Membrane fouling in reverse osmosis _RO_ systems is as all by malj



The recent evolution of spiral-wound membrane element (sepralator) design brings an
increased need to understand the intricacies of testing, operating and predicting performance of
this most popular crossflow membrane design. After over a decade of stagnant development,
the last five years have produced several innovations, which diversify and expand the
applications as well as improve the performance of sepralators.

These innovations include improved materials of construction for specialized applications such
as ultrapure water reclamation (1). Another innovation is the Full-Fit design, which eliminates
the outer cover/brine seal combination. This design reduces microbial growth and provides
better chemical rinse-out for cleaning/sanitizing in medical/pharmaceutical applications. Also by
improving fluid dynamics the Full-Fit design reduces fouling when operating on particle laden
and/or gel-forming solutions. Clear channel "tubular spiral" sepralators reduce axial pressure
drop on viscous solutions, and are able to handle high "plugging" streams (2). Permeate
channel design innovations allow increased temperature capability even at high pressures (3).
Tapered shape spirals conform exactly to standard cartridge filter sumps (4).

Sepralator designs are now expanding and improving to complement the membrane
improvements, which are regularly commercialized now as crossflow membrane technology
comes of age. To fully utilize these design improvements, sepralator operation and performance
knowledge must also improve to keep pace.

Some operational and test technique nuances are not widely known. "Standard" test methods
alone are inadequate, or in some cases inaccurate, and improved understanding must be more
widely communicated. This paper presents some guidelines and illustrations for testing
sepralators and predicting performance in three areas:

1. Separation

2. Production rate (flux)

3. New characterization concerns

Separation Measurement

The "industry accepted standard" test for reverse osmosis (RO) sepralators is fairly simple, but
one often overlooked detail (the feedwater) can make a significant performance rating
difference. There is no "industry accepted standard" separation test for ultrafiltration (UF)
sepralators, but polyethylene glycols (PEGs) and dextrans of varying molecular weights are
commonly used as marker molecules. Different operating conditions, especially crossflow
velocities and solute concentrations, can yield radically different separation results. No accepted
test method to judge microfiltration (MF) sepralators exists.

Most RO sepralator manufacturers state similar test conditions to achieve their solute
                                                                   °    °
separation specification; 225 or 400 psig (15.5 or 27.6 bar), 77 F (25 C), 10-15% recovery and a
solution of 2000 ppm NaC1 (5,6,7). However, the more specific nature of that test solution is a
significant factor in the separation outcome. If tap water is used as the NaCl solvent, and that
tap water has high levels of divalent salts, higher rejection will be seen since conductivity is the
normal measurement method. For example, Colorado River Water (CRW) is the basis for the
municipal water supply in much of southern California - where RO "grew up" and where many
spiral RO manufacturers are currently located.

To quantify this effect, we tested a sepralator containing medium rejection, cellulose-acetate
based membrane, on 0.2% NaC1 solution based both on simulated CRW and RO treated tap
water (Table 1). While the difference between 94.8 and 95.9% rejection may not appear
significant at a casual glance, the passage value of 5.2% on the RO treated tap water solution
represents 27% more salt in the permeate than the 4.1% CRW-based test solution yields. This
could be an important difference when evaluating the economics of RO as pretreatment to a
large Deionization (DI) system.

                              Table 1: Analysis of Water Sources

                                                       Colorado        Reverse
                                                         River         Osmosis
                                                         Water          treated
                                                        (CRW)2          Water

                               Calcium                    130            25

                             Magnesium                     75            10

                               Sodium                     160              9

                               Alkalinity                 120            20

                                Sulfate                   130            20

                               Chloride                   110              6

                             Conductivity                 690            70
                                                         umhos          umhos
                          Percent Passage                4.1%           5.2%

1.   2000 mg/l added to feedwater, Pop=27.6 bar (400 psig), Recovery (Qp) = 8%
     Concentrations were measured using conductivity.                   Qf

2.   All values are in ppm as CaCO3 except as noted.

3.   Produced from Minnetonka, Minnesota tap water.

4.   Percent Passage = Cc + Cf x l00.

Crossflow velocity is also an important parameter to consider for separation. It is usually
measured practically in terms of feed flow (Qf), or more accurately as the average of the feed
and concentrate flows [Qavg = (Qf +Qc)/2].

