Fouling Behaviour and Radioactive Retention Properties of Inorganic Crossflow Microfiltration Membranes
A Thesis Submitted for the Degree of Doctor of Philosophy of the University of London
by
Manouchehr Asaadi, B.Sc., M.Sc., D.I.C.
Department of Chemical Engineering and Chemical Technology Imperial College of Science, Technology and Medicine London SW7 2BY
February 1990
1 Abstract
Variation of the filtration rate was examined for different types of tubular inorganic microfiltration membranes. The steady state filtration rate was examined with change in the membrane operating conditions such as, crossflow velocity, transmembrane pressure and feed concentration. A suspension of magnesia particles in water was used for the purpose of these experiments. When a run was started using a clean membrane the permeate declined sharply in the first few minutes of the operation and reached a near constant rate. The flux was related to the hydrodynamic properties of the flow and the feed concentration. The experimental results were successfully correlated by a mathematical model based on the resistance in series approach. The model may be used to estimate the flux for a given feed concentration, crossflow velocity and transmembrane pressure. An expression was also derived for the estimation of the period during which flux changes continuously in the initial stages of microfiltration. This model is particularly useful for estimation of the permeability of the fouling film.
To examine the radionuclide retention characteristics of the membranes, a number of active runs were carried out using radioisotopes of interest. Apart from the membrane type other variables examined were the feed pH, the solute concentration and influence of absorbers in the feed. The resulting decontamination factors showed that in general the membranes were capable of retaining substantial amounts of radiocolloids such as Am-241, Eu-152 / Eu-154, Th-230, Pu-239, Co-60 and U-235 especially at high pH ranges. However, those species such as Sr-85, Cs-137 and Sb-125 which remained in aqueous phase were only retained partially or not at all even in the presence of magnesia suspension and high pH conditions.
2 Acknowledgements
My sincerest thanks to Dr. D. A. White for his supervision, invaluable advice, guidance and interest throughout this project
Thanks are also due to the members of staff, in particular Professor M. Streat for applying for my work permit, Dr. P. G. Clay for reading the manuscript and his useful suggestions and students at the Nuclear Technology Research Section for their friendship and advice in the course of this work. Valuable assistance was received from Mr. R. King ( departmental workshop) and Mr. P. Amrit ( postgraduate workshop ).
Finally, the financial support of British Nuclear Fuels Pic. is gratefully acknowledged.
3 To m y Covingwife A zar
and my son Aryan Table of Contents
Page Title Page 1 Abstract 2 Acknowledgements 3 Table of Contents 5 List of Tables 10 List of Figures 13
CHAPTER ONE Introduction 17 1.1 Crossflow Microfiltration 17 1.2 Microfiltration in the Nuclear Industry 17 1.2.1 Type of Waste and Process Streams 18
1.3 Objectives 19
CHAPTER TWO Literature Survey 20 2.1 Introduction 20 2.2 Development and Application of Microfiltration 20
Membranes 20 2.2.1 Early Membranes 20 2.2.2 Inorganic Membranes 21 2.2.2.1 Manufacturing Methods 2.2.2.2 Application 2.3 Separation Properties of MF Membranes 22 2.3.1 Membrane Retention 22 2.3.2 Colloidal and Particulate Retention 22
5 2.4 Effect of Operating Parameters on Flux 25 2.4.1 Crossflow Velocity Effect 25 2.4.2 Transmembrane Pressure Effect 26 2.4.3 Feed Concentration Effect 26 2.5 Variation of Flux with Time 27 2.6 Temperature Effect 28 2.7 Influence of the Membrane Microstructure on Flux 28 2.8 Anti - fouling Measures 29
CHAPTER THREE Theory 31 3.1 Introduction 31 3.2 Fundamental Equations for Flow Through Porous Mass 31 3.3 Crossflow Microfiltration Process 33 3.3.1 Gel Polarisation - Formation of Gel Polarised Layer 34 3.6 Mathematical Models for Fouling 37 3.6.1 Semi - empirical Models 37 3.6.1.1 Exponential Relationship 3.6.1.2 Resistance in Series 3.6.2 Mechanistic Models 38 3.6.2.1 Hydrodynamic Resegregation Model 3.6.3 Aggregation Model 41 3.7 Retention Properties of Membranes 42
CHAPTER FOUR Experimental 44 4.1 Summary 44 4.2 Apparatus 44 4.2.1 Flow Rig 44 4.2.2 Filtration Module 45 4.2.3 Membranes 50
6 4.2.3.1 PALL Sintered Stainless Steel 4.2.3.2 DOULTON Ceramic 4.2.3.3 FAIREY Stainless Steel 4.3 Feed and Its Characteristics 53 4.4 Experiments 53 4.4.1 Determination of the Hydraulic Permeabilities of the Membranes 53 4.4.2 Preliminary Fouling Experiments 53 4.4.3 Experiments to Investigate Variation of Flux with Time and the Operating Conditions 56 4.4.3.1 Variation of Flux with Time 4.5 Radioactive Retention Experiments 57 4.5.1 MILLIPORE Tests 57 4.5.1.1 Feed Preparation and Experimental Technique 4.5.2 Retention Characteristics of M4 CARBOSEP Membrane 59 4.5.2.1 Apparatus and Membrane Used 4.5.2.2 Experiments 4.5.3 Radioactive Retention Properties of PALL, DOULTON 63 and CARBOSEP Membranes 4.5.3.1 Apparatus and Feed 4.5.3.2 Experiments
CHAPTER FIVE Results and Discussion 65 5.1 Summary 65 5.2 Results 66
5.2.2 Hydraulic Resistance of the Membranes 6 6
5.2.3 Fouling Experiments 6 6 5.2.3.1 Variation of Flux with Time and Cross Row Velocity at Constant Pressure
7 5.2.3.2 Variation of Flux Time and Transmembrane Pressure at Constant Velocity 5.2.3.3 Further Investigation of the Variation of Flux with the Operating Variables 5.3 Model Development 74 5.3.1 Derivation of a Relationship Between Rf and Hydrodynamic Properties of Flow 75 5.3.2 Application of the Model to the Experimental Data 80 5.3.3 Further Points About the Model 85 5.4 Transient Filtration Period 88
5.4.1 Model Development 8 8 5.4.2 Application of the Model 93 5.4.2.1 Estimation of the Film Thickness 5.5 Results from the Active Experiments 98 5.5.1 MELLIPORE Filters 98 5.5.1.1 Effect of pH and Fouling Agent 5.5.1.2 Effect of Pore Size 5.5.2 M4 CARBOSEP Membrane 102 5.5.2.1 Retention of Nuclides at Different pH 5.5.2.2 Retention of Nuclides at Constant pH 5.5.2.2.1 Europium Run 5 5.2.2.2 Europium and Cobalt Run 5.5.3 M4,M6 CARBOSEP and PALL and DOULTON Membranes 109
CHAPTER SIX Conclusions and Recommendations 113 6.1 Conclusions 113 6.2 Recommendations for Further Work 115
8 Nomenclature 117 References 120
Appendices
Appendix A 127 Operating Conditions and Steady State Flux Values from the Fouling Experiments at Different Crossflow Velocities and Feed Concentrations When Transmembrane Pressure is Kept Constant. Appendix B 130 Operating Conditions and Steady State Flux Values from the Fouling Experiments at Different Transmembrane Pressures and Feed Concentrations When Crossflow Velocity is Kept Constant. Appendix C 134 Application of the Schock Relationship to the Experimental Data Appendix D 140 Calculation of Transient Filtration Period, Cake Permeability and the Film Thickness
9 List of Tables
CHAPTER 2 Table 2.1 Position of Microfiltration Process in Relation to Other Separation Techniques Where Porous Membranes are Employed.
