Effect of membrane configuration on bench-scale MF and UF fouling experiments

ARTICLE IN PRESS WAT E R R E S E A R C H

41 (2007) 3842– 3849

Available at www.sciencedirect.com

journal homepage: www.elsevier.com/locate/watres

Effect of membrane configuration on bench-scale MF and UF fouling experiments Kerry J. Howea,, Ashish Marwahb, Kuang-Ping Chiuc, Samer S. Adhamc a

Civil Engineering, MSC01 1070, University of New Mexico, Albuquerque, NM 87131-0001, USA Boyle Engineering Corporation, 1131 West Sixth Street, Suite 350, Ontario, CA 91762, USA c MWH Americas, Inc., 301 N. Lake Avenue, Suite 600, Pasadena, CA 91101, USA b

ar t ic l e i n f o

abs tra ct

Article history:

Hollow fiber and flat sheet membranes were compared in side-by-side bench-scale

Received 31 January 2007

experiments to evaluate whether the configuration has an impact on the rate of membrane

Received in revised form

fouling. Both microfiltration (MF) and ultrafiltration (UF) membranes were evaluated. In

10 May 2007

general, flat sheet membranes fouled more rapidly than hollow fiber membranes.

Accepted 15 May 2007

Pretreatment such as coagulation generally affected both configurations similarly, but in

Available online 23 May 2007

some cases coagulation reduced fouling on hollow fiber membranes but increased fouling

Keywords: Microfiltration Ultrafiltration Hollow fiber Bench testing

on flat sheet membranes. Prefiltration to remove foulants above 1 mm in size had a consistent effect on both configurations. A bench-scale apparatus employing a single-fiber module that allows testing over multiple filter runs with integral backwashing capabilities was demonstrated to provide more detailed information about fouling, which can be applied to full-scale applications. When bench-scale tests are to be used to screen treatment options for full-scale applications, the use of a backwashable hollow fiber system is recommended. & 2007 Elsevier Ltd. All rights reserved.

1.

Introduction

Membrane filtration with microfiltration (MF) and ultrafiltration (UF) membranes has become a common treatment technology for drinking water treatment. Membrane filtration is viable for particle removal in treatment plants of all sizes treating a wide range of source waters. Operating plants range from standalone plants where membrane filtration is the only significant unit process to plants where membrane filtration follows several pretreatment processes, which can include coagulation, flocculation, sedimentation, softening, oxidation, adsorption, and/or granular filtration. The use of membrane filtration has grown significantly over the past 15 years, and bench-scale research has contributed to the expansion of this technology. Fouling has been identified as a significant issue (AWWA Membrane TechnolCorresponding author. Tel.: +1 505 277 2702; fax: +1 505 277 1988.

E-mail address: [email protected] (K.J. Howe). 0043-1354/$ - see front matter & 2007 Elsevier Ltd. All rights reserved. doi:10.1016/j.watres.2007.05.025

ogy Research Committee, 2005). Bench-scale research has provided some understanding of fouling with respect to interactions between constituents in water and membrane surfaces, and the importance of membrane hydrophobicity, charge, and morphology (Laıˆne´ et al., 1989; Jucker and Clark, 1994; Elimelech et al., 1997; Laabs et al., 2006). MF and UF membranes can be fabricated in either flat sheet or hollow fiber configurations. For bench-scale membrane research, flat sheet membranes are commonly used. Flat sheet membranes offer a number of advantages. First, the equipment is readily available, inexpensive, and easy to set up and operate. A large variety of membrane materials and pore sizes are available from scientific suppliers. The cells can be used in a stirred (representative of inside-out crossflow filtration) or unstirred (representative of outside-in or deadend filtration) mode. The shear rate at the membrane surface

