Open Collections

UBC Theses and Dissertations

UBC Theses Logo

UBC Theses and Dissertations

Modeling irreversible fouling in submerged hollow fiber membrane systems for drinking water treatment Lin, Hong 2007

Your browser doesn't seem to have a PDF viewer, please download the PDF to view this item.

Item Metadata

Download

Media
831-ubc_2007-0484.pdf [ 5.49MB ]
Metadata
JSON: 831-1.0063234.json
JSON-LD: 831-1.0063234-ld.json
RDF/XML (Pretty): 831-1.0063234-rdf.xml
RDF/JSON: 831-1.0063234-rdf.json
Turtle: 831-1.0063234-turtle.txt
N-Triples: 831-1.0063234-rdf-ntriples.txt
Original Record: 831-1.0063234-source.json
Full Text
831-1.0063234-fulltext.txt
Citation
831-1.0063234.ris

Full Text

MODELING IRREVERSIBLE FOULING IN SUBMERGED HOLLOW FIBER MEMBRANE SYSTEMS FOR DRINKING WATER TREATMENT  By  Hong Lin  A THESIS SUBMITTED IN PARTIAL F U L F I L M E N T OF THE REQUIREMENTS FOR T H E D E G R E E OF  M A S T E R OF APPLIED SCIENCE  in The Faculty of Graduate Studies (Civil Engineering)  THE UNIVERSITY OF BRITISH COLUMBIA August, 2007 © Hong Lin, 2007  ABSTRACT The use of low pressure membrane filtration processes in water and wastewater treatment fields has been increasing rapidly due to evolving health concerns and the development of new and lower-cost membranes. Among diverse types of operating mode and membrane modules, submerged hollow fiber membrane systems are very competitive because of their simpler operation, greater ability to resist against fouling, lower maintenance cost and smaller footprint. However, membrane fouling still remains the main disadvantage and limitation of these systems. Membrane fouling decreases the permeate flux which in turn increased the capital and operating costs of membrane systems, and also affects pretreatment needs, cleaning requirements and operating conditions. Therefore, fouling control is an important consideration in the design and operation of a membrane filtration system. A lot of research has been conducted to seek effective methods to control fouling. Based on the understanding of fouling mechanisms and the influence of operating parameters on membrane fouling, numerical models have been developed to quantitatively predict fouling. Nevertheless, most of the studies and models developed to date combine reversible and irreversible fouling together. Irreversible fouling, the main cause of the long term fouling has received limited attention. As a result, there remains a knowledge gap in terms of the mechanisms that govern irreversible fouling as well as the fouling behavior. This research was undertaken to investigate irreversible fouling, and moreover, attempt to develop a simple and reliable model to accurately predict irreversible fouling in submerged hollow fiber membrane filtration systems for drinking water treatment.  The study results revealed that even though all experiments were performed with an operating flux that was less than the critical flux, a substantial amount of fouling was observed when filtering over extended periods of time. The extent of fouling was observed to be related to both the operating permeate flux and the system hydrodynamic conditions. Irreversible fouling observed in this study was due to both extensive internal/pore fouling and surface/cake fouling. Internal fouling was the predominant mechanism that governed irreversible fouling. ii  A semi-empirical relationship was developed to model the extent of fouling when filtering over extended periods of time for conditions where the operating permeate flux was less than the critical flux. The relationship was based on the membrane characteristics, the extent of surface/cake fouling and the extent of internal/pore fouling, respectively.  The extent of surface/cake fouling was determined to be governed by the operating permeate flux and the system hydrodynamic conditions (i.e. the cross-flow velocity). The extent of internal/pore fouling was determined to be governed only by the operating permeate flux. In addition, the results from the present study indicated that when operating the membrane filtration system below critical flux conditions, for a given volume of permeate filtered, the extent of overall fouling increased as the operating permeate flux increased and decreased as the cross-flow velocity increased.  in  TABLE OF CONTENTS  ABSTRACT  ii  T A B L E OF CONTENTS  iv  LIST OF T A B L E S  vii  LIST OF FIGURES  viii  LIST OF S Y M B O L S  x  ACKNOWLEDGEMENTS  xii  C H A P T E R 1 INTRODUCTION A N D R E S E A R C H OBJECTIVES  1  CHAPTER 2 LITERATURE REVIEW  7  2.1 Permeate Flux Decline and Membrane Fouling  7  2.1.1 The phenomenon of permeate flux decline  7  2.1.2 Mechanisms of membrane fouling  8  2.1.3 Behavior of particle back-transport  at membrane surface  9  2.1.4 The concept of critical flux  10  2.2 Strategies of Fouling Control  12  2.2.1 A review of general methods for fouling control  12  2.2.2 Air sparging in hollow fiber ultrafiltration  14  membranes  2.3 Irreversible Fouling  15  2.3.1 Causes of irreversible fouling and corresponding prevention strategies  16  2.3.2 Operating factors that influence irreversible fouling  17  2.3.3 Quantification  19  of irreversible fouling  2.4 Modeling of Fouling in Submerged Hollow Fiber Membrane Systems 2.4.1 Models  that account for  the effect of different fouling  mechanisms  20 either  separately and or in succession  20  2.4.2 Models that account for the effect of different fouling mechanisms in parallel....  21  2.4.3 Models that combined reversible and irreversible fouling together  24  2.5 The Knowledge Gap: Mechanisms occurring in irreversible fouling and a reliable model iv  to predict irreversible fouling  25  CHAPTER 3 M A T E R I A L S A N D M E T H O D S  26  3.1 Bench-scale Membrane System  26  3.1.1 Submerged membrane system  27  3.1.2 Aeration system  30  3.1.3 Permeate flux generation, collection and measurement system  32  3.1.4 Vacuum measurement system  33  3.2 Routine Membrane Integrity Testing and Maintenance  34  3.2.1 Membrane integrity testing  34  3.2.2 Membrane cleaning  35  3.3 Source Water  36  3.4 Experimental Program  37  3.4.1 Experimental conditions considered  37  3.4.2 Experimental monitoring  39  3.4.3 Trans-membrane pressure correction  40  CHAPTER 4 E X P E R I M E N T A L RESULTS  42  4.1 Evolution of Trans-membrane Pressure for Clean Water  42  4.2 Evolution of Trans-membrane pressure over Time for Mixed Source Waters  43  CHAPTER 5 M O D E L D E V E L O P M E N T , DATA ANALYSIS A N D DISCUSSION 5.1 Model Development for Irreversible Fouling  46 46  5.1.1 Model 1: Irreversible fouling is only due to internal fouling  46  5.1.2 Model 2: Irreversible fouling is due to internal fouling  53  + surface fouling  5.2 Data Analysis and Parameter Estimation of Irreversible Fouling Model  55  5.2.1 Approach J: Combining numerical approaches to estimate Ki, K2 and K3  55  5.2.2 Approach 2: Estimate K/, K2 and K3 separately  56  5.3 Model Parameter Quantification and Discussion 5.5.7 Quantification 5.3.2 Quantification  61  ofKi of K2 and the impact  61 of operating  permeate flux and bulk  cross-flow velocity on the extent of surface fouling  62  5.3.3 Quantification ofKj and the impact of the operating permeate flux on the extent of internal fouling  64  5.3.4 Relative contribution of internal fouling and surface fouling to irreversible  fouling 66  5.4 Comparison of Model Predictions and Experimental Data  68  5.4.1 Conditions for which model agrees with experimental trans-membrane pressure data  69  5.4.2 Conditions when model overestimates trans-membrane pressure data  74  C H A P T E R 6 CONCLUSIONS A N D SIGNIFICANCE TO E N G I N E E R I N G  75  6.1 Conclusions  75  6.2 Significance to Engineering  76  REFERENCES  78  A P P E N D I X A : CODES FOR M A T L A B C A L C U L A T I O N  85  vi  LIST OF TABLES  Table 3. 1 Physical characteristics of the membrane used in the present study  29  Table 3. 2 Operating bulk cross-flow velocities and air flow rate for dual phase system  32  Table 3. 3 TOC of raw water and their mixtures  37  Table 3. 4 Corresponding pseudo-steady-state permeate flux to system hydrodynamics  38  Table 3. 5 Experimental program  39  Table 5. 1  for regression of model 1 and estimated K i " , K 3 values  50  Table 5. 2 Estimated K ] " and K 3 values at a bulk cross-flow velocity of 0.4m/s  58  Table 5. 3 Estimated K i ' and K 2 values  60  vii  LIST O F FIGURES  Figure 1. 1 Inside-out and outside-in hollow fiber module Figure 1. 2 Simplified schematic of membrane system operating modes  3 4  Figure 2. 1 Four classical fouling models (Adapted from Hermia, 1982) Figure 2. 2 Comparison of filtration resistance after 90 min filtration with  8 and W)  without air sparging (Adapted from Chang and Fane, 2000 a)  19  Figure 3. 1 Schematic of the bench-scale submerged membrane system  26  Figure 3. 2 Picture of laboratory bench-scale experimental setup  27  Figure 3. 3 Schematic diagram of system tank with cylindrical baffle  28  Figure 3. 4 Schematic of membrane module  29  Figure 3. 5 Membrane module position inside the system tank  30  Figure 3. 6 Schematic of center aerator  31  Figure 3. 7 Schematic of dual phase cross-flow aeration system  31  Figure 3. 8 Schematic of the membrane integrity testing apparatus  35  Figure 3. 9 Water viscosity vs Temperature  41  Figure 4. 1 Trans-membrane pressure for filtered and deionized water  42  Figure 4. 2 Linear relationship between operating permeate flux and trans-membrane pressure 43 Figure 4. 3 Trans-membrane pressure vs Volume of water filtered at permeate flux 40 L/m h 2  44 Figure 4. 4 Trans-membrane pressure vs Volume of water filtered at permeate flux 50 L/m h 2  44 Figure 4. 5 Trans-membrane pressure vs Volume of water filtered at permeate flux 60 L/m h 2  45 Figure 5. 1 Unit length membrane pore and its cross section (Where ro is the initial pore radius, and r' is the pore radius after filtering a unit volume of liquid) Figure 5. 2 K 3 vs Jv at different cross-flow velocities  47 51  Figure 5. 3 S E M images for clean and fouled membranes  52  Figure 5. 4 Active surface areas of clean membrane and after filtering unit volume of liquid (where A is the initial active surface area of a given membrane module) 0  53  Figure 5. 5 The relationship between K and Jv as well as C F V  63  Figure 5. 6 Linear relationship between K3 and Jv  64  Figure 5. 7 Non-linear relationship between K3 and Jv (forcing intercept = 0)  65  Figure 5. 8 The change of surface fouling over volume of water  67  2  filtered  Figure 5. 9 The change of internal fouling over volume of water filtered  68  Figure 5. 10 Predicated against measured data for 40 L/m h operating permeate flux  71  Figure 5. 11 Predicated against measured data for 50 L/m h operating permeate flux  72  Figure 5. 12 Predicated against measured data for 60 L/nrh operating permeate flux  73  2  2  IX  LIST OF SYMBOLS  J  the permeate flux [m /m s] or [L/m hr]  AP  the trans-membrane pressure [N/m ]  \i  the water viscosity [Ns/m ],  Rt  the total filtration resistance [m /m ]  Rm  the intrinsic membrane filtration resistance [m / m ]  Rr  the reversible resistance [m /m ]  Rir  the irreversible filtration resistance [m / m ]  ki  the fouling strength of dissolved organic matter [m/g]  3  2  2  2  2  3  3  2  3  the dissolved organic matter concentration [g/m ] 3  S  T  the total volume of the filtrate per unit area of the membrane [m /m ] 3  A  the rate of pore blockage,  Rpo  the initial resistance of the protein deposit [m /m ]  f'R'  the rate of cake growth  k  T  the total fouling coefficient  k  R  the reversible fouling coefficient  2  2  3  k,  the irreversible fouling coefficient  Pv  the trans-membrane pressure after filtering a given permeate volume [psi],  Po  the initial trans-membrane pressure [psi]  p  0  2  20°C  the trans-membrane pressure value at temperature 20.0 C [N/m ]  PT  the trans-membrane pressure value at temperature T [N/m ]  T  the operating temperature [°C]  M"20°c  water viscosity values at temperature of 20.0 C [Ns/m"]  r  water viscosity values at temperature of T [Ns/m ] 2  Jv  the operating permeate flux [nr /rrfs] or [L/m hr] X  xL  f l o w path length [m]  N  n u m b e r o f pores [ N o . / m ]  R  m e m b r a n e pore radius [m]  ro  i n i t i a l m e m b r a n e pore radius [m]  r'  m e m b r a n e radius after filtering V o f water [m]  A  2  0  the i n i t i a l active surface area o f a g i v e n m e m b r a n e m o d u l e  Ki  a constant related to m e m b r a n e m o d u l e a n d r a w water characteristics [ N m h / L ]  Ki"  a constant related to m e m b r a n e m o d u l e and raw water characteristics [ N m h / L ]  Ki'  a constant related to m e m b r a n e m o d u l e and raw water characteristics [m ]  K2  rate o f pore n u m b e r reduction (i.e. rate o f surface f o u l i n g ) [1/m]  K3  rate o f pore radius r e d u c t i o n (i.e. rate o f internal f o u l i n g ) [1/m]  CFV  c r o s s - f l o w v e l o c i t y a p p l i e d to the m e m b r a n e system [m/s]  V  v o l u m e o f water filtered [nr ]  6  4  xi  ACKNOWLEDGEMENTS  I would like to express my sincere gratitude to my supervisor, Dr. Pierre R. Berube, for his guidance, enthusiasm, trust, and continuous support throughout this research. His inspiring instruction helped me to develop my research and technical writing skills, and his devotion to research/teaching encouraged me to try my best to serve the community in the near future.  I thank Dr. Eric Hall, the second reader of this thesis, for his insight, constructive comments and advice on this research and the thesis writing.  I would also like to thank Susan Harper and Paula Parkinson for their talented analytical assistance and advice in the Environmental laboratory, Bill Leung for building my reactors and checking my experimental systems, Scott Jackson for setting up my data recording system, and Ph. D. student Colleen Chan for her suggestions and proof reading of this thesis.  I am also grateful to summer student Jackie Wong, Ph. D. student Kazi Parvez Fattahmy, and fellow graduate student Guo Feng for collecting source water from Capilano Reservoir and Jericho Pond.  I appreciate all my friends in Canada and China, for their precious friendship, understanding and care during the period of this research.  xii  CHAPTER 1 INTRODUCTION AND RESEARCH OBJECTIVES  The use of low pressure membrane filtration processes in water and wastewater treatment fields has been increasing rapidly due to evolving health concerns and the development of new and lower-cost membranes. According to the definition given by USEPA, a low pressure membrane filtration process is "a pressure or vacuum-driven separation process in which particulate matter larger than 1 \im is rejected by an engineered barrier primarily through a size exclusion mechanism and which has a measurable removal efficiency of a target organism that can be verified through the application of a direct integrity test" (USEPA, 2005).  Based on sieving mechanisms, low pressure membrane filtration is capable of removing suspended or colloidal particles larger than the membrane pore size. Compared with traditional granular media filtration processes used in drinking water treatment, low pressure membrane filtration removes more disinfection by-product precursors (USEPA, 2005) and pathogens, uses reduced amount of treatment chemicals, has a smaller footprint and a lower labor requirement (Metcalf and Eddy, 2003). Four main membrane technologies exist: microfiltration (MF), ultrafiltration (UF), nanofiltration (NF), and reverse osmosis (RO). M F and U F are low pressure membranes while N F and RO are typically classified as high pressure membranes. Consequently, in this document, low pressure membranes will hereafter be referred to simply as membranes. M F and U F were firstly commercialized in the early 1990's and have now become widespread in water and wastewater treatment applications. The pore sizes for M F membranes range from 0.1 to 0.2 Jim (nominally 0.1 |im), while the pore size for U F membranes range from 0.01 to 0.05 um (nominally 0.01 Jim) or less (USEPA, 2005).  Membrane properties such as surface charge, hydrophobicity, pH and oxidant tolerance, strength and flexibility are mainly determined by the materials used for manufacturing the membranes (USEPA, 2005). Currently, most of the membranes used for drinking water production are typically made of low-cost polymeric material (i.e. polyacrylonitrile (PAN)).  Membrane modules are the smallest operational units in membrane filtration systems. Usually, a membrane system consists of the membrane modules, bulkheads into which the membranes are potted, pressure support structures, feed inlet, concentrate outlet ports and permeate collection points (Malleialle et al., 1996). There are four major types of membrane modules: hollow fiber, plate and frame, spiral wound, and tubular modules. The application of hollow fiber modules is becoming the most popular type of membranes because of the lower energy cost and back-flushing capability of these modules. Hollow fiber modules are generally bundled together (100 to 10,000 fibers) and both ends are potted using resin into bulkheads to form membrane systems. Each fiber usually has an outside diameter of 0.5 to 2.0 mm, an inside diameter of 0.3 to 1.0 mm, a wall thickness of 0.1 to 0.6 mm and a fiber length of 1 to 2 meters (USEPA, 2005). A large surface area per unit volume and a high shear rate can be achieved by compactly packing membrane modules together. Large surface area leads to large permeate flow, while the very small flow channels between the compactly packed modules provides a high shear rate (Malleialle et al., 1996). A l l these contribute to lower the cost of hollow fiber modules when compared to other types of membrane modules. In addition, the ability to back-flush hollow fiber modules typically results in better fouling controlling (Malleialle et al., 1996). Hollow fiber membrane modules may operate in either "inside-out" or "outside-in" mode (see Figure 1.1). In inside-out configurations, feed water from the inside (i.e. fiber lumen) permeates through the wall of membranes to the outside; while in outside-in mode, feed water from the outside of the membranes permeates along and passes the membrane surface, and filtrate is collected inside the lumen (USEPA, 2005). The driving force for hollow fiber modules is the trans-membrane pressure difference, which is either supplied by providing a positive or a negative (vacuum) pressure.  