CHAPTER 4 Table 4.1 Characteristics of the Membranes Used in Fouling Experiments. Table 4.2 Characteristics of the Membranes Used in the Active Experiments.
CHAPTER 5 Table 5.1 Hydraulic Resistance and Clean Water Flux of the Membranes. Table 5.2 Initial, 1 Hour and 2 Days Flux of Runs with Magnesia Suspension at Various Feed Concentrations, Crossflow Velocities and Transmembrane Pressures. Table 5.3 Summary of the DF Ranges for Different pH and Feed Conditions Using MILLIPORE Filters. Table 5.4 Variation of DF's with Feed pH for Different Elements in the Experiments with M4 CARBOSEP Membrane. Table 5.5 Variations of Activities and DF's with Time for Europium Feed at pH 11.5 Using M4 CARBOSEP Membrane. Table 5.6 Variations of Activities and DF's with Time for Europium and Cobalt Feed at pH 11.5 Using M4 CARBOSEP Membrane. Table 5.7 Summary of the DF Ranges of the Radionuclides of Interest for Different Membranes. Table 5.8 Ratio of Activity of Feed at the Beginning and the End of the Runs Using Different Membranes.
10 APPENDIX A Table A.l Variation of Steady State Flux with Crossflow Velocity and Feed Concentration for PALL Membrane at Constant Transmembrane Pressure. Table A.2 Variation of Steady State Flux with Crossflow Velocity and Feed Concentration for DOULTON and FAIREY Membranes at Constant Transmembrane Pressure.
APPENDIX B Table B.l Variation of Steady State Flux with Transmembrane Pressure and Feed Concentration for PALL Membrane at Constant Crossflow Velocity. (Conc.6.0 and 18.5 % WAV) Table B.2 Variation of Steady State Flux with Transmembrane Pressure and Feed Concentration for PALL ( Cone. 28.5 % W/W ) and DOULTON ( Cone. 6.25 and 28.5 % WAV ) Membranes at Constant Crossflow Velocity. Table B.3 Variation of Steady State Flux with Transmembrane Pressure and Feed Concentration for DOULTON ( Cone. 17.5 % WAV ) and FAIREY (Cone. 13.0% WAV) Membranes at Constant Crossflow Velocity.
APPENDIX C Table C.l Various parameters of the Runs with PALL Membrane for 6.0 % W/W Magnesia Suspension and Parameters Evaluated from Schock Equation, p = 1042.0 Kg m-3 , p =1.582e-3 Pa s Table C.2 Various Parameters of the Runs with PALL Membrane for 18.5 % W/W Magnesia Suspension and Parameters Evaluated from Schock Equation, p = 1128.0 Kg m-3, p = 4.837 e-3 Pa s, ty = 1.64 Pa Table C.3 Various Parameters of the Runs with PALL Membrane for 28.5 % W/W Magnesia Suspension and Parameters Evaluated from Schock Equation. p =1200.0 Kg m-3, p = 9.692 e-3 Pa s, ty = 11.2 Pa Table C.4 Various Parameters of the Runs with DOULTON Membrane for 6.25% WAV Magnesia Suspension and Parameters Evaluated from Schock Equation,
p = 1043.7 Kg m-3 , p = 1.621 e - 3 Pa s Table C.5 Various Parameters of the Runs with DOULTON Membrane for 12.0% WAV Magnesia Suspension and Parameters Evaluated from Schock Equation, p = 1084.0 Kg m-3 , p = 2.825 e -3 Pa s
Table G. 6 Various Parameters of the Runs with DOULTON Membrane for 17.5 % WAV Magnesia Suspension and Parameters Evaluated from Schock Equation,
p = 1122.8 Kg m-3, p = 4.561 e-3 Pa s, ty = 1.02 Pa Table C.7 Various Parameters of the Runs with FAIREY Membrane for 13.0 % W/W Magnesia Suspension and Parameters Evaluated from Schock Equation,
p = 1091.1 Kg m-3, p = 3.10e-3 Pas
12 List of Figures
CHAPTER 3 Figure 3.1 Schematic Representation of Crossflow Microfiltration. Figure 3.2 Formation of Gel Polarisation Layer in the Mass Transfer Controlled Region. Figure 3.3 Influence of Hydrodynamic and Diffusive Controlling Effects as a Function of Particle Size.