ARTICLE IN PRESS WAT E R R E S E A R C H

4 1 (200 7) 384 2 – 384 9

can be easily controlled by varying the stirring speed. Second, analysis of fouling data is relatively straightforward. Typically, the feed water is maintained at constant pressure and the flux declines as the membrane fouls. In this mode of operation, the rate of flux decline between tests can be compared graphically, with parameters such as resistant coefficients, or with the well-known filtration blocking laws (Hermia, 1982). Third, for studies involving the development of new materials, flat sheet membranes are easier to fabricate in the laboratory than hollow fiber membranes. Fourth, physical and chemical examination of the surface of flat sheet membranes is relatively straightforward. Analyses such as contact angle for hydrophobicity, streaming potential for surface or pore charge, atomic force microscopy for roughness, attenuated total reflectance Fourier transform infrared spectrometry for functional chemistry, scanning electron microscopy for visual imaging, and energy-dispersive spectrometry and X-ray photoelectron spectrometry for chemical composition are all easier to perform on flat sheet membranes than on hollow fiber membranes. Thus, the use of flat sheet membranes has been critical for developing a fundamental understanding of the nature of membrane filtration. While bench-scale testing has indeed been useful for understanding fundamental aspects of membrane performance, it would also be desirable to use bench testing to explore treatment options for specific membrane installations, such as identifying the best membrane to use, backwashing or cleaning frequency, or the best pretreatment strategy for optimizing performance. A key disadvantage of the use of flat sheet membranes for this purpose, however, is that the hollow fiber configuration is most commonly used in full-scale MF and UF systems. Hollow fibers membranes are used in full-scale systems because they can be backwashed, have good structural integrity, and can be put into industrialscale modules that have a high packing density while still allowing easy removal of solids during backwashing. The difference in configuration leads to several issues for using bench tests to explore site-specific applications. First, with few exceptions, the manufacturers of industrial-scale hollow fiber membrane filtration systems do not make flat sheet membranes. Thus, it is necessary to test with a surrogate membrane from a scientific supplier. Even though membranes made of the same base material can be obtained, the actual chemical composition may not be identical. Hollow fiber manufacturers have been known to develop multiple versions of a particular membrane material, and scientific suppliers also sell more than one membrane product of the same material (for instance, the Durapore GVWP from Millipore is a ‘‘hydrophilic PVDF’’ membrane and the Durapore GVHP is a ‘‘hydrophobic PVDF’’ membrane). Thus, there is no guarantee that a PVDF membrane from Millipore would be chemically identical to a PVDF membrane from, say, Memcor. The difference in chemical composition may lead to differences in fouling characteristics. Furthermore, even if the chemical composition was identical, the difference in the fabrication process could lead to differences in surface roughness, porosity, pore geometry, or thickness, all of which could affect the hydrodynamics of water flow through the membrane and the propensity of fouling.

3843

Another limitation of flat sheet membranes is the inability to test over multiple filter runs with backwashing as part of the filter cycle, since the typical filter holders only support the membrane for flow in one direction. To facilitate comparison between bench and full scale, and to allow testing over multiple filtration runs, some bench-scale research has been done with hollow fiber membranes. Chang and Benjamin (1996) developed a backwashable hollow fiber module convenient for use at bench scale and other researchers have subsequently used a similar approach (Serra et al., 1999; Jacangelo et al., 2005; Kim and DiGiano, 2006; Heijman et al., 2007). It has not specifically been established, however, whether flat sheet and hollow fiber membranes tested under similar conditions on bench scale will yield similar results with respect to fouling. The objective of this manuscript is to describe experiments where flat sheet and hollow fiber membranes were tested side-by-side with the same source waters and similar operating conditions, to determine whether performance was similar. Tests were done as part of a larger study, and several source waters were compared in this manner. In addition, the results of subsequent tests with a more extensive testing protocol that allowed backwashing and testing over multiple filtration runs are presented. The manuscript offers insight into the differences between testing with flat sheet and hollow fiber membranes and offers guidance for future bench-scale membrane research.

2.

