Effect of Cake Layer Structure on the Backwash ...

J. Environ. Sci. Eng. (2001), 3, 50-58

Effect of Cake Layer Structure on the Backwash Efficiency of Hollow Fiber MF Membranes Do Hee Kim*, Seungkwan Hong1 and Jaeweon Cho Department of Environmental Science and Engineering, Kwangju Institute of Science and Technology (K-JIST) Oryong-dong, Buk-gu, Gwangju 500-712, Korea 1 Civil and Environmental Engineering Department, University of Central Florida P.O. Box 162450, Orlando, Florida 32816-2450, USA A series of filtration experiments were performed to investigate physical and chemical factors affecting the efficiency of backwashing during MF filtration of colloidal suspensions. In this study, all experiments were conducted in dead-end filtration mode utilizing an outside-in, hollow-fiber module with nominal pore size of 0.1 µm. The silica particles (mean diameter = 0.14 µm) were used as model colloids. As expected, fouling experiments clearly showed that the severity of particle fouling increased with an increase in particle concentration. Under identical particle loading, particle fouling was enhanced with increasing operating pressure and solution ionic strength. Using a flux decline model based on the Happels cell for the hydraulic resistance of the particle layer, the cake structure was determined from fouling experimental data and then correlated to backwash efficiency. Modeling of experimental data revealed no changes in cake layer structure when feed particle concentration was increased. However, the packing density of the cake layer (1cake porosity) increased slightly with operating pressure and drastically with ionic strength. Lastly, the results of backwashing experiments demonstrated that the efficiency of backwashing deceased with increasing transmembrane pressure and solution ionic strength, while there was no clear trend observed between backwash efficiency and particle concentration. This finding clearly suggested that the backwash efficiency is closely related to the structure of the cake layer formed during particle filtration. More densely packed cake layers were formed under high ionic strength and operating pressures, and consequently less flux was recovered per given backwash volume during backwashing. Key words : Colloidal Fouling, Microfiltration, Backwashing Efficiency, Cake Layer Structure, Water Treatment

Introduction The use of membrane technologies has increased dramatically over the recent years. While this technology was previously sparse, membranes are now commonplace in numerous industrial processes (Tran-Ha and Wiley, 1998; Wang and Song, 1999; Marchese et al., 2000), wastewater treatment plants (Kennedy et al., 1998; Bourgeous et al., 2000; Decarolis et al., 2001), and drinking water treatment plants (Jonsson and Jonsson, 1996; Boerlage et al., 1998; Hagen, 1998; Hillis et al., 1998; Lipp et al., 1998; Vos et al., 1998; Yuan and Zydney, 2000). The process by which MF/ UF membranes operate is elegantly simple. They present a physical barrier to the suspended particles contained in the feed stream, whereby all particles larger than the pore are retained on the feed side of the membrane. However, it is *Corresponding author Tel. 062-970-2449, Fax. 062-970-2434 E-mail:[email protected]

the very simplicity of this operation that also presents a significant drawback for this separation technology. With time, the retained particles accumulate on the surface of the membrane and increase the resistance to water flow across the membrane (Wang et al., 1999). As a result of fouling, the MF/UF processes must be periodically stopped to reverse the direction of flow through the membrane. This process, called backwashing, recovers a portion of productivity lost through operation by removing materials deposited on the surface of the membrane (Chellam et al., 1988) (Fig. 1). Some of the most common causes of fouling include deposition of a cake layer (Belfort et al., 1994; Bacchin et al., 1995), pore blocking (Koltuniewicz et al., 1996; Timmer et al., 1997; Giorno et al., 1998; Huang et al., 1998), and particle adsorption (Clark et al., 1991; Kim et al., 1992; Crozes et al., 1993 Jucker et al., 1994; Tracey et al., 1994). Amongst these, cake formation is perhaps the most dominant mode of long term membrane fouling (Belfort et al., 1994; Song, L.F., 1998). Based on kinetic models of adsorption or pore blocking, it is generally observed that fouling by these processes is rapid, and is jointly responsible with

EFFECT OF CAKE LAYER STRUCTURE ON THE BACKWASH EFFICIENCY OF HOLLOW FIBER MF MEMBRANES 51

Fig. 1. Typical UF/MF membrane operations for various industrial separation processes.