CHAPTER 4 Figure 4.1 Flow Diagram of Microfiltration Rig. Figure 4.2 Detailed Drawing of Filtration Module Used for PALL and DOULTON Membranes. Figure 4.3 Detailed Drawing of Filtration Module Used for FAIREY Membrane. Figure 4.4 Particle Size Distribution of Magnesia Suspension. Figure 4.5 Shear Stress Against Shear Rate for Different Concentration of Magnesia Suspension. Figure 4.6 Variation of Viscosity of Magnesia Suspension with its Solid Concentration. Figure 4.7 Variation of Density of Magnesia Suspension with its Solid Concentration. Figure 4.8 Flow Diagram of Microfiltration Rig Used in Active Experiments. Plate 1 Photograph of the Microfiltration Rig Used in the Fouling Experiments. Plate 2 Photograph of the Tubular membranes Used in the Experiments.
CHAPTER 5 Figure 5.1 Variation of Flux with Time and Crossflow Velocity at Constant Transmembrane Pressure of 64.0 KPa and Feed Concentration of% 6.0 WAV for PALL Membrane.
13 Figure 5.2 Variation of Flux with Time and Crossflow Velocity at Constant Transmembrane Pressure of 76.0 KPa and Feed Concentration of 6.0 % WAV for PALL Membrane. Figure 5.3 Variation of Flux with Time and Crossflow Velocity at Constant Transmembrane Pressure of 94.0 KPa and Feed Concentration of 18.5 % WAV for PALL Membrane. Figure 5.4 Variation of Flux with Time and Crossflow Velocity at Constant Transmembrane Pressure of 60.0 KPa and Feed Concentration of 28.5 % WAV for PALL Membrane. Figure 5.5 Variation of Flux with Time and Crossflow Velocity at Constant Transmembrane Pressure of 60.0 KPa and Feed Concentration of 6.25 % WAV for DOULTON Membrane. Figure 5.6 Variation of Flux with Time and Crossflow Velocity at Constant Transmembrane Pressure of 100.0 KPa and Feed Concentration of 12.% WAV for DOULTON Membrane. Figure 5.7 Variation of Flux with Time and Crossflow Velocity at Constant Transmembrane Pressure of 50.0 KPa and Feed Concentration of 17.5 % WAV for DOULTON Membrane. Figure 5.8 Variation of Flux with Time and Crossflow Velocity at Constant Transmembrane Pressure of « 60 KPa and Feed Concentration of 13.0 % WAV for FAIREY Membrane. Figure 5.9 Variation of Flux with Time and Transmembrane Pressure at Constant Crossflow Velocity of 1.343 m s“l and Feed Concentration of 6.0 % WAV for PALL Membrane. Figure 5.10 Variation of Flux with Time and Transmembrane Pressure at Constant Crossflow Velocity of 1.02 m s’^ and Feed Concentration of 12.0 % WAV for DOULTON Membrane. Figure 5.11 Plot of Equation (5.6 ) Using Experimental Data for PALL Membrane with the Gradients of Lines Being Equal to the Constant B in Equation (5.6).
14 Figure 5.12 Relationship Between the Constant A in Equation ( 5.6 ) and Magnesia Slurry Concentration. Figure 5.13 Comparison of the Experimental Flux vs. Transmembrane Pressure Data at Different Crossflow Velocities with the Results Predicted by Equation (5.8) ( Continuous Lines). Membrane: PALL, Feed Concentration : 6.0 % W/W. Figure 5.14 Comparison of the Experimental Flux vs. Transmembrane Pressure Data at Different Crossflow Velocities with the Results Predicted by Equation (5.8) ( Continuous Lines). Membrane: PALL, Feed Concentration : 18.5 % W/W. Figure 5.15 Comparison of the Experimental Flux vs. Transmembrane Pressure Data at Different Crossflow Velocities with the Results Predicted by Equation (5.8) ( Continuous Lines). Membrane : PALL, Feed Concentration : 28.5 % W/W. Figure 5.16 Comparison of the Experimental Flux vs. Transmembrane Pressure Data at Different Crossflow Velocities with the Results Predicted by Equation (5.8) (Continuous Lines). Membrane: DOULTON, Feed Concentration : 6.25% W/W. Figure 5.17 Comparison of the Experimental Flux vs. Transmembrane Pressure Data at Different Crossflow Velocities with the Results Predicted by Equation (5.8) (Continuous Lines). Membrane: DOULTON, Feed Concentration : 12.0% W/W. Figure 5.18 Comparison of the Experimental Flux vs. Transmembrane Pressure Data at Different Crossflow Velocities with the Results Predicted by Equation (5.8) (Continuous Lines). Membrane: DOULTON, Feed Concentration : 17.5 % W/W. Figure 5.19 Comparison of the Experimental Flux vs. Transmembrane Pressure Data at Different Crossflow Velocities with the Results Predicted by Equation (5.8) (Continuous Lines).Membrane: FAIREY, Feed Concentration: 13.0W/W.
15 Figure 5.20 Experimental Values of Flux Against Those Calculated Using the Present Model. Figure 5.21 Application of the Schock Model to the Experimental Results. Figure 5.22 Application of the Modified Schock Model to the Experimental Results. Figure 5.23 Cumulative Flux Against Filtration Time Representing the Transient Filtration Model. Figure 5.24 Cumulative Flux as a Function of Filtration Time for a Run Carried Out with PALL Membrane at V= 0.7 m s _1, AP* = 98.6 KPa and C = 2.25 % WAV. Figure 5.25 Cumulative Flux as a Function of Filtration Time for a Run Carried Out with PALL Membrane at V= 1.34 m s _1, APt = 76.17 KPa and C = 2.25 % WAV. Figure 5.26 Variations of Decontamination Factors ( DF ) with Pore Diameters of MILLIPORE Filters for Various Radionuclides at Feed pH of 9.0. Figure 5.27 Variations of Decontamination Factors ( DF ) with Pore Diameters of MILLIPORE Filters for Various Radionuclides at Feed pH of 11.5. Figure 5.28 Variations of Decontamination Factors ( DF ) with Pore Diameters of MILLIPORE Filters for Various Radionuclides at Feed pH of 10.10 in the Presence of Magnesia Suspension. Figure 5.29 Variations of Decontamination Factors ( DF ) with Pore Diameters of MILLIPORE Filters for Various Radionuclides at Feed pH of 11.5 in the Presence of Magnesia Suspension. Figure 5.30 Variation of Decontamination Factors ( DF ) with Feed pH for Various Radionuclides When M4 CARBOSEP Membrane is Used.