Experimental methods

Bench-scale testing was performed with both flat sheet and hollow fiber membranes using natural surface waters from four different sources in North America and Europe (differences in fouling of the same membrane material by different source waters is evident in figures later in this manuscript, but the focus of this manuscript is differences in fouling of flat sheet and hollow fiber membranes when filtering the same water source). The tests presented here were part of a larger project investigating membrane performance at bench, pilot, and full scale (Adham et al., 2006). Flat sheet membrane testing was done in an unstirred 47-mm diameter filtration cell, which provided a membrane surface area of 17.3 cm2. Hollow fiber membrane modules were hand fabricated in the laboratory using polyethylene tubing with an inside diameter of 9.5 mm (3/8 in). T-joints placed in the tubing provided the water inlet and outlet to the shell side of the membrane fiber. The modules had a total length of 63–81 cm (25–32 in). The membrane fiber was placed inside the tubing, with 10–20 cm of fiber hanging out at each end. Both ends of the tubing were sealed with epoxy. One end of the membrane fiber was also sealed with epoxy but the other end was left open, which was the entrance to the lumen of the fiber. As fabricated, the hollow fiber membrane modules had a membrane surface area between 13.0 and 18.1 cm2. All modules were integritytested before use with air-pressure hold tests. Both MF and UF membranes were tested. MF membranes were made of PVDF and had nominal pore sizes of 0.22 mm for the flat sheet membranes and 0.1 mm for the hollow fiber membranes. The UF membranes were polysulfone and had a

ARTICLE IN PRESS WAT E R R E S E A R C H

41 (2007) 3842– 3849

molecular weight cut-off (MWCO) of 100 kDa for both flat sheet and hollow fiber membranes. Both membranes are slightly hydrophilic (contact angle 50–701) and neutrally charged, but the PVDF membrane is rougher than the polysulfone membrane. During testing, the membrane cell or module was connected to a pressurized reservoir containing sample water. Both membrane configurations were tested in the dead-end filtration mode, i.e., there was no continuous retentate flow from the hollow fiber modules, and the flat sheet cell was not stirred during filtration. The tests used constant-pressure, declining-flux filtration. Gas pressure regulated from a nitrogen tank was used for pressurizing the reservoir. Membrane flux was determined by weighing the permeate on an electronic balance at timed intervals with computerized data acquisition. The test pressure was between 0.70 and 1.0 bar (10.1 and 14.5 psi) and temperature was between 20 and 24 1C. Pressure was measured with a digital pressure gage that had a resolution of 0.1 kPa (0.014 psi) and water temperature was measured with a certified thermometer that had a resolution of 0.1 1C. All flux values were normalized to a standard temperature (20 1C) and pressure (1 bar) using the equation (Crittenden et al., 2005) JSP ¼ JM

1:03ð20TM Þ ; DPM

(1)

where JSP is specific flux (L/m2 h bar), JM is measured flux (L/m2 h), DPM is measured transmembrane pressure (bar), and TM is measured temperature (1C). Clean-membrane permeability was determined by filtering deionized water (i.e., type I reagent-grade water, Standard Method 1080 (APHA et al., 2005) for a minimum of 30 min or until the flux was constant. Some hollow fiber membranes were tested over multiple filtration runs. These modules were similar to the hollow fiber modules described above, but did not have T-joints. Instead, pieces of 3.2 mm (1/8 in) ID tubing were epoxied into the ends of the 9.5 mm tubing to provide the inlet and outlet to the shell side of the membrane fiber. In addition, the filtration apparatus incorporated a timer, solenoid valves, and separate reservoirs for feed and backwash water. During backwash, the feed, backwash, and waste valves were actuated so that the feed tank and permeate collection system were isolated and water flowed from the backwash tank, backwards through the membrane fiber, and to a waste collection system. The backwash frequency and duration and a short rinse cycle were controlled by the timer. The feed and backwash water were supplied at the same pressure. Data quality and reproducibility were ensured by preliminary testing, performance criteria, and duplicate tests. It was noted in preliminary testing that reproducibility was poorer if the flux of clean water declined significantly during the initial permeability testing; therefore, subsequent tests were terminated and started over if the clean water flux changed by more than 5 percent over 30 min. In addition, about 10–15 percent of the flux decline tests were run in duplicate. The final normalized specific flux of duplicate tests typically agreed within 5 percent of each other. Additional details of the experimental methods are available elsewhere (Adham et al., 2006).

Resistance coefficients were calculated using (Crittenden et al., 2005) JSP ¼

DP ; mðkM þ kF Þ

(2)

where m is the viscosity (kg/m s), kM is the membrane resistance coefficient (m1), and kF is the fouling resistance coefficient (m1).

3.