concentration polarization for the rapid initial flux decline (Boyd, R.F. and Zydney, 1998; Song, L.F., 1998). Apart from this, there is a considerable doubt regarding the actual mechanism of protein or macromolecular fouling in UF and MF processes, in which the macromolecular species are known to foul the membranes by adsorption as well as by formation of a cake layer (Clark et al., 1991; Kim et al., 1992; Marshall, A.D., 1993; Boyd, R.F. and Zydney, 1998). In this study, we will focus on fouling by cake formation. The dynamics of cake layer growth and evaluation of the cake layer properties, like structure and permeability, have received considerably greater attention over the recent years. There has been a significant resurgence of studies that attempt to directly incorporate colloidal interactions in predicting the structure and permeability of the cake layer deposits. Such studies were pioneered by a set of experimental observations by McDonogh et al. (1984). Further studies were pursued to assess the influence of particle charge, background electrolyte concentration, and other physico-chemical conditions on the permeability of the filter cake (McDonogh, R.M., et al., 1989; McDonogh, R.M., et al., 1992). There have been a variety of studies pertaining to elucidation of the structure of the cake layer under the influence of colloidal and hydrodynamic interactions, some even involving computer simulations (Fu, L.F. and Dempsey, B.A., 1998). In these studies, Cake porosity was assumed for maximium random packing of spheres (i.e., ε = 0.36). However, it is not true sometimes. Through the classical DLVO theory, the stability of a colloidal suspension is a function of solution chemistry such as ionic strength and pH. Several studies were conducted to investigate the effect of these parameters on colloidal fouling. As the ionic strength of a solution is increased, the magnitude of the repulsive electrical double layer (EDL) is reduced. This reduces the stability of particles and results in a more

densely packed cake layer. As a result, the resistance to permeate flow is increased and membrane productivity is decreased (Jonsson and Jonsson, 1996; Yiantsios and Karabelas, 1998). The same trend is observed with respect to pH. As the pH is decreased, the magnitude of the repulsive EDL is reduced, thereby reducing the stability of the colloidal particles and increasing membrane fouling (Jonsson and Jonsson, 1996). Therefore, significant change will be expected with the physical and chemical properties of the feed solution containing particle. Despite its importance, no literature was found the effect of the chemical and physical effect of the feed solution to the cake structure. The purpose of this study was to examine the effect of feed water quality and operational parameters on the efficiency of backwashing. In addition, Increased system efficiency and reduced operational costs can be achieved with improved performance of backwashing and reducing the frequency of chemical cleaning. Several researches have studied about the backwashing in the MF/UF filtration. These studies can be classified 3 categories: (1) the effect of backwashing in the filtration, (2) the effect of physical conditions of backwashing, and (3) additional step for improving backwashing step. All studies about backwashing showed that backwashing step recovered at least 90% initial flux, increased the life of membrane, and decreased the operation coast. Kennedy et al. (1998) showed the effect of backwashing conditions such as pressure and duration. The efficiency of backwashing was more dependent on the backwashing time than time. Several researchers reported the effect of backwashing cycle. In these studies, backwash frequency (shorter operation times between backwash cycles) and an increase in backwash duration (volume backwashed) reduced membrane fouling (Xu et al., 1995; Chellam et al., 1998; Hillis et al., 1998; Decarolis et al., 2001). Serra and coworker (1999) descried that the air sparing during backwashing can improve backwashing efficiency. However, there is no fundamental study about the backwashing (i.e. principle and mechanism of the backwashing). In this study, It is hypothesized that backwashing efficiency can be affected by the cake structure near membrane. The relation between cake structure and backwashing efficiency was studied. Therefore, in this study, a series of fouling experiments were first performed under various operating conditions. The primary operating parameters varied throughout the experiments were particle concentration, operating pressure and solution ionic strength. Furthermore, a theoretical model for flux decline due to cake formation was evaluated and utilized to determine the structure of the cake layer formed during fouling experiments. Following each fouling exper-

52

KIM, DO HEE ET AL.

iment, the membrane was backwashed in order to relate the backwash efficiency to both physical and chemical parameters. Finally, a theoretical value for the particle packing density of the cake layer calculated from the model was correlated to the backwash efficiency in order to elucidate the effect of cake structure on the efficiencies of backwashing.

and NaCl (Fischer Scientific, Pittsburgh, PA). These salts were dissolved in DI water (LD5A and MegaPure, Barnstead/Thermolyne, Dubuque, IO). Adjustments in pH, for both zeta potential measurements and fouling studies, were made with ACS grade HCl. Lastly, all cleaning solutions were made with USP grade sodium hydroxide and citric acid dissolved in DI water.

