, and D. T. Bray, p. 315 in Advances in Water Pollution
Research, Vol. 3, ed. "by J. Paz Maroto and F. Josa, Water Pollution
Control Federation, 1966.
^. Savage, H. C., N. E. Bolton, H. 0. Phillips, K. A. Kraus, and J. S.
Johnson, Water and Sewage Works, p. 102, March 1969.
5. Kraus, K. A., Application of Hyperfiltration to Treatment of
Municipal Sewage Effluents, Final Report to Federal Water Quality
Administration, Department of the Interior, from Oak Ridge National
Laboratory, January 1970.
6. Hauck, A. R., and S. Sourirajan, Environmental Science and Technology,
3: 1269 (1969).
7. Smith, J. M., A. N. Masse, and R. P. Miele, Renovation of Municipal
Wastewater by Reverse Osmosis, U. S. Department of the Interior,
Federal Water Quality Administration, Municipal Treatment Research
Program, Advanced Waste Treatment Research Laboratory, May 1970.
8. Nusbaum, I., J. H. Sleigh, Jr., S. S. Kremen, "Study and Experiments
in Waste Water Reclamation by Reverse Osmosis," WPCR-170iKD-05/70,
May 1970.
9. Beckman, J. E., E. Bevege, J. E. Cruver, S. S. Kremer, and I. Nusbaum,
"Control of Fouling of Reverse Osmosis Membranes when Operated on
Polluted Surface'Waters," Gulf Environmental Systems Report GA-10232,
Final Report to Office of Saline Water, August 1, 1970.
10. Environgenics Company, "New Technology for Treatment of Wastewater
by Reverse Osmosis," 'Final Report for Federal Water Quality Adminis-
tration, U. S. Department of the Interior, September 1970.
11. Feurstein, D. L., and T. A. Bursztynsky, Chem. Eng. Progress,
Symposium Series, 67(107): 568 (1970).
12. Conn, W. M., "Raw Sewage Reverse Osmosis," Presented at 695th Annual
Meeting, A.I.Ch.E., Cincinnati, Ohio, May 17, 1971.
79
-------
13. Sheppard, J. D. , and D. G. Thomas, Applied Polymer Symposia, No. 13,
pp. 121-138, "Membranes from Cellulose and Cellulose Derivatives,"
ed. by A. F. Turbak, John Wiley and Sons, 1970.
Ik. Keilin, B. , "The Mechanics of Desalination by Reverse Osmosis," Office
of Saline Water Research and Development Progress Report 117, EB l66
395, August
15. Michaels, A. S., H. J. Bixler, and R. M. Hodges, Jr., MIT Department
of Chemical Engineering Report 315-1 DSR 9^09,
l6. Sheppard, J. D. , and D. G. Thomas, A.I.Ch.E. Journal, Vol. 17, p. 910,
(1971).
17. Sheppard, J. D. , and D. G. Thomas, "Effect of Flux on Limiting
Deposit Velocity for Particulated Foulants in Hyperfiltration, "
unpublished.
18. Thomas, D. G. , A.I.Ch.E. Journal, 7:^23 (1961).
19. Thomas, D. G. , A.I.Ch.E. Journal, 10:303 (196^).
20. Thomas, D. G. > A.I.Ch.E. Journal, 10:517 (196U).
21. Thomas, D. G. , A.I.Ch.E. Journal, 8:373 (1962).
22. Sheppard, J. D. , P. H. Hayes, W. L. Griffith, R. M. Keller, and
D. G. Thomas, "A Study of Hydrodynamic Aspects of Reverse Osmosis
(Hyperfiltration)," Final Report to Office of Saline Water,
January 1972.
r U. S. GOVERNMENT PRINTING OFFICE: 1 973--5Ht-1 55/318
80
-------
SELECTED WATER
RESOURCES ABSTRACTS
INPUT TRANSACTION FORM
Report Vfo.
3. Accession. No.
w
4. Title
Hydrodynamic Flux Control for Waste Water
Application of Hyperfiltration Systems
7. Author(s)
David G. Thomas and Richard B. Gallaher
9. Organization
Oak Ridge National Laboratory
Oak Ridge, Tennessee 37830
S, Report Date
6.
8. Rerfartmng Organization
Report ffo.
10. Project No.
17020-FEV
11. Contract/Grant No,
Type s>f Repot t and
Period Covered
12. S-itmsorioi
IS. Supplementary Notes
Environmental Protection Agency Report No. EPA-R2-73-228, May 1973
16. Abstract
The effect of hydrodynamics on flux decline of cellulose acetate hyperfiltration
membranes was studied using primary effluent from the Oak Ridge East Sewage Plant
as feed. The system contained multiple, annular-geometry housings which could be
operated simultaneously at different velocities with the same feed. The range of
fluxes studied was from 2 to 150 gal. /ft3»day and axial velocities ranged from 3
to 30 ft/sec. System pressure was 600 psig.
The results demonstrate that there is a threshold velocity above which flux decline
is markedly smaller than at lower velocities. Visual inspection of the membranes
after a test indicates that operation above the threshold velocity markedly reduces
accumulation of solids during the first 200-300 hours of a test. In scouting
studies addition of a flocculant appeared to markedly reduce the threshold velocity.
During the initial 200-300 hours of those tests with primary sewage effluent as
feed and in which the axial velocity was above the threshold velocity, the flux
decline parameter, b, [b = — (A log flux)/(A log time)] was directly proportional
to the six-tenths power of the flux and inversely proportional to the square root
of the velocity.
17a. Descriptors
Particulate fouling, hydrodynamic control, threshold velocity, flux decline,
hyperfiltration, reverse osmosis, waste water treatment.
17b. Identifiers
Particulates, flux control, hydrodynamics.
17 c. COWRR Field & Group
18. Availability
19. Security dim,
(Report)
•'0. Security Clast*
(Page)
21. Ho. of
Pages
23. JPwt*
Send To:
WATER RESOURCES SCIENTIFIC INFORMATION CENTER
U.S. DEPARTMENT OF THE INTERIOR
WASHINGTON. D. C. 2O24O
Abstractor David G. Thomas
I institution Oak Ridge National Laboratory
------- |