2  Permeate  Permeate  Feed  Fep.fi  Inside-out  Outside-in  Figure 1. 1 Inside-out and outside-in hollow fiber module  Based on the operating mode, membrane filtration systems can be typically divided into two categories: submerged and external, as seen in Figure 1. 2. External systems typically use the inside-out flow configuration and pressure is used as the driving force, while submerged systems typically use an outside-in flow configuration and a vacuum is used as the driving force. The high shear rates needed to minimize fouling are typically achieved through applying a high cross-flow velocity along the membrane surface in external systems, while a high shear rate is typically achieved by sparging air along the membrane surface in submerged systems. Typical operating trans-membrane pressure ranges from 21 kPa to 276 kPa (3 to 40 psi) in a pressure-driven hollow fiber membrane system; while a vacuum of approximately -21 kPa to -83 kPa (-3 to -12 psi) is typically applied in a vacuum-driven hollow-fiber membrane system (USEPA, 2005).  3  Reactor Membrane Influent  Permeate  Cross-flow  E x t e r n a l operating mode  Reactor Permeate  Influent Membrane  Submerged operating mode  Figure 1. 2 Simplified schematic of membrane system operating modes  Submerged hollow fiber systems are very competitive when compared to other types of operating mode and membrane modules. The advantages of these systems include simpler operation, greater ability to resist against fouling, lower maintenance cost and smaller footprint (Fane et al, 2002; Yoon et al, 2004). The present thesis deals exclusively with submerged hollow fiber membrane systems. Consequently, in this document, submerged hollow fiber membranes will hereafter be refereed to simply as membranes.  Although there are a number of advantages associated with membrane systems, when compared to other filtration processes such as conventional sand filtration, membrane fouling 4  still remains the main disadvantage and limitation of membrane systems. Membrane fouling, which is caused by the deposition and accumulation of materials on the membrane surface, or inside the membrane pores, decreases the permeate flux which in turn increases the capital and operating costs of membrane systems. In addition, membrane fouling affects pretreatment needs, cleaning requirements and operating conditions (Metcalf & Eddy, 2003). Therefore, fouling is an important consideration in the design and operation of a membrane filtration system. A lot of research has been conducted to seek effective methods to control fouling. Based on the understanding of fouling mechanisms and the influence of operating parameters on membrane fouling, a number of numerical models (Ho and Zydney, 2000; Yuan et al, 2002; Katsoufidou et al, 2005; Bolton et al, 2006a; Berube and Lei, 2006; etc) have been developed to quantitatively predict fouling. Corresponding strategies have also been developed and applied to control and minimize membrane fouling (Bellara et al, 1996; Smith et al, 2002; Smith et al, 2005; Cui et al, 2003; Berube and Lei, 2006; Hilal et al, 2005; etc). Operating membrane filtration systems under critical permeate flux, back-washing and controlling the system hydrodynamic conditions are some strategies that have been widely used to control fouling. However, there remains a knowledge gap in terms of the mechanisms that govern irreversible fouling. Most of the studies and models developed to date combine reversible and irreversible fouling together. Irreversible fouling, which is typically the main cause of long term fouling has received limited attention. As a result, the mechanisms that govern irreversible fouling, as well as the fouling behavior, are poorly understood. Due to this knowledge gap, there is currently no accurate model to predict irreversible fouling, or to develop effective strategies to prevent and control irreversible fouling.  This research studied irreversible fouling in submerged hollow fiber membrane filtration systems for drinking water treatment. The objectives of this research were to firstly gain a better understanding of the mechanisms of irreversible fouling, secondly quantify the relationships between the extent of irreversible fouling and the operating parameters, thirdly develop a mechanistic model to accurately predict irreversible fouling, and ultimately propose practical suggestions to the engineering application of membrane filtration systems. 5  A l l experiments were performed at sub-critical flux conditions. As discussed in Chapter 2, when operating below critical flux, only irreversible fouling, that is fouling which is not impacted by back-transport mechanisms, should theoretically occur. The specific hypotheses that were tested are:  1. that irreversible fouling, when operating below critical flux, is not impacted by back-transport mechanisms (i.e. extent of air sparging), and 2. that for a given permeate volume filtered, the extent of irreversible fouling is not affected by the operating permeate flux  6  CHAPTER 2 LITERATURE REVIEW  There are two obstacles that prevent the widespread application of membrane filtration systems: concentration polarization and membrane fouling. Concentration polarization is the increase in the concentration of retained material at the membrane surface, while membrane fouling refers the adsorption or deposition of material in or on the membranes. Both concentration polarization and membrane fouling increase the membrane resistance by limiting mass transfer at membrane surface, and cause a decrease in the permeate flux. They are also interrelated: severe concentration polarization can aggravate the extent of membrane fouling (Ghosh, 2006). The negative impact of concentration polarization on the permeate flux is relatively constant. On the other hand, the negative impact of membrane fouling increases with time, and is generally more significant than that from concentration polarization. Membrane fouling is generally regarded as the main cause of the decrease in the permeate flux in membrane filtration systems. Fouling and strategies for fouling control are reviewed in the follow sections.  2.1 Permeate F l u x Decline a n d M e m b r a n e F o u l i n g  2.1.1 The phenomenon ofpermeate flux decline During the membrane filtration process, the permeate flux decreases with respect to time. The general form of Darcy's law (Equation 2.1) describes the overall characteristics of flux reduction.  |iR  t  where J is the permeate flux [m /m s], A P is the trans-membrane pressure [N/m ], (I is the water 3  2  2  viscosity [Ns/m ], R, is the total filtration resistance [m /m ]. The driving force for the transport 2  2  3  of water across membranes is the pressure gradient from one side of the membrane to the other (i.e. trans-membrane pressure). The total filtration resistance, which increases over time as fouling progresses, limits the transport of water across the membrane.  Membrane fouling is due to the accumulation of material on or in the membrane skin and can be classified into two categories: reversible fouling and irreversible fouling. Reversible fouling is the fraction of fouling that can be impacted by the rate of mass transfer away from the membrane surface, while irreversible fouling is not impacted by the rate of mass transfer away from the membrane surface (Berube and Wong, 2007). Reversible fouling can be controlled hydraulically. However, irreversible fouling can only be removed by chemical cleaning. The total fouling resistance is the sum of the resistance due to the membrane itself, which is constant, the resistance due to reversible fouling and the resistance due to irreversible fouling as presented in Equation 2.2 (Aimar and Howell, 1989).  R. = R + R  (2. 2)  +R,  where R is the intrinsic membrane filtration resistance [m /m ], R is the reversible resistance 2  3  m  r  [irr/m ], Rj is the irreversible filtration resistance [m7m]. Darcy's law can then be written as r  Equation 2.3.  J =  AP  (2. 3)  |i(R +R + R ) m  r  ir  2.1.2 Mechanisms of membrane fouling As illustrated in Figure 2.1, there are four classical models that have been popularly accepted and applied to explain membrane fouling mechanisms.  |0§ laa 8 c C  Cake fouling  6 Q  18  Standard blocking  Complete blocking  Intermediate blocking  Figure 2. 1 Four classical fouling models (Adapted from Hermia, 1982)  8  Cake fouling occurs when a permeable cake layer forms on the membrane surface with increasing thickness that creates an additional layer of resistance to permeate flow. Standard blocking assumes material accumulates and absorbs inside the membrane wall, leading to the reduction of pore radius and an increase in the membrane resistance. Complete blocking assumes pores are completely plugged by particles and no flow can go through the plugged pores. Intermediate blocking assumes that some of the particles seal off the pores and the rest accumulate on the top of other deposited particles. The reduction of effective pore size and pore numbers caused by standard blocking and cake fouling are typically the main causes for the permeate flux decrease (Minegish et al, 2001). The four models are also commonly referred to as pore blocking (complete blocking), direct adsorption (standard blocking), long term adsorption (intermediate blocking) and boundary layer resistance (cake fouling ) (Bowen et al, 1995).  Each of these models has been used individually or in combination to explain experimental observations. However, membrane fouling is an extremely complicated physico-chemical phenomenon, which can involve several fouling mechanisms occurring simultaneously.  2.1.3 Behavior ofparticle back-transport at membrane surface While the permeate flux carries particles towards a membrane surface through convection, this mass flux is balanced by a number of back-transport mechanisms that carry particles away from a membrane surface (Belfort et al, 1994). Four particle back-transport mechanisms (i.e. Brownian diffusion, shear-induced diffusion, lateral migration by inertial lift and surface transport) are typically used to explain the behavior of particulate and macromolecular solutions at membrane surfaces. Brownian diffusion is only significant for submicron particles (less than 1 (im) and is typically not a significant back-transport mechanism in low pressure membrane (i.e. M F and UF) systems. Shear-induced diffusion on the other hand is significant for larger micron-sized particles (from 1 Jim to 30-40 |J,m) and is proportional to the shear rate at a membrane surface. Inertial lift is strongest for large particles (greater than 30 to 40 (Xm), and is due to pressure differences that arise at the extremities of large particles when exposed to a 9  velocity gradient. Surface transport can be significant when a membrane is exposed to a relatively high tangential flow, causing retained particles to slide along the membrane surface (Belfort et al, 1994).  A l l four back-transport mechanisms depend strongly on the shear rate and/or particle size. The shear rate or shear stress near a membrane surface is an important hydrodynamic parameter in describing the particle back-transport. Induced surface shear forces are regarded as a major strategy to control fouling because they increase the mass transfer away from the membrane surface (Mallevialle et al, 1996).The experimental results from Bian et al. (2000) indicated that the velocities of back-transport away from a membrane surface increased as the shear rate increased, resulting in a decrease in membrane fouling.  It should be noted that of the four fouling models, standard blocking, which is due to the adsorption of foulants within the membrane pores, is not impacted by the back-transport mechanisms discussed above, since these mechanisms manifest themselves at the membrane surface and not within the membrane pores.  2.1.4 The concept of critical flux  The concept of critical flux hypothesizes that there exists a flux below which fouling does not occur; above this flux, fouling is observed (Field et al, 1995). At critical flux, the rate of mass transfer towards a membrane surface, due to particle convection towards the membrane surface, is balanced by the rate of mass transfer away from the membrane surface, due to one or more back-transport mechanisms. Therefore, the critical flux for a particular application depends on the system hydrodynamics and the particle size.  It should be noted that true critical flux conditions are typically never achieved, because solutions typically contain particles of various sizes, each with different back-transport rates, and 10  because the critical flux concept assumes that fouling due to adsorption within the membrane pores (i.e. standard blocking) does not occur.  A number of researchers have demonstrated that the concept of critical flux can not be applied directly (Wisniewski et al, 2000; Cho and Fane, 2002; Ognier et al, 2004; Ye et al, 2006). Wisniewski et al. (2000) found that for solutions of particles of non-uniform size as well as the presence of the soluble compounds, which are not influenced by hydrodynamic conditions, the critical flux concept could not be applied. A number of researchers have reported a two-stage increase in trans-membrane pressure when operating bioreactors at a permeate flux substantially below the critical flux (Ognier et al, 2002a; Ognier et al, 2004; Cho and Fane, 2002; Nagaoka et al,  1996). The first stage consisted of a long term slow and gradual increase in the  trans-membrane pressure, while the second stage was characterized by a rapid rise in the trans-membrane pressure. The gradual fouling in the first stage was reported to be "hydraulically irreversible" and could not be avoided even when operating the system under critical flux conditions (Ognier et al, 2004). Cho and Fane (2002) suggested that the fouling that occurred when operating below critical flux was due to non-homogeneous flux conditions at a membrane surface. They suggested that due to fouling, the localized fluxes in some membrane areas were reduced (i.e. first stage). Therefore, for membrane systems operated with a fixed permeate flux, the reduced localized flux in some areas must be accompanied by an increase in the localized flux in other membrane areas that were not yet fouled. When the increase in the localized flux exceeded the critical flux, fouling occurred and was accompanied by a sudden rise in the trans-membrane pressure (i.e. second stage). The length of the first fouling stage has been observed to be related to the operating flux. When applying lower sub-critical fluxes, the first fouling stage was observed to last longer (Cho and Fane, 2002; Ognier et al, 2002b; Le Clech et al, 2003).  It should also be noted that the critical flux measured in the studies presented above was usually determined based on the flux stepping experiments which only lasted a few hours. Le ii  Clech et al. (2003) indicated that critical flux values might be affected by the period of each flux step and the time delay between flux steps. Therefore, care must be taken when interpreting critical flux values and the use of critical flux to assist in investigating fouling (Ye et al, 2006).  As an alternative, the concept of pseudo-steady-state permeate flux has been proposed. For a constant trans-membrane pressure system, the magnitude of the permeate flux after the initial period of rapid decrease in the flux was defined as pseudo-steady-state flux. The pseudosteady-state permeate flux was found to be governed by the hydrodynamic conditions applied to the membrane systems. Hydrodynamic studies conducted by Berube and Lei (2006) indicated that membranes operated with dual-phase cross-flow (i.e. with air sparging) could increase the pseudo-steady-state permeate flux by 20-60% when compared to membranes operated with single phase cross-flow (i.e. without air).  2.2 Strategies of Fouling Control  Strategies to avoid fouling or decrease the extent of fouling have been explored for decades. The methods employed for the control of fouling in low pressure membranes was reviewed comprehensively by Hilal et al. (2005). Pretreatment of the feed water, membrane material/surface modification, optimization of operating parameters/conditions and recovery cleaning are the four main approaches to control membrane fouling.  2.2.1 A review of general methods for fouling control Pretreatment typically involves a physical and/or chemical process designed to remove solute, colloidal and/or particulate material that can foul the membrane. Typically, physical processes are used to remove large suspended particles that may block the membrane pores or accumulate on the membrane surface. Typically, chemical processes are used to reduce internal adsorption inside membranes pores through coagulation and flocculation to convert solute and colloidal foulant into larger particles (Hilal et al, 2005). 