TOWARDS OPTIMAL SAVING IN MEMBRANE OPERATION
The development of
process inspection and
feedwater characterization tools.
ii
This work was financially supported by the Dutch Science Foundation, STW, project
6115, Optimal Saving in Membrane Operation
Wilhelmus Johannes Cornelis van de Ven
Towards Optimal Saving in Membrane Operation
The development of process inspection and feedwater characterization tools
ISBN 978-90-9022992-8
© W.J.C. van de Ven, Enschede, 2008.
All rights reserved.
Cover design by Wilbert van de Ven.
The background picture shows crystallized sodium alginate from a calcium solution. The picture was taken by Koen van ‘t Sant and the colors were digitally adjusted.
iii
TOWARDS OPTIMAL SAVING IN MEMBRANE OPERATION
PROEFSCHRIFT
ter verkrijging van
de graad van doctor aan de Universiteit Twente,
op gezag van de rector magnificus,
prof. dr. W.H.M. Zijm
volgens besluit van het College voor Promoties
in het openbaar te verdedigen
op donderdag 24 april om 15.00 uur
door
Wilhelmus Johannes Cornelis van de Ven
geboren op 14 mei 1977
iv
Dit proefschrift is goedgekeurd door de promotor
Prof. Dr.-Ing. M. Wessling
en assistent promotor
v
Voor mijn ouders
vi
Chapter 1
Introduction
1
1 Drinking water 2
2 Membrane technology in water treatment 3
2.1 Filtration processes 3
2.2 Market size of Ultrafiltration and Microfiltration 4
2.3 Modes of operation 5
3 Filtration theory and definitions 7
3.1 Darcy’s Law 7
3.2 Filtration cycles and definition of (ir)reversible fouling 8
3.3 Initial, final and backwash resistances 9
4 The OSMO Project 10
4.1 Project description 10
4.2 Aim of the OSMO- Inspector project 12
5 Scope of this thesis 13
Literature 15
Chapter 2
A critical flux concept for dead-end ultrafiltration
17
1 Critical flux 18
1.1 Definitions of critical flux 19
1.2 Methods to determine the critical flux in crossflow filtration 20 1.3 Reversible fouling in dead-end filtration 21
1.4 Concluding remarks 23
2 Determination of critical parameter in dead-end filtration 24
2.1 Standard step-method 24
2.2 The transition point 25
3 Experimental 27
3.1 Materials 27
3.2 Membrane 27
3.3 Experimental setup 28
3.4 Layout of a single filtration-backwash cycle 29
4 Results and discussion 31
4.1 Initial studies 31
4.2 Short-term experiments 32
5 Discussion 46
Appendix 1 - Alternating Step method 49
Protocol 49
Results 49
Standard vs. Alternating step method 51
Appendix 2 - Results of Twente Canal water analysis 52
vii
Chapter 3
Hollow fiber dead-end ultrafiltration:
55
Influence of ionic environment on filtration of alginates
1 Introduction 56
1.1 Background 56
1.2 Membrane filtration of alginates 57
1.3 Fouling phenomena in dead-end filtration 57
2 Materials and Methods 59
2.1 Materials 59
2.2 Membrane and modules 59
2.3 Filtration set-up 60
2.4 Experimental procedure 61
3 Results & Discussion 64
3.1 Sodium alginate 64
3.2 Sodium alginate with calcium chloride 66
3.3 Sodium alginate with potassium chloride 70
3.4 Sodium alginate with potassium chloride and calcium chloride 72 3.5 Comparison of the various feed solutions 73
3.6 Start-up effects 78
3.7 Axial variations of transport 78
4 Conclusions 79
Literature 80
Chapter 4
Hollow fiber dead-end ultrafiltration:
83
Axial variations during humic acid filtration
1 Introduction 84
2 Material and Methods 85
2.1 Chemicals 85
2.2 Membranes and modules 86
2.3 Setup 86
2.4 Fouling reversibility in cyclic operation 87
2.5 Retention in dead-end filtration 88
2.6 UV absorption and TOC measurements 89
2.7 Local variance in crossflow velocity and flux 90
3 Results 91
3.1 Reversible operation 91
3.2 Retention during humic acid filtration 94
3.3 Axial features in filtration 96
4 Discussion 99
4.1 Fouling reversibility 99
4.2 Humic acid retention 99
4.3 Possible applications 103
5 Conclusions 104
Appendix 1 - Flat sheet inspector 105
viii
The concept of partial backwashing
1 Introduction 112
2 Experimental 115
2.1 Materials 115
2.2 Membranes 115
2.3 Experimental Setup 116
2.4 Layout of a filtration-backwash cycle 116
3 Results and Discussion 118
3.1 Sodium alginate and Humic acid 118
3.2 The influence of calcium 119
3.3 Twente Canal water 122
3.4 Parameters affecting successful implementation of partial backwashing 124
4 Conclusion 125
Appendix 1 - Filtration of humic acid with added calcium 126
Literature 127
Chapter 6
Unraveling Ultrafiltration of Polysaccharides with
129
flow Field-Flow Fractionation
1 Introduction 130
2 Theory 131
2.1 Flow-FFF 131
2.2 MALS 134
2.3 Filtration analysis with flow-FFF experiments 136
3 Experimental 138
3.1 Material 138
3.2 Equipment 138
4 Results and Discussion 141
4.1 Flows and fluxes 141
4.2 Sodium alginate in water 142
4.3 Sodium alginate in a salt solution 146
4.4 Conformation of sodium alginate in pure water and in a salt solution 150
5 Conclusions 152
ix
Chapter 7
Ultrafiltration - Miscellaneous topics
155
1 Online streaming potential measurements 156
1.1 Introduction 156
1.2 Theory 156
1.3 Methods 157
1.4 Results 160
1.5 Concluding remarks 166
2 The effect of temperature on the filtration of alginates 167
3 The first filtration cycles 171
4 Virus filtrations 173
4.1 Introduction 173
4.2 Material and Methods 174
4.3 Results 175
5 Star-shaped Ultrafiltration hollow fibers 182
6 Conclusion 185
Literature 186
Chapter 8
Summary & Outlook
187
1 Summary 188
1.1 Introduction 188
1.2 Critical process parameters 188
1.3 Feed water characterization: Sodium alginate filtration 190 1.4 Axial variations of fouling and partial backwashing 192
2 Outlook 194
2.1 Fouling Inspector 194
2.2 Flow-FFF as a fingerprinting tool 194
2.3 Fouling modeling 195
2.4 Improved observation of membrane fouling 196
3 Nederlandse samenvatting 197
Literature 201
Nomenclature
202
Acknowledgements/Dankwoord 204
Chapter 1
In this chapter the topic of water filtration with the use of membrane technol-ogy is introduced. The market size and need for membrane technoltechnol-ogy in water treatment are briefly discussed. The chapter continues with an over-view of the main concepts used throughout the thesis and a description of the optimal saving in membrane operation (OSMO) project.