Experimental Microfiltration Membranes All experiments conducted during this study utilized a bench-scale, outside-in, hollow-fine fiber, MF module (SK Chemicals, Seoul, South Korea). Manufacturer’s specifications revealed the following physical characteristics of the membrane: nominal pore size of 0.1 µm, inside fiber diameter of 0.7 mm (2.3×10−3 ft), outside fiber diameter of 1.0 mm (3.3×10−3 ft), and a fiber length of 520 mm (1.7 ft). Containing a total of 150 hollow-fine fibers, the MF module provided approximately 0.25 m2 (2.7 ft2) of membrane area. The average specific flux of this membrane was estimated at 3.55±0.20 lmh/kPa (14.43 ± 0.83 gfd/psi) under given operating conditions. Colloidal Particles Silica (SiO2) particles from Nissan Chemical Industries (Houston, TX) were used as model colloids for all of the fouling and backwashing experiments. The particles were received as stable concentrated (40.7% by weight) aqueous suspensions at an alkaline pH. The manufacturer’s certificate reported a mean particle diameter of 0.10±0.03 µm (as determined by the centrifugal method) and a specific gravity of 1.301 at 20°C. The size and shape of these model colloids were further verified by a scanning electron microscope (JEOL Model 400, JEOL, Peabody, MA) and by a dynamic light scattering (Nicomp Model 380, Particle Sizing Systems, Santa Barbara, CA). The SEM images and DLS analyses revealed that the model silica particles are monodispersed with a mean particle size of 140 nm. Lastly, the zeta potential of the colloidal silica was determined from electrophoretic mobility measurements (Zeta PALS, Brookhaven Instruments Corp., NY). The results showed the zeta potentials of silica particles to be in the range of −27 to −30 mV at the pH 8 and 10−2 M NaCl, which was the solution environments employed in the majority of fouling and backwash experiments. More detailed properties of these silica particles are well documented in a recent paper by Vrijenhoek et al. (2001). Standards and Reagents All solutions were prepared with ACS grade NaHCO3,

Bench-Scale Membrane Filtration Unit The colloidal suspensions were prepared and stored in a magnetically stirred high-density polyethylene 20 L (5.3 gal). The temperature of this suspension was maintained at 20°C (68°F) by a Neslab CFT-33 (Portsmouth, NH) digital refrigerated recirculator. The feed suspension was delivered to the MF module by a 6.83 lpm (1.8 gpm) constant flow diaphragm pump (Hydracell, Wanner Engineering, Inc., Minneapolis, MN) with a maximum pressure of 3,447 kPa (500 psi). Initial operating conditions (e.g. filtrate flux and crossflow velocity) were set and maintained through the careful manipulation of feed, concentrate, and bypass needle valves (Swagelok, Solon, OH). Feed pressure was monitored by an analog pressure gauge (Dresser Industries, Stratford, CT). Filtrate and concentrate flows were measured both by an in-line flowmeter (Blue-White Industries, Westminster, CA) and by the timed collection of filtrate in a graduated cylinder. Sequence of Fouling and Backwash Experiments Prior to each fouling experiment, the bench-scale MF unit was thoroughly cleaned by recirculating sodium hydroxide (pH 11) and citric acid solutions (pH 3) for a minimum of one hour, respectively. In addition, the module was backwashed with these solutions at a pressure of 68.9 kPa (10 psi). After chemical cleaning was completed, the system was rinsed and flushed with DI water. An initial clean water test was performed to determine membrane productivity prior to each fouling experiment. Each clean water test was conducted in a dead-end mode of operation at a feed pressure of 41.4 kPa (6 psi) with a background electrolyte solution identical to that which would be used for the ensuing fouling study (e.g. 10−3 M NaHCO3 and 10−2 M NaCl). A total of 5 L of filtrate was collected, and a stopwatch was used to measure the collection times associated with 1, 2, 3, 4, and 5 L of filtrate accumulated. Following the initial clean water test, feed, concentrate, and bypass valves were manipulated to achieve the desired initial operating conditions for the fouling study. Once stable operation was attained, the predetermined volume of concentrated silica particles was added to the feed solution to achieve the desired particle concentration. Immediately