12  Membrane material/surface modification aims to reduce the adsorption of foulant on the membrane surface by changing the properties of the membrane surface. The methods of membrane material/surface modification consist of either reducing the attractive forces or increasing the repulsive forces between the solute (i.e. foulant) and the membrane (Belfort et al, 1994). It has been generally known that membranes with a hydrophilic surface have a lower fouling potential and fouling of hydrophilic surfaces is often reversible. Since many commercial membranes are made from hydrophobic polymers, because of the superior chemical resistance and thermal and mechanical properties of these polymers, there is interest in coating hydrophobic polymers with hydrophilic materials (Hilal et al, 2005). Techniques used for surface modification include coating and photo-induced grafting.  Membrane type and configuration also affect fouling (Hilal et al, 2005). For example, hollow fiber membranes are back-washable. In addition, submerged hollow fiber systems are relatively simple, low cost (Cui et al, 2003) and require a relatively small footprint (Yoon et al, 2004).  A number of operating parameters/conditions can be modified to reduce membrane fouling. Commonly used approaches include back-flushing, back-pulsing, back-washing, air sparging and the use of high cross-flow velocities, mixers, turbulence promoters and vortex generators. Back-flushing and back-pulsing consist of periodically reversing the flow through the membrane to remove foultants that have plugged membrane pores. Back-washing is similar to back-flushing except that the reversed flow contains chemical cleaning agents. Air sparging, high cross-flow velocities, mixers, turbulence promoters and vortex generators generate high shear forces at the membrane surface and as a result increase the extent of back-transport of foultants away from the membrane surface. In addition, operating the membrane system under critical flux conditions can also reduce membrane fouling (Hilal et al, 2005).  13  Recovery cleaning procedures usually involve physical and/or chemical cleaning.  Physical  scouring, using pressurized water or air, can typically be used to effectively remove surface fouling. Chemical cleaning is necessary to remove foulants from membrane pores and to remove heavily consolidated surface foultants.  2.2.2 Air sparging in hollow fiber ultrafiltration membranes Air sparging is an effective way to generate high surface shear at membrane surfaces (Cui et al, 2003). By injecting air directly in the concentrate compartment during the filtration process, air spaging creates a gas-liquid two-phase flow along a membrane surface, increasing the surface shear, and reducing fouling. A i r sparging has been recognized as a very effective approach for fouling control and permeate flux enhancement. The mechanisms by which air sparging reduces fouling in a cross-flow tubular membrane system have been summarized by Cui et al. (2003). However, the mechanisms by which air sparging reduces fouling in submerged hollow fiber membranes are relatively poorly understood.  It has been well recognized that fouling in submerged hollow fiber membrane systems is greatly reduced by air sparging (Bellara et al, 1996; Smith et al, 2002; 2005; Cui et al, 2003; Berube and Lei, 2006; etc). However, the mechanisms of the flux enhancement as well as the performance of the sparged air in hollow fiber membranes are different from those in tubular membranes.  In a review paper by Cui et al. (2003), the main mechanisms of permeate flux enhancement by air sparging in submerged hollow fiber membrane systems were identified to be due to (1) scouring induced by the falling film that can surround a confined rising bubble (Bellara et al, 1996; Smith et al, 2005), (2) shear forces induced by oscillating flows that are present in the wake of a rising bubble (Bellara et al, 1996; Smith et al, 2005; Berube and Lei, 2006), (3) high bulk cross-flow velocities induced by the rising bubbles, and (4) pressure instabilities induced by the rising bubbles (Cui et al, 2003). 14  Cui et al. (2003) argued that gas flow rate, fiber orientation, fiber size and flexibility all had an impact on the flux enhancement in submerged hollow fiber membranes. The benefit of air sparging could be measured either by the degree of flux increase or the degree of deposit resistance decrease. Cui et al. (2003) reported that increasing the intensity of air sparging typically increased the flux and reduced the magnitude of resistance (also see Figure 2.2). However, a plateau is typically observed above which an increase of air sparging intensity does not result in an increase of the permeate flux (Chang and Fane, 2000b; Berube and Lei, 2006). Cui et al. (2003) reported that smaller fibers responded more positively to air sparging because they were more flexible and could move laterally and remove the deposited material on the membrane surface.  Berube and Lei (2006) suggested that both the lateral velocity of a swaying fiber and the physical contact between swaying fibers could generate high shear forces at a membrane surface. Promoting lateral swaying could potentially increase the physical contact between membrane fibers, and mechanically remove the foulant layer on the membrane surface. The magnitude of the surface shear forces was related to the configuration of the membrane system and the air sparging practice (Berube et al, 2006).  2.3 Irreversible Fouling  To minimize fouling in membrane filtration systems, physical membrane cleaning methods such as back-washing, high cross-flow velocity or scouring with air bubbles are routinely used in many existing full-scale membrane plants. Back-washing changes the direction of the flow and removes foultants that plug membrane pores and/or accumulate on membrane surfaces. High cross-flow velocities and air sparging create shear along the membrane surface, thus removing the material accumulating on the surface. However, membrane fouling can not be totally prevented by the above physical methods. This is partly because fouling due to adsorption within the membrane pores, cannot be controlled by physical membrane cleaning methods. As a 15  result, regardless of the physical membrane cleaning methods applied, a membrane becomes irreversibly fouled over time (Katsoufidou et al, 2005; Kimura et al, 2004).  Berube and Wong  (2007) defined irreversible fouling as the fraction of the total fouling that is not impacted by the rate of mass transfer away from a membrane surface.  Since physical membrane cleaning methods cannot control irreversible fouling, chemical cleaning is eventually needed to eliminate irreversible fouling. However, membrane systems need to be shut down during chemical cleaning, leading to a reduction in plant production. The waste produced from chemical cleaning also needs to be disposed of. Also, frequent chemical cleaning can reduce the membrane life. In addition, the costs of the cleaning chemicals are relatively high. As a result, compared with the physical cleaning methods applied to reduce reversible fouling, chemical cleaning is more expensive and complicated. Consequently, irreversible fouling poses a great challenge when optimizing the operation of membrane filtration systems (Crozes et al, 1997; Kimura et al, 2004).  2.3.1 Causes of irreversible fouling and corresponding prevention strategies To date, there has been no comprehensive research study conducted to investigate irreversible fouling. The very few studies that have focused on irreversible fouling have focused more on water chemistry aspects such as the interactions between natural organic matter and membrane materials. Crozes et al. (1993) found irreversible fouling was caused mainly by the adsorption of low molecular weight natural organic matter inside the membrane pores. This study also revealed that adsorption occurred more severely in a hydrophobic membrane than a hydrophilic membrane,  when filtering surface water. Kimura et a l (2004) found that  polysaccharide-like organic matter was the main cause for irreversible fouling in hydrophobic membrane systems. In addition, iron and manganese also contributed to irreversible fouling to some extent. Kimura et al. (2006) concluded that the degree of irreversible fouling was a function of the membrane properties and the types of organic matter in the solute. Although hydrophilic natural organic matter was observed to produce irreversible fouling regardless of the type of 16  membrane used, the exact composition of the organic material that caused irreversible fouling was reported to vary for different membrane materials. Irreversible fouling was also found to be proportional to the volume of solute filtered. When operated at the same trans-membrane pressure, U F membranes, which had a lower permeate flux, showed less irreversible fouling than M F membranes which had a higher permeate flux (Kimura et al, 2006).  Pretreatment of the water matrix to remove material that can potentially adsorb inside membrane pores and change the physicochemical properties of the membrane surface has been regarded as the major strategy for the prevention of irreversible fouling (Crozes et al, 1997; de Amorim and Ramos, 2006). However, research conducted by Kimura et a l (2005) found that although pre-coagulation/sedimentation did reduce reversible fouling significantly, it was ineffective at reducing irreversible fouling. This was because pre-coagulation/sedimentation was not able to remove polysaccharide/protein-like organic matter which was reported to cause irreversible fouling in this study. The effectiveness and efficiency of other pre-treatment methods need to be investigated in future research.  In terms of modification of the physicochemical properties of membrane surfaces, blending Pluronic F127 with poly (ether sulfone) (PES) increased the hydrophilicity of U F membranes, and was reported to be less prone to irreversible fouling (Wang et al, 2005). de Amorim and Ramos (2006) observed that the irreversible fouling of hydrophilic P V D F membranes was reduced when treated by a hydrophilic polymer solution. Kimura et a l (2006) observed that the irreversible fouling of polyacrylonitrile (PAN) membranes by naturally occurring organic matter was relatively low, while polyvinylidenefluoride (PVDF) and polyethylene (PE) suffered severe irreversible fouling.  2.3.2 Operating factors that influence irreversible fouling Crozes et a l (1997) investigated the impact of operating parameters on irreversible fouling. The cross-flow velocity, the operating permeate flux, the back-washing frequency and the 17  trans-membrane pressure all influenced irreversible fouling. Both reversible and irreversible fouling could be controlled by increasing the cross-flow velocity, reducing the permeate flux, and increasing the back-washing frequency. It was empirically observed that irreversible fouling in L E D (membrane developed by Lyonnaise des Eaux) U F membranes could be avoided by limiting the increase of trans-membrane pressure to a critical value. Above the critical value, irreversible fouling increased exponentially with the increase of trans-membrane pressure.  The extent of irreversible fouling related to permeate flux, back-washing frequency and time was measured by K i m and DiGiano (2006). The rate of irreversible fouling decreased substantially when the flux decreased from 65 to 50 L/m h and back-washing frequency increased 2  from every 15 to 10 minutes. Increase of back-washing time did result in a slight decrease in the fouling rate, but an increase of back-washing frequency was more effective at decreasing the fouling rate. Nevertheless, irreversible long-term fouling could not be completely avoided by the increase of back-washing frequency and back-washing time. Both internal fouling and cake formation were observed to be responsible for irreversible long-term fouling.  Air sparging was observed to be able to reduce irreversible fouling to some extent (Chang and Fane, 2000a). The resistance associated with reversible and irreversible fouling during filtration with and without air spaging is shown in Figure 2.2. As presented, the air sparging could reduce both reversible and irreversible filtration resistance. The reversible resistance was reduced significantly with an increase of superficial velocity. On the other hand, the irreversible resistance was much less significantly reduced with an increase of superficial velocity.  18  * r  *ir  Figure 2. 2 Comparison of filtration resistance after 90 min filtration with  C ) and  (®)  without air sparging (Adapted from Chang and Fane, 2000 a)  Other research indicates that reversible fouling and irreversible fouling are related to some extent. Berube and Wong (2007) observed that material that initially reversibly fouled a membrane surface could consolidate and generate irreversible fouling. This conclusion is somewhat consistent with those observed by other researchers. Nagaokaer al. (1998) suggested that some irreversible fouling could result from the consolidation/compaction of accumulated material on the membrane surface. They suggested that the rate of material accumulation on the membrane surface was proportional to the trans-membrane pressure. Hong et al. (2002) observed that the extent of the consolidation was a function of the duration of the filtration cycle. Katsoufidou et al. (2005) also observed that although reversible fouling could be removed by back-washing, the combined influence of pore blocking and cake formation could result in long term flux decline and eventually the formation of irreversible fouling.  2.3.3 Quantification of irreversible fouling Liang et al. (2006) developed a relationship to determine the magnitude of irreversible fouling resistance Rj based on the fouling strength of dissolved organic matter kj [m/g], the r  dissolved organic matter concentration S [g/m], and the total volume of the filtrate per unit area T  of the membrane j Jdt [m /m ] as presented in Equation 2.4. 3  2  0  19  The fouling strength factor of dissolved organic matter kj was regarded to be operation-specific, and a function of the characteristics of dissolved organic matter and the membrane materials (Liang et al, 2006).  2.4 Modeling of Fouling in Submerged Hollow Fiber Membrane Systems  Using Darcy's law and the four classical fouling models described in Section 2.1.2, individually or in combination, a number of studies have been conducted to attempt to explain the mechanisms that occur during membrane fouling. Usually, only one of the four fouling models is applied at one time. However, fouling kinetics suggest more than one fouling mechanism occurs at one time, and therefore, two or more of the four classical fouling models need to be considered successively or simultaneously (Jacob et al, 1998).  2.4.1 Models that account for the effect of different fouling mechanisms either separately and or in succession Most early studies expressed fouling behavior by identifying a dominant mechanism contributing to the evolution of flux decline (Katsoufidou et al, 2005). Tracy and Davis (1994) observed that during the filtration of bovine serum albumin (BSA) solutions  through  microfiltration membranes, both the standard blocking model and the pore blocking model initially fitted the experimental data well. However, with time, the cake filtration mechanism eventually dominated fouling behavior. Bowen et al. (1995) reported that all four classical fouling models could dominate the fouling process, but at different times, depending on the material in the solution being filtered and the membrane pore size. Theoretically, if the size of most membrane pores was larger than the material in the solution being filtered, the sequence of fouling started with standard blocking, complete blocking, followed by intermediate and cake fouling. If the size of most membrane pores was smaller than material in the solution being 20  filtered, the sequence of fouling started with complete blocking, followed by intermediate blocking and cake fouling. However, in practice, the fouling models are likely superimposed due to the membrane pore size distribution. In addition, although the initial and final steps could be identified, the transition between each dominant fouling model was gradual and complex. For membrane pore sizes studied in this research (smaller than material being filtered), the initial step was the complete blocking of the smallest pores, while the final step was the beginning of cake fouling. Nevertheless, for the smallest pore size membrane (nominal pore diameter 0.2 urn), the initial step seemed to be the combination of standard blocking and intermediate blocking (Bowen et al, 1995).  Minegish et a l (2001) observed that when hollow fiber U F membranes fouled, the decrease in the pore diameter and the decrease in the pore number occurred simultaneously. The decrease in the pore diameter was due to the adsorption of the low molecular weight substances (standard blocking model) inside the pore wall. The decrease in pore numbers was due to the accumulation of the high molecular weight humic substances (complete blocking model) on membrane surface. However, the mathematic model developed in their study only considered the effect of the primary fouling mechanism: the decrease of pore numbers by the high molecular weight humic substances.  