1
Drinking water
In the Netherlands, drinking water traditionally is prepared from ground water (roughly two thirds) or surface water (roughly one third) [1]. For ground water, it is typically sufficient to use aeration followed by a disinfection method to produce water according to drinking water standards. However, in large parts of the Netherlands and throughout the rest of the world, the use of ground water from aquifers is not possible. Excessive usage of ground water and global climate change allow penetration of sea water into aquifers [2, 3]. Population growth, not surprisingly, leads to more pollution of aquifers [4], making the water quality unsuitable for drinking water purposes without extensive polishing.
Preparing drinking water from low quality aquifers or from surface water may require additional cleaning steps, like aeration, extensive sand filtration, adsorption processes, etc. In the Netherlands, surface water is classified into one of three distinct quality categories, determining the basic cleaning strategy to be undertaken. Conventional cleaning technologies can consist of the following steps [1]:
• Coarse filtration to remove suspended solids • Coagulation/flocculation
• Aeration and degassing • Fast and slow sand filtration • Adsorption with activated carbon • Disinfection
• Correction of the pH • Removal of hardness
Additionally to these technologies, membrane technology is rapidly introduced in water treatment to prepare drinking water [5]. Membrane technology can cope with a wide variety of feed water sources and can produce any desired water quality. Reverse osmosis has been used throughout the last four decades to desalinate seawater for the production of drinking water. For the treatment of freshwater for the production of drinking water, mainly microfiltration/ultrafiltration combined with nanofiltration/reverse osmosis are used.
Chapter 1
2
Membrane technology in water treatment
2.1 Filtration processes
Pressure driven membrane technologies for water purification are classified by their sieving mechanism [6]. Porous membranes are on one side of the spectrum. Sieving is achieved mainly by size separation and, in the limit of small pore size, charge. The separation mechanism for dense membranes is based on a solution diffusion mechanism. Porous membrane processes are microfiltration (MF) and ultrafiltration (UF), where the distinction between the two processes is based on the size of the pores. MF features pore sizes of 100nm up to a few micrometers. Microfiltration membranes can be symmetric or asymmetric in structure. In symmetric membranes, porosity and poresize are constant over the thickness of the membrane, while in asymmetric membranes the membrane becomes denser towards the separation layer. Symmetric membranes are stronger, while asymmetric membranes feature higher fluid permeability. Ultrafiltration (UF) membranes feature smaller pores, from 1-100 nm. They are always asymmetric. Size sieving is the main separation mechanism for ultrafiltation membranes, but charge may play a role as well, especially at low pore size. Nanofiltration (NF) and reverse osmosis (RO) membranes are considered dense. They are usually prepared by coating a thin top layer on an ultrafiltration membrane. Contrary to porous membranes, NF and RO retain dissolved salts in water. Nanofiltration will typically have a high retention for bivalent ions and a moderate retention for monovalent ions. RO retains both mono- and bivalent ions. Figure 1 shows the applicability of porous and non porous membrane processes ranked by the
Figure 1 Various membrane processes sorted by their size sieving ability. Membranes are applied from molecular level to particles with a size of several micrometers.
constituents of the feed that can be retained.
2.2 Market size of Ultrafiltration and Microfiltration
The use of membrane filtration in the preparation of drinking water and treatment of municipal waste water was accelerated by the introduction of low pressure ultra- and microfiltration systems. For water treatment purposes, especially ultrafiltration (UF) membranes are interesting, as they form a solid barrier against waterborne pathogens. The acceptation and implementation of membrane technology was accelerated by the outbreaks of cryptosporidium in Milwaukee [7]. Ultrafiltration features a greater than log 4 removal of viruses and bacteria.
Since the mid-nineties application of membrane technology has taken a flight. Numbers from the United Kingdom have shown that the cumulative capacity of membrane plants for water treatment increased from almost nothing in 1998 to 106 m3day-1 in 2004, making up 6% of the
total supplied water in the United Kingdom [8]. The same paper shows the predicted trends from Osmonics (predictions from 2002) for the global market value of UF and MF for water treatment. They are expected to increase exponentially till 2010, reaching a global market value of about $1.5-1.8 billion for combined use of membranes in water and waste water treatment. Other predictions show a worldwide growth of the combined market for ultrafiltration and microfiltration modules (not limited to water treatment applications) from $3.8 to 5.7 billion between 2003 and 2008 [9]. In Ontario, Canada, cost effective implementation of membrane plants started in 1998 [10]. From 1998 to 2006, the total number of plants increased from 4 to 37, corresponding to a growth in capacity of 50 to over 400·103 m3day-1.
An upcoming application of UF is in the pretreatment of sea water for desalination purposes [11]. Sea water is often of too low quality to directly use in RO plants. The improvement of UF membranes and technology has made using their use as a pretreatment economically viable and large scale plants are designed. UF provides a constant, very high quality feed water to the RO plant in comparison to traditional in-line coagulation and two-stage sand filtration [12, 13]. The seawater desalination market is huge. Plants have been commissioned since the 1970’s.
Chapter 1 2.3 Modes of operation
2.3.1 Crossflow vs. Dead-End Filtration
Most of the ultrafiltration membrane plants that are used for the preparation of drinking water use hollow fiber membranes. In contrast to flat sheet membranes, hollow fibers feature a higher packing density and module design is simpler (there is no need for feed and permeate spacers in the modules). Thousands of these fibers are packed in large filtration modules (Figure 2). Depending on the properties of the feed water, two different operating regimes can be used. For feedwater with a high suspended solids content, crossflow filtration is applied, for feedwater with a low suspended solids content, dead-end filtration is used.
In classical inside-out crossflow filtration, the feed is fed to the inside of the fiber. Part of the feed solution crosses the membrane and forms the permeate (product). The concentrated waste stream (called retentate or concentrate) is either recycled or discharged. In the early days of membrane filtration with ultra- and microfiltration membranes, crossflow filtration was the only applied mode of operation. In crossflow filtration, the feed solution is circulated at a high speed over the membrane surface, while (part of the) retentate is purged. The flow over the membrane is called the crossflow, and the aim of this is to prevent material from the feed solution to deposit on the membrane surface. However, the typically high crossflow velocity (~2-4 m·s-1) is a large
Figure 2 8” Xiga ultrafiltration module from Norit X-Flow. The permeate is collected from the central tube. The feedwater is fed to the inside of the fibers (picture from:
energy sink.
For feed water having a low solids content, dead-end filtration is much more economical. Dead-end filtration with hollow fibers closely remembers regular filtration. All feedwater passes the membrane as permeate. As material is retained by the membrane, it accumulates inside the fiber, leading to an increase in pressure (when the permeate flow rate is kept constant). Due to the absence of the crossflow, and the generally low pressures that are used, dead-end filtration requires very low amounts of energy. The process is also simpler and smaller pumps can be used. To remove the accumulated matter from the membrane module, periodic backwashing and chemical cleaning are performed.