Results and discussion

3.1.

Comparison of MF membranes

A comparison between flat sheet and hollow fiber MF membranes treating the same water is shown in Fig. 1. In these tests, the flux decline through the flat sheet membrane is significantly more rapid than through the hollow fiber membrane. Both were PVDF membranes, but the flat sheet membrane was from Millipore and the hollow fiber was from an industrial membrane manufacturer. As noted earlier, Millipore has more than one PVDF membrane product. In addition, the industrial membrane manufacturer provided more than one PVDF membrane product over the course of this project, and the different products were found to be susceptible to different amounts of fouling when treating the same water (Adham et al., 2006). Each product has surface modifications that may affect its propensity for fouling. The specific chemical formulation and/or surface modifications are, in most cases, proprietary, so it would be nearly impossible to guarantee that the flat sheet and hollow fiber membranes were of identical material. It is possible that differences in chemical composition contributed to the difference in performance, even though both were PVDF membranes and the pore size was similar. One notable difference between these membranes was the initial permeability. For the tests shown in Fig. 1, the hollow fiber membrane had an initial permeability of 313 L/m2 h bar and the flat sheet membrane had a permeability of 3070 L/m2 h bar, nearly an order of magnitude higher. This

1.0 Hollow fiber Flat sheet

Normalized specific flux

3844

0.8

0.6

0.4

0.2

0.0 0

100

200 300 Specific volume (L/m2)

400

Fig. 1 – Comparison of normalized specific flux of flat sheet and hollow fiber PVDF MF membranes (source water A) (Adham et al., 2006).

ARTICLE IN PRESS WAT E R R E S E A R C H

3845

4 1 (200 7) 384 2 – 384 9

3000 1.0

2500

Flat sheet

Normalized specific flux

Specific flux (L/m2.h.bar)

Hollow fiber

2000 1500 1000 500 0 0

100

200 300 Specific volume (L/m2)

0.8

Hollow fiber

0.6

Flat sheet

0.4

0.2

0.0

400

0

Fig. 2 – Comparison of specific flux of flat sheet and hollow fiber PVDF MF membranes (source water A).

100

3.2.

Comparison of UF membranes

A comparison between flat sheet and hollow fiber UF membranes is shown in Fig. 3. For these tests, the flat sheet membranes were specially fabricated for this project and were from the same supplier as the hollow fiber membranes, so there was greater assurance of similarity in the material chemistry. The nominal MWCOs were identical. The initial permeabilities were also more consistent; 674 L/m2 h bar for the flat sheet membrane versus 306 L/m2 h bar for the hollow fiber membrane. Despite the greater similarity between the flat sheet and hollow fiber products, the flat sheet membrane

400

700 Hollow fiber

Specific flux (L/m2.h.bar)

600 substantial difference in permeability may have been because of structural differences such as porosity, pore geometry, specific surface area, or thickness. Experience has shown that membranes with higher permeability consistently foul faster than membranes with lower permeability (Howe, 2001). The water velocity through the membrane may influence hydrodynamic conditions that affect how foulants come into contact with the membrane. At slower velocities, foulants may follow the water streamlines through the membrane and have less opportunity for contact with the membrane, whereas higher velocity may lead to more transportation of foulants to the membrane surface. The importance of initial permeability on the rate of fouling is shown in Fig. 2, where the specific flux has not been normalized against the initial permeability. The high permeability of the flat sheet membrane caused rapid fouling, but at the end of the filter run, the flux is nearly the same as the hollow fiber membrane. In fact, the specific flux at the end of the run is 196 L/m2 h bar for the hollow fiber membrane and 200 L/m2 h bar for the flat sheet membrane. In other words, the flux was nearly identical at the end of the filter run, and the effect of the higher fouling in the flat sheet membrane was merely to bring the flux to the level of the hollow fiber membrane. Ultimately, the important issue in membrane filtration is long-term flux, and high initial fouling of a membrane with high permeability may not be as significant in long-term operation.