2.4.2 Models that account for the effect of different fouling mechanisms in parallel Recently, several studies considered more than one fouling mechanism to explain and model fouling behavior (Ho and Zydney, 2000; Yuan et al, 2002; Katsoufidou et al, 2005; Bolton et al, 2006a; Bolton et al, 2006b; Ye et al, 2006).  Ho and Zydney (2000) modeled the flux decline over time in membranes by taking both pore blockage and cake fouling into account. The filtration flow rate could be determined by three parameters: the rate of pore blockage a, the initial resistance of the protein deposit R  p 0  [m /m ] and the rate of cake growth f ' R ' . The values of these parameters were estimated from 2  3  21  the experimental data and provided excellent agreement between the data and model predictions. The model also demonstrated a smooth transition from pore blockage, as the initial principal mechanism, to cake fouling, which eventually dominated fouling. For this study, one mathematical equation was able to explain fouling behavior regardless of the different phenomena and mechanisms occurring during the filtration process. This was regarded as the first model to combine the effect of two fouling mechanisms together (Bolton et al, 2006a).  Yuan et al. (2002) modified the combined pore blockage and cake fouling model developed by Ho and Zydney (2000) and found it was in good agreement with the flux decline behavior during microfiltration of Aldrich humic acid through 0.2 um pore size membranes. Since the three parameters a, R o and f'R' were all estimated from independent measurements of p  the physical properties of the humic acid systems, the changes of the solution condition could be reflected in the model calculation results. The model thus had the ability to analyze humic acid fouling under different conditions. In addition, a flux was observed below which the cake filtration could be negligible. This is similar to the critical flux concept developed for colloidal filtration (Yuan et al, 2002).  Katsoufidou et al (2005) took the simultaneous effect of all fouling mechanisms: pore construction, pore blocking and cake development into account and developed a new model based on the previous study by Bowen et al. (1995) and Yuan et al. (2002). The model successfully described the experimental data collected when filtering humic acid solutions through a single hollow fiber U F membrane. Their results suggested that the internal pore adsorption caused relatively rapid irreversible fouling. In parallel, pore blocking became important with time and a fouling cake developed on the membrane surface. Although back-washing could partly reverse the effect of pore blocking and cake formation, the combined influences of these two mechanisms could persist for a long time and eventually contribute to irreversible fouling.  22  Based on the classical fouling models, Bolton et al. (2006a) generated five new models that accounted for the combined effect of individual fouling mechanisms. These models, such as caking and complete blockage model, assumed that cake formation and complete pore blocking occur simultaneously.  When pores were blocked, although still permeable,  the  resistance for solutions to go through the membrane increased, and more cake would build upon the membrane surface. Eventually, more cake formation and less pore blockage occurred. Explicit equations were also derived from Darcy's Law under constant pressure and constant flow conditions. The equation could also be reduced to a single form of either the complete blocking model or a cake fouling model. Experimental results from the filtration of sterile IgG demonstrated that the combined caking and complete blockage model was the most useful since it provided good fits for both data sets, and was confirmed to be useful for a wide range of applications. The combined cake-standard and caking-intermediate models could give good fit to data as well and would also be effective in certain applications (Bolton et al, 2006a).  Bolton et al. (2006b) also developed a new adsorptive model that incorporated the effects of the permeate flux on irreversible fouling. While the standard model assumed all foulants entering into membrane pores deposit and accumulate uniformly along the pore wall, the adsorptive model allowed some foulants to deposit at the wall following zero order kinetics. This new adsorption model was combined with cake fouling, intermediate blocking and complete blocking models to predict fouling behavior. The combined intermediate-adsorption model was observed to be the best at matching the experimental data. Because this new model accounted for the effect of the flow rate and adsorption, it could be applied to membrane systems operated under different flow rates.  When using a membrane bioreactor for wastewater treatment, Ye et al. (2006) observed that fouling, when operating at subcritical flux conditions, was initially due to the standard pore blocking, and subsequently due to cake fouling. They also observed that the rate of standard pore blocking and cake fouling increased exponentially with respect to the permeate flux.  23  2.4.3 Models that combined reversible and irreversible fouling together Very few studies were found that considered the effect of reversible fouling and irreversible fouling in parallel.  A series of studies are being undertaken at the University of British Columbia to attempt to develop a model that can be used to accurately predict the evolution of the trans-membrane pressure over time in a submerged hollow fiber membrane system. These studies investigate the effects of the hydrodynamic conditions and the system configurations on membrane fouling. The results to date indicate that the increase in the trans-membrane pressure during one filtration cycle, can be modeled using a simple exponential relationship as presented in equation 2.5 (Berube and Watai, 2005).  P =P e' v  0  [k ]V T  =P e  [k  R  +k,]v  (2.5)  L,i  0  where k is the total fouling coefficient, which is the sum of the reversible (k ) and irreversible T  R  (ki) fouling coefficients, P is the trans-membrane pressure after filtering a given permeate v  volume [psi], Po is the initial trans-membrane pressure [psi], which is equivalent to the clean water flux. The reversible fouling coefficient was reported to be proportional to the different between the operating permeate flux and the pseudo-steady-state permeate flux (Berube and Watai, 2005). The irreversible fouling was initially estimated to be proportional to the inverse of the permeate flux (Berube and Lei, 2006). However, this initial estimate was derived based on experiments with heavily fouled membranes operated over relatively short periods of time (i.e. 8 hours) and with a constant trans-membrane pressure and variable permeate flux. Therefore, it is not clear whether this initial estimate can be applied to a properly operated (i.e. not extensively fouled) membrane with a constant permeate flux and variable trans-membrane pressure.  Liang et al. (2006) also described membrane fouling in submerged membrane bioreactor systems for wastewater treatment with a relationship that included both reversible and 24  irreversible fouling parameters. Reversible fouling was attributed to the mixed liquor suspended solids in solution, while long-term irreversible fouling was attributed to dissolved organic matter. Based on their study, reversible fouling was determined to be responsible for the initial rapid increase in the membrane resistance, and irreversible fouling was responsible for the subsequent and relatively slow increase in the membrane resistance.  2.5 The Knowledge Gap: Mechanisms occurring in irreversible fouling and a reliable model to predict irreversible fouling  As discussed above, membrane fouling has been extensively studied for decades. While the previously developed models have provided insight into the fouling mechanisms, they have certain limitations. Most of these studies and the models developed to date, combine reversible and irreversible fouling together. Irreversible fouling, the main cause of long-term fouling, has received limited attention. As a result, the mechanisms that govern irreversible fouling, as well as the fouling behavior, are poorly understood. Particularly, better knowledge of the impact of back-transport mechanisms (i.e. extent of air sparging) and the operating permeate flux on irreversible fouling is needed. Due to this knowledge gap, there is currently no accurate model to predict irreversible fouling, or to develop effective strategies to prevent and control irreversible fouling.  25  CHAPTER 3 MATERIALS AND METHODS  This study was designed to investigate the impact of system hydrodynamics on irreversible fouling in a submerged hollow fiber membrane system. A l l experiments were conducted at a constant permeate flux and variable trans-membrane pressure conditions. The specific experimental conditions investigated are presented in the following sections.  3.1 Bench-scale M e m b r a n e System  The bench-scale membrane system used in the present research is illustrated in Figure 3.1. It consisted of a submerged membrane system (system tank and membrane module), an aeration system (air line, air flow controller, central aerators), a permeate flux generation, collection and measurement system (Masterflex variable speed pump and digital scale), and a vacuum measurement system (digital pressure gauge, pressure transducer and data logger).  Pressure Transducer Peristaltic Pump  Computer  Pressure Gauge Permeate Line Thermometer  Mesh  Scale  Cylindrical 'Tank System Tank  , Cylindrical Baffle M embrane M odule  Air Flowmeter Compressed A i r Line  Source 'Water  Valve  ^  — D r a i n  Air Sparging System  Figure 3. 1 Schematic of the bench-scale submerged membrane system 26  Trans-membrane pressure for the system was defined as the difference between the magnitude  of the  vacuum inside the  permeate line  and  the  atmospheric  pressure.  Trans-membrane pressure was automatically measured by a transducer and recorded through a data-logger once every minute.  The permeate flux was periodically (once every 4-5 hours  during daytime hours) monitored by a digital scale to ensure that it remained constant over time.  Figure 3. 2 Picture of laboratory bench-scale experimental setup  3.1.1 Submerged membrane system  The submerged membrane system consisted of a system tank with flow control baffle inside, and a membrane module.  System tank  The system tank consisted of an open top cylindrical tank with an open ended concentric cylindrical baffle inside. The cylindrical tank had a diameter of 0.142 m and a height of 1.4 m. The working depth of the system tank was 1.0 m, resulting in a total working volume of 16 L. A 27  valve was located at the bottom of the tank to allow the fluid used in the experiment to be drained after each experiment.  The cylindrical baffle, with a diameter of 0.07 m and a height of 0.8 m, was located along the vertical center line of the cylindrical tank, and 0.025m from the bottom of the tank, as shown in Figure 3.3. The cylindrical baffle enabled the hydraulics and air sparging in the submerged membrane system to be controlled by using the aeration system, as described in Section 3.1.2.  Figure 3. 3 Schematic diagram of system tank with cylindrical baffle  28  Membrane module  The membranes used in the present research were outside-in flow hollow fiber membranes, produced by Zenon Environmental Inc (Oakville ON). The physical characteristics of the membrane are presented in Table 3.1.  Table 3. 1 Physical characteristics of the membrane used in the present study Configuration  Outside-in hollow fiber  Outside  Surface  N o m i n a l pore  Typical  diameter  properties  diameter  operating T M P  1.77 mm  Non-ionic &  0.04|Jm  1-8 psi  hydrophilic  For the present research, single membrane fiber modules were used. Membrane modules were produced by potting the membrane fibers into a top and bottom bulkheads as illustrated in Figure 3.4. Epoxy was used to pot the membrane fibers into the bulkheads. While the top bulkhead was open to allow for the permeate collection, the bottom bulkhead was sealed. Permenr.e  Figure 3. 4 Schematic of membrane module  29  The membrane fiber was 0.42 m long. It was held tightly and located inside the cylindrical baffle along the vertical center line, and was placed 0.195 m from the bottom of the system tank as illustrated in Figure 3.5. The total membrane surface area was 0.0023 m . 2  Figure 3. 5 Membrane module position inside the system tank  3.1.2 Aeration system The aeration system was used to control the hydrodynamic conditions in the system tank by changing the bulk cross-flow pattern and cross-flow velocities along the membrane surface.  According to previous studies, higher permeate flux could be achieved under dual-phase flow (air and water cross-flow along membrane surface) (Bellara et al, 1996, Smith et al, 2002, Smith et al, 2005, Cui et al, 2003, Berube and Lei, 2006, etc). Therefore, dual-phase bulk cross-flow conditions were adopted throughout the present research. Dual phase cross-flow was generated by aerating the system tank with the central aerator. 30  The central aerator, located at the bottom of the cylindrical tank, consisted of a perforated plate with eight 2 mm diameter holes arranged in a circle (see Figure 3.6). Air was added to the system tank through the central aerator.  0  2 nun Aerator  Base of system tank  Figure 3. 6 Schematic of center aerator The rising air bubbles, which were confined to the inside of the cylindrical baffle, entrained water upwards along the inside of the cylindrical baffle. The air and the entrained water flowed upwards along the membrane surface, on the inside of the cylindrical baffle, thus forming a dual phase cross-flow condition. A schematic of the flow path for the dual phase cross-flow is illustrated in Figure 3.7.  .  I I  O  i  l  Central Aerator Figure 3. 7 Schematic of dual phase cross-flow aeration system (Arrows indicate the direction of listed flow) 31  The velocities of bulk cross-flow were controlled by varying the air flow rate. For the present study, bulk cross-flow velocities of 0.2, 0.3 and 0.4 m/s were considered. The corresponding air flow rates (and air flow meter readings), to achieve these bulk cross-flow velocities for the dual phase cross-flow system, are presented in Table 3.2, and were obtained as part of a pervious study which focused on the impact of different hydrodynamic conditions on submerged membrane systems (Lei, 2005).  Table 3. 2 Operating bulk cross-flow velocities and air flow rate for dual phase system (Lei, 2005) B u l k cross-flow  A i r flow rate (ml/min)  A i r flow meter readings  0.2  3200  25  0.3  5300  42  0.4  7500  60  velocities (m/s)  The air flow rate to the aerator was controlled by using a valve and an air flow meter (model: Cole-Parameter: N034-39(ST)) connected to a compressed air line, as illustrated in Figure 3.1.  3.1.3 Permeate flux generation, collection and measurement system A peristaltic pump was used to provide the driving force that enabled the permeate to flow through the membrane. The pump was connected to the membrane module via a quarter inch hard tubing (1/4" OD food Grade Polyethylene Green Line G800-04). Hard tubing was used to prevent any collapse in the tube since the system was operated under vacuum (see Section 3.4). A Masterflex pump (model number 7521-50) with two Masterflex pump heads (model number 7013-21) operated in parallel was adopted. Each of the pump head rotors were offset to minimize pressure fluctuations. An additional set of two pump heads were used to pump raw water to the system tank to maintain the constant working volume of 16 L during every experiment. The 32  pump speed was adjusted to achieve the required permeate flux (see Section 3.4).  Permeate flux was collected in a 5L flask. The change in weight of permeate in the flask was monitored over time using an electronic scale (Scout Pro 4000 g Scale) to ensure that a constant permeate flux was maintained during every experiment. The permeate flux was regularly measured every 4 to 5 hours (day time).  3.1.4 Vacuum measurement system The use of the pump created a vacuum inside the membrane fibers, providing the driving pressure differential (i.e. trans-membrane pressure) that enabled the permeate to flow through the membrane.  Pressure transducer and data logger  A pressure transducer (Omega PX240 M0263/1100) was connected to the permeate line to measure the trans-membrane pressure. The transducer was connected to a data logger (National Instrument USB-6009) that conveyed data to a computer once every 1 minute using a custom Labview application (version 7.0).  Since the output from the pressure transducer was a voltage signal, a relationship was developed and used to convert the data from the pressure transducer to a pressure reading. The relationship was developed by applying several known vacuums to the transducer and recording the voltage signal obtained by the data logger for each vacuum. Prior to each experiment, the transducer was calibrated. Since the pressure transducer measured the relative pressure, all of the pressure readings obtained corresponded to pressure differences between the inside of the permeate line and the atmosphere (i.e. trans-membrane pressure).  Pressure gauge  A compound digital pressure gauge (Cole-Parameter, K-68950-00) was used to 33  periodically monitor the vacuum that was applied to the pressure transducer and ensure that the trans-membrane pressure measured by both the pressure transducer and pressure gauge were in agreement. Since all pressure measurements were relative measurements with respect to atmospheric pressure, prior to each experiment, the pressure gauge reading was set to zero.  3.2 Routine Membrane Integrity Testing and Maintenance  3.2.1 Membrane integrity testing Before starting each experiment, the integrity of the membrane module used was verified using a pressure hold (bubble) test.  