2.3.2 Filtration and backwashing
The accumulated material in dead-end filtration is removed by a periodic hydraulic cleaning, which is often referred to as ‘backwashing’ or ‘backflushing’. During filtration the material in the feed solution that can not pass the membrane accumulates on the membrane surface. When constant permeate flux filtration is applied, the feed pressure will increase over time (or the flux declines when the pressure is kept constant). A periodic backwash removes (most of) this accumulated material (Figure 3). Periodic chemical cleaning is necessary to remove the hydraulically irreversible fouling from the membranes.
Chapter 1
3
Filtration theory and definitions
3.1 Darcy’s Law
The performance of porous membranes in aqueous separations can conveniently be described by the well-known Darcy’s Law. Darcy’s Law describes the flux (J, L·h-1m-2) as a function of
permeability (Lp, L·h-1m-2bar-1) and applied transmembrane pressure (ΔP, bar) taking the
osmotic pressure difference between feed and permeate (Δπ, bar) into account.
(
π)
= p Δ − Δ
J L P (1)
Membrane manufacturers generally report fluxes in L·h-1m-2 and pressures in bar. The main
disadvantages of using permeabilities to assess fouling are the small change in permeability at high fouling and the fact that there is no correction for temperature changes. More convenient is the use of the filtration resistance. The resistance encountered during filtration is equal to
η 1
p
L with η the fluid’s viscosity [Pa·s].
This results in the following expression for the total membrane resistance:
(
π)
η Δ − Δ = t P R J (2)With ΔP, the transmembrane pressure, in Pa, and the flux, J, in m3m-2s-1, the resistance is
obtained in m-1. The viscosity corrects for the effect of temperature on the hydraulic filtration
resistance. A simple empirical relation can be used to calculate the viscosity of water [Pa·s] as a function of temperature T [°C] [14].
(
)
η −
= + 1.5
0.497 T 42.5 (3)
As a result of membrane fouling, the resistance of the membrane towards mass transport will increase, resulting in an increase in the required transmembrane pressure to maintain the set flux. The resistance-in-series approach can be used to determine the contributions of various types of membrane fouling. The total resistance in Equation (2) is regarded as the sum of the membrane resistance, and the extra resistances as a result of concentration polarization, cake layer, adsorption, and pore blocking phenomena [6]:
= + + + +
t m cp c a pb
The fouling rate is defined as the change in resistance as a function of the filtered volume,
f
dR dV .
The filtered volume is normalized for the membrane area, m3m-2.
3.2 Filtration cycles and definition of (ir)reversible fouling
Dead-end ultrafiltration operation consists of subsequent filtration and backwash steps. The definitions of the various cycles that are used throughout this thesis, as well as the definitions of reversible and irreversible fouling used in this work are graphically depicted in Figure 4.
A filtration-backwash cycle consists of a filtration with a given set of parameters followed by a backwash. During the filtration phase, the resistance towards water transport through the membrane will increase as a result of membrane fouling. After a predefined time, filtered amount of water, or when a certain pressure is reached, the flow is reversed (backwash) and the fouling is flushed from the membrane. Fouling that is removed using backwashing alone is, in this work, defined as reversible fouling. Fouling strongly adsorbed to the membrane surface or embedded within the porous structure of the membrane may not be removed by the applied backwash and is defined as irreversible fouling. When the amount of irreversible fouling becomes too high, or after a predetermined amount of filtration-backwash cycles, the membrane is chemically cleaned. All filtration-backwash cycles and the following chemical cleaning form one chemical cleaning
Chapter 1 cycle. Ideally, chemical cleaning removes all irreversible fouling from the membrane. Although
not discussed in this research work, an additional cycle exists, the membrane lifetime cycle. This cycle contains all chemical cleaning-cycles from the startup until replacement of the membranes.
The irreversible fouling rate can be calculated from the change in the values for the resistance at the end of a backwash, the initial resistance, or the resistance at the end of filtration (final resistance), with progressing cycle number or total cumulatively produced permeate.
3.3 Initial, final and backwash resistances
The evolution of initial, final, or backwash resistances is extensively used in this work. The initial resistance is the resistance found during filtration before the onset of fouling. The initial resistance is calculated from extrapolation of the filtration curve to zero filtered volume. The fraction of the data that is used for the extrapolation depends on the used feed solution. The final resistance is the value for the filtration resistance at the end of the filtration. The backwash resistance is calculated from the backwash flux and the pressure measured by the pressure sensor in the backwash line (Chapter 2, this thesis).
4
The OSMO Project
4.1 Project descriptionThe OSMO project aims at optimizing dead-end ultrafiltration of surface water for drinking water preparation. OSMO is an acronym for Optimal Saving in Membrane Operation. The complete cycle of filtration and backwash, as well as the effects of chemical cleaning and membrane aging are studied, and an economically most profitable strategy is the end result. The project is divided into three parts: Two Process Optimizers and one Process Inspector. The Optimizers work on the short and long term optimization. The goal of short term optimization is to derive an optimal scheduling of filtration-backwash cycles, including variables as the filtration flux, recovery, backwash operation, and coagulant dosing. The results of this work are described by Blankert [15]. Zondervan studied long term optimization [16]. The subjects of these studies are the effects of membrane cleaning and aging on the economics of the entire process. The models that are used for the short term and long term optimization are black box models, using a modeling approach taken from chemical process control theory [17] (see Figure 5).
In membrane filtration, manipulated variables are for example the pump speeds, the coagulant dosing, and the backwash settings. External disturbances can sometimes be easily measured, like the temperature or the pH of the feed water, but can also be difficult to assess. Similarly, some outputs of the filtration process can be measured directly (the filtration flux for example). Others can not be measured in such a way to be part of process control, and estimates based on mathematical models are required (a good example for this is the use of the pressure decay test to
Figure 5 Schematic overview of the modeling strategy used in chemical process control. Manipulated variables can be adjusted by a control mechanism or an operator. External disturbances affect the system, but can not be adjusted by a control mechanism. Output variables are either measured or not (or cannot be) measured.
Chapter 1 estimate virus removal of membranes [18] – the virus removal is an unmeasured output that can
be estimated by the measured pressure decay, a measured output). The exact feed composition is generally unknown when surface water is concerned. The inspector part of the OSMO-project, described in this thesis, aims to develop methods to assess the filtration characteristics of the feed water a priori, assisting the model structure developed by the two optimizers. Schematically, the OSMO project is depicted in Figure 6.
The Inspector project contains two major sub-projects. The first topic is the development of a method to determine the filterability of any given feedwater in combination with a given membrane. The outcome of this method should assist the short term optimizer to decide the optimal process conditions for both the filtration and the backwash. The method aims at replacing conventional fouling indices, like the silt density index (SDI) [19] and the various modified fouling indices (MFI) [20-22]. These tests were mainly developed to assess the pretreatment for RO systems, and use constant pressure filtration in flat sheet cells. The analysis assumes cake filtration.
The second part of the inspector project that is subject of this thesis is the development of a feed
Figure 6 Schematic overview of the OSMO projects. The two optimizers develop models to find the economically most feasible process operation strategy. The process inspector develops method to quantify external disturbances.
water characterization tool. This tool aims at making a fingerprint of feedwater samples and at correlating the fingerprint to fouling behavior. This fingerprinting tool is mainly of use for the short term optimizer.