When sepralators were only available with an impervious outer cover/brine seal design, a
recovery range of 10-15% was adequate to obtain consistently acceptable fluid dynamics test
results. The advent of the Full-Fit design (sans outer cover and seal) pointed to the need for
testing at a critical minimum flow, despite this recovery ratio. Figure 1 presents the different
NaC1 passage values obtained on the same sepralator at varying Q avg rates. Two Osmo® 416
Full-Fit sepralators were tested to determine the effect of crossflow velocity or flow on
performance. Two types of cellulose acetate membrane were tested. The operating pressure
was held constant over a range of average flows (Q avg) and the two most popular pressures
were tested. From this data it is obvious that for Full-Fit sepralators a critical crossflow is
required to obtain optimum rejection and valid comparative data.
                                               Fig. 1

Figure 2 shows the calculated crossflow velocity at measured Q avg rates for both the "standard"
outer cover/brine seal and the new Full-Fit sepralator designs. Velocity [ft/sec or m/sec] is
                                           3         3
calculated by dividing the channel flow [ft /sec or m /sec] by the cross sectional flow area in the
sepralator [ft or m ].
                                              Fig. 2

The Full-Fit design has permitted the use of spiral sepralators for ultrafiltration of
electrodeposition paint. Hollow fiber and tubular membrane modules had dominated this
application because spirals with standard designs and materials of construction fouled easily.
The Full-Fit design allows operation at higher crossflow velocity without risking damage due to
differential pressure.

Turbulence has always been desired in sepralators to disrupt the concentration polarization
layer, but had to be estimated empirically from DP data since the convoluted flow paths
presented by the standard mesh spacer make classical Reynolds Number (N Re) calculations
extremely complex. With the introduction of the new "tubular spiral" feed spacer material (clear-
channel), calculations closer to classical NRe formula may be applied. Figure 3 demonstrates
that an equivalent sepralator with this spacer requires five to ten times more Qavg than a
traditional mesh spacer sepralator to achieve turbulence region Reynolds Numbers. Whether
turbulence is required to maintain acceptable flux values with this spacer must be determined
for each application.
                                               Fig. 3

Permeate Rate and Long Term Performance

Accurate measurement of permeate rate values would seem to be straightforward and simple.
In the worst case, the large-pore UF sepralators might be subject to test equipment and
operator inaccuracy at the low pressures and high flows typically employed. Then use of low-
range gauges, careful gauge calibration, and adequate consideration of actual hydraulic
pressures throughout the test system should provide accurate data.

However, the potential for permeate rate reduction from fouling, even in short-term testing,
cannot be ignored. Consider the view of the membrane community toward membrane
"compaction." There are widely used rules of thumb for estimating permeate rate output at 1, 2
and 3 years of membrane service. For CA membrane operated at 350-450 psig (166.3-213.9
bar), these values usually are based on an 8, 13 and 22% rate loss at 1, 2 and 3 years
respectively. For the PA membranes operated at 200-250 psig (13.8-118.9 bar), the lower
values of 10, 15 and 20 are used. These losses are assigned to the membrane compaction
effect; densification of the membrane polymer which results in reduced void volume, hence
increased resistance to hydraulic flow. Compaction is assumed because this flux loss is seen in
controlled tests where "pure water" and clean test systems are used, so no fouling is assumed.

Yet, note in Figure 4 the difference in flux reduction over 1000 hours when Rudie et. al. placed
20K MWCO (Molecular Weight Cutoff) ultrafiltration sepralators immediately ahead of the RO
test cells to ultrafilter the feed, despite the use of an apparently clean test system. (8) Even at
the high pressure of 600 psig (41.4 bar) less flux loss is seen than typically estimated for CA
systems in the 400 psig (27.6 bar) range. Increasing the pressure above 600 psig (41.4 bar),
                                                °     °
and increasing the temperature above 104 F (40 C) both had dramatic effects on flux loss,
indicating these parameters in these ranges do compact the membrane.)
                                             Fig. 4

Therefore, fouling is more of a factor and compaction less of a factor, than previously assumed
by the industry. (Even with the UF pretreatment, microscopic examination of the test
membranes showed significant foulant deposition.)