200 300 Specific volume(L/m2)

Flat sheet

500 400 300 200 100 0 0

100

200 300 Specific volume (L/m2)

400

Fig. 3 – Comparison of flux of flat sheet and hollow fiber polysulfone UF membranes. (A) Normalized specific flux, (B) specific flux (source water B).

still fouled worse than the hollow fiber membrane when compared on a normalized specific flux basis, as in panel A of Fig. 3. In fact, the fouling of the flat sheet membrane appears more dramatic when the specific flux is not normalized, as shown in panel B of Fig. 3. The flux at the end of the test was 162 L/m2 h bar for the flat sheet membrane versus 248 L/m2 h bar for the hollow fiber membrane, even though the flat sheet membrane started with a higher permeability. These results clearly show a greater degree of fouling by flat sheet membranes, even when both configurations are the same material and produced by the same supplier. These results were consistent in nearly all tests for both MF and UF membranes over a range of source waters and operating conditions throughout the project (Adham et al., 2006).

3.3.

Effect of source water pretreatment

Although this research demonstrated that flat sheet membranes nearly always foul more rapidly than hollow fiber membranes, bench-scale flat sheet membrane testing may still be useful for screening treatment options if the change in fouling induced by treatment options is consistent between

ARTICLE IN PRESS 3846

WAT E R R E S E A R C H

41 (2007) 3842– 3849

Hollow fiber - raw water

1.0

Hollow fiber - prefiltered Flatsheet - raw water

0.8

0.6

Normalized specific flux

Normalized specific flux

1.0

Hollow fiber - raw Hollow fiber - coagulated Flat sheet - raw

0.4

Flat sheet - coagulated

0.2

0.8

Flatsheet - prefiltered

0.6 0.4 0.2 0.0

0.0 0

100

200 300 Specific volume (L/m2)

0

400

100

200 Specific volume

300

400

(L/m2)

Fig. 4 – Comparison of flux of flat sheet and hollow fiber PVDF MF membranes with and without coagulation (source water B).

Fig. 5 – Comparison of flux of flat sheet and hollow fiber polysulfone UF membranes with and without prefiltration (source water C) (Adham et al., 2006).

flat sheet and hollow fiber membranes; i.e., if the trends are consistent despite the differences in the absolute extent of fouling. To explore this possibility, side-by-side tests with flat sheet and hollow fiber membranes were conducted with and without coagulation pretreatment. In most cases, the relative change in fouling was similar for both configurations. With one source water (a different source than the source for Figs. 1 and 2), however, opposite trends were observed, as shown in Fig. 4. In this group of tests, coagulation pretreatment caused a decrease in fouling through the hollow fiber membrane and an increase in fouling through the flat sheet membrane. Based on these results, flat sheet membranes cannot be relied upon to always predict the relative change in performance for full-scale systems that will be using hollow fiber membranes. Consistent trends were observed, however, for pretreatment that involved physical changes to the feed water quality, such as prefiltration to remove particles above a particular size. The flux through flat sheet and hollow fiber membranes with and without prefiltration to remove particles larger than 1 mm is shown in Fig. 5. These results indicate that the removal of particles larger than 1 mm caused a slight improvement in membrane performance, and the change was similar for both membrane configurations. The fact that the majority of membrane foulants in this source water were smaller than 1 mm is a result that can be observed with either membrane configuration. Throughout this project, tests with prefiltration of the feed water consistently gave similar results for flat sheet and hollow fiber membranes. The difference between the results demonstrated in Figs. 4 and 5 may be due to the nature of changes induced by the pretreatment. Prefiltration is a physical change, only removing particles above a specific size, and a physical removal of foulants may be expected to produce similar changes in fouling regardless of the membrane configuration. Coagulation, however, is a chemical pretreatment that changes both the particles and the dissolved matter in a water sample. Particles are aggregated into larger sizes and removed by settling prior to membrane filtration, and dissolved matter is partially removed by adsorption onto the floc. Coagulation

preferentially removes the higher-molecular-weight constituents of natural organic matter (Randtke, 1988), which have also been shown to preferentially contribute to membrane fouling (Yuan and Zydney, 2000; Howe and Clark, 2002). Previous work has demonstrated that coagulation can cause either an increase or a decrease in membrane fouling, depending on the coagulant dose (Howe and Clark, 2006a). These complex interactions occur because coagulation can change the size distribution of materials that contribute to the fouling of an MF membrane. After coagulation with alum, the materials that contribute most to fouling of MF membranes shift from materials smaller than 1 mm to materials larger than 1 mm in diameter (Howe et al., 2006b). In this particular source water, the dissolved organic carbon concentration was less than 1 mg/L, indicating a low concentration of foulants in the natural water, and consequently a relatively low level of fouling of the raw water. The formation of foulants larger than 1 mm apparently had a greater effect on the flat sheet membrane.