Tools for the long-term optimizer should give information on the effectiveness of chemical cleaning operations and the effect of those on the aging of the membranes. A toolbox for the latter part has been described by Zondervan et al. [23], which consists of rapid aging tests of fouled and clean membranes in the presence of either clean water or a chlorine solution. Characterization tools for the cleanliness of the membranes are briefly discussed in this work in chapter 7 (streaming potential method).
4.2 Aim of the OSMO- Inspector project
The aim of the project is best visualized by Figure 7. The OSMO inspector will run parallel to a full-scale filtration plant and determine optimal process settings for this full plant. The inspector will determine filterability of the raw feed water on a regular basis, depending on the frequency with which the water quality changes.
Figure 7 The aim of the OSMO inspector project. A full-scale filtration plant equipped with an OSMO Inspector box, combined with the work from the long and short term optimizers, should yield improved economics for the system and make the filtration process more robust with regard to changes in the feed water quality.
Chapter 1
5
Scope of this thesis
This thesis describes the work performed on the development of a filtration inspector for the production of drinking water from natural water sources.
Chapter 2 describes a critical flux concept that can be applied in dead-end filtration. The literature on critical flux in cross-flow filtration is summarized and from this starting point, a concept for dead-end filtration was developed. The chapter describes the basic method and two analysis methods to determine the transition point between critical and non-critical operation. The chapter also shows that the same method can be used to study other critical parameters, like the recovery or backwash flux. The method was applied for sodium alginate solutions and natural water.
Sodium alginate solutions showed completely different filtration behavior depending on the ions present in the feed solution. Since the filtration characteristics can be adjusted by adding only small amounts of salt, these solutions are ideal for validation of the OSMO inspector concept. In
chapter 3, a detailed study is presented on the filtration of sodium alginate solutions. The method from chapter 2 is used to determine the critical flux for irreversibility, and the filtration curves are explained using a description of the structure of the sodium alginate in solution.
Humic acid is an important component of natural organic matter present in surface water. In
chapter 4 the filtration behavior of a commercial humic acid is tested. The observed filtration behavior led to the conclusion that the flux and retention are non-uniformly distributed over the length of the fibers, opening possible changes to the filtration system to increase the effectiveness dramatically.
One of these options is discussed in chapter 5. The observation that humic material accumulates primarily in the end of the modules is exploited by the partial backwash concept. As a proof-of-principle, a process is described where only 20% of the active filtration surface is backwashed. The results for various solutions are presented in this chapter.
In chapter 3 we presented hypothetical polymer conformations and a qualitative model for membrane-polymer interactions that can explain the observed filtration results. In chapter 6 flow-field flow-fractionation (Flow-FFF) coupled to multi-angle light scattering (MALS) is introduced as an analysis method that can determine the particle size and molecular weight distribution of complex samples. The method can also be used to study membrane-polymer
interactions, and to perform conformational studies of sodium alginates. Both topics will be discussed in chapter 6.
A number of small studies that can form the basis of future research activities are presented in
chapter 7. In the chapter the effect of temperature on the filtration performance, determination of removal efficiency for viruses, streaming potential measurements, and initial fouling behavior are discussed.
Chapter 8 gives the overall evaluation of the thesis and finishes with an outlook for future research.
Chapter 1
Literature
1. van Deventer, W.T., Milieutechnologie; Van schoonmaaktechnologie naar schone
technologie (Dutch). 1997, Alphen a/d Rijn: Samsom.
2. Sherif, M.M. and V.P. Singh, Effect of climate change on sea water intrusion in coastal
aquifers. Hydrological Processes, 1999. 13(8): p. 1277.
3. Grassi, S. and R. Netti, Sea water intrusion and mercury pollution of some coastal aquifers
in the province of Grosseto (Southern Tuscany - Italy). Journal of Hydrology, 2000. 237(3-4): p. 198.
4. Schot, P.P. and J. van der Wal, Human impact on regional groundwater composition
through intervention in natural flow patterns and changes in land use. Journal of Hydrology, 1992. 134(1-4): p. 297.
5. Escobar, I.C., et al., Committee Report: Recent advances and research needs in membrane
fouling. Journal / American Water Works Association, 2005. 97(8): p. 79.
6. Mulder, M., Basic principles of membrane technology. 1997, Dordrecht, the Netherlands: Kluwer Academic Publishers.
7. Mac Kenzie, W.R., et al., A massive outbreak in Milwaukee of cryptosporidium infection
transmitted through the public water supply. New England Journal of Medicine, 1994. 331(3): p. 161.
8. Rachwal, T. and S. Judd, A synopsis of membrane technologies in UK municipal potable
water treatment: History, status and prospects. Water and Environment Journal, 2006. 20(3): p. 110.
9. Sutherland, K., Profile of the International Membrane Industry - Market Prospects to 2008. 2004: Oxford.
10. Sahely, B., Hundreds of millions of membrane fibres in Ontario, but who’s counting? Environmental Science & Engineering, 2005. 9.
11. www.worldwater.org, The world's water - Information on the World's freshwater resources.
2008.
12. Pearce, G.K., The case for UF/MF pretreatment to RO in seawater applications. Desalination, 2007. 203(1-3): p. 286.
13. Wolf, P.H., et al., UF membranes for RO desalination pretreatment. Desalination, 2005. 182(1-3): p. 293.
14. Roorda, J.H. and J.H.J.M. Van Der Graaf. New parameter for monitoring fouling during
ultrafiltration of WWTP effluent. 2001: IWA Publishing.
15. Blankert, B., Short to medium term optimization of dead-end ultrafiltration 2007, Rijksuniversiteit Groningen: Ph.D. Thesis.
16. Zondervan, E., Intermediate to long term optimization of dead-end ultrafiltration. 2007, Rijksuniversiteit Groningen: Groningen.
17. Stephanopoulos, G., Chemical Process Control - An Introduction to Theory and Practice. PTR Prentice Hall International Series in the Physical and Chemical Engineering Sciences, ed. N. Admundson. 1984, Englewood Cliffs, New Jersey: Prentice-Hall, Inc. 18. Cote, P., et al., Monitoring and maintaining the integrity of immersed ultrafiltration
membranes used for pathogen protection. Water Science and Technology: Water Supply, 2002. 2(5-6): p. 307.
19. Walton, N.R.G., Some observations on the considerable variability of silt density index
results due to equipment, filter and operator variables. Desalination, 1987. 61(3): p. 201-210.
20. Boerlage, S.F.E., et al., The MFI-UF as a water quality test and monitor. Journal of Membrane Science, 2003. 211(2): p. 271-289.
21. Khirani, S., et al., Improving the measurement of the Modified Fouling Index using
nanofiltration membranes (NF-MFI). Desalination, 2006. 191(1-3): p. 1-7.
22. Schippers, J.C. and J. Verdouw, The modified fouling index, a method of determining the
fouling characteristics of water. Desalination, 1980. 32: p. 137.