The overall message here is that fouling cannot be dismissed from consideration when
predicting flow rate performance in practically any crossflow membrane system.

How to predict permeate rate over time remains more art than science. Figure 5 shows
permeate rate production as a percentage of original flow vs. time for three ultrafiltration
applications. Sepralators operated on ultrapure water generally show less flow loss than on
other applications, but fouling can and does occur. Losses can be minimized by monitoring
parameters outlined in the next section (TOC, bacteria, and pyrogens) and correcting problems
as they occur.
                                              Fig. 5

The best and sometimes only way to predict long term performance is to run a pilot test. The
unfiltered water of application B in Figure 5 was from a city water supply with only 30 ppm TDS.
SDI readings were about five, but not so high as to predict the flow loss actually seen. This flow
vs. time data, exhibited by applications B and C is essential to properly size a machine and
recommend cleaning chemicals and schedules.

New Test Parametres

With the expanded range of applications for spirals, made possible by both an increased
understanding of their correct operation and the design innovations mentioned previously, come
new areas of concern for their characteristics. When spirals were first applied for microchip
water reclamation at AT&T in the mid 1970's, no one had thought to measure total organic
carbon (TOC) rinse-down and resistivity rinse-up values first. Now these characteristics are
commonly evaluated and quantified for this industry.

As well as optimizing the materials of construction, design innovations have helped these rinse-
up characteristics. Figure 6 shows the markedly superior rinse down of the Full-Fit design,
where both types of sepralators contain the same materials of construction (except for the lack
of outer cover and brine seal on the Full-Fit ).
                                              Fig. 6

Permeate particle count is also a new concern, especially for the electronics industry where
particle shedding and time to rinse out a spiral is a concern, particularly for replacement
sepralators and point of use (POU) systems. Particles are also a concern in the beverage

The materials a sepralator adds to the fluid it treats are of even broader concern to the
pharmaceutical and medical industries. Here not only particles generated, but any microbes
added and the nature and amount of extractables (leached solutes) may be requested in
quantified terms.

It can be assumed many of the methods employed for cartridge filters, already established in
these applications, will be adapted for sepralators. To date, no ASTM, HIMA, etc. test methods
exist to test sepralators for these characteristics.

The use of sepralators in production of Water for Injection, in general depyrogenation and
microbial reduction applications has also led to a need for test methods with the element in
place in its housing/system. Some manufacturers use the vacuum-hold test method. This
method is specified for RO elements in an ASTM Standard Reference D 3923-80. A similar
standard for ultrafiltration spirals is going through the final phases of approval now in ASTM
Committee D-19, and should be "on the books" by 1990. Since the vacuum hold test is not
sensitive, but functions as a go/no-go test for seal and element integrity, improved techniques
will probably follow. Forward flow diffusion testing, similar to the procedures used for pleated MF
membrane cartridges, is on the horizon.

1. Gach, G.J. et al "A Case For Point-of-Use UF-RO Ultrapure Water Systems." Ultrapure
     Water Journal, Apr 87
2. Paulson, D.J. et al "Design Innovations For Processing High Fouling Solutions With
     Spiral-Wound Membrane Elements" Proceedings of the Third North American Chemical
     Congress - 6-10 June 88, Toronto, Published by National Research Council of Canada
     Press, NRCC #29895, 1989
3.   U.S. Patent #4,411,787 Reverse Osmosis Apparatus
4.   U.S. Patent #4,839,037, June 13, 1989, Tapered, Spirally-Wound Filter Cartridge and
     Method of Making Same. Inventors, Bertelsen, Paulson
5.   Osmonics® Product Bulletin dated Nov 1986
6.   FilmTec/Dow Product Bulletin dated Jan 1988
7.   Fluid Systems Division/UOP Product Bulletin dated Jan 1987
8.   Rudie, B.J. et al "RO and UF Membrane Compaction and Fouling Studies Using UF
     Pretreatment, Reverse Osmosis and Ultrafiltration" -281-ACS Symposium Series, 1985
     (Presented Aug 84)

To top