EFFECT OF CAKE LAYER STRUCTURE ON THE BACKWASH EFFICIENCY OF HOLLOW FIBER MF MEMBRANES 53

following the addition of silica particles, 5 L of filtrate was collected, and collection times were measured and recorded for 1, 2, 3, 4, and 5 L. Once the fouling study was completed, 1 L of DI water was backwashed through the MF module at a pressure of 68.9 kPa (10 psi), and a final clean water test was conducted. Similar to the initial clean water test, the final clean water test was conducted in a dead-end mode of operation at a feed pressure of 41.4 kPa (6 psi) with a background electrolyte solution identical to that which was used for the previous fouling study. Again, a total of 5 L of filtrate was collected, and a stopwatch was used to measure the collection times associated with 1, 2, 3, 4, and 5 L. Feed and filtrate samples for each fouling study were collected and analyzed for conductivity, pH, and turbidity. Conductivity and pH measurements were made with an Accumet AR-50 (Fischer Scientific, Pittsburgh, PA) and turbidity was determined using a Hach Ratio Turbidimeter (Loveland, CO). Due to the fact that the particles used throughout the experiments were larger than the membrane pores (approximately 0.14 µm and 0.10 µm, respectively), filtrate turbidity was always below the detection limit (≈0.01 ntu) regardless of experimental conditions, indicating that all particles were retained on the feed side of the membrane and formed a particle cake layer on the membrane surface. Evaluation of Flux Decline and Backwash Efficiency In order to compare multiple data sets obtained under various experimental conditions, it was necessary to analyze all experiments on a dimensionless basis. Two parameters of particular interest throughout this study were the normalized flux (Jn) and the backwash efficiency (η). The normalized flux was calculated based on Jw J n = -----Jo

(1)

where Jw was the flux after the collection of 5 L of filtrate and Jo was the initial filtrate flux. Similarly, the backwash efficiency (η) was estimated by t η = ---i tf

(2)

where ti was the time required to collect 5 L of filtrate during the initial clean water test and tf was the time required to collect the same 5 L during the final clean water test.

Results and Discussion Cake Layer Structure In pressure driven membrane filtration of colloidal

suspensions, particles are transported to the membrane surface by the filtrate flow and form a cake layer on the membrane surface. Particle accumulation in the cake layer provides an additional resistance to filtrate flow and, hence, reduces flux. Resulting pressure drops in the membrane system can be expressed as ∆P = ∆P m + ∆P c

(3)

As shown in Equation 3, the applied (transmembrane) pressure drop (∆P) is equal to the sum of the pressure drops across the membrane (∆Pm) and the cake layer (∆Pc). The pressure drop across the membrane is simply the product of membrane resistance Rm and filtrate flux Jw: ∆P m = J w R m

(4)

The pressure drop in the cake layer is associated with the frictional drag resulting from the flow of filtrate through the dense layer of accumulated particles: kT ∆Pc = ------ A S ( θ )J w M c D

(5)

Here, kT/D (= 6πµap) is the frictional drag coefficient, k is the Boltzmann constant, T is absolute temperature, D is the particle diffusion coefficient, µ is the solvent viscosity, ap is the particle radius, and Mc is the total number of particles (per unit area) accumulated in the cake layer. The AS(θ) is a correction function accounting for the effect of neighboring retained particles and can be evaluated from Happel’s cell model [Song and Elimelech, 1995]: 2 1 + --- θ 5 3 A S = ------------------------------------------3--3 1 – θ + --- θ 5 – θ 6 2 2

(6)

where θ = (1 − ε)1/3 is a porosity dependent variable, with ε being the porosity of the cake layer of accumulated particles. As shown in Equations 4 to 5, the pressure drop across the cake layer is primarily influenced by cake layer structure as well as particle size and concentration. The flux decline in membrane filtration of colloidal suspensions can be estimated based on Happel’s cell model for the hydraulic resistance of the particle cake layer formed: 3kTAS ( θ ) ∆PC 0 –1 / 2 Jw  - J n = ------ =  1 + -------------------------------------3 2 J0  2 π a p DRm J w A 

(7)

Here J0 is the initial filtrate flux, C0 is the bulk (feed) particle concentration, A is the membrane area, and V is the filtrate

54

KIM, DO HEE ET AL.