23. Zondervan, E., et al., Statistical analysis of data from accelerated ageing tests of PES UF
Chapter 2
In this chapter the concept of ‘critical process parameters for irreversibility in dead-end filtration’ is introduced. The concept of critical flux in cross flow filtration is used as the basis to develop a method to determine critical process conditions in dead-end ultrafiltration.
The method is comprised of a step-wise increase of one process setting until membrane fouling is no longer fully reversible by backwashing and two meth-ods to determine the transition point from the data.
The method was tested for various process parameters with model feedwater solutions. We could distinguish between two forms of the critical process con-ditions in dead-end ultrafiltration, a weak and a strong form. The strong form is valid if no irreversible fouling of the membranes takes place. If the weak form is valid, some irreversible fouling may take place, but the irreversible fouling rate is not influenced by a change in the parameter under investigation .
Critical process conditions in
Dead-End Ultrafiltration
1
Critical flux
The critical flux concept has been used extensively in crossflow filtration. Without denoting it as the critical flux, Cohen and Probstein [1] measured a threshold flux when using a reverse osmosis process to filter solutions of ferric hydroxide. Depending on the stability of the solutions, a flux was reported below which the fouling layer did not grow. The threshold flux was attributed to double layer interactions between the colloids in the bulk and the initial fouling layer formed on the surface. Over the years, the concept gained additional interest till, in 1995, three key papers were published that presented the first definitions of critical flux. Field et al., proposed the definition for critical flux in the form of a hypothesis:
“The critical flux hypothesis for MF is that on start-up there exists a flux below which a decline of flux
with time does not occur; above it fouling is observed This flux is the critical flux and its value depends on the hydrodynamics and probably other variables.” [2]. Howell described a number of experimental methods to determine critical process parameters and the possible consequence for plant design [3]. Bacchin et al. managed to explain in detail [4] the ‘colloid flux paradox’ found by Cohen and Probstein [1].
The critical flux depends on the applied process conditions and feed water properties. In general, a higher crossflow velocity results in a higher value for the critical flux [5-7]. For small particles however (<0.1μm), the influence of the shear induced diffusion is not so significant [8]. The pore size of the membrane does not influence the value for the critical flux much [9]. The magnitude of the resistance increase when operated above the critical flux does however strongly depend on the pore size. This can be attributed to internal fouling of the membrane when the pores are larger than the foulants, and suggests that operating with membranes having a smaller pore size is preferred [10].
Particle size has a pronounced influence on the value for the critical flux. In the previously mentioned work by Lee and Clark [8], sensitivity analysis on their numerical model showed that particle size predominantly determines the steady-state permeate flux in cross flow filtration. Their model furthermore showed that there exists a critical particle size at which filtration performance is worst. Both above as well as below that size, the steady state permeate flux increased. This can be attributed to a transition between Brownian controlled and shear-induced diffusion. The critical particle size reported in literature varies between 0.1μm [9, 11] and 0.5μm [12]. The critical flux decreases with increasing concentration of foulants in the feed solution [5,
Chapter 2 9]. Since interactions between particles and membrane play a major role in fouling phenomena,
the pH and ionic strength of the fouling solution influence the critical flux.
1.1 Definitions of critical flux
According to Field et al. two forms of critical flux exist [2]. First, there is the so-called strong
form of critical flux. It is the highest flux at which the permeability of a fouling solution is the same as the permeability for the pure solvent (water) at the same flux. Subcritical operation in the weak form is said to exist when the permeability of the fouling solution is lower than that for the pure solvent, but independent of the imposed flux. For real world feeds, the strong form of the flux is rarely observed [13]. In a recent review [14] two additional definitions of critical flux are mentioned (Table 1).
The critical flux for irreversibility refers to the first process flux which results in irreversible fouling. A transition from dissolved to dispersed solutes can occur at the membrane at a certain flux. To restore the permeability of the membrane some cleaning action might be necessary. In case of the critical flux for irreversibility, the nature of the cleaning action should be part of the definition and tightly controlled in the experiments. Above the critical flux for irreversibility, the cleaning action is not sufficient to completely remove the foulants from the membrane.
In the same review, the concept of sustainable flux is explained. In sustainable operation, fouling itself is accepted. In many real world applications, complete prevention of fouling is not possible or viable. However, the rate of fouling is kept strictly under control to avoid excessive cleaning actions. Since sustainable operation depends on many factors, including an economic evaluation comprising the actual costs of cleaning and filtration, the sustainable flux is not so much a solid number as the critical flux.
Table 1 Definitions of critical flux adapted from [14]
Definition Discrimination between
Strong form No fouling Any kind of fouling
Weak form Fouling independent of solvent transfer
Fouling driven by solvent transfer
Critical flux for
irreversibility Reversible fouling Irreversible fouling Sustainable flux Sustainable flux No sustainable flux
1.2 Methods to determine the critical flux in crossflow filtration
There are three methods to determine the critical flux that are applicable to fouling of submicron particle matter. The first method focuses on the relationship between flux and pressure. In its most simple form, the flux is held constant while the trans membrane pressure is measured (see for example [15]). The pressure will be constant at sub-critical flux. Then the flux is stepwise increased and the pressure is monitored again. Above the critical flux, the pressure will increase with time. When the clean water permeability is known, both the weak and strong form of the critical flux can be determined. The accuracy of the method is limited by the precision of the equipment used to measure pressure variations and to set the flux. Alternatively, one can also set the pressure and follow the flux decline in time. The pressure and cross flow velocity are set to constant values, and the permeate flux is measured over time. Apart from determining the critical pressure using this method, the limiting flux is also determined. The limiting flux is the maximum flux that can be achieved, regardless of an increase in operating pressure. The standard step-method to determine the critical flux was improved by Espinasse et al. [16] and Mantarri et al. [7]. In this method an alternating step-pattern is applied. The method is shown in Figure 1. The flux is recorded at the different pressures. As long as the flux has the same value as the clean water flux, the critical flux (strong form) is not reached. The flux at pressure (3) is clearly above the critical flux, as the permeabilty is lower. The strong point of the method is that
Figure 1 Schematic overview of the TMP method adapted from [16]. The method can find the strong or weak form of the critical flux as well as determine the reversibility of fouling beyond the critical flux. When operated at pressure 3, the critical flux is passed. The reversibility of the fouling is determined by the flux measured at pressure 4. If the flux has a value of a (same as at pressure 1) the fouling is completely reversible. If the value is b, it is fully irreversible.
Chapter 2 it also determines the fraction of irreversible fouling that built up at pressure (3) with the help of
the flux found at pressure (4). If all fouling is reversible, the flux will be the same as at pressure (1), when the fouling is completely irreversible, the flux will have a value corresponding to ‘b’ in Figure 1.