A schematic of the membrane integrity testing (pressure hold test) apparatus is illustrated in Figure 3.8. For the integrity testing, fully wetted and submerged membrane fibers were pressurized to 41 kPa (6 psi) using a compressed air line, the membrane was then isolated from the compressed air line by closing an isolation valve, and the pressure was monitored for 2 minutes. According to the information from the manufacturer, at a pressure of 41 kPa (6 psi), a breach of approximately 3 um in diameter in the integrity of membrane surface should be detected without damaging the membrane fibers. The membrane was considered breached if a significant decline in the pressure (10% of the testing pressure) as observed during the test. The membrane was also considered breached if bubbles were observed to escape from the membrane module.  If the integrity of the membrane was compromised, the membrane module was discarded.  34  Pressure Gauge  Compressed Air Water Surface  Distilled Water  Isolation valve  4L Beaker  Membrane Module  Figure 3. 8 Schematic of the membrane integrity testing apparatus 3.2.2 Membrane cleaning After each experiment, the membrane module was cleaned using a sodium hypochlorite (NaCIO) solution (diluted from 6.0% Domestic Miroclean Bleach). As recommended by the membrane manufacturer, the membrane cleaning procedure used during the present study was as follows: 1. the used membrane fiber was soaked in a 750 ppm solution of sodium hypochlorite (NaCIO) solution (diluted from 6.0% Domestic Miroclean Bleach) for 16 hours; 2. the 750 ppm solution of NaCIO was then filtered through the membrane fiber at a -28 kPa (-4. lpsi) vacuum for a period of 20 minutes; 3. the membrane fiber was transferred into a fresh 50 ppm solution of NaCIO, and the solution was then filtered through the membrane fiber at -28 kPa (-4.1 psi) for a period of 20 minutes; 4. the cleaned membrane fiber was then stored by soaking it in a 50 ppm solution of NaCIO.  Before a cleaned membrane fiber was used for a filtration test, the membrane fiber was rinsed three times using distilled water. Distilled water was then filtered through the membrane fiber at -4. lpsi for a period of 20 minutes. 35  3.3 Source Water  The source water used in the present research was a mixture of raw water from the Capilano Reservoir and Jericho Pond. The Capilano Reservoir is located in North Vancouver and supplies 40 percent of the Great Vancouver Regional District's drinking water. Raw water from the Capilano Reservoir is characterized by low turbidity (<1 NTU) and low organic content (approximately 2 ppm as Total Organic Carbon). Because of its high quality, raw water from the Capilano Reservoir had a very low fouling potential. Preliminary tests indicated that it would take a number of days to weeks for the membrane to foul. To enable a greater number of experimental conditions to be considered with a shorter time frame, raw water from the Capilano Reservoir was mixed with raw water from Jericho Pond, located in Vancouver. The resulting raw water mixture had a TOC of approximately 6 ppm, and a relatively higher fouling potential.  Approximately 80 L of raw water from Jericho Pond was collected on June 5 , 2005 and th  approximately 120 L on Aug 4 , 2005 and stored in a refrigerator at 4 °C in the Environmental th  Engineering Lab at U B C . Raw water from Capilano Reservoir was collected on the same dates and stored in the same refrigerator as the Jericho Pond water. Approximately 300 L of Capilano water was collected on June 5 , 2005 and approximately 600 L was collected on Aug 4 , 2005. th  th  Prior to each experiment, Capilano and Jericho water was brought to room temperature and then mixed in a 5:1 ratio (5 parts Capilano : 1 part Jericho).  The TOC concentrations of different raw waters and their mixtures are summarized in Table 3.3. TOC tests revealed that even though the raw waters were stored in a refrigerator at 4 °C, the TOC of the raw water changed slightly over time as presented in Table 3.3.  36  Table 3. 3 TOC of raw water and their mixtures Second collection event  First collection event Water  Test date  Volume  TOC  collected  (ppm)  Test date  Volume  TOC  Test  TOC  collected  (ppm)  date  (ppm)  Capilano*  2005/06/05  300 L  2.62  2005/08/05  600 L  2.69  2006/01/27  2.02  Jericho*  2005/06/05  80 L  12.46  2005/08/05  120 L  27.50  2006/01/27  17.32  Mixture  2005/06/05  4.26  2005/08/05  6.82  2006/01/27  4.56  represents Capilano reservoir water and Jericho pond water, respectively  TOC tests were performed by the Persulfate-Ultraviolet Oxidation Method 5310 (APHA, AWWA and WEF) with the aid of a Dohrman Phoenix 8000 UV- Persulfate analyzer (Dohrman).  3.4 Experimental Program  3.4.1 Experimental conditions considered The overall objective of the present research was to develop a simple and reliable relationship to accurately model irreversible fouling in a submerged membrane system under different hydrodynamic conditions. The impact of different hydrodynamic conditions on irreversible fouling was assessed by monitoring the evolution of the trans-membrane pressure over time when  operating  the system  at a permeate flux  that was less than the  pseudo-steady-state (i.e. critical) permeate flux (J s) and when operating the system under S  different hydrodynamic conditions. The hydrodynamic conditions considered were dual-phase cross-flow at bulk cross-flow velocities of 0.2, 0.3 and 0.4 m/s.  As previously discussed, irreversible fouling was defined as the fouling that is not 37  impacted by the rate of mass transfer away from the membrane surface. Therefore, irreversible fouling can be observed when the operating permeate flux was less than the pseudo-steady-state permeate flux. Table 3.4 lists the pseudo-steady-state permeate fluxes for different operating conditions. These were obtained from previous studies using a similar source water and experimental apparatus (Lei, 2005). To ensure that the system was operating below pseudo-steady-state flux for all hydrodynamic conditions considered, operating fluxes of 40, 50, and 60 L/m h were considered. 2  Table 3. 4 Corresponding pseudo-steady-state permeate flux to system hydrodynamics B u l k d u a l phase  0.3 m/s  0.2 m/s  0.4 m/s  cross-flow velocities Pseudo-steady-state  65.0±1.30L/m h  91.8±4.24L/m h  2  2  93.7±1.0L/m h 2  permeate flux  Two series of experiments were performed to assess the impact of the system hydrodynamics on irreversible fouling. Series 1 was conducted using filtered and deionized water for all experimental conditions to estimate the clean water trans-membrane pressure. The source water used in Series 1 was filtered and deionized tap water (tap water was treated using a Reverse Osmosis Filtration System (model*: GXRM10G, Serial*: 04-1647). Series 2 was conducted using the 5:1 Capilano and Jericho raw water mixture. A l l experiments performed as part of Series 2 were performed in duplicate. The experimental series and experiment names are summarized in Table 3.5.  All experiments were performed at room temperature (range between 15-20 °C). Prior to the start of each experiment the waters were brought to room temperature. Water temperature was monitored throughout  the experiment using a thermometer  (Fisher 14-1997). A l l  experiments were conducted until the trans-membrane pressure increased above 35 kPa (5 psi).  38  Table 3. 5 Experimental program Permeate flux  B u l k cross-flow  (LmV )  velocity (m/s)  Series 1 (clean water)  Series 2  40  0.2  1-T-l  2-T-C, 24-T-G  40  0.3  l-T-2  24-T-E, 24-T-F  40  0.4  1-T-l  2-T-D, 24-T-D  50  0.2  l-T-2  2-T-H, 24-T-B  50  0.3  l-T-2  2-T-G, 24-T-H  50  0.4  l-T-2  2-T-J, 2-T-K  60  0.2  1-T-l  2-T-F, 2-T-F'  60  0.3  l-T-2  24-T-A, 24-T-C  60  0.4  1-T-l  2-T-E', 2-T-I  1  E x p e r i m e n t names  3.4.2 Experimental monitoring The experiments in Series 1 (i.e. clean water tests) lasted approximately 5 minutes. However, experiments in Series 2 lasted 6 to 15 days each. During this time, the permeate flux, the air flow meter readings, the pressure gauge readings and the water temperature were periodically monitored every 4-5 hours (day time) to ensure the system worked properly. The pressure transducer data were recorded automatically every minute as described in Section 3.1.4.  At the beginning of each experiment, a lag phase of up to 5 minutes was observed when the trans-membrane pressure increased as a vacuum was being formed inside the membrane and the tubing. The duration of the lag phase was relatively insignificant with respect to the duration of the entire experiments. To simplify the analysis of the data from the experiments in Series 2, the monitoring time was set to zero (i.e. time = 0) when the measured trans-membrane pressure reached the steady-state clean water trans-membrane pressure.  39  3.4.3 Trans-membrane pressure correction After the experiments were completed, the data were collected from the data logger and converted into a Microsoft Excel file.  Factors such as temperature, atmospheric pressure, consistency of the system set up (e.g. length of tubing, height of the instruments), can all have effects on trans-membrane pressure measurements.  Variations caused by system set up were avoided by keeping system  configuration consistent throughout the study period. The effect of ambient atmospheric pressure changes could be neglected since a relative pressure transducer was used in the study. However, the effect of temperature changes needed to be considered. Variations in the temperature affect the viscosity of the liquid and directly impact the measured trans-membrane pressure, as presented in Equation 3.1.  AP = JuR  (3.1)  t  where AP is the trans-membrane pressure [N/m ], J is the operating permeate flux [m /m s], | l is 2  3  the water viscosity [Ns/m ], and R is the total filtration resistance [m"7nr].  2  To remove the  t  effect of variation in the temperature on trans-membrane pressure, the measured trans-membrane pressure was standardized to a reference temperature of 20.0 °C using Equation 3.2.  P | i 20°C  (3.2)  T  1  20°C  where P o [N/m"] and P [N/m ] are trans-membrane pressure values at temperatures of 20.0 -  20  C  T  °C and T, respectively, and T is the operating temperature [°C]. | l  [Ns/m ] and fl [Ns/m ] 2  2 ( r c  2  T  are water viscosity values at temperatures of 20.0 °C and T, respectively. The temperature range in the present study was within 15-25 °C, and fi „ 20  Ns/m",  (i 5° 2  = c  c  = 1.002 x 10  3  Ns/m , u: o = 1.139x 10" 2  3  l5  c  0.890x 10"' Ns/m". Assuming the relationship between temperature and water 40  viscosity is linear within a temperature range of 15-20 °C, |0, can be calculated by Equation T  | i = -2.74x 10" T +1.55xl0" (see Figure 3.9 a), within temperature 20-25 °C, [i 5  3  T  T  calculated by Equation \i  = -2.24x 10" T+1.45x 10" (see Figure 3.9 b). To simplify the 5  T  can be  3  calculation within the 15-25 °C range, average values were adopted, where the water viscosity at a temperature T (\i ) can be calculated by Equation 3.3. T  | i =-2.49xl0~ T + 1.50xl0" 5  (3.3)  3  T  Water viscosity within temperature 15-20 C  3  u = -2.24E-05T+ 1.45E-03  u = -2.74E-05T + I.55E-03  r  T  0.00 II5 ^  Water viscosity within temperature 20-25 "C 0.00105  R- = I.00E+00  0.0011  ^  0.00105  R"= 1.00E+00  0.00!  3 0.00095  it  I"  g  0.001  f-  0.0009  t0.00095  0.00085 5  10 15 , Water Viscosity ur (Ns/nf)  20  25  10 15 20 , Water Viscosity ur (Ns/m")  (a)  25  (b) Figure 3. 9 Water viscosity vs Temperature  The standardized trans-membrane pressure at a temperature of 20.0 °C can therefore be calculated using Equation 3. 4.  P r  20°C  =  P | i 20°C  (3.4)  T  •2.49xlO T + 1.50xlO" _5  3  41  C H A P T E R 4 E X P E R I M E N T A L RESULTS  Two series of experiments, totaling 24 experiments were performed to obtain the data necessary to model irreversible fouling over time (see Table 3.5 for Experimental Program). Series 1 used a filtered and deionized water to measure clean water trans-membrane pressure under different operating permeate fluxes and cross-flow velocities. Series 2 used mixed source water to conduct a similar experimental program as series 1 to measure the evolution of trans-membrane pressure over time.  4.1 E v o l u t i o n of Trans-membrane Pressure for C l e a n W a t e r  When using clean water (i.e. filtered and deionized water), the trans-membrane pressure remained constant for a given operating permeate flux, regardless of the system hydrodynamic conditions (i.e. cross-flow velocity, and the volume of water filtered). However, the trans-membrane  pressure  increased when the operating permeate flux  increased. The  experimental results of series 1 are presented in Figure 4.1.  8000 7 0 0 0  ~g 6000  3  </>  $> 5000 D. § 4000 |  3000  §  2000  J v = 5 0 IVrrrh  Jv=601Vm-h i»ll»*llllH%»*Mliyilll*IIM«»myi»lllfc«l  Jv = 40l7rrfh  1000 0 20  40  60 ! Time (mins)  100  120  Figure 4. 1 Trans-membrane pressure for filtered and deionized water 42  A linear relationship was observed between trans-membrane pressure and the operating permeate flux when filtering clean water as presented in Figure 4.2. The linear relationship (i.e. AP = 127.79 Jv) was expected since for clean water, the trans-membrane pressure is theoretically proportional to the permeate flux as presented in Equation 3.1. 8000  7500  0=  7000  I  D.  | X)  6 5 0 0  E  U ?  6000  H  5500  5000 0  10  20  30  40  50  60  70  Operating permeate flux (17m h) 2  Figure 4. 2 Linear relationship between operating permeate flux and trans-membrane pressure 4.2 E v o l u t i o n of Trans-membrane pressure over T i m e for M i x e d Source Waters  As discussed in Section 3.4.1, 18 experiments were conducted in series 2. At least ten thousand trans-membrane pressure values were collected for each of the experiment. A l l the experiments showed a similar trend of increasing trans-membrane pressure over time. The trans-membrane pressure initially increased slowly. However, as more volume of water was filtered, the trans-membrane pressure increased at a faster rate. For a given operating permeate flux, the trans-membrane pressure increased more slowly at higher cross-flow velocity than at lower cross-flow velocity.  Figure 4.3, 4.4 and 4.5 show experimental trans-membrane pressure values with respect to the volume of water filtered (V) when operating at a permeate flux of 40, 50 and 60 L/m h. A l l 2  43  the data were normalized to 20 C to account for the impact of temperature on the change of source water viscosity as discussed in Section 3.4.3.  0  0.005  0.01  0.015  0.02  0.025  0.03  0.035  0.04  0  0.005  0.01  Volume of water filtered (m ) 3  0.015  0.02  0.025  0.03  0.035  0.04  Volume of water filtered (m ) ?  (a)  (b)  Figure 4. 3 Trans-membrane pressure vs Volume of water filtered at permeate flux 40 L/m h 2  (The experiments in (b) were duplicates of experiments in (a); C F V standards for cross-flow velocity)  0.005  0.01  0.015  0.02  0.005  Volume of water filtered (rrf ) 1  (a)  0.01 0.015 0.02 Volume of water filtered (m )  (b)  Figure 4. 4 Trans-membrane pressure vs Volume of water filtered at permeate flux 50 L/m h 2  (The experiments in (b) were duplicates of experiments in (a); C F V standards for cross-flow velocity)  44  0  0.002  0.004  0.006  0.008  0.01  0.012  0.014  0.016  0.002  Volume of water filtered (ni"')  0.004  0.006  0.008  0.01  Volume of water filtered (m ) 3  Figure 4. 5 Trans-membrane pressure vs Volume of water filtered at permeate flux 60 L/m h 2  (The experiments in (b) were duplicates of experiments in (a); C F V standards for cross-flow velocity) Three major observations can be made from the experimental results. First, for all experimental conditions, the trans-membrane pressures increased over time even though the experiments were performed with operating permeate fluxes that were less than the critical permeate flux. Second, for a given volume of permeate filtered, the extent of fouling (i.e. increase in trans-membrane pressure) was greater at a higher operating permeate flux. And finally, for a given volume of permeate filtered, the extent of fouling was lower at a higher cross-flow velocity.  45  C H A P T E R 5 M O D E L D E V E L O P M E N T , DATA ANALYSIS A N D DISCUSSION  5.1 M o d e l Development for Irreversible F o u l i n g  Fouling is typically assumed to result from the accumulation of material at the membrane surface (cake fouling), the plugging of membrane pores (intermediate/complete blocking) and adsorption of material within membrane pores (standard blocking, i.e. internal fouling) (Field, et al., 1995; Hermia, 1982; Bolton et al., 2006a). When the operating permeate flux in a membrane system is less than the critical permeate flux, the extent of fouling that is due to cake fouling and/or intermediate/complete blocking is expected to be minimal (Field et al., 1995). Therefore, when operating below critical flux, fouling should be predominately due to standard blocking, i.e. internal fouling.  5.1.1 Model 1: Irreversible fouling is only due to internal fouling The experiments performed in the present study were all conducted below critical flux conditions. Therefore, it was assumed that the sole mechanism responsible for fouling was material adsorption inside the membrane pore. Material adsorption causes a reduction in the pore radius. Based on this assumption, Model 1 was derived.  