Fig. 2. Effect of colloidal concentration on flux decline of hollow fiber MF membranes. Symbols represent experimental results and solid lines represent theoretical predictions. Silica particles with a mean diameter of 140 nm were used as model colloids. Experiments were conducted at dead-end filtration mode. Filtration conditions employed were temperature = 20°C, transmembrane pressure = 41.4 kPa, pH = 8 and ionic strength = 10−2 M NaCl. The correction function (As) was varied as a fitting parameter in the theoretical calculations. The membrane resistance (Rm) for each run was determined based on initial flux (at t = 0) and transmembrane pressure.

volume. Detailed theoretical development is well presented in a paper by Hong et al. (1997). By utilizing Equation 6, structural characteristics of the cake layer formed under various operating conditions can be determined from filtration experiments. Specifically, the correction function, AS(θ), is estimated first by fitting flux decline experimental data and then the particle packing density (i.e. 1-cake porosity) is calculated based on Happel cell model. Particle Loading As expected from Equation 7, the extent of flux decline in dead-end filtration of colloidal suspensions is directly related to cumulative particle loading to the membrane system (= C0×V). In this study, particle volume concentrations were varied from 0.005% to 0.045% under identical physical and chemical operating conditions. The representative results are schematically presented in Fig. 2. Each experiment was repeated up to maximum of five times to ensure statistical reliability. The majority of experiments showed good reproducibility with a few exceptions, especially those performed with low particle concentrations. As shown in Fig. 2, the extent of membrane fouling increased with the concentration of colloidal particles. Specifically, after the filtration of 5 L, the averaged normalized flux values were 0.87, 0.79, 0.62, and 0.56 for feed particle concentrations of 0.005%, 0.015%, 0.030%, and 0.045%, respectively. The larger mass of particles was transported to

Fig. 3. Effect of colloidal concentration on particle packing density (1-cake porosity) for hollow fiber MF membranes. Experimental conditions identical to those described in Fig. 2.

the surface of the membrane for a suspension with a high feed particle concentration, resulting in the formation of a thicker cake layer on the membrane surface. This cake layer provided a greater resistance to filtrate flow across the membrane, caused significant pressure drop, and hence flux decline. In order to determine the structure of the cake layer formed under different particle concentrations, a plot of the average particle packing density (1-cake porosity) versus feed particle concentration was generated by following the steps described in the previous section. All of the operating parameters required for the model (refer to Equation 7), such as temperature, pressure, and flux, were carefully monitored throughout the fouling experiments. The resistance of the membrane for each experiment was determined from the initial filtrate flux and pressure drop across the clean membrane. The modeling results are presented in Fig. 4. As shown, the structure of the cake layer for each particle concentration was relatively consistent, with particle packing density factors ranging from 0.661 to 0.671. This observation suggested that the structure of the cake layer was independent of particle concentration. It is not surprising since all of the operating parameters remained constant except particle concentration, which only affected cake layer thickness, not structure. In addition to fouling experiments, backwashing studies were also conducted to determine the effectiveness of the backwashing under various feed particle concentration. The results of these experiments are summarized in Fig. 4. The average efficiency of all backwashing procedures ranged from 0.969 to 0.985, however, no clear correlation was observed between particle concentration and backwash efficiency considering variations. The apparent lack of

EFFECT OF CAKE LAYER STRUCTURE ON THE BACKWASH EFFICIENCY OF HOLLOW FIBER MF MEMBRANES 55

Fig. 4. Effect of colloidal concentration on backwash efficiency for hollow fiber MF membranes. Experimental conditions identical to those described in Fig. 2.

dependence of backwash efficiency on particle concentration was in accordance with no changes in particle packing density with feed particle concentrations. Thus, under the given range of particle loading to the membrane systems, the backwash efficiency was more closely related to cake structure than particle mass accumulated on the membrane surface. Operating Pressure Filtration experiments were also conducted to determine

Fig. 5. Effect of operating pressure on flux decline of hollow fiber MF membranes. Symbols represent experimental results and solid lines represent theoretical predictions. Silica particles with a mean diameter of 140 nm were used as model colloids. Experiments were conducted at dead-end filtration mode. Filtration conditions employed were temperature = 20°C, silica particle volume concentration = 0.005%, pH = 8 and ionic strength = 10−2 M NaCl. The correction function (As) was varied as a fitting parameter in the theoretical calculations. The membrane resistance (Rm) for each run was determined based on initial flux (at t = 0) and transmembrane pressure.