16 CHAPTER ONE
Introduction
1.1 Crossflow Microfiltration
Microfiltration ( MF ) is a process where a semi-permeable barrier, known as a membrane, is used to separate particles of a few microns in size from soluble species and water. The membrane pore size range of interest in MF processes is between 0.1 - 10 |im. This allows the retention of macromolecules, colloids and small particles while allowing micromolecules and soluble ions to pass into the permeate. In crossflow microfiltration the process suspension moves in parallel to the membrane surface. Such a flow helps to remove entrapped particles adjacent to the membrane surface, thus, minimising the resistance to permeation.
1.2 Microfiltration in the Nuclear Industry
In general, for treatment of radioactive waste and process streams conventional physical and chemical unit operations are employed. Chemical methods include coagulation, flocculation and ion exchange. The physical group of processes are centrifugation, evaporation, filtration, ultrafiltration ( UF), and reverse osmosis ( RO). Microfiltration falls into the latter class of operation. It may be considered as an extension of filtration processes which are often used in the nuclear waste treatment. Application of MF in the nuclear field is relatively recent RO, UF and MF are a group of operations that are traditionally employed in agricultural and biotechnological fields and in the food and dairy industries. Typically polymeric membranes have been used in these areas. Such membranes are liable to degradation and it has been established that they can be unstable
17 at high temperatures or in the presence of ionising radiation. In addition, all membranes suffer from loss of flux as a result of fouling. In theory it is possible to regenerate organic membranes by introducing a chemical wash cycle, but in practice they have often been found to be prone to attack by chemicals used in chemical washing. However, with recent development of inorganic or mineral membranes which possess a high degree of resistance to chemical abrasion and degradation and can be cleaned quite readily, they have found wider application in the nuclear and other industries. Before defining the task of this project, the types of waste which might be encountered in the nuclear field are discussed.
1.2.1 Type of Waste and Process Streams
The process and waste streams which are of interest here are known as wet wastes which arise during the treatment of process and dilute waste streams and cleaning up operations. They consist of emulsions, solutions, slurries and sludges of materials contaminated with low levels of activity. Liquid streams may contain dissolved and suspended solids of varying amount of activity. Radioactive contaminants such as Co-58, Co-60, Mn-54, Cr-51, Ni-58 and Fe-59 could come from the corrosion of a reactor core while Sr-90, Cs-134, Cs-137, 1-131 and Kr-85 are radioactive fission products. Contaminants due to defective fuel and uranium present in the cladding of fuel elements also contribute to all these. Therefore, in general, a significant amount of suspended solids in the wet waste is due to the radioactive corrosion products ( such as iron and nickel ) while many fission products tend to be present predominantly in soluble form. It is obvious thatAmicrofiltration process can be used in a number of areas to deal with such wastes for more efficient waste treatment and management Therefore, it may be used for decontamination and removal of particulate waste and isotopes present in colloidal form and for dewatering and waste volume reduction. Such applications, though, require knowledge ofjnembranes' abilities in retaining radioactivity and their behaviour under operational conditions. This project is concerned with such
18 investigations.
1.3 Objectives
Radioactive decontamination properties of inorganic membranes are not well understood and it is not clear how they behave in ionising radiation. The presence of particulate matter in the process stream may cause the suspension to act as a sorption surface and build up of particles on the membrane surface could enhance the membrane decontamination properties. For study of such behaviour a more detailed examination of inorganic membranes is required and mathematical models are needed to predict and extend the experimental investigations where it is necessary.
By and large, the majority of the work carried out in membrane separation processes use pre-filtered feeds containing only a small amount of particulate matter which behaves as Newtonian fluids. However, in the present research project a slurry of high solid content will be used as the feed which, above a certain concentration, behaves as a non-Newtonian fluid. Different inorganic tubular membranes will be used to examine the effect of build up of particulate matter on the membrane when a simulated radioactive slurry feed is microfiltered. The effect of fouling on the rate of production of permeate (flux) and its variations with change in the operating conditions will be examined. These included crossflow velocity, transmembrane pressure and feed concentration. A suspension of magnesia particles in water is used for the purpose of such fouling experiments. In a separate class of experiments, a number of radioisotopes of interest will be used to examine the radioactive retention capabilities of the membranes under investigation.
19 CHAPTER TWO
Literature Survey
2.1 Introduction
With recent development and improvement in manufacturing of inorganic membranes, a whole new area of their application is envisaged. Although most research work has so far been directed towards the investigation of organic membranes there are signs that, due to their enhanced properties, work on inorganic membranes is progressively increasing. The basis for this literature survey is the extensive investigations carried out on organic membranes, particularly those related to their basic theory and operation. These seem to apply well to inorganic membranes. In this Chapter particular attention is paid to the factors influencing the flux as a result of membrane fouling which is of interest in this work.
2.2 Development and Application of Microfiltration Membranes 2.2.1 Early Membranes
It is only in the last fifty years that microfilters have been used in industry. The first application of microfilters, which were of organic type and made of finely porous cellulose nitrate or cellulose acetate, was for bacteriological analysis of water. These membranes were essentially surface filters on which bacteria were retained and grew as the nutrient was supplied. These were prepared by casting a polymer solution and a mixture of solvents on a suitable surface and then gelling the liquid film slowly by exposing it to humid air ( Lonsdale, 1982 ). Other methods of producing organic
20 microporous are also reported in the literature ( Fletcheret al, 1969). The membranes so developed had limited applications due to their low resistance to high temperature
(maximum 50 ° C ) and limited pH range ( 4 8to ). Later on development of other organic membranes made of synthetic polymers of polyamide, polyacrylonitrile and polysulphonic derivatives led to the possibility of operation at higher temperature and pH.