Derivation of Model 1  From Darcy's law, the flow through a clean membrane can typically be expressed using Equation 5.1. nr 7tAP 4  Jv=—  —  (5.1)  Where Jv:  the operating permeate flux [m /m s]  OUTL  3  xL: flow path length [m] n: number of pores [No./m ] 46  2  r: membrane pore radius [m] Rewriting Equation 5.1 to isolate the trans-membrane pressure yields Equation 5.2. AP  =  8)i x L J v  (5.2)  4  nr 7i Converting the units of operating permeate flux to L / m h yields Equation 5.3. 2  A T  ,  AP  =  8UTLJV  .  4  nr n  x  10 "  3  (5.3)  3600  Where Jv: the operating permeate flux [L/m h] 2  For a given membrane, the pore radius was assumed to become smaller when a given volume of liquid (V) had been filtered as illustrated in Figure 5.1  Accumulated foulant  (a) Unit length membrane pore  (b) Cross section of the pore after filtering a unit volume of liquid  Figure 5. 1 Unit length membrane pore and its cross section (Where ro is the initial pore radius, and r' is the pore radius after filtering a unit volume of liquid).  The reduction of pore radius can be related to the volume of water filtered. The greater the volume of water filtered, the greater the amount of material adsorbed inside the pore (i.e. reduction membrane pore size). For a specific membrane pore, M v was defined (see Equation 5.4) as the ratio of the volume of material adsorbed inside the pore wall to the volume of liquid that had passed through a pore of unit length.  47  Volume of materials adsorbed inside the pore wall  Mv =  (5.4)  Volume of liquid through the pore  Based on this assumption, the change in radius for a unit length of membrane pore, with respect to the volume of permeate filtered can be expressed using Equation 5.5. dr  -Mv  dV  2m(l)  (5.5)  Where V: volume of water filtered [nr ]  Integrating Equation 5.5 with r = ro at V = 0 and r = r' at V = V yields Equation 5.6.  Combining Equation 5.6 into Equation 5.3 yields Equation 5.7. AP=  8  x i ° -  ^ V ;  n ^ - ^ V )  1  3  6  0  (5.7)  0  K  For a given membrane module and raw water matrix, (I, xL, n and n are constants and can be grouped to into a single parameter K i " . Also,  MV/TC  can be grouped into a single parameter K 3 ,  which is proportional to the extent to which material adsorbs inside the pores. As a result, Equation 5.7 can be rewritten as presented in Equation 5.8 (i.e. Model 1). J v K ," A  P  = —  (r  l 2  0  —  (5. 8)  -K V) 3  -3  Where  K,"=  x—— Tin 3600  represents all the constants in Equation 5.7 [Nm h/L] 4  Mv =  • 71  is the volumetric reduction rate of pore radius [l/m]  ro: initial membrane pore radius = 2e-8 m in the present study  48  Significance of coefficients  The parameter K i " is not expected to be affected by the experimental conditions investigated in the present study, since K i " i s function of the raw water characteristics  and the  membrane module characteristics (TL, n). Recall that both the raw water composition and the membrane module configuration were constant during the different experiments. Therefore, in the present study, Ki"was expected to be constant for all experiments. The parameter K  3  describes the extent to which material adsorbs on the inside of a membrane pore. Considering that the adsorption of concern occurs within the pores themselves, K 3 is not expected to be affected by the hydrodynamic conditions outside of the pores (e.g. bulk cross-flow velocity). However, adsorption is expected to be affected by the hydrodynamic conditions within the pores themselves, and these conditions are governed by the permeate flow rate (i.e. operating permeate flux). Adsorption is also expected to be affected by the raw water matrix. Considering that all experiments were performed with the same raw water, K 3 was therefore expected to be affected only by the permeate flux for all experiments.  By using regression wizard in Sigmaplot software, the relationship presented in Equation 5.8 was fitted to the experimental data, and K , " , K values could be subsequently estimated. 3  Table 5.1 shows the estimated K i " , K 3 values and R for the regression. 2  49  Table 5. 1 R for regression of model 1 and estimated K i " , K values 2  3  Operating  Cross-flow  permeate  Velocity  flux(L/m h)  (m/s)  40  0.2  1#  40  0.2  40  Membrane  R  K i "values ( N m h / L )  K 3 values (1/m)  0.993  2.617E-29±2.906E-32  1.136E-14± .744E-18  4#  0.966  2.835E-29±6.830E-32  9.608E-15±1.543E-17  0.3  3#  0.975  3.094E-29±5.692E-32  6.540E-15±8.167E-18  40  0.3  3#  0.961  3.202E-29±6.794E-32  6.892E-15±1.044E-17  40  0.4  1#  0.992  2.279E-29±2.433E-32  7.248E-15±4.212E-18  40  0.4  4#  0.946  2.462E-29±2.888E-32  5.124E-15±3.307E-18  50  0.2  4#  0.968  2.853E-29±4.152E-32  1.394E-14±1.120E-17  50  0.2  4#  0.997  2.715E-29±1.454E-32  1.327E-14±4.065E-18  50  0.3  3#  0.953  3.702E-29±7.169E-32  9.264E-15±1.527E-17  50  0.3  3#  0.975  4.170E-29±7.598E-32  1.200E-14±1.873E-17  50  0.4  3#  0.972  3.177E-29±6.253E-32  7.640E-15±1.244E-17  50  0.4  3#  0.968  3.242E-29±9.841E-32  8.752E-15±2.388E-17  60  0.2  3#  0.978  3.110E-29±1.080E-31  2.397E-14+7.642E-17  60  0.2  3#  0.946  3.718E-29±1.495E-31  1.962E-14±6.287E-17  60  0.3  3#  0.972  3.226E-29±9.249E-32  1.550E-14±3.369E-17  60  0.3  3#  0.974  3.422E-29±6.879E-32  1.439E-14±2.614E-17  60  0.4  3#  0.905  3.750E-29±1.283E-31  1.085E-14±4.759E-17  60  0.4  3#  0.948  3.813E-29±1.201E-31  1.373E-14±4.226E-17  2  module  2  4  ± values are based on 90% confidence interval of the estimated parameters The fit of Equation 5.8 to the data was consistently good (i.e. R typically greater than 2  0.9). As expected, for a given membrane module, the parameter K i " was relatively constant for all experiments (Table 5.1). Recall that the parameter K i " corresponds to the raw water viscosity and clean membrane characteristics (see Equation 5.8), which are not expected to change over  time. However, as shown in Figure 5.2, the parameter K3 was impacted by not only the operating flux, but the bulk cross-flow velocity (i.e. hydrodynamic conditions outside of the membrane pores) as well. As previously discussed, parameters K was not expected to be impacted by the 3  bulk cross-flow velocity. K,@ 0.3m/s CFV  K,@ 0.2 m/s CFV K, = 5.654E-16Jv- I.298E-I4  3E-14 2.5E-14  £  K, = 4.1130E-16Jv-9.8010E-15  1.8E-14  R" = 0.8659  R = 0.9379  1.5E-14  2E-14  1.2E-14  1.5E-14  9E-15  IE-14  6E-15  5E-15  3E-15 0  0 10  20  30  40  50  60  10  70  20  30  40  50  60  70  Jv(L/m h)  Jv(L/m h)  2  2  (a)  (b)  K @ 0.4 m/s CFV K, = 3.0510E-16J V-6.3650E-15| 3  R = 0.8147 2  1.5E-I4 1.2E-I4 9E-15 6E-I5 3E-15 0 0  10  20  30  40  50  60  70  Jv(L/m h) 2  (c) Figure 5. 2 K3 vs Jv at different cross-flow velocities  Scanning Electron Microscope (SEM) images were taken of clean and fouled membrane surfaces to provide additional insight into the fouling mechanisms at the membrane surface. Presented in Figure 5.3 are 3 S E M images: Figure 5.3 a shows a clean membrane; Figure 5.3 b shows an intermediate fouled membrane (i.e. trans-membrane pressure reached 17.5 kPa); Figure 5.3 c shows an extensively fouled membrane (i.e. trans-membrane pressure reached 35 kPa). 51  A s presented, in Figure 5.3 b, there was no surface fouling for conditions representative of an intermediate amount of fouling. For these conditions, fouling appeared to be due mainly to the constriction of the pore structure. However, surface fouling was observed  for  conditions  representative of extensive fouling (see Figure 5.3 c). A s previously discussed, surface fouling is impacted by the bulk cross-flow velocity. Therefore, the parameter K 3 is likely not impacted by the bulk cross-flow velocity, but rather, the bulk cross-flow velocity likely impacts surface fouling, which is not considered in Equation 5.8.  (a) Clean Membrane  (b) Partially fouled Membrane (Pore size reduced)  (c) Extensively fouled membrane (With surface fouling)  Figure 5. 3 S E M images for clean and fouled membranes  A s previously discussed, when the operating permeate flux in a membrane system is less than the critical permeate flux, the extent of surface fouling (i.e. cake fouling and/or intermediate/complete blocking) is expected to be minimal (Field et al,  1995). However, as  presented in Figure 5.3, even though all of the operating permeate fluxes utilized in the present study were below critical permeate flux conditions, surface fouling was observed, for extensively fouled membranes.  52  5.1.2 Model 2: Irreversible fouling is due to internal fouling + surface fouling To account for the observed surface fouling, the relationship presented in Equation 5.8 was modified. A new model incorporating both internal fouling and surface fouling was developed.  Derivation of Model 2  For a given membrane module, the number of pores n (i.e. the active membrane surface area) was assumed to become smaller when a given volume of liquid had been filtered as illustrated in Figure 5.3. Fouled  Initial Active Area (A )  Surface  Active Surface Area  0  (a) Area of clean membrane  (b) Active surface area after filtering a unit volume of liquid  Figure 5. 4 Active surface areas of clean membrane and after filtering unit volume of liquid (where Ao is the initial active surface area of a given membrane module)  The reduction in the active surface area could be related to the volume of water filtered. The greater the volume of water filtered, the greater the amount of surface fouling (i.e. reduction in active surface area). For a specific membrane, Sv was defined (see Equation 5.9) as the ratio of the area impacted by surface fouling to the volume of liquid filtered through the membrane.  Sv-  Area impacted by surface fouling Volume of liquid filtered through membrane_  . ^ (5.9)  /r  Based on this assumption, the change in the active surface area with respect to the 53  volume of permeate filtered could be expressed using Equation 5.10.  Active area = A - Sv V  (5.10)  0  Where Ao: membrane initial active surface area = 0.0023 m" in the present study Renaming Sv as K , for consistency in terminology, substituting Equation 5.10 for the 2  number of filtration pores (n) in Equation 1, and applying a similar development as presented in Equation 5.2 to 5.8 yields Equation 5.11(i.e. Model 2)  a p  =  (A -K V)(r 0  2  \  2  0  (5.1D  r  -K V) 3  -3  Where K, = ^M-tL ^ JTJ— represents all constants in Equation 5.7 [Nm h/L] it 3600 6  K : rate of pore number reduction (i.e. rate of surface fouling) [1/m] 2  K : rate of pore radius reduction (i.e. rate of internal fouling) [1/m] 3  Significance of coefficients  Model 2 represents irreversible fouling comprehensively, by combining surface fouling and internal fouling together. Again, the parameter K , was not expected to be affected by the experimental conditions investigated in the present study, and K 3 was not expected to be affected by the bulk cross-flow velocity, but rather by the permeate flux. The parameter K describes the 2  extent of surface fouling, and its value was expected to be related to the difference between the rate of mass transfer away from the membrane surface and the rate of mass transfer toward the membrane surface. The rate of mass transfer away from a membrane surface is governed by the hydrodynamic conditions (i.e. cross-flow velocity) and system configuration (Malleviale et al., 1996), while the rate of mass transfer toward a membrane surface is governed by the operating permeate flux. Therefore, the parameter K was expected to be affected by the bulk cross-flow 2  velocity and the operating permeate flux. 54  5.2 Data A n a l y s i s and Parameter E s t i m a t i o n of Irreversible F o u l i n g M o d e l  Since Model 2 was a mathematically complex model, the regression software (i.e. Sigmaplot) could not be used to compute the parameters K | , K and K3 directly for Model 2. 2  5.2.1 Approach 1: Combining numerical approaches to estimate Kj, K2 and K3 Initially, a combination of analytic software: Matlab, Excel and regression  software:  Sigmaplot, was used to attempt to estimate parameters K), K and K 3 . 2  The first step consisted of inverting both sides of Equation 5.11 yielding Equation 5.12 1 _ A / AP  (-2A r K -K r ) 2  {  0  0  JvK,  4  3  2  0  v  |  A K 0  JvK,  2 3  + 2r K K 2  0  JvK,  2  3 y  2  <  (-K K 2  2 3  ) 3 y  JvK,  Equation 5.12 can be simplified by adopting the definitions presented in Equations 5.13 to 5.18.  Y=AP  (5.13)  X =V  (5. 14)  Y  a  r  A  0  0  4  = ^ JvK,  (5.15)  (-2A r 'K,-K r:>  =  0  0  2  JvK,  b  _  A  0  K  3  2  +  2  r  Q  2  K  2  K  3  ,5  JvK,  JvK, 55  With the above definition, Equation 5.12 could be rewritten as presented in Equation 5.19 for which the transition coefficients yo, a, b, c, could be solved using polynomial equation fitting in regression wizard of Sigmaplot. Y = Y +aX + bX +cX 2  (5.19)  3  0  The second step consisted of using Matlab (see Appendix A for the editing code) to compute parameters K i , K and K from the coefficients y , a, b, c obtained using Sigmaplot. 2  3  0  Unfortunately, due to the high dependency between parameters K  and K , their values  2  3  calculated from the coefficients obtained from Sigmaplot were not real numbers, but rather complex numbers. When using Matlab to solve K i , K and K , the imaginary part was not taken 2  3  into account, leading to significant errors in the estimation of K and K values. In addition, it 2  3  was also not possible to obtain the standard errors of the estimated parameters K and K using 2  3  this approach. It was concluded that a more robust approach was needed to estimate the parameters K | , K and K 2  3  5.2.2 Approach 2: Estimate Kj, K2 and K3 separately A more robust approach was developed where by the relationship presented in Equation 5.11 was separated into two parts as presented in Equation 5.20. (5. 20)  A  B  Where K, = K i ' K i "  (5.21)  K i ' : a constant related to membrane module and raw water characteristics [m ] 2  In Equation 5.20, Part A represented surface fouling and Part B represented internal fouling. This more robust approach enabled the highly dependent parameters K and K to be 2  estimated. Details of the estimation procedure are presented below.  56  3  Step 1 Estimating the parameters K/' and Ki  The first step consisted of rewriting Equation 5.8 which is similar to Part B of Equation 5.20, as presented in Equation 5.22. VJVV 2  VAP =  \  (5.22)  i + (-%v r  o  Equation 5.22 can be simplified by adopting the definitions presented in Equations 5.23 to 5.26.  Y = VAP  (5.23)  x=v  (5. 24)  a=  VJvK," —  (5. 25)  b = i (5. 26) Combining Equations 5.23 to 5.26 yields Equation 5.27 Y =  ^  1 + bX  (5. 27)  Where a, b could be solved using rational equation fitting in regression wizard of Sigmaplot. Then parameters K i " and K 3 can be computed using Excel. Note that with this approach, the standard error of the estimates for the parameters K i " and K 3 could be obtained.  Equation 5.27 was fitted (i.e. using rational equation in regression wizard of Sigmaplot) to the experimental data obtained when operating the membrane system with a high bulk cross-flow velocity (0.4 m/s). At a high bulk cross-flow velocity, the impact of surface fouling was expected to be minimal (Field et al, 1995). This was confirmed using S E M images (i.e. for 57  bulk cross-flow velocities of 0.4 m/s, no surface fouling was observed).  The estimated K j " and K3 values for the different experiments conducted with a bulk cross-flow velocity of 0.4 m/s are presented in Table 5.2.  Table 5. 2 Estimated K i " and K3 values at a bulk cross-flow velocity of 0.4m/s Operating  Cross-flow  permeate flux  Velocity  Membrane Ki"values (Nm h/L)  K3 values ( l / m )  4  module  (L/m h)  (m/s)  40  0.4  1#  2.279E-29±2.433E-32  7.248E-15±4.212E-18  40  0.4  4#  2.462E-29±2.888E-32  5.124E-15±3.307E-18  50  0.4  3#  3.177E-29±6.253E-32  7.640E-15±1.244E-17  50  0.4  3#  3.242E-29±9.841E-32  8.752E-15±2.388E-17  60  0.4  3#  3.750E-29±1.283E-31  1.085E-14±4.759E-17  60  0.4  3#  3.813E-29±1.201E-31  1.373E-14±4.226E-17  2  ± values are based on 90% confidence interval of the estimated parameters  Table 5.2 shows that K | " values are relatively constant for a given membrane module (e.g. #3). Table 5.2 also shows that for a given operating permeate flux, K 3 values are relatively constant as expected (see Section 5.1.1).  Step 2 Estimating the parameters Ki' and K?  The second step consisted of estimating parameters K i ' and K from the residuals 2  obtained from fitting Equation 5.8, with the parameters ( K , " and K3) obtained from step 1, to all of the experimental data as presented in Equation 5.28. AP  K ' /c  experimanetal  JvK," (r -K V) 2  0  ~A -K,V n  2  3  58  Oft^  WhereAPexperimentai: the increase of trans-membrane pressure values from experimental data  K i " and K 3 were estimated from step 1  Equation 5.28 can be modified as presented in Equation 5. 29 ^^experimental  (r  0  2  O H \  AQ  -K,V)  A0  Equation 5.29 can be simplified by adopting the definitions presented in Equations 5.30 to 5.33. AP ,r  experimental  JvK," (r -K V) 2  0  P  orw ]  2  3  a' = ^  (5.31)  b' = - ^  (5.32)  X =V  (5.33)  Combining Equations 5.30 to 5.33 yields Equation 5.34  a 1 + b'X  (5. 34)  Where a', b' can be estimated using rational equation fitting in regression wizard of Sigmaplot. The parameters K i ' and K could be subsequently calculated. The standard error of 2  the estimates for the parameters K i ' and K could also be obtained. 2  59  The estimated K | ' and K values from different experiments conducted are presented in 2  Table 5.3. Table 5. 