the effect of operating pressure (or initial flux) on particle fouling. It should be noted that, unlike the previous series of experiments, particle loading per unit membrane surface area was kept constant throughout the experiments. Thus, the number of particles transported to the surface of the membrane at any given volume of filtrate would be the same, regardless of the value of operating pressures. Results are presented for 5 values of operating pressures in Fig. 5. After the collection of 5 L of filtrate, operating pressure of 20.7, 34.5, 41.4, 65.6 and 41.4 kPa resulted in average normalized flux values of 0.920, 0.916, 0.867, 0.915, and 0.912, respectively, indicating that particle fouling increased slightly with increasing operating pressure except 41.4 kPa. The trend illustrated in Fig. 5 may be explained by the phenomenon of filtrate drag. As the value of operating pressure (initial flux) is increased, the force that transports suspended particles to the surface of the membrane is also increased. As the filtrate drag is increased, the particles carried to the membrane surface form a more densely packed cake structure which presents a greater resistance to permeate flow and results in a greater reduction of membrane productivity. However, the effect of permeation drag is not significant with operating range in this study as shown in the Fig. 5, that is, the same cake structure maintained during each different operating pressure experiment. Further verification of this hypothesis was made by plotting 1-cake porosity versus operating pressure as shown in Fig. 6, based on Equation 7. As shown, the structure of the cake layer for each particle concentration was relatively consistent, with particle packing density factors ranging from 0.661 to 0.671. Even though no definitive conclusion can be drawn from these data as all data points are within one standard deviation of each other, the average data points contained in this figure seem to indicate that the density of the cake

Fig. 6. Effect of operating pressure on particle packing density (1-cake porosity) for hollow fiber MF membranes. Experimental conditions identical to those described in Fig. 5.

56

KIM, DO HEE ET AL.

Fig. 7. Effect of operating pressure on backwash efficiency for hollow fiber MF membranes. Experimental conditions identical to those described in Fig. 5.

structure were almost same with increasing operating pressure. At the end of fouling runs, backwashing experiments were performed to evaluate the reversibility of fouling experienced during these fouling studies. Fig. 7 presents the results of these experiments. The average efficiency of the backwashing procedure for each operating pressure value was relatively consistent, with ranging from 0.98 to 0.99. This observation may suggest that the structure of the cake layer formed during each filtration was almost same within operation pressure range in this study, that is, permeation drag is not significantly effect to the cake structure within operating pressure ranging in this study when mass loading is same. Ionic Strength The effect of solution chemistry, specifically ionic strength, on the fouling of a hollow fiber MF module was also investigated. A total of three ionic strengths, spanning two orders of magnitude, were tested; 0.001 M, 0.01 M, and 0.1 M NaCl. Once again, the mass loading of particles on the surface of the membrane was held constant throughout all experiments conducted in this series, as the particle concentration was fixed at 0.005%. Furthermore, the operating pressure was set at 41.4 kPa for all experimental runs. The results of these experiments are shown in Fig. 8 and clearly showed that membrane fouling became more severe as the ionic strength of the solution was increased. The average normalized flux values after 5 L of filtrate was collected were 0.97, 0.87, and 0.81, for ionic strengths of 0.001 M, 0.01 M, and 0.1 M NaCl, respectively. Namely, cake structure can be affected significantly by ionic strength. These can be explained by the DLVO theory and cake structure. According to the classic DLVO theory of colloidal interactions, the

Fig. 8. Effect of ionic strength on flux decline of hollow fiber MF membranes. Symbols represent experimental results and solid lines represent theoretical predictions. Silica particles with a mean diameter of 140 nm were used as model colloids. Experiments were conducted at dead-end filtration mode. Filtration conditions employed were temperature = 20°C, transmembrane pressure = 41.4 kPa, silica particle volume concentration = 0.005%, and pH = 8. The correction function (As) was varied as a fitting parameter in the theoretical calculations. The membrane resistance (Rm) for each run was determined based on initial flux (at t = 0) and transmembrane pressure.