2.2.2 Inorganic Membranes
The early development and manufacture of inorganic membranes are reported in a number of patents ( Boomanet a l, 1974; Berger et a l, 1970 and Rajanet a l, 1969). However, according to Charpinet al ( 1987 ), inorganic microfilters in their present form appeared only in early 1980's.These membranes are basically made of macroporous structure which may or may not be coated with a thin microporous layer. The first of these was CARBOSEP membranes made of carbon support coated with a thin layer of zirconium oxide ( Veyre, 1984). These membranes come both in UF and MF pore size ranges. Other membranes include DOULTON porous ceramic made of
aluminium silicate with a base pore size of8 pm and skin pore size of1 . 0 pm ( Bank et al ), FAIREY sintered stainless steel with a macroporous support and skin pores size of 1 and 2 pm ( private communication with company). Membranes with uniform structure are also available. One such membrane is the PALL sintered stainless steel with an average pore size of 2.5 pm. Another MF membrane made of composite alumina is the CERAFLO membranes with pore diameters of 0.2,0.45 and 1 pm .
2.2.2.1 Manufacturing Methods
The most widely used method to manufacture inorganic membranes is the use of powdered material of specific size. The composite membranes, for instance, are made of two very different grain sizes. Intergranular spaces which are produced by fusion of
21 these grains specify the pore size of membranes. In manufacture of a composite mineral membrane the following three steps are taken. ( a ) preparation of the macroporous substrate, ( b ) preparation of the microporous mineral and ( c ) preparation of the microporous layer. These procedures are very involved and often modified to suit the manufacturer's requirements. Detailed description of the various manufacturing methods are given by Charpin et al (1987), Leenaars et al ( 1984 ) and ( 1985), Bergez et al (1984) and Strathmann (1986 and 1981).
2 2 2 2 Application
Due to their high pressure and high temperature tolerances (10 bars and 500 ° C and higher) inorganic membranes have found numerous applications and other new areas are under study. The high temperature and pressure properties make them ideal for processing liquids which are very viscous and hardly flow at ambient temperature. Such applications are under development in the petroleum industry for recovery of de-asphalting solvents and direct de-asphalting of crude oils ( Charpinet a l, 1987 ). Other high temperature applications include treatment of emulsions, recovery of lignin in the paper industry, for recovery of colouring matters in the textile industry and for vapour sterilisation in the pharmaceutical industry . Their enhanced mechanical properties and minimum maintenance and their long life ( six years in the case of CARBOSEP membranes) make them ideal for use in the nuclear industry. CARBOSEP membranes have already been employed for the uranium isotope enrichment and in the BNFL EARP process for treatment of reprocessing plant effluents.
2.3 Separation Properties of MF Membranes 2.3.1 Membrane Retention
MF membranes in principal are screen filters. Table 2.1 shows the position of MF processes in relation to other membrane separation techniques where porous membranes
22 are employed.The pore size range of interest is between 0.01 - 2.5 Jim. It is expected that with increasing pore radius the retention of small particles will decease. A pore size distribution which exists in nearly all membranes could affect the retention especially in the early stages of operation. The relationship between membrane pore characteristics and retention of UF membranes is examined by Jonssonet al (1986). They noted that in some UF membranes true retention, expressed as the rejection coefficient, was dependent on the molecular weight of macromolecules in solution. In the case of MF membranes, however, membrane retention is less dependent on the pore size of membranes since MF membranes are more liable to fouling and, in general, the retention in MF membranes isonlyslightlydependent on the pore size.The work by Jaffiinet al (1989) on microfiltration of wine confirms this. They showed that a similar permeate turbidity resulted when two different pore size ceramic membranes were used. Therefore, in MF membranes the true retention is of little significance and most sizes of colloidal species, macromolecules and particulate matter are retained while all soluble species are passed to the permeate.
2.3.2 Colloidal and Particulate Retention
Existence of species in colloidal form in the optimum environment may be used for their separation. In the acidic region solutes tend to be in ionic form and remain in solution. As the pH is raised they precipitate and form colloids or, in the presence of impurities or other species, form pseudocolloids or mixed species colloids of varying sizes. Thus, higher pH conditions are more favourable for physical separation of colloids. Reports on the retention of colloidal and particulate species from solution are abundant. Although such data are the results of small scale laboratory experiments, they give a useful insight to the viability of large scale operation. Workers such as Kepak (1971), Grebenshchikova et al (1961), Tsvetaeva et al (1986) and Olofssonet al (1982 and 1983 ) have utilised this property and investigated the colloidal presence of a number of species and reported significant retention of pseudocolloids of varying sizes by screen
23 Membrane Process Membrane Type Size Range (pm) Driving Force Method of Separation Microfiltration Symmetric or Colloidal Particles Transmembrane Sieving Mechanism Asymmetric of 0.01 - 50 pm Pressure, and Absorption Microporous Size 0 . 1 - 1 bar Membrane Ultrafiltration Asymmetric Macromolecules Transmembrane Sieving Mechanism Microporous of 0 . 0 1 - 0 .1 pm Pressure, Membrane Size 0.5 - 5 bar Reverse Osmosis Asymmetric Ionic Range of Transmembrane Solution Diffusion Skin Type 0 .0 0 0 1 -0 . 0 0 1 Pressure, Mechanism Membrane pm Size 2 0 - 1 0 0 bar Dialysis Symmetric Ionic Range of Concentration Diffusion Microporous 0 .0 0 0 1 -0 . 0 0 1 Gradient Mechanism Membrane pm Size Electrodialysis Cation and Ionic Range of Electrical Electrical Charge Anion Exchange 0 .0 0 0 1 -0 . 0 0 1 Potential of Particle and Membrane pm Size Gradient Size Gas Separation Homogeneous or Micromolecular Transmembrane Separation of Gas Porous Polymer Size Range Pres, and Cone. Mixture Gradient
Table 2.1 Position of Microfiltration Process in Relation to Other Separation Techniques Where Porous Membranes Are Employed. filters. Olofsson et al (1983 ), for instance, used screen filters to estimate colloidal sizes of Am-241 species. They reported the presence of Am-241 colloids of 100 |im in size at
pH 5 - 8 and at the pH 12.
2.4 Effect of Operating Parameters on Flux
In this work variation of flux with the operating parameters such as crossflow velocity, transmembrane pressure and feed concentration is of particular interest. These are generally referred to as hydrodynamic properties. A suspension containing particles of a few micrometers in size which are expected to form a secondary cake layer is used. Therefore, the following sections are discussed with this in mind.