The third method to determine the critical flux is to make a particle balance over the membrane feed and retentate [9, 10]. As long as the concentration of particles in the feed and retentate is the same, deposition on the membrane does not occur and therefore the critical flux has not been reached. Since the method does not rely on measurement of the pressure or flux, a distinction between the strong and weak form of the critical flux can not be made. The limitation of the method is that it requires that all particles are retained by the membrane. Furthermore, particles are lost in the system by adsorption to tubing etc. so this has to be corrected for. Gesan-Guizou et al. [10] combined the method with the pressure-step method and the first unstable fluxes were found when deposition was observed which shows the validity of the method.
1.3 Reversible fouling in dead-end filtration
In dead-end filtration, an increase in filtration resistance with time under constant pressure or constant flux conditions is always expected. As long as the membrane is at least partly retaining the components in the feed solution, the concentration above the membrane or inside the fiber will increase. Inevitably, this will result in a decline in flux or an increase in the transmembrane pressure. In that sense, it is not possible to determine critical parameters as described by the definitions given in Table 1. Fouling is generally removed from the membrane surface by either a chemical and/or hydraulic cleaning step. As such, the reversibility of the fouling can be studied by examining the performance of the membrane before and after the cleaning step. A number of variables might influence the reversibility of the fouling:
• Filtration flux or pressure • Fouling load during filtration
• Hydraulic cleaning settings (backwash flux and duration, forward flush, etc) • Chemical cleaning settings (applied chemicals, soak time, temperature, etc)
If the filtration flux is the only variable in an experiment, and the other settings are kept constant, the maximum flux at which the fouling is reversible would satisfy the criterion set in
Table 1 for ‘Critical flux for irreversibility’. Eventually the concept can be used to determine a complete critical operation matrix for irreversibility. This matrix describes critical values for other settings as well like process recovery and backwash settings.
Some authors explored the limits of fouling reversibility in dead-end filtration. Critical operating conditions were shown for natural organic matter in the work by Bessière et al. [17] and for skimmed milk by Gésan-Guiziou et al. [18]. Bessière et al. described dead-end filtration of real waste water using flat sheet membranes at constant pressure. The experiments consisted of dead-end filtration, and a rinse with pure water. After the rinse, the pure water permeability of the membrane was determined. Following this, the membrane was stored in pure water for 24hrs to examine the long term reversibility of the fouling. Depending on the filtered volume, the fouling was completely reversible, mostly reversible or significantly irreversible. Furthermore, experiments performed at higher pressure resulted in a lower degree of reversibility and a lower critical filtered volume.
Gésan-Guiziou et al. used the same experimental procedure to study the reversibility of fouling by casein micelles [18]. The aim of the study was to understand cake formation in order to optimize cross-flow filtration. In a subsequent publication the concept was extended to irreversible fouling in dead-end ultrafiltration [19]. Instead of the rinse procedures applied in their previous work, a more standardized backwash procedure was used. Furthermore, filtrations were performed at constant flux. The results were explained in terms of a model combining concentration polarization and deposition. Their hypothesis states that at certain conditions (flux and filtered volume) a critical osmotic pressure is reached at the membrane interface at which the particles form a deposit. This was incorporated in a slightly modified version of Darcy’s Law.
Recently the concept of sustainable flux (which was in their work defined as the weak form of the critical flux) was described for the treatment of tunneling waste water [20]. Their method described the estimation of a sustainable flux from single filtration runs at varying flux. The filtration time was kept constant which does make the results somewhat difficult to judge since the fouling load varied. Judging from the experiments with backwashing that were presented at the end of the paper, the method however worked quite well. The use of the weak form of the critical flux as a definition of the sustainable flux might not be suitable however, as a clear increase in pressure can be observed even when operating under this threshold flux.
Chapter 2 1.4 Concluding remarks
The use of the concept of critical flux is wide spread in cross flow filtration. Various definitions exist, and various methods are available to determine the critical flux. With the recent publication of the review paper by Bacchin et al. [14] the definitions have been formalized. In dead-end filtration, a critical flux in the strong or weak form as such does not exist, as an increase in resistance will always occur due to the increasing concentration of foulants. However, the critical flux for irreversibility can be a valid tool if it incorporates backwash procedures. Also, the concept for sustainable flux can be used in dead-end filtration, as it refers to control of the fouling rate over time, rather than occurrence of fouling at all. An experimental method to determine these parameters in dead-end ultrafiltration will be presented in the next sections.
2
Determination of critical parameter in dead-end filtration
2.1 Standard step-methodSeveral methods are presented here to determine critical parameters for sustainability as well as the transition point between sub- and super critical performance. The core of the method consists of performing a number of filtration cycles and recording the initial (Ri), backwash (Rbw)
and final resistances (Rf) for each cycle. The initial resistance can be determined by extrapolation
of the resistance as a function of filtered volume to zero volume in the first stable part of the curve. The backwash resistance is determined by measuring the backwash pressure and the backwash flow at the end of the backwash. The final resistance is the average of the last three data points in the resistance vs. filtered volume curve, or the result of a linear fit to the last section of the curve. After a certain number of filtration-backwash cycles, the parameter for which the critical value is to be determined, is increased and the resistances are once again recorded. In case of sub-critical behavior, the resistances are constant with increasing filtration-backwash cycles, while at super-critical operation, the resistances will increase with increasing cycle number. The number of cycles that is performed at one value of the parameter is important. Although a large number of cycles allows clearer judgment of the reversibility of the fouling, the time required to perform the experiments obviously increases with an increasing number of cycles. In initial experiments (e.g. Figure 5 in the results section), up to 200 cycles were performed at a given setting. This gives a good indication of the reversibility of the fouling, but the entire experiment takes several weeks. When studying the results from these initial tests, 10-15 cycles at each parameter step was found as a good compromise between accuracy and the
Figure 2 The method used to determine the critical flux for reversibility in the flux range
between 20 and 110 L·h-1m-2 with a precision of 10 L·h-1m-2. 15 cycles are
Chapter 2 time required to perform an experiment.
The standard method to determine the critical flux for irreversibility is shown in Figure 2. In the experiments the filtered volume is kept constant by changing the filtration time for different fluxes. Under the assumption of constant feed water properties, ideal cake filtration, and retention being independent of flux, the resistance that builds up is independent of the applied flux. As long as the fouling is fully reversible, the three measured resistances will have the same value at each cycle. If the operational flux is higher than the critical flux, some of the fouling will remain on the membrane, resulting in increasing values for the resistances. The degree of reversibility of the fouling at one flux is assessed by studying the rate at which the final, initial, or backwash resistances increase with increasing cycle number.
2.2 The transition point
The fouling rate was defined as the increase in either the initial, backwash or final resistance as a function of cumulatively produced permeate, or cycle number. Since the changes were small at times, especially compared to the absolute value of the resistance, two methods were developed to visualize the fouling rate. The first method determines the deviation from an average resistance with increasing cycle number, while the other method uses linear regression combined with the statistic significance of the calculated slope. The methods are explained with the assumption that the critical flux for irreversibility is determined.