Fig. 9. Effect of ionic strength on particle packing density (1cake porosity) for hollow fiber MF membranes. Experimental conditions identical to those described in Fig. 8.

magnitude of the electrical double layer (EDL) is inversely proportional to the ionic strength of the solution. As the ionic strength of a solution increases, the EDL becomes more compressed and hence the EDL repulsion among particles is greatly reduced. This reduction in repulsive forces results in the formation of a more densely packed cake layer on the surface of the membrane, which offers a greater resistance to filtrate flow. This finding was further validated through the determination of cake structure at various ionic strengths based on the model previously

EFFECT OF CAKE LAYER STRUCTURE ON THE BACKWASH EFFICIENCY OF HOLLOW FIBER MF MEMBRANES 57

Fig. 10. Effect of ionic strength on backwash efficiency for hollow fiber MF membranes. Experimental conditions identical to those described in Fig. 8.

described in Section 3.1. The modeling results are summarized in Fig. 9 and clearly demonstrated that the density of the cake layer increased with the ionic strength of the particle suspension. Lastly, backwashing studies were performed to investigate if a relationship exists between the ionic strength of the solution and the efficiency of backwashing events. The results indicated that the average efficiency of backwashing events decreased as the ionic strength of the feed solution increased, as presented in Fig. 10. The average backwash efficiency for solution ionic strengths of 0.001 M, 0.01 M, and 0.1 M NaCl were estimated at 0.986, 0.982, and 0.962, respectively. Correspondingly, it is shown that the cake layers formed during the filtration of high ionic strength suspensions were more densely packed than those formed during the filtration of low ionic strength suspensions, as predicted by the theory of colloidal interactions. These finding further suggest that the efficiency of backwashing events is a function of the structure of the cake layer formed during the filtration process. Specifically, the efficiency of the backwashing decreased with an increase in the density of the cake layer. At high ionic concentrations, a more densely packed cake layer is formed due to repressed electrostatic interactions among particles, and consequently less flux was recovered per given backwash volume.

Conclusions Conclusions Primary inferences from this research are summarized as follows: 1. An increase in particle concentration resulted in a reduced normalized flux under identical operating conditions, which was attributed to the formation of a thicker

cake layer caused by an increase in particle loading. Despite varying degrees of cake thickness, the structure of the cake layer, as determined by the Happels cell model, did not vary with particle concentration. In addition, the efficiency of backwashing procedure remained relatively constant throughout all particle concentration experiments, which may suggest close relationship between backwash efficiency and cake structure. 2. Under identical particle loading, a slight decrease in normalized flux was observed with increasing operating pressure, which may partially be ascribed to the compression of the cake structure at elevated operating pressures. The effect of the operating pressure, however, is not significant in this study. The modeling of experimental data showed evidence of slightly reduced cake porosity at higher operating pressures. The backwash efficiency was also slightly decreased as operating pressure increased, but the variations of both cake structure and backwash efficiency were very small, indicating that operating pressure had a marginal impact on cake structure and hence backwash efficiency, at least for the range of operating pressures studied. 3. The normalized flux was significantly reduced as the ionic strength of the feed solution increased even when particle loading to the membrane system was kept constant. This was explained by the formation of a more compact cake layer at higher salt concentrations. Modeling data clearly demonstrated a direct relationship between the density of the cake layer (i.e. 1cake porosity) and ionic strength, as predicted by the classical colloidal interactions among the particles accumulated in the cake layer. Furthermore, an inverse relationship was observed between backwash efficiency and the ionic strength of the feed solution, which indicates that the efficiency of backwashing deteriorated as the packing density of the cake layer increased. Recommendations: 1. Further research into the various other particle types and sizes is needed to ascertain impact of the same on backwashing efficiency. 2. It is recommended that more in depth study be conducted to investigate the impact of pressure changes on the operation of microfiltration membrane operations. 3. The impact of solution chemistry such as pH on the MF membrane performance should be studied intensively. 4. The volume of backwash water increase or any other chemical anti-foulant usage should also be investigated for practical purposes. 5. The impact of natural waters on the backwashing efficiency should also be studied in depth to practically

58

KIM, DO HEE ET AL.

apply these kinds of studies. 6. The efficiency of chemical backwashing using surfactant and/or oxidants should also be investigated with various types of source waters.

Acknowledgements The authors would like to acknowledge the American Water Works Association Research Foundation (AwwaRF) for financial support. Additional funding for this project was provided by Brain Korea 21 (BK21). Furthermore, SK Chemicals and Nissan Chemical Industries, Ltd. respectively provided the hollow fiber MF module and model silica colloids used throughout this study.