2.4.1 Crossflow Velocity Effect
In general, a higher crossflow velocity results in higher fluxes. It also reduces the rate of the flux decline. This is particularly the case with feeds of high fouling potential.The crossflow velocity tends to shear off the deposited particles and entrain them in the bulk feed. The crossflow velocity effect has been investigated by a number of investigators, though, the majority have used RO and UF membranes. Sheppardet al ( 1972 ) used tubular hyperfiltration membranes and river water feed report that up to a certain velocity, which they called the threshold velocity, the rate of flux decline varied with feed velocity as V"0-5- Above the threshold velocity the flux decline was less than that predicted by the half - power relationship. Patelet al ( 1987 ) also report lower fouling rates at higher velocity in microfiltration of yeast cell suspensions. Other workers also reported similar observations in MF and UF operations(Kuo et a l, 1983 and Probstein etal, 1981).
25 2.4.2 Transmembrane Pressure Effect
Ideally for a non-fouling feed flux should increase as the transmembrane pressure increases. However, in microfiltration and most of UF processes this is not quite the case. Investigations of Pateletal (1987) show that higher fouling rates occur at higher transmembrane pressure. In fact, provided that other operating parameters i.e., temperature and velocity are kept constant flux becomes independent of transmembrane pressure. This region of operation, which will also be looked at in Chapter Three, is known as the mass transfer controlled region. Under such conditions the flux is controlled by the interaction between particles from the bulk fluid and gel or boundary layer formed on the membrane surface and its mass transfer characteristics. Presence of such a region has been verified experimentally by numerous authors ( Porter, 1972 and 1972; Blatt et a l, 1970 and Cheryanet a l, 1984 ). The pressure dependency of flux seems to be in the early stages of UF and MF operations. This favours low feed concentrations and high operating flow velocities.
2.4.3 Feed Concentration Effect
It is expected that higher feed concentrations result in lower fluxes. Experimental investigation on concentration effect is extensive. Porter ( 1972 ) found that flux decreased as concentration of an albumin feed increased. Kroneret al (1986) reported an asymptotic relationship between flux and concentration. They found a 50% drop in the flux as concentration of a feed consisting of copolymers of polypropylene glycol was b C C n increased. The influence of concentration on flux Has^modelled mathematically by a number of authors. One such model due to Lundqvistet al ( 1986 ), based on the resistance in series approach, simulated an asymptotic decline in the steady flux of a fouled UF membrane as the feed concentration increased. The model compared well with their experimental data. On the other hand, not every worker believes that concentration is of great significance. Murkes etal ( 1988 ), for instance, report
26 obtaining a high quality filtrate regardless of the quality of the feed water.
2.5 Variation of Flux with Time
When a suspension is microfiltered a fouling film is quickly formed on the feed side of the membrane. Properties of the fouling layer and its behaviour as a function of time affects the flux greatly. As soon as a film is formed on the membrane, its permeability changes and the total permeability of the membrane continues to vary until a stable film thickness is reached. Even at that time, depending on the cake layer properties and its surroundings, the fouling layer might undergo further compaction and an aging process. Faneet al (1987) believes that changes in the cake layer properties with time occurs in three phases. Initially with no film present, flux declines due to the passage of pure solvent. This period is referred to as fouling compaction. The second stage is the flux decline during the build up of a concentration polarisation layer where boundary layer plugging and adsorption take place.The duration of these stages are quite short (typically two minutes). The third stage is a long-term flux decline over hours, days or weeks of operation. During this period fouling deposition or adsorption and cake consolidation occurs. In practice the flux may not continue to decrease continuously and at some stage reach a steady value. However, such behaviour is very much dependent on the characteristics of feed, membrane properties and operating conditions. There are numerous papers which report that flux decline becomes steady sometime after the start up of operation while others mention a continuous but steady decline of the flux after the initial sharp drop. For instance, Faneet al (1980 )reports achieving a steady state flux after 60 hours operation of a thin channel UF using activated-sludge containing dissolved macrosolutes and bacterial cells. Lundqvistet al ( 1986 ) also observed a rapid flux decline during the first five minutes of their UF experiments with oil - water emulsion as the feed. During the period following this, a steady state was obtained which lasted throughout their experimental runs of 10 to 12 hours. The steady state flux behaviour was also reported by Asaadiet al ( 1989 ) during microfiltration of magnesia
27 - water suspension.
2.6 Temperature Effect
was Higher temperatures affect the rheological properties of slurries and* expected to result in higher permeate fluxes. In the mass transfer controlled region higher temperatures increase diffusivity and mass transfer coefficient of the gel polarised layer and, hence, enhances the filtration rate. Of those workers reporting such a result Madsen ( 1977 ) could be mentioned. He reported a three fold increase in the flux of skim milk when the temperature was varied from 15 °C to 75 °C. In cases where a fouling film is present the temperature effect is quite the opposite. Higher temperatures may decrease the solubilities of some solutes or hasten their coagulation, thus, increasing the hydrodynamic resistance to the flow of the filtrate as the fouling layer thickens (Cheryan, 1986 ). An example of such behaviour is reported by Maubois ( 1980 ). He found that the calcium salt solubility decreased as the temperature of the feed increased resulting in the precipitation and crystallisation of calcium in the membrane pores and decreasing the flux.
2.7 Influence of the Membrane Microstructure on Flux
The relationship between membrane surface pore characteristics and flux of UF membranes was examined by Faneet al (1981 ). Electromicroscopy of a number of membranes including AMICON XM100A and XM300 showed a distribution of pore sizes and porosities of around 1%. The flow through the membrane surface was found to be non - homogeneous and regions of differing permeability were found. UF experiments with protein solutions confirmed that for low permeability membranes the gel polarised flux was dependent on the initial membrane permeability while for higher permeability membranes, gel polarisation flux became independent of the water flux and hence surface properties. The latter phenomenon is usually the case in microfilters which
28 possess large pore sizes and high porosity. The work by Jafftinet al ( 1989 ) on the flux - time behaviour of tubular inorganic microfilters showed that the steady flux values from two different pore size ceramic membranes were the same. When a different set of membranes made of zirconium oxide with porous carbon support were used, a similar flux behaviour occurred although, this time the flux was much lower than the previous ceramic membranes. Jaffrin et al ( 1989 ) state that this behaviour could not be explained by the difference in the pore size alone but may have been caused by factors such as the higher porosity of the ceramic support and other solution - membrane interactions. These interactions include electrostatic interactions, hydrophobic effects, charge transfer or combination these between solution species and membrane material (Orchard, 1989).