2.2.1 Differential Resistance Method
For the first method, the data was manipulated as follows to allow for a quick, visual assessment of the reversibility. For each flux the average of the 15 final resistances is determined. Next, the difference between the observed final resistance and the average resistance is calculated for all of the 15 filtration cycles at one flux:
,
Δ =Ri Rf i − Rf (1)
Assuming a standard step method, ΔRi values for all 150 cycles (10 different fluxes) are plotted
in one graph. Reversible fouling shows random values for ΔRi with increasing cycle number
Irreversible fouling (at one flux) expresses itself as a continuous increase in the value for ΔRi,
starting at a negative value for the first cycle and increasing to a positive value at the last one. Reporting the values for ΔRi together with the graph showing the final, initial and backwash
resistances, gives a complete picture of the effect of a change in an operational parameter on the filtration performance. It shows the magnitude of the filtration resistance, as well as the reversibility of this resistance. It can be quickly performed for all operating parameters and yields a membrane performance matrix for the membranes in combination with a given feed water. The downside of the method is that it still relies on the judgment of the scientist to distinguish between reversible and irreversible fouling.
2.2.2 Linear Regression Method
The most straightforward way to assess reversibility over multiple cycles is to compare the values for the final, backwash, or initial resistances found in different cycles. The irreversible fouling rate can be found by plotting the final, initial or backwash resistances vs. cycle number or cumulatively filtered volume. When the fouling is fully reversible, the slope of this graph is, in principal, zero. When the fouling is irreversible, the slope is positive. The critical flux for irreversibility is then determined from a graph showing the fouling rate as a function of the filtration flux as the first flux where the fouling rate is unequal to zero.
Although in theory sound, in practice, the determination of the fouling rate is not straightforward, especially at high reversibility of the fouling. The slope of the graph is then determined by the experimental error in the data, and the results for linear regression are unreliable. The squared regression coefficient,Rreg2 , which gives an indication of the goodness of
the fit, is regularly as low as 0.05. Taking both the regression coefficient and the calculated slope into account gives a clearer image of the fouling reversibility than the fouling rate alone.
Algorithm to calculate the significant fouling rate
For all cycles performed at one flux Ji, either the final, initial, or backwash resistance is determined. The resistances are plot vs. the cumulative filtered volume and the irreversible fouling rate is determined as the slope (αi in [m
-2]) of the line calculated by linear regression.
Together with the slope, the regression coefficient (Rregression i2 ,) is determined and the slope and
regression coefficient are multiplied, to give the significant fouling rate (ξi[m
-2]) at flux i
J .
This is repeated for all fluxes and the values for ξi are plotted vs. Ji.
2 ,
· i i Rregression i
Chapter 2 From this graph, the critical flux for reversibility can be determined. At high fouling reversibility,
the significant fouling rate is close to zero. At very pronounced, low reversibility of the resistance
2 ,
reg i
R is close to one and ξi is almost equal to the fouling rate αi.
3
Experimental
3.1 Materials
Feed solutions were prepared from sodium alginate (low viscosity, Sigma) or humic acid (Aldrich). Solutions were prepared in batches of 25L by dissolving the material in ultrapure water (MilliQ, Millipore, >18.2MΩ). The ionic strength of the solutions was adjusted with potassium chloride (Merck). The influence of bivalent ions on the filtration performance was studied by adding CaCl2·2H2O (Acros organics). The required amount of Calcium chloride for
the 25L feed vessel was dissolved in about 500mL of ultrapure water. This was gradually added to the feed tank while it was filled. This prevents local ‘overdosing’ of the calcium chloride, potentially leading to instantaneous gel formation. All chemicals were used without further purification.
Cleaning solutions were received from Aquacare Europe. The cleaning solution, a mixture of Car 11 (sodium form of EDTA) and Fer 12 (based on iminodisuccinic acid and non-ionic surfactants), was prepared according to the directions of the manufacturer. The pH of the resulting alkaline solution was adjusted to a pH of 12.5-13 with sodium hydroxide. In some cases, approximately 500 ppm sodium hypochlorite was added to this solution. After soaking and rinsing with water, the membranes were thoroughly flushed with a 0.1M nitric acid solution. Ultrapure water was used to prepare all cleaning and fouling solutions.
3.2 Membrane
The membranes used in the experiments were supplied by X-Flow, the Netherlands. 10 membrane hollow fibers (PES/PVP blend, UFC-M5, inner diameter 8·10-4m) were potted in 8
mm PVC tubes with a polyurethane resin. The cut-off of the membranes is 200 kDa, and the clean water permeability of the membranes is around 600-700 L·h-1m-2bar-1 (manufacturer’s
data). The streaming potential coefficient through the (clean) membranes is -2·10-7 V·Pa-1, at
neutral pH measured using a 10 mM potassium chloride solution. This indicates that the membrane is negatively charged. Various modules were used throughout the experiments and for all modules initial clean water experiments were performed.
3.3 Experimental setup
The filtration setup, shown in Figure 3, was designed to mimic the filtration of large plants as closely as possible without compromising the accuracy and reproducibility of the experiments. To enable long-term and reproducible measurements, automated process operation was a key-issue during design. The feed and permeate vessels have a volume of 25L each. The system contains a feed and backwash pump (Ismatec, Reglo-Z). By means of solenoid switching valves (Plast-o-matic), the water from the backwash tank can be fed either to the feed or permeate side of the membrane, enabling the integration of automatic clean water flux experiments. Since most ultrafiltration plants operate at constant flux, two high precision mass flow controllers (Bronkhorst, NL, Cori-Flow M53C and M54C) are installed to measure the feed and backwash fluxes. Simultaneously with the measurement of the fluxes, constant flux operation is achieved by means of the internal PID controllers of the two instruments, which directly control the pumps. The PID controller allows for different control settings during the setpoint trajectory. Close to the setpoint, the settings are different compared to when there is a large difference between measured and set value. This results in very constant flow at the setpoint (deviations are less than 0.5% at medium flow rates), while the setpoint is achieved in a reasonable time-span (typically < 20 seconds for the feed pump and <5 seconds for the backwash pump). The precision of the measured flow rate is 0.1% of the mass flow controller’s full scale. The full scale
Figure 3 Flowsheet of the filtration setup. BW and Feed are backwash water and feed water tanks respectively. The flow is controlled by two mass flow controllers (F). Pressures (P) are measured in the feed and permeate. The pH, temperature (T) and conductivity (κ) of the feed water are monitored online. The equipment can be outfitted with two Ag/AgCl electrodes to measure the streaminge potential (Δφ).
Chapter 2 of the feed side mass flow controller is 4 L·h-1, and 10 L·h-1 for the backwash mass flow
controller.
Pressures (± 0.006 bar) are measured in the feed and permeate lines. The pH, conductivity and temperature (±0.2 °C) of the feed water are measured continuously. The signals from the sensors (apart from the mass flow controllers) are fed to a Fieldpoint (FP1600) system from National Instruments. All signals are filtered and digitized with 16 bit accuracy, apart from the temperature signals which are processed with 12 bit accuracy. Communication between the PC and the Fieldpoint system is done via a local area network.