References 1. Boerlage, S. F. E., M. D. Kennedy, M. P. Aniye, E. M. Abogrean, G. Galjaard, and J. C. Schippers. 1998. Monitoring particulate fouling in membrane systems. Desalination. 118: 131-142. 2. Bourgeous, K. N., J. L. Darby, and G. Tchobanoglous. 2001. Ultrafiltration of wastewater: effects of particles, mode of operation, and backwash effectiveness. Water Research. 35: 77-90. 3. Chellam, S., and M. Wiesner. 1997. Particle back-transport and permeate flux behavior in crossflow membrane filters. Environmental Science and Technology. 31: 819-824. 4. Chellam, S., J. G. Jacangelo, and T. P. Bonacquisti. 1998. Modeling and experimental verification of pilot-scale hollow fiber, direct flow microfiltration with periodic backwashing. Environmental Science and Technology. 32: 7581. 5. Decarolis, J., S. Hong, and J. Taylor. Submitted 2001. Effect of operating conditions on fouling behavior and solid retention during ultrafiltration of wastewater for reclamation. Journal of Membrane Science. 6. Hagen, K. 1998. Removal of particles, bacteria and parasites with ultrafiltration for drinking water treatment. Desalination. 119: 85-91. 7. Hillis, P., M. B. Padley, N. I. Powell, and P. M. Gallagher. 1998. Effects of backwash conditions on out-to-in membrane microfiltration. Desalination. 118: 197-204. 8. Hong, S., R. S. Faibish, and M. Elimelech. 1997. Kinetics of permeate flux decline in crossflow membrane filtration of colloidal suspensions. Journal of Colloidal and Interface Science. 196: 267-277. 9. Hong, S., J. Taylor, and T. Tak. 2000. Application of mem-

brane technologies for drinking water treatment in Florida. Korean Membrane Journal. 1: 26-35. 10. Jonsson, A. S., and B. Jonsson. 1996. Colloidal fouling during ultrafiltration. Separation Science and Technology. 31: 2611-2620. 11. Kenned y, M., S. M. Kim, I. Mutenyo, L. Broens, and J. Schippers. 1998. Intermittent crossflushing of hollow fiber ultrafiltration systems. Desalination. 118: 175-188. 12. L. Wang, and L. Song. 1999. Flux decline in crossflow microfiltration and ultrafiltration: experimental verification of fouling dynamics. Journal of Membrane Science. 160: 41-50. 13. Lipp, P., G. Baldauf, R. Schick, K. Elsenhans, and H. H. Stabel. 1998. Integration of ultrafiltration to conventional drinking water treatment for a better particle removalefficiency and costs?. Desalination. 119: 133-142. 14. Ma, H., C. N. Bowman, and R. H. Davis. 2000. Membrane fouling reduction by backpulsing and surface modification. Journal of Membrane Science. 173: 191-200. 15. Marchese, J., N. A. Ochoa, C. Pagliero, and C. Almandoz. 2000. Pilot-scale ultrafiltration of an emulsified oil wastewater. Environmental Science and Technology. 34: 2990-2996. 16. Serra, C., L. Durand-Bourlier, M. J. Clifton, P. Moulin, J. C. Rouch, and P. Aptel. 1999. Use of air sparging to improve backwash efficiency in hollow-fiber modules. Journal of Membrane Science. 161: 95-113. 17. Song, L., and M. Elimelech. 1995b. Theory of concentration polarization in crossflow filtration. J. of Chem. Soc. Faraday Trans. 91: 3389-3398. 18. Tran-Ha, M. H., and D. E. Wiley. 1998. The relationship between membrane cleaning efficiency and water quality. Journal of Membrane Science. 145: 99-110. 19. Vos, G., Y. Brekvoort, H. A. Oosterom, and M. M. Nederlof. 1998. Treatment of canal water with ultrafiltration to produce industrial and household water. Desalination. 118: 297303. 20. Xu, Y., J. Dodds, and D. Leclerc. 1995. Optimization of a discontinuous microfiltration-backwash process. The Chemical Engineering Journal. 57: 247-251. 21. Yiantsios, S. G., and A. J. Karabelas. 1998. The effect of colloid stability on membrane fouling. Desalination. 118: 143152. 22. Yuan, W., and A. L. Zydney. 2000. Humic acid fouling during ultrafiltration. Environmental Science and Technology. 34: 5043-5050. 23. Zhang, M., and L. Song, 2000. Mechanisms and parameters affecting flux decline in cross-flow microfiltration and ultrafiltration of colloids. Environmental Science and Technology. 34: 3767-3773.

(Received October 8, 2001)