3.4.

Resistance coefficients

The calculation of resistance coefficients from flux data can provide additional insight into graphical flux data. Resistance coefficients calculated from the flux data in Figs. 1–5 are shown in Table 1. In Figs. 1 and 2, the membrane resistance of the hollow fiber membrane is an order of magnitude greater than the flat sheet membrane. The fouling resistance of the flat sheet membrane, however, is only slightly greater than the fouling resistance of the hollow fiber membrane, indicating that the dramatic difference in flux decline curves in Figs. 1 and 2 is mostly due to the higher initial permeability of the flat sheet membrane rather than a dramatic difference in the extent of fouling. In contrast, the difference in fouling shown in Fig. 3 is not an artifact of the initial permeability. Table 1 demonstrates that the membrane resistance of the hollow fiber is about 2 times greater than the flat sheet membrane, but that the fouling resistance of the flat sheet membrane is about 8 times greater than the hollow fiber membrane.

ARTICLE IN PRESS WAT E R R E S E A R C H

3847

4 1 (200 7) 384 2 – 384 9

Table 1 – Resistance coefficients Membrane resistance, kM (  1011 m1)

Fouling resistance, kF (  1011 m1)

Figs. 1 and 2 (PVDF MF membranes) Flat sheet Hollow fiber

1.17 11.5

16.8 12.2

Fig. 3 (polysulfone UF membranes) Flat sheet Hollow fiber

5.3 11.8

22.1 2.8

Fig. 4 (PVDF MF membranes) Flat sheet, raw water Flat sheet, coagulated Hollow fiber, raw water Hollow fiber, coagulated

1.2 1.3 23.8 22.4

0.3 14.1 10.5 4.9

Fig. 5 (polysulfone UF membranes) Flat sheet, raw water Flat sheet, prefiltered Hollow fiber, raw water Hollow fiber, prefiltered

4.7 4.1 9.7 9.8

35.6 10.8 6.2 5.4

Fig. 6

16.0

Fig. 7

13.3

20.7 (kIR) 10.7 (kR) 12.1 (kIR) 3.2 (kR)

Table 1 shows an increase in fouling resistance after coagulation for the flat sheet membrane, and a decrease in fouling resistance after coagulation for the hollow fiber membrane. These changes are consistent with the flux data in Fig. 4. Similarly, the changes in fouling resistance after prefiltration to remove 1-mm particles are consistent with Fig. 5.

3.5.

Fouling over multiple filter runs

As noted earlier, a disadvantage of bench-scale membrane experimentation is that fouling is often evaluated over a single filter run, starting with a new, unused membrane. The initial irreversible fouling of a new membrane may not be representative of longer-scale fouling trends, particularly if the initial fouling is high due to a high permeability, as shown in Fig. 2. Because of the inherent structural integrity of hollow fiber membranes, it is easier to incorporate a backwash cycle into a bench-scale hollow fiber configuration than a flat sheet apparatus. Fig. 6 demonstrates data generated from a bench-scale system designed to operate over multiple filter runs. These data include the flux decline over a single filter run, as well as the longer-term fouling over multiple filter runs. It is evident from the data that the first filter run has more significant fouling than individual filter runs later in the test. Therefore, the first filter run (using a clean membrane) is less representative of the fouling expected in a full-scale system than individual filter runs with a used filter. Furthermore, the rate of fouling experienced over any individual filter run is more pronounced than the longer term fouling trend.

0.8 0.6 0.4 0.2 0.0 0

500

1000

1500

2000

Specific volume (L/m2)

Fig. 6 – Normalized specific flux of raw water in bench-scale hollow fiber testing with backwashing between filter runs (source water D).