The control software was developed specifically for this setup using National Instruments Labview 7.0. It logs the data and controls the setup. Control from the software is either manual (switching valves, changing mass flow controller setpoints and data logging options) or automatic. In automatic mode a sequence of commands, including loops and conditional commands, can be programmed to allow for long term unattended operation. Continuous experiments could be performed for extended periods of time without problems. All filtration data is saved with a one second interval, and each separate filtration-backwash cycle (see Figure 4) is saved in one log file. Matlab (The Mathworks, v7.0) scripts were developed to automatically extract relevant data from the log files, to perform calculations, and create graphs.
3.4 Layout of a single filtration-backwash cycle
Main subject of study is the influence of subsequent filtration-backwash cycles and the effect of various operating parameters on the amount of reversible and irreversible fouling. In initial tests it was found that the changes were often very small, requiring a very well defined experimental protocol for the filtration and backwash experiments.
The layout of a filtration-backwash cycle is shown in Figure 4. It consists of a forward flush with feed water, the actual filtration, a flush of the shell side of the module with clean backwash water, and the backwash. A typical pressure profile is given as well. The forward flush aims to fill the module with feed water. It is performed for 45 seconds at a flow rate of 0.15 mL·s-1. In a typical
module, this corresponds to ±3 times the inside volume of the fibers. Before the filtration-backwash cycle, the fibers are filled with clean water. Consequently, if the forward flush was omitted, the filtration step would start with clean water, instead of feed water. The downside of applying this step is that some adsorption of fouling matter may take place. The actual filtration is performed at a constant set flux. The end-point of the filtration is usually a set filtered volume,
but a set time, maximum pressure, or resistance are also possible. Before the backwash is performed, the shell side of the module is flushed with backwash water at 1 mL·s-1 for 60
seconds. This is to make sure that during the actual backwash the same water quality is used for the entire duration of the backwash.
0
Chapter 2
4
Results and discussion
4.1 Initial studies
Figure 5 shows the results of the filtration of a 50 ppm sodium alginate in a solution of 10 mM KCl. Between 150 and 200 filtration cycles were performed at varying fluxes to study the influence of the flux on the fouling reversibility and to reveal the ‘critical flux for reversibility’. The filtered volume in each filtration run was 9.2 L·m-2. Backwashing was performed at 250 L·h -1m-2. The module contained 12 fibers with an active filtration length (total fiber length minus the
potting) of 0.38m. The membrane area was 1.15·10-2 m-2.
The experiments ran for more than three weeks. Initially, an increase in both the backwash and final resistance was observed. This could be due to initial adsorption or blocking phenomena by the alginate [21]. After about 15 cycles, the resistance stabilizes. It remains fairly constant at a flux of 20 L·h-1m-2, although the resistance is influenced by the temperature (after correcting for
viscosity changes). Temperature variations are a result of day-night cycles. After 200 cycles, some alterations were made to the setup, including a new mass flow controller to control the
Figure 5 Results of long term filtration experiments. The final resistance is depicted with circles, while the backwash resistance is shown as the triangles. The dashed line shows the flux. Prior to
the filtrations at 30 L·h-1m-2 (a), some new equipment was
installed. At (b) one setting in the control software was
erroneously changed. Before the experiments at 50 L·h-1m-2
(c) the module was chemically cleaned.
Cycle number 0 200 400 600 800 1000 1200 1400 Fil trat ion flux , J [L ·h -1 m -2 ] 0 20 40 60 80 100 120 Resi st anc e, R [1 0 12 m -1 ] 0.0 2.0 4.0 6.0 Final resistance Backwash resistance Flux a b c
backwash flow. As a result, resistance is initially quite high at a flux of 30 L·h-1m-2, over time
however it reduces to more stable values. The following experiments at 40 and 30 L·h-1m-2
showed stable values for the final and backwash resistance. When the flux is increased to values equal to or higher than 50 L·h-1m-2, irreversible fouling is evident. The resistance is no longer
constant and steadily increases. After a few cycles at 110 L·h-1m-2 the maximum system pressure
of 2 bar was reached, automatically interrupting the experiment. The experiments depicted in Figure 5 clearly demonstrate that the reversibility of fouling depends on the applied flux. The transition between reversible and irreversible fouling appears between 40 and 50 L·h-1m-2. 4.2 Short-term experiments
4.2.1 Sodium alginate
The module contained 10 fibers with an active filtration length of 0.36±0.02m. Before carrying out the fouling experiments, the clean water permeability was measured by filtration at 5 different fluxes between 50 and 250 L·h-1m-2. After the flux was set, the pressure was allowed to
stabilize for a few minutes. Next, the pressure was measured 3 times with intervals of 30s. The permeability was determined by fitting a straight line through the data points. The measured permeability corrected to 20°C was 566±5.1 L·h-1m-2bar-1 (error from regression, 95% certainty
Figure 6 The results of the short-term step experiment to determine the critical flux for irreversibility. The figure shows the final resistances of filtration cycles at 10 different fluxes between 20 and 110 L·h-1m-2. Cycle number [-] 0 15 30 45 60 75 90 105 120 135 150 Fi ltrat ion fl ux [ L· h -1 m -2 ] 0 20 40 60 80 100 120 140 Fi ltra tion re sist ance, R [10 12 m -1 ] 0 1 2 3 4 5 6 7 Final resistance Flux
Chapter 2 interval). The membrane resistance corresponding to this clean water permeability is 6.38·1011
m-1.
After the determination of the initial clean water resistance, the fouling experiments were started with a 50 ppm sodium alginate solution. The standard step method described in paragraph 2.1 was used. The parameter under investigation was the critical flux for irreversibility. The software was programmed to perform a total of 150 filtration-backwash cycles at 10 different fluxes. At each flux 15 filtration-backwash cycles were scheduled. The filtered volume during filtration was 9.2 L·m-2. The backwash was performed at 250 L·h-1m-2 for 1 minute. The feed side of the
module was flushed prior to the filtration with feed water at a flow of 0.1 g·s-1 for 30 seconds.
The shell side was flushed at 1 g·s-1 for one minute. The final resistance was determined by
averaging the last 3 data points of the filtration (sample rate is 1 Hz). Linear regression is applied to the resistance values between Vf=3·10-4-5·10-4 m3m-2, and extrapolated to zero filtered volume
to give an estimation for the initial resistance. The results for the standard step method with this feed solution are given in Figure 6 and Figure 7.
Figure 6 shows the final resistance at the end of the filtration as a function of cycle number. It is apparent from this figure that there is a strong influence of the flux on the absolute values for the final resistance at the end of each filtration step. The resistance starts of high at a flux of 20 L·h
-Figure 7 Results for the backwash (circles) and initial (triangles) resis-tances for the experiment with sodium alginate in 10 mM KCl. Cycle number [-] 0 15 30 45 60 75 90 105 120 135 150 Filtr a tio n f lu x, J [L ·h -1 m -2 ] 0 20 40 60 80 100 120 Fi ltra tion resi st ance, R [1 0 12 m -1 ] 0.0 0.5 1.0 1.5 2.0 Backwash resistance Initial resistance Flux