2.8 Anti - fouling Measures
In crossflow membranes the main factor controlling the fouling layer thickness is flow velocity and the presence of turbulence. Therefore, to minimise fouling the crossflow velocity is increased or the channel diameter decreased. However, it is not always possible to keep the flow velocity very high. A popular way of reducing fouling is to use turbulence promoting devices. These may be built as an integral part of structure of the membrane in shape of rings or twists in the tubes ( Charpinet a l, 1987 ) or by using inserts and glass beads which decrease the volume hold-up and increases crossflow velocity. Increased turbulence was investigated by Porter (1981) who found that power dependency on flow velocity increased from 0.33 to 0.7 - 0.8 when strips were positioned in the membrane channel. Another method of reducing fouling is use of electricity in inorganic membranes of metallic structure. Wakemanet al (1986) reported enhanced flux when a DC electric field was applied across the^membrane. Pulsing current also restored the flux by dislodging fouling layer from the membrane surface. Workers at UKAEA Harwell ( 1989 ) have also patented an electrolytic filter cleaning technique which makes use of pulse of electric current applied across a conductive steel
29 membrane. The gas bubbles generated on the inside of the membrane surface A claimed to float off the fouling layer.
Other techniques for slowing down the rate at which membrane fouling occur are use of chemicals, such as flocculants and surfactants. It has been reported that correct adjustment of feed pH and feed pretreatment can reduce fouling. Bedwellet al (1988 ) reduced the fouling by increasing the feed pH and addition of chemicals to promote flocculation. Membrane pretreatment with surfactants or polymers are other methods which could reduce fouling. These chemicals increase the homogeneity and hydrophilicity of the surface, allowing water to contact the membrane preferentially. Back-washing particularly with air is also used for flux restoration.
30 CHAPTER THREE
General Theory of Microfiltration
3.1 Introduction
Most of the theoretical developments and models which are reported for organic ultrafiltration membranes are fundamentally applicable to inorganic microfilters. In this Chapter the basic theory of MF is presented and equations relating flux to the properties of flow, feed and membrane are quoted. The fouling mechanism is discussed in detail and a number of mathematical models relevant to this investigation are presented.
3.2 Fundamental Equations for Flow Through Porous Mass
Microfiltration membranes and substantial fouling layers may be considered as porous media. Flow through such structures are investigated extensively in many chemical engineering fields. The most fundamental equation relating the flowrate to the properties of a porous medium is the Hagen - Poiseuille model
f _ edp APt (3 .1 ) 3 2 A X f i
f = permeate flowrate, m3 nr2 s_1 e = porosity dp = pore diameter, m AP^ = applied transmembrane pressure, Pa
[L = viscosity of permeating fluid, Kg m^s*1
31 AX = thickness of membrane or porous medium, m
Equation ( 3.1) assumes that all pores are circular cylinders and flow through them is laminar. In reality porous media have a more complicated structure. Flow of solvent through such structures can also be given by the Carman - Kozeny relationship
f = AP, (3 .2 ) 1 8 0AX ( 1 - 6 ) 2 |x
where d0 = diameter of particle, m or AP, f = (3.3)
where Rm = hydraulic resistance of medium
( or membrane), Kgm-2 s-1
-1 f dpa 1 £3 ^180 AX (1 -e)2 [i
Rm is related to permeability and thickness of the membrane by
Rm = £ ...... (3.4) k where h = membrane thickness, m
k = membrane permeability, m 2 Pa_1 s _1
In the presence of a fouling layer, the membrane resistance is altered and resistance due to the fouling layer has to be taken into account. In such a case the resistance due to the
32 fouling layer is added to the hydraulic resistance of membrane. Thus flux is given by
APt f = (3.5) Rm + Rf
where Rf = fouling layer resistance, Kg m• 2 s -1
In the following sections specific relationships are discussed in more detail.
3.3 Crossflow Microfiltration Processes
Crossflow microfiltration processes may be represented schematically by Figure 3.1. A pressurised feed stream is brought to the membrane where particles and macromolecules smaller than the pore size of the membrane pass to the permeate side, while those larger than the pore size are transported out of the membrane as the concentrate stream. Large particles also tend to concentrate near the membrane surface or precipitate on the membrane itself forming a gelatinous layer or a fouling film. These phenomena are known as concentration polarisation, gel polarisation and fouling. These invariably result in a drop in the permeation rate and in tubular membranes under certain conditions cause the eventual blockage of the tube. Concentration polarisation is of minor interest here since, it occurs where only low molecular weight solutes are present in the feed and where the membrane is a low permeability one. These situations mostly occur in reverse osmosis membranes. Gel polarisation is a more realistic situation which occurs in UF and MF membranes. Fouling, on the other hand, is a more time dependent process and usually occurs when a highly particulate feed is used and where the membrane is a high permeability one. The following sections discuss these occurrences in detail.
33 3.3.1 Gel Polarisation - Formation of Gel Polarised Layer
Particles and molecules are transported to the membrane surface through the boundary layer adjacent to the membrane, while the rejected and accumulated species are transported back to the bulk fluid by means of diffusion through the boundary layer. At some point of the process the convective transport to the membrane and back diffusion from it reach an equilibrium ( concentration polarisation). If the equilibrium between the convective transport to the membrane and back diffusion from it no longer exists then trapped the larger particles tend to beK in the boundary layer, forming a gelatinous or cake layer on the membrane surface as shown in Figure 3.2. At this stage increasing the transmembrane pressure can only result in transport of more solute to the membrane without additional back diffusion from the membrane to the bulk solution. This results in more precipitation and thickening of the fouling layer and increases the resistance to the permeate flow. Thus, the flux is no longer affected by the pressure and is controlled by the hydrodynamic properties of the flow and the fouling layer thickness. Figure 3.2 shows the flux variation with transmembrane pressure over the two different controlling regions. In the mass transferred control region flux may be estimated using Equation ( a) which is known as gel polarisation model. The model has been verified experimentally by a number of investigators including Porter ( 1972 ) and Schweitzer (1979 ). Using experimental values, according to Equation(a), a linear plot of f against In C should be a straight line with the gradient as the mass transfer coefficient. Mass transfer coefficients may also be estimated by the usual mass transfer correlations relating Sh with Re and Sc numbers for the appropriate flow regime. More detail on the values of the constants may be found in^Ultrafiltration Handbook" (Cheryan, 1986).