3 Estimated K i ' and K values 2  Operating  Cross-flow Membrane  permeate flux  Velocity  K i 'values (m )  K 2 values ( l / m )  2  module  (L/m h)  (m/s)  40  0.2  1#  2.102E-03±1.635E-06  6.914E-02±4.315E-05  40  0.2  4#  2.277E-03±4.368E-06  4.963E-02±1.142E-04  40  0.3  3#  2.300E-03±3.356E-06  1.935E-02±1.207E-04  40  0.3  3#  2.413E-03±4.760E-06  2.231E-02±1.559E-04  50  0.2  4#  2.192E-03±2.552E-06  7.728E-02±1.073E-04  50  0.2  4#  2.088E-03±1.330E-06  6.941E-02±6.130E-05  *50  0.3  3#  3.229E-03±5.296E-06  5.131E-02±2.315E-04  60  0.2  3#  2.477E-03±7.280E-06  1.528E-01±4.685E-04  60  0.2  3#  2.914E-03±1.097E-05  1.085E-01±6.990E-04  60  0.3  3#  2.445E-03±6.194E-06  6.682E-02±4.394E-04  60  0.3  3#  2.673E-03±5.779E-06  4.398E-02±4.560E-04  2  ± values are based on 90% confidence interval of the estimated values *The results of K i ' and K from one of the experiments with a cross-flow velocity of 0.3 m/s 2  and an operating permeate flux of 50 L/m h were discarded because the model did not fit the 2  experiment data (R = 0.202 from Sigmaplot regression) 2  It should be noted that there are no results for K | ' and K for a cross-flow velocity of 0.4 2  m/s, since at this cross-flow velocity, the extent of surface fouling was assumed to be minimal as discussed in Step 1.  Table 5.3 shows that K i ' values are relatively constant for the same membrane module, 60  and all the K i ' values from different membrane modules are similar. Table 5.3 also shows that for a given operating permeate flux and cross-flow velocity, K values are relatively constant as 2  expected (see Section 5.1.2).  Step 3 Combining Steps 1 and 2 to yield K\_  K, = K , ' K , "  The product of K | " and  (5.35)  (i.e. K i = K i ' K i " ) was relatively constant for all the  experiments. This was expected since the parameter Kj is only a function of the membrane module and the raw water characteristics, both of which were constant through the study. Thus the average values of K i " and K i ' were used to estimate parameter K\. The resulting Kj was 7.080E-32±2.745E-37 for the membrane module and raw water used in the present study.  5.3 M o d e l Parameter Quantification a n d Discussion  5.3.7 Quantification ofK] As discussed previously, K i is only a function of the characteristics of the specific membrane module used and the raw water characteristics. As expected, the parameter K] was relatively constant for a given membrane module. Recall that the parameter K i corresponds to the raw water viscosity and clean membrane characteristics (see Equation 5.20), which are not expected to change over time. Since the same length and material of membrane modules were used for all experiments and the same raw water was used throughout the study, K i values should be constant for a given membrane module. Thus the average value of K i could be presented as K i for the membrane module and raw water used in the present study. As shown in Step 3 in 5.2.2, the average K i was estimated to be K, = 7.080E-32±2.745E-37.  61  5.3.2 Quantification of K2 and the impact of operating permeate flux and bulk cross-flow velocity on the extent of surface fouling Surface fouling is governed by both the mass transfer towards and away from the membrane surface. The rate of mass transfer towards the membrane surface is largely governed by the operating permeate flux, while the rate of mass transfer away from the membrane surface is largely governed by the bulk cross-flow velocity. Therefore, the extent of surface fouling, which is described using the parameter K was expected to be a function of both the operating 2  permeate flux and the bulk cross-flow velocity.  The relationship between the parameter K and the operating flux as well as the bulk 2  cross-flow velocity is presented in Figure 5.5. A n empirical relationship was developed in Sigmaplot to describe the impact of cross-flow velocity and operating permeate flux on K as 2  presented in Equation 5.36. Equation 5.36 could be used to accurately estimate the parameter K  2  from the cross-flow velocity and operating permeate flux (i.e. R = 0.865 for the exponential 2  relationship was the highest R among all the regressions).  K  2  =133.0exp  •0.5x  Ov-387.lY 87.73  +  f CFV-0.2  V  0.08004  Where CFV: cross-flow velocity applied to the membrane system [m/s]  62  (5. 36)  Figure 5. 5 The relationship between K and Jv as well as C F V 2  As expected, increasing the cross-flow velocity decreases the rate of surface fouling, defined based on the parameter K . Higher cross-flow velocities create stronger shear force 2  along a membrane surface, and remove and/or prevent the accumulation of material on the membrane surface, thus reducing surface fouling.  In addition, increasing the operating permeate flux increases the rate of surface fouling K . At higher operating flux, the mass transfer of material towards the membrane surface is 2  higher, and therefore, more material is likely to accumulate on the membrane surface. Surface fouling occurs more quickly at higher operating flux.  It should be noted that the empirical relationship developed to estimate the parameter K? (Equation 5.36) is only valid for conditions relatively similar to those used in the present study. More specifically, Equation 5.36 cannot be used to estimate the parameter K when the 2  63  cross-flow velocity is less than 0.2 m/s and when the operating permeate flux is greater than 387.1 L/m h. Additional research is required to develop a mechanistic relationship that can be 2  used to model the experimental results presented in Figure 5.4.  5.3.3 Quantification of K3 and the impact of the operating permeate flux on the extent of internal fouling Internal fouling is governed by the hydrodynamic conditions inside the membrane pores, which in turn is determined by the permeate flow rate (i.e. operating permeate flux). Therefore, the parameter K3 is only expected to be affected by the operating permeate flux.  Linear relationship between Kumd Jv  A relatively linear relationship was observed between the parameter K and the 3  operating permeate flux (Jv) from the experimental data as illustrated in Figure 5.6. As expected, even for operation below the critical flux, increasing the operating permeate flux (Jv) increases the rate of internal fouling.  1.6e-14-|  1  1.4e-14-|  4.0e-15 -I  1  1  1  1  ,  1  35  40  45  50  55  60  65  Operating permeate flux (L/m h) 2  Figure 5. 6 Linear relationship between K3 and Jv 64  The fit of K3 values to the linear relationship (i.e. K = (3.051E-16)Jv -6.365E-15) was 3  good (R = 0.8147). However, this linear relationship was only effective over the range of 2  experiments performed in this study. Theoretically, the rate of internal fouling K should be 3  minimal when there is no flux. But according to the above relationship, when operating at a low permeate flux (e.g. Jv = 20L/m h), the value associated with K would be negative. Thus the 3  linear relationship is not fully compatible with the definition of K . 3  Non-linear  relationship between Kjand Jv  For model development purpose, it was assumed that K should be zero at an operating 3  permeate flux of zero. Based on this assumption, a non-linear regression was conducted to reveal the relationship between K and the operating permeate flux. A non-linear relationship was 3  observed between K a n d the operating permeate flux as presented in Figure 5.6. The quadratic 3  polynomial relationship presented in Equation 5.37 could be used to accurately estimate the parameter K from the operating permeate flux (i.e. R = 0.9311). 2  3  K =2.727E-18 Jv +3.769E-17Jv  (5. 37)  2  3  1.6E-14 1.4E-14 1.2E-14 IE-14 £  8E-15 6E-15 4E-15 2E-15 0 0  10  20  30  40  50  60  2  Operating permeate flux (L/m h)  Figure 5. 7 Non-linear relationship between K and Jv (forcing intercept = 0) 3  65  70  Non-linear  relationship and adsorption kinetics  The non-linear relationship between K3 and operating permeate flux is consistent with the hypothesis that internal pore fouling occurs due to adsorption of foulants to the pore surface. The rate of adsorption is greatest when the concentration gradient of the material being adsorbed at the surface of the adsorbent is high. Increasing the operating permeate flux (i.e. liquid flow through the pore), increases the concentration gradient at the surface of the adsorbent (i.e. membrane surface).  The results from the present study appear to contradict those reported by Bolton et al. (2006b). Base on their data analysis, they reported that membrane capacity could increase with increasing flow rate (i.e. extent of fouling decreases with increasing flow rate). However, their estimate of capacity was based on the volume of permeate filtered after a given time. When considering the extent of fouling (i.e. pressure increase) based on a given volume of permeate filtered, the extent of fouling increased with increasing flow rate, contradicting their conclusion that membrane capacity can increase with increasing flow rate.  Further research is required to comprehensively investigate adsorption kinetics within membrane pores. Nonetheless, the results from the present study, and those from other studies (Bolton et al. 2006b) confirm that for a given volume of permeate filtered, increasing the operating permeate flux increases the extent of fouling.  5.3.4 Relative contribution of internal fouling and surface fouling to irreversible fouling It should be noted that the model developed in the present study (Equation 5.11) is only valid when K V is less than Ao and K 3 V is less than r . Equation 5.11 can be rewritten as 2  0  presented in Equation 5.38:  AP = JvK,(  1 A  o - K V 2  (5. 38)  ) ( — - J -) (r -K V) 2  0  3  66  Where Part C =  , and A  o"  K  2  Part D =  T  (r - K V ) 2  V  0  3  For a given membrane module, Part C represents surface fouling, which is impacted by cross-flow velocity as well as operating permeate flux, while Part D represents internal fouling, which is impacted by operating permeate flux. Both surface fouling and internal fouling contribute to the increase of trans-membrane pressure. However, the part that changes the most when filtering a given volume of water has a greater impact on the increase of trans-membrane pressure. Taking conditions of cross-flow velocity of 0.3 m/s and operating permeate flux 50 L/m h as an example, the changes of surface fouling and internal fouling (i.e. Parts C and D) 2  over time are presented in Figure 5.8 and 5.9.  800  600 u  V  4  00  200  0.005  0.01  0.015  0.02  Volume of water filtered (m") Figure 5. 8 The change of surface fouling over volume of water filtered  67  2.5E+31 2E+31 Q Cu  1.5E+31 1E+31 5E+30 0 0  0.005  0.01  0.015  0.02  Volume of water filtered(m ) 3  Figure 5. 9 The change of internal fouling over volume of water filtered  Figure 5.8 shows that the magnitude of Part C increases less than 100% over the volume of water V filtered, while Figure 5.9 shows that the magnitude of Part D increases more than 200% over the volume of water V filtered. These results indicate that internal fouling has a much bigger effect on fouling. This also indicates that for operation below the critical flux, even though surface fouling and internal fouling both contribute to the extent of fouling, internal fouling plays a much more important role than surface fouling. Extensive internal fouling occurs along with small amount of surface fouling over the volume of water filtered, when operating the system at less than critical flux.  5.4 C o m p a r i s o n of M o d e l Predictions a n d E x p e r i m e n t a l D a t a  By quantifying the parameters K ] , K  2  and K 3 in model 2 (i.e. Equation 5.11),  trans-membrane pressure over time can be predicted. Figures 5.10 to 5.12 present a comparison between the modeled and measured  trans-membrane  pressure values for the  different  experiments performed with operating permeate fluxes of 40 L/m h, 50 L/m h and 60 L/m h 2  2  2  respectively. To model the evolution of the trans-membrane pressure over time, the parameter K | was estimated to be K i = 7.080E-32, and the parameters K and K3 were estimated from 2  Equations 5.36 and 5.37, respectively. In these figures, data points that are located directly on the 68  dotted line correspond to conditions under which the measured trans-membrane pressure was identical to the modeled trans-membrane pressure.  5.4.1 Conditions for which model agrees with experimental trans-membrane pressure data For an operating permeate flux of 40 L/m h, as presented in Figure 5.10, the model 2  predictions were typically in relatively close agreement with the experimental measurements. For experiment 24-T-D (Figure 5.10 f), the measured trans-membrane pressure was consistently and substantially less than the modeled trans-membrane pressure. However, it should be noted that experiments 24-T-D (Figure 5.10 f) and 2-T-D (Figure 5.10 e) were performed under similar experimental conditions (i.e. cross-flow velocity 0.4 m/s and operating permeate flux of 40 L/m h), and that the measured and modeled trans-membrane pressures for experiment 2-T-D (Figure 5.10 e) were in relatively close agreement. The discrepancies observed between the modeled and measured values presented in Figures 5.10 f are likely due to non-controllable experimental variations (i.e. experimental error).  For an operating permeate flux of 50 L/m h, as presented in Figure 5.11, the model 2  predictions were typically in relatively close agreement with the experimental measurements. For experiment 24-T-H (Figure 5.11 d), the measured trans-membrane pressure was consistently and substantially less than the model predicted trans-membrane pressure. However, it should be noted that experiments 24-T-H (Figure 5.11 d) and 2-T-G (Figure 5.11 c) were performed under similar experimental conditions (i.e. cross-flow velocity 0.3m/s and operating permeate flux of 50 L/nTh), and that the measured and modeled trans-membrane pressure for experiment 2-T-G (Figure 5.10 c) were in relatively close agreement. The discrepancies observed between the modeled and measured values presented in Figures 5.11 d are likely due to non-controllable experimental  variations  (i.e.  experimental  error).  In  addition,  the  model  predicted  trans-membrane pressures of experiments 24-T-B (Figure 5.11 b) and 2-T-J (Figure 5.11 e) were more than the measured trans-membrane pressure after certain amount of water had been filtered. However, for experiments 2-T-H (Figure 5.11 a) and 2-T-K (Figure 5.11 f), which were 69  conducted under the similar conditions as experiments 24-T-B and 2-T-J respectively, the model predictions were in relatively close agreements with the measured trans-membrane pressures. Again, the discrepancies observed between the modeled and experimental values presented in Figures 5.11 c and d, as well as Figure 5.11 b and e, are likely due to non-controllable experimental variations (i.e. experimental error).  70  2-T-C (0.2 m/s, 40 L/mTi)  24-T-G (0.2 m/s. 40 L/m'h)  41XXX) '—' 15000 < G 30000 CL tU c 173  25000  X)  20000  E  <D  E  15000  C  is -o  OJ  KXXX)  0>  T3 O  5000  2  5000  I (XXX)  15()(X)  2IXXX)  251XX)  3IXXX)  5000  35000  10000  151XX)  20000  (a)  (Pa  «  CL  4(XXX>  *aj  gj 5? 35(XX> aj Cl 30(XX)  45000  45000 4(M)(X)  =3 t/3  35000 K a. 3(M)(X)  c  .0  25(HX) 2(X)(X)  15000  ans -me  c  KXXX)  •o  KMXX)  x> 25(XX) E 2IXXX) ans -me  40000  24-T-F (0.3 m/s, 40 L/m"h)  45(XX)  •o o S  35(X)()  (b)  24-T-E (0.3 nVs,40 L/m"h)  h:  30000  Measured trans-membrane pressure(Pa)  Measured trans-membrane pressure (Pa)  •O u <U  25(X)0  E  < u HJ T3 O  5(XX)  15000  5(HX) 0  SIMM)  KXXX)  50(X)  15000 2(XXX) 25IXX) 3(XXX> 35(XX) 4(XXX) 45000 5(XXX)  KXXX)  15000  2IKKX)  25000  3(XXX)  350IM)  4()(X)() 450IX)  Measured trans-membrane pressure (Pa)  Measured trans-membrane pressure (Pa)  (c)  (d)  2-T-D (0.4 m/s,40 L/mTi)  24-T-D(0.4m/s.40L/mTi)  35000 .KXXX) 25000 2IXXX) 15000 KXXX) 5000 0 0  5000  10000  15000  20000  25000  .KXXX)  35000  40000  0  Measured trans-membrane pressure (Pa)  5000  KXXX)  15000  20000  25000  30(XX)  35000  Measured trans-membrane pressure (Pa)  (e)  (f)  Figure 5. 10 Predicated against measured data for 40 L/m h operating permeate flux 2  71  40000  2-T-H (0.2 m/s,50 Um h)  24 T-B(0.2 nVs,50 Urn h)  50000  ;sure  ,—.45000  a a. a) 2  40000 35000 30000  £ 25000 £ 20000 c  fa15000  13 JU  10000  Mo  -a  5000 0 5000  10000  1500O  2«)00  25000  30000  350(H)  5001)  10000  15000  20000  (a)  40000  45000  35000  40000  (b)  2-T-G (0.3 m/s .50 Um h) 2  24-T-H(0.3 m/s,50 Until)  45(X)0  40000 CL  w 4OO0O  iT  35000  =>  35000  < G 30000 Q1) C  ° - 30000 J  35000  Measured trans-membrane pressure(Pa)  Measured trans-membrane pressure (Pa)  |  251X10 30000  40000  25000  25000 E 20000 OJ E 15000  trans  £ | 20000  c  s  •a  •g I0(XX)  10000 5000  o 5(XX1  I  0 0  5000  10000  15000  20000  25000  30000  35000  0 5(X)0  40000  10000  15000  2(XXX)  25000  30000  Measured trans-membrane pressure (Pa)  Measured trans-membrane pressure (Pa)  (c)  (d) 2-T-K(0.4 m/s,50 Um"h)  2-T-J(0.4 m/s.50 Um h)  70000  6001X1  50000  400(X)  30000  20000  10000  0 5000  10000  15000  20000  25000  30000  35000 40000  10000  45000  15000  2IXXX)  25000  30000  35000  Measured trans-membrane pressure (Pa)  Measured trans-membrane pressure (Pa)  (e)  (f) 2  Figure 5. 