1.0 Normalized specific flux

Figure

Normalized specific flux

1.0

0.8 0.6 0.4 0.2 0.0 0

500

1000 1500 Specific volume (L/m2)

2000

Fig. 7 – Normalized specific flux of pretreated water in bench-scale hollow fiber testing with backwashing between filter runs (source water D).

A test evaluating the fouling of the same water and same operating conditions, but after pretreatment consisting of oxidation, lime softening, recarbonation, and adsorption is shown in Fig. 7. Several differences between Figs. 6 and 7 are evident. First, the extent of fouling of the first filter run is different. Second, the fouling of each individual filter run is much less after pretreatment than in the raw water. Finally, the rate of fouling over multiple runs appears to be less than with raw water. As with single filter runs, resistance coefficients can provide insight into the data, except that now three resistance coefficients can be calculated for each test: the membrane resistance (kM), the irreversible fouling resistance due to the long-term adsorption of foulants that are not removed by backwashing (kIR), and the reversible fouling resistance (kR), which is removed at the end of each backwash cycle. Comparison of the resistance coefficients in Table 1 for Figs. 6 and 7 indicates that the pretreatment reduced the irreversible fouling resistance by about 40 percent and the

ARTICLE IN PRESS 3848

WAT E R R E S E A R C H

41 (2007) 3842– 3849

reversible fouling resistance by about 70 percent. The ability to assess both irreversible and reversible fouling, using a membrane configuration identical to the configuration used at full scale has greater utility for evaluating treatment options at site-specific applications. Kim and DiGiano (2006) found reasonable correspondence between bench and pilot data using a bench hollow fiber system similar to that presented here. In contrast, a comparison of bench and pilot data using a bench flat sheet system found that relative trends in performance were often similar, but the absolute extent of fouling differed significantly between the bench and pilot scales (Adham et al., 2006). It should be noted that the Kim and DiGiano (2006) bench configuration operated in a constant flux, rising pressure mode, an additional improvement for comparison with full-scale systems.

washing incorporated as an integral part of the filtration cycle.

Acknowledgments Hollow fiber membranes for the various tests were provided by Koch, US Filter, and Aquasource. This manuscript is partially based on work funded by the American Water Works Association Research Foundation. The comments and views detailed herein may not necessarily reflect the views of AWWARF, its officers, directors, affiliates, or agents. Additional support for a portion of the work was provided by Carollo Engineers; Daniel Hugaboom is recognized for his support of the project. R E F E R E N C E S

4.

Conclusions

Although bench-scale testing with either flat sheet or hollow fiber membranes is present in the literature, an explicit comparison between these configurations has not previously been available. This research provides the needed comparison between these membrane configurations and arrives at the following conclusions: 1. Flat sheet bench-scale membrane testing has been instrumental in developing a mechanistic understanding of membrane filtration and fouling. 2. Flat sheet membranes typically fouled more than hollow fiber membranes in this study. This can complicate the use of flat sheet membranes for assessing treatment options for full-scale applications that will use hollow fiber membranes. The exact mechanistic cause of differences in fouling between configurations cannot be deduced from these experiments, but it is possible that differences in the fabrication process could lead to differences in surface roughness, porosity, pore geometry, or thickness, all of which could affect the hydrodynamics of water flow through the membrane and the propensity of fouling. 3. Although pretreatment typically causes the same relative changes in performance in flat sheet and hollow fiber membranes, cases have been observed where coagulation caused an increase in fouling on flat sheet membranes and a decrease in fouling on hollow fiber membranes. 4. Resistance coefficients can provide insight into the differences between the fouling of flat sheet and hollow fiber membranes. As presented here, bench-scale tests using hollow fiber membranes that operate over multiple filter runs have the potential to provide more robust data regarding the fouling expected in larger scale systems. Thus, when benchscale testing is to be used to evaluate treatment options for site-specific applications, it is recommended that the benchscale apparatus incorporate hollow fiber membrane modules that can be operated over multiple filter runs, with back-

Adham, S.S., Chiu, K.-P., Lehman, G., Howe, K.J., Marwah, A., Mysore, C., Clouet, S.J., Do-Quang, Z., Cagnard, O., 2006. Optimization of Membrane Treatment for Direct and Clarified Water Filtration. AWWA Research Foundation, Denver, CO. APHA, AWWA and WEF, 2005. Standard Methods for the Examination of Water and Wastewater, 21st ed. APHA, Washington, DC. AWWA Membrane Technology Research Committee, 2005. Committee report: recent advances and research needs in membrane fouling. J. Am. Water Works Assoc. 97 (8), 79–89. Chang, Y.J., Benjamin, M.M., 1996. Iron oxide adsorption and UF to remove NOM and control fouling. J. Am. Water Works Assoc. 88 (12), 74–88. Crittenden, J.C., Trussell, R.R., Hand, D.W., Howe, K.J., Tchobanoglous, G., 2005. Water Treatment Principles and Design, second ed. Wiley, Hoboken, NJ. Elimelech, M., Zhu, X., Childress, A.E., Hong, S., 1997. Role of membrane surface morphology in colloidal fouling of cellulose acetate and composite aromatic polyamide reverse osmosis membranes. J. Membr. Sci. 127 (1), 101–109. Heijman, S.G.J., Vantieghem, M., Raktoe, S., Verberk, J.Q.J.C., van Dijk, J.C., 2007. Blocking of capillaries as fouling mechanism for dead-end ultrafiltration. J. Membr. Sci. 287 (1), 119–125. Hermia, J., 1982. Constant Pressure Blocking Filtration Laws— Application to Power-law Non-Newtonian Fluids. Trans. Inst. Chem. Eng. 60, 183–187. Howe, K.J., 2001. Effect of coagulation pretreatment on membrane filtration performance. Ph.D. Thesis, University of Illinois at Urbana-Champaign, Urbana, IL. Howe, K.J., Clark, M.M., 2002. Fouling of microfiltration and ultrafiltration membranes by natural waters. Environ. Sci. Technol. 36 (16), 3571–3576. Howe, K.J., Clark, M.M., 2006a. Effect of coagulation pretreatment on membrane filtration performance. J. Am. Water Works Assoc. 98 (4), 133–146. Howe, K.J., Marwah, A., Chiu, K.P., Adham, S.S., 2006b. Effect of coagulation on the size of MF and UF membrane foulants. Environ. Sci. Technol. 40 (24), 7908–7913. Jacangelo, J.G., Madec, A., Schwab, K.J., Huffman, D.E., Mysore, C.S., 2005. Classification of low-pressure, hollow fiber membranes through bacterial and viral profiling. In: 2005 AWWA Membrane Technology Conference, Phoenix, AZ, American Water Works Association. Jucker, C., Clark, M.M., 1994. Adsorption of aquatic humic substances on hydrophobic ultrafiltration membranes. J. Membr. Sci. 97, 37–52.

ARTICLE IN PRESS WAT E R R E S E A R C H

4 1 (200 7) 384 2 – 384 9

Kim, J., DiGiano, F.A., 2006. A two-fiber, bench-scale test of ultrafiltration (UF) for investigation of fouling rate and characteristics. J. Membr. Sci. 271 (1–2), 196–204. Laabs, C.N., Amy, G.L., Jekel, M., 2006. Understanding the size and character of fouling-causing substances from effluent organic matter (EfOM) in low-pressure membrane filtration. Environ. Sci. Technol. 40 (14), 4495–4499. Laıˆne´, J.-M., Hagstrom, J.P., Clark, M.M., Mallevialle, J., 1989. Effects of ultrafiltration membrane composition. J. Am. Water Works Assoc. 81 (11), 61–67.

3849

Randtke, S.J., 1988. Organic contaminant removal by coagulation and related process combinations. J. Am. Water Works Assoc. 80 (5), 40–56. Serra, C., Durand-Bourlier, L., Clifton, M.J., Moulin, P., Rouch, J.C., Aptel, P., 1999. Use of air sparging to improve backwash efficiency in hollow-fiber modules. J. Membr. Sci. 161 (1-2), 95–113. Yuan, W., Zydney, A.L., 2000. Humic acid fouling during ultrafiltration. Environ. Sci. Technol. 34 (23), 5043–5050.