11 Predicated against measured data for 50 L/m h operating permeate flux 72  2-T-F(0.2 m/s,60 Unfh)  2-T-F(0.2 m/s,60 Um h)  40000 35000  <L)  5? <D CL  c  2  30000 25000  x> E 20000 <u  E  c £  15000 10000  — O  y  5000  5  0  10000  20000  .10000  40000  50000  10000  Measured trans-membrane pressure(Pa)  20000  30000  40000  Measured trans-membrane pressure(Pa)  (a)  (b)  24-T-A (0.3 nVs.60 L/m h)  24-T-C (0.3 m/s .60 Um h)  :(Pa  45000  a  40000 35(X)0  u!  bra  a. 30000 c 25(X)0 E  OJ  E  20000  c 15(X)0 t:  •o 1 (XXX) o  s  5000  50(X)  10000 15000 20000 25000 30000 35000 40000 45000 50000  0  5000  KXXX) 15000 20000 25000 30000 350<)0 40000 45000 50000  Measured trans-membrane pressure (Pa)  Measured trans-membrane pressure (Pa)  (c)  (d) 2-T-I (0.4 m/s,60 Unfh)  2-T-E' (0.4 m/s,60 L/mTi) 45000  ~40000 I 35000 30000 C J 25000 E  | 20000 1 15000 13 10000 1  s  5000 0  5000  10000  15000 20000  25000 30000 35000 40000 45000  5000  Measured trans-membrane pressure (Pa)  10000  15000 20000 25000  30000 35000 40000 45000  Measured trans-membrane pressure (Pa)  (e)  (f)  Figure 5.12 Predicated against measured data for 60 L/m h operating permeate flux 2  73  For an operating permeate flux of 60 L/m h, as presented in Figure 5.12, the model predictions were typically in relatively close agreement with the experimental measurements. For experiment 2-T-I (Figure 5.12 f), the measured trans-membrane pressure was consistently and substantially less than the model predicted trans-membrane pressure. However, it should be noted that experiments 2-T-I (Figure 5.12 f) and 2-T-E' (Figure 5.12 e) were performed under similar experimental conditions (i.e. cross-flow velocity 0.4m/s and operating permeate flux of 60 L/m h), and that the measured and modeled trans-membrane pressure for experiment 2-T-E' 2  (Figure 5.12 e) were in relatively close agreement. In addition, the model predicted trans-membrane pressures of experiments 2-T-F (Figure 5.12 a) were less than the measured trans-membrane pressure after certain amount of water had been filtered. However, for experiments 2-T-F' (Figure 5.12 b), which was conducted under the similar conditions as experiments 2-T-F (Figure 5.12 a), the model predictions were in relatively close agreements with the measured trans-membrane pressures. Therefore the discrepancies observed between the model predicted and measured values presented in Figure 5.12 f and e, as well as Figure 5.12 a and b, are likely due to non-controllable experimental variations (i.e. experimental error).  5.4.2 Conditions when model overestimates trans-membrane pressure data For all experiments, the model predictions were greater than the experimental measurements at low trans-membrane pressure. Low trans-membrane pressures occur at the start of filtration cycles when the membrane is not extensively fouled. It is not clear why this discrepancy was observed. It could be hypothesized that as foulants adhere to membrane material, the effective characteristics of the membrane surface change. As a result, a partially fouled membrane could subsequently foul more easily than a clean membrane. Evidence for such behavior was also provided by the S E M images (Figure 5.3). For partially fouled membranes, no surface fouling was visually observed, while for heavily fouled membranes, extensive surface fouling was observed. Further studies are required to investigate this hypothesis.  74  C H A P T E R 6 CONCLUSIONS A N D SIGNIFICANCE T O E N G I N E E R I N G  6.1 Conclusions  1. Even though all experiments were performed with an operating flux that was less than the critical flux in this study, a substantial amount of fouling was observed when filtering over extended periods of time.  2. The extent of fouling was observed to be related to both the operating permeate flux and the system hydrodynamic conditions (i.e. the cross-flow velocity).  3. Irreversible fouling observed in this study was due to both extensive internal/pore fouling and surface/cake fouling. Internal fouling was the predominant mechanism that governed irreversible fouling.  4. A semi-empirical relationship (see Equation 6.1) was developed to model the extent of fouling when filtering over an extended period of time for conditions under which the operating permeate flux is less than the critical flux. The relationship is based on three parameters K i , K and K which correspond to the membrane characteristics, the extent of 2  3  surface/cake fouling and the extent of internal/pore fouling, respectively.  JvK, A  =  P  \  (A -K V)(r 0  Where  2  T 2  0  (6.D  -K V) 3  K i : a constant related to membrane module and raw water characteristics [Nm h/L] 6  K : rate of pore number reduction (i.e. rate of surface fouling) [1/m] 2  K : rate of pore radius reduction (i.e. rate of internal fouling) [1/m] 3  5. The extent of surface/cake fouling was determined to be governed by the operating permeate 75  flux and the system hydrodynamic conditions (i.e. the cross-flow velocity) as presented in Equation 6.2. K  2  =133.0exp^-0.5x  Ov-387.lY  fCFV-0.2^  87.73  0.08004  2  (6.2)  Where CFV: cross-flow velocity applied to the membrane system [m/s] Jv: the operating permeate flux [L/m"h]  6. The extent of internal/pore fouling was determined to be governed only by the operating permeate flux as presented in Equation 6.3.  K =2.7268E-18 Jv +3.769E-17Jv  (6.3)  2  3  7. The study indicated that when operating a membrane filtration system below critical flux conditions, for a given volume of permeate filtered, the extent of irreversible fouling increases as the operating permeate flux increases and decreases as the cross-flow velocity increases.  6.2 Significance to E n g i n e e r i n g  As concluded in Section 6.1, when operating the membrane filtration system below the critical flux conditions, increasing the operating permeate flux can increase the extent of fouling. Also, increasing the cross-flow velocity to the system can decrease the extent of fouling. Therefore, for the design and operation of membrane filtration system, high air sparging intensity and low operating permeate flux can be applied to the systems operated below critical flux to control the extent of fouling.  However, high air sparging intensity can increase the operating cost of a membrane system. In addition, operating at a lower operating flux, can increase the membrane surface area 76  needed to maintain a given permeate flux, and therefore can increase the capital cost of a membrane system. A n engineering economic analysis would have to be performed to establish the optimal air spaging intensities and operating permeate flux to minimize the cost associated with membrane systems.  77  REFERENCES  Aimar, P. and Howell, J.A. (1989) Effect of concentratrion boundary layer development on the flux limitations in ultrafiltration, Chemical Engineering Research and Design, 67, 255-261,  Belfort, G., Davis, R . H . , Zydney, A . L . (1994) Review: The behavior of suspensions and macromolecular solutions in crossflow microfiltration, Journal of Membrane Science, 96, pl-58  Bellara, S., Cui, Z. F. and Pepper, D. S. (1996) Gas sparging to enhance permeate flux in ultrafiltration using hollow fibre membranes, Journal of Membrane Science, 121, pl75 - 184  Berube, P. R. and Waitai, Y (2005) Impact of operational mode, hydrodynamic conditions and system configuration on the permeate flux in a submerged membrane system, Proceedings, A W W A Membrane Technology Conference, Phoenix, USA  Berube, P. R. and Lei, E. (2006)  The effect of hydrodynamic conditions and system  configurations on the permeate flux in a submerged hollow fiber membrane system,  Journal of  Membrane Science, v 271, n 1-2, p 29-37  Berube, P. R., Afonso, G., Taghipour, F. and Chan, C.C.V. (2006) Quantifying the shear at the surface of submerged hollow fiber membranes, Journal of Membrane Science, v 279, p 495-505  Berube, P. R. and Wong, J. (2007) Evolution of the trans-membrane pressure during successive filtration cycles in a submerged hollow fiber membrane system, Proceedings, AWWA Membrane Technology Conference, Tampa, USA  Bian, R., Yamamoto, K . , Watanabe, Y . (2000) The effect of shear rate on controlling the concentration polarization and membrane fouling, Desalination, 78  131, p225-236  Bolton, G. R., LaCasse, D. and Kuriyel, R. (2006a) Combined models of membrane fouling: Development and application to microfiltration and ultrafiltration of biological fluids, Journal of Membrane Science, v 277, p 75-84  Bolton, G. R., Boesch, A.W. and Lazzara, M . J. (2006b) The effects of flow rate on membrane capacity: Development and application of adsorptive membrane fouling models, Journal of Membrane Science, v 279, p625-634  Bowen, W. R., Calvo, J. I. and Henandez, A. (1995) Steps of membrane blocking in flux decline during protein microfiltration, Journal of Membrane Science, 101, p i 53-165  Chang, S. and Fane, A . G. (2000a) Characteristics of microfiltration of suspensions with inter-fiber two-phase flow, Journal of Chemical Technology and Biotechnology,  v 75, n 7, p  533-540  Chang, S. and Fane, A . G (2000b) Filtration if biomass with axial inter-fiber upward slug-flow: performance and mechanism, Journal of Membrane Science, 180, p 57 - 68  Cho, B. D. and Fane, A. G. (2002) Fouling transients in nominally sub-critical flux operation of a membrane bioreactor, Journal of Membrane Science, 209, p391-403  Crozes, G.F., Anselme, C. and Mallevialle, J. (1993) Effect of adsorption of organic matter on fouling if ultrafiltration membranes, Journal of Membrane Science, 84, p61  Crozes, G.F., Jacangelo, J.G., Anselme, C , Laine, J.M. (1997) Impact of ultrafiltration operating conditions on membrane irreversible fouling, Journal of Membrane Science, v 124, n 1, p 63-76  Cui, Z. E , Chang, S. and Fane, A . (2003) Review: The use of gas bubbling to enhance membrane 79  processes, Journal of Membrane Science, 221, p 1 - 35  de Amorim, M.T. P. and Ramos, I. R. A . (2006) Control of irreversible fouling by application of dynamic membranes, Desalination,  v 192, n 1-3, p 63-67  Fane, A. G , Chang, S., and Chardon (2002) Submerged hollow fiber membrane module - Design options and operational considerations, Desalination,  146, n 1-3, p 231-236  Field, R.W., Wu, D., Howell, J.A., Gupta, B.B. (1995) Critical flux concept for microfltration fouling, Journal of Membrane Science, v 100, p259-272  Ghosh, R. (2006) Enhancement of membrane permeability by gas-sparging in submerged hollow fibre ultrafiltration of macromolecular solutions: Role of module design, Journal of Membrane Science, v 274, n 1-2, p 73-82  Hermia, J. (1982) Constant pressure blocking filtration  laws-application to power-law  non-Newtonian fluids, Trans. IchemE. 60, 183  Hilal, N . , Ogunbiyi, O. O., Miles, N . J., Nigmatullin, R. (2005) Methods employed for control of fouling in M F and U F membranes: A comprehensive review, Separation Science and Technology, v 40, n 10, p 1957-2005  Ho, C.C. and Zydney, A . L . (2000) A combined pore blockage and cake filtration model for protein fouling during microfiltration, Journal of Colloid and Interface Science, 232, p299-389  Hong, S.P., Bae, T.H., Tak, T.M., Hong, S., and Randall, A . (2002) Fouling control in activated sludge submerged hollow fiber membrane bioreactors, Desalination,  80  143: 219-228  Katsoufidou, K . , Yiantsios, S.G. and Karabelas, A.J. (2005) A study of ultrafiltration membrane fouling by humic acids and flux recovery by backwashing: Experiments and modeling, Journal of Membrane Science, v 266, p40-50  Kim, J. and DiGiano, F. A . (2006) A two-fiber, bench-scale test of ultrafiltration (UF) for investigation of fouling rate and characteristics, Journal of Membrane Science, v 271, n 1-2, p 196-204  Kimura, K., Hane, Y., Watanabe, Y., Amy, G , Ohkuma, N . (2004) Irreversible membrane fouling during ultrafiltration of surface water, Water Research, v 38, n 14-15, p 3431-3441  Kimura, K., Hane, Y , Watanabe,Y. (2005) Effect of pre-coagulation on mitigating irreversible fouling during ultrafiltration of a surface water, Water Science & Technology, v 51, n 6-7, p93-100  Kimura, K . , Yamamura, H . , Watanabe, Y. (2006) Irreversible fouling in M F / U F membranes caused by Natural Organic Matters (NOMs) isolated from different origins, Separation Science and Technology, v 41, n 7, p 1331-1344  Jacob, J., Pradanos, P., Calvo, J.I., Hernandez, A., Jonsson, G. (1998) Fouling kinetics and associated dynamics of structural modifications, Colloids and Surfaces A: Physicochemical and Engineering Aspects, v 138, n 2-3, p 173-183  Le Clech, P., Jefferson, B., Chang, I.S., Judd, S.J. (2003) Critical flux determination by the flux-step method in a submerged membrane bioreactor, Journal of Membrane Science, 227, p81-93  Lei, X . E. (2005) Impact of hydrodynamic conditions and membrane configuration on the 81  permeate flux in submerged membrane systems for drinking water treatment, Master thesis  Liang, Sh., Song, L . , Tao, G , Kekre, K . A . , Seah, H . (2006) A modeling study of fouling development in membrane bioreactors for wastewater treatment, Water Environment Research, v 78, n 8, p 857-863  Mallevialle, J., Odendaal, P. E. and Wiesner, M . R. (1996) Water Treatment  Membrane  Processes, McGraw-Hill  Minegish, S., Jang, N.Y. and Watanabe, Y., Hirata, S and Ozawa, G. (2001) Fouling mechanism of hollow fiber ultrafiltration membrane with pretreatment by coagulation /sedimentation process, Water Science and Technology: Water Supply Vol No 4 pp 48-56  Nagaoka, H . , Ueda, S., Miya, A . (1996) Influence of bacterial extracellular polymers on the membrane separation activated sludge process, Water Science Technology, 34, pl65-172  Nagaoka, H . , Yamanish, S. and Miya, A . (1998) Modeling of biofouling by extracellular polymers in a membrane separation activated sludge system, Water Science and Technology, 38(4-5):497-504  Ognier, S., Wisniewski, C., Grasmick, A . (2002a) Characterization and modeling of fouling in membrane bioreactors, Desalination,  146, pl41-147  Ognier, S., Wisniewski, C , Grasmick, A . (2002b) Membrane fouling during constant flux filtration in membrane bioreactors, Membrane Technology, 7, p6-10  Ognier, S., Wisniewski, C , Grasmick, A . (2004) Membrane bioreactor fouling in sub-critical filtration conditions: a local critical flux concept, Journal of Membrane Science, 229, p!71-177 82  Smith, S., Taha, T. and Cui, Z. F. (2002) Enhancing hollow fiber ultrafiltration using slug-flow -a hydrodynamic study, Desalination,  146, p 69-74  Smith, S., Cui, Z. F. and Field, R.W. (2005) Upper- and lower-bound estimation of flux for gas-sparged ultrafiltration with hollow fiber membranes, Journal of Membrane Science, 44, p 7684-7695  Tchobanoglous, G , Burton, F. L., and Stensel, H . D. (2003) Wastewater Engineering, Treatment and Reuse, McGraw Hill, New York  Tracey, E . M . and Davis, R . H . (1994) Protein fouling  of track-etched  polycarbonate  microfiltration membranes, Journal of Colloid and Interface Science, 161, p l 0 4 - l 16  USEPA (2005) Membrane filtration guidance manual, Office of Water, Cincinnati, USA  Wang, Y , Wang, T., Su, Y , Peng, F. Wu, H . , Jiang, Zh. (2005) Remarkable reduction of irreversible fouling and improvement of the permeation properties of poly(ether sulfone) ultrafiltration membranes by blending with pluronic F127, Langmuir, v 21, n 25, p 11856-11862  Wisniewski, C , Grasmick, A . , Cruz, L . (2000) Critical particle size in membrane bioreactors. Case of denitrifying bacterial suspension, Journal of Membrane Science, 178, 141-150  Ye, Y., Chen, V . and Fane, A. G. (2006) Modeling long-term subcritical filtration of model EPS solutions, Desalination,  191, p318-327  Yoon, S.H., Kimb, H.S. and Yeomb, I.T. (2004) Optimization model of submerged hollow fiber membrane modules, Journal of Membrane Science,, 234, pl47-156  83  Yuan, W., Kocic, A . and Zydney, A . L . (2002) Analysis of humic acid fouling during microfiltration using a pore blockage-cake filtration model, Journal of Membrane Science, 198 p51-62  84  APPENDIX A: CODES FOR MATLAB CALCULATION function f = computeKs(yO,a,b,c,Jv)  A0=2.3e-3; R=2e-8; %Jv=3.83e-8; %Jv=3.19e-8 %Jv= 2.56e-8;  kl=AO*R 4/(Jv*yO); A  %if  (((a*kl*Jv) 2-3*A0*b*R 4*kl*Jv)>=0) A  A  k3_l=(-a*kl*Jv+sqrt((a*kl*Jv) 2-3*A0*b*R 4*kl*Jv))/(3*A0*R 2); A  A  A  k3_2=(-a*k 1 *Jv-sqrt((a*k 1 *Jv) 2-3*A0*b*R 4*k 1 *Jv))/(3*A0*R 2); A  A  A  k2_l =-(a*Jv*k 1 +2*A0*k3_ 1 * R 2 ) / R 4 ; A  A  k2_2=-(a*J v*k 1 +2*A0*k3_2*R 2)/R 4; A  A  k3_ll=c*Jv*kl/k2_l 2; A  k3_12=c*Jv*kl/k2_2 2; A  fprintf('\n Group 1: \n'); ! printf('\nkl= %d\n', k l ) ; :  fprintf('\nk2_l= %d\n', k2_l); fprintf('\nk3_l= %d\n', k3_l);  fprintf('\n Group 2: \n'); fprintf('\nkl=%d\n',kl); fprintf('\nk2_2= %d\n', k2_2); fprintf('\nk3_2= %d\n', k3_2); %else %  fprintf('\n b2-4ac<0, no resolvation! Jv= %d\n', Jv);  %end clear all;  85  

Cite

Citation Scheme:

        

Citations by CSL (citeproc-js)

Usage Statistics

Share

Embed

Customize your widget with the following options, then copy and paste the code below into the HTML of your page to embed this item in your website.
                        
                            <div id="ubcOpenCollectionsWidgetDisplay">
                            <script id="ubcOpenCollectionsWidget"
                            src="{[{embed.src}]}"
                            data-item="{[{embed.item}]}"
                            data-collection="{[{embed.collection}]}"
                            data-metadata="{[{embed.showMetadata}]}"
                            data-width="{[{embed.width}]}"
                            async >
                            </script>
                            </div>
                        
                    
IIIF logo Our image viewer uses the IIIF 2.0 standard. To load this item in other compatible viewers, use this url:
http://iiif.library.ubc.ca/presentation/dsp.831.1-0063234/manifest

Comment

Related Items