Other models which may be used to estimate the flux of a gel polarised membrane are s modified filtration or resistance modelAand the osmotic model. Their detai^treatment are given in the literature ( Gekas, 1987; Blatt, 1976 and Cheryan, 1986).
34 PI V P2
concentrate stream
PI = inlet pressure P2 = outlet perssure P3 = permeate perssure
Pi+P2 APt = - L - ^ - P 3= P 1 + P2 2 V = crossflow velocity
Figure 3.1 Schematic Representation of Crossflow Microfiltration.
35 gel or cake layer
f=-5 h f - ^ l
where — = k = mass transfer coefficient x
/
Figure 3.2 Formation of Gel Polarisation Layer in the Mass Transfer Controlled Region.
36 3.6 Mathematical Models for Fouling
During filtration of a suspension particulate matter deposit on the membrane surface in the form of a cake layer. This causes the flux to decline drastically as the filtration progresses. Fouling occurs if the flux decline is not reversible when the operating conditions, such as crossflow velocity and transmembrane pressure are changed. Build-up of a fouling film would be favoured under conditions of laminar flow and high solids content in the feed side, and for membranes with a high surface permeability. Mathematical models to predict the flux of fouled membranes are widely reported. Most of these models are for RO and UF membranes. In the following sections different approaches to modelling of fouling mechanism are presented.
3.6.1 Semi - empirical Models 3.6.1.1 Exponential Relationship
These models relate the fouled flux to the initial flux by an exponential relationship
f = f 0t‘n ...... (3.6)
where f = fouled flux
f 0 = initial flux t = time n = exponent which varies with crossflow velocity ( also known as fouling constant)
The n values determine the extent of fouling, with higher values representing higher fouling rates. This model predicts that f finally approaches zero which is not always the case in practice. Thomas et al ( 1972 ) and Probsteinet al ( 1981 ) are some of the workers who have used similar types of exponential relationships to explain their experimental results.
37 3.6.1.2 Resistance in Series
A popular method for modelling of fouling is to use the resistance in series approach. Here the fouled flux is related to the transmembrane pressure and various resistance contributions. Thus,
APt f = (3.7) Rm + Rf+^b+ •••
where Rf = fouling or deposit resistance Rb = boundary layer resistance and other symbols are as before.
This approach is used by a number of workers including Gutman ( 1977 ), Sukiet al (1984), Fane et al (1980) and Lundqvistet al (1986).
3.6.2 Mechanistic Models
Apart from these semi-empirical models which are based on the experimental data, there are other models which use a more mechanistic approach to estimate the flux of fouled membranes. In a typical model development it is assumed that particles move out of the bulk stream convectively towards the membrane surface. There is also a lateral migration of particles away from the wall to the bulk. The later is a function of crossflow velocity. Particles which reach the membrane gradually build up a cake which causes flux to decline. The fouling layer decreases the cross sectional area of the channel and increases the cross flow velocity. At some stage convective particle movement and the lateral migration reach an equilibrium where the fouling film thickness no longer increases and the flux remains constant. An Example of this kind of model which was recently
38 published is due to Schock ( 1986 ) and Rautenbachet al ( 1987 ). This model is specifically derived for microfiltration of suspensions where tubular microfilters are used and will be looked at in more detail.
3.6.2.1 Hydrodynamic Resegregation Model
In development of this model it is assumed that the particle is carried from the bulk fluid "VK«. towards the membrane surface by the permeate flowing inradial direction. There is also a resistance force acting on the particle due to the feed flow in the laminar boundary layer in parallel to the membrane surface. The mean velocity acting on the particle in the direction of flow will be proportional to the distance of the centre of the particle from the membrane surface. Therefore, as the particle size decreases, the effective velocity parallel to the membrane will increase, while the radial velocity remains constant. Therefore, particles of larger size will move towards the centre of the channel into the bulk flow. This behaviour is known as tubular pinch effect and is utilised in development of this model. A model combining the tubular pinch effect and the back-diffusion mass transfer for estimation of the flux is suggested by Riesmeier (1987). That is
+
diffusive term
pinch effect where U = mean crossflow velocity rp = particle radius R = tube radius r = radial position of the particle in the tube
39 r * = equilibrium radial position of the particle a = constant b = constant
The back-diffusive mass transfer of particles away from the membrane wall may be neglected due to the particle sizes dealt with in microfiltration ( as shown graphically in Figure 3.3 ). Thus, the flux becomes a function of the hydrodynamic properties of the flow. Schock ( 1986 ) used this assumption to derive a model which is of the general form
0.44 f = C Re126 [2 - (3.9)
where f = flux Re = Reynolds Number v = kinematic viscosity d = channel diameter dp = particle diameter C = constant
C was found to vary according to the state of the flow. The C values are
C = 0.02 for laminar flow C = 0.22 for laminar and turbulent flow and C = 1.25 for turbulent flow
Equation ( 3.9 ) is used in Chapter Five to correlate the experimental results in this work.
40 1000
Figure 3.3 Influence of Hydrodynamic and Diffusive Controlling Effects as a Function of Particle Size.
3.6.3 Aggregation Model
The above approach, however, ignores the physico-chemical effects, such as flocculation and aging which may greatly influence the behaviour of the fouling layer especially in colloidal suspensions. A model which considers these effects and takes into account the solute - solute interaction in the polarised layer makes use of the following approach. The film gradually builds up on the membrane surface where solute and colloidal species flocculate and aggregate, increasing the film thickness layer by layer. Progression of flocculation results in a decrease in permeability of the layer and reduction in the flux. The fouling layer thickness is assumed to increase in increments of x where the average voidage of each layer may be estimated from the aggregate size distribution if the electrokinetic parameters of the sample is known ( Sukiet a l, 1986).
41 For each layer increment its corresponding resistance is found from the Carman - Kozeny relationship and added up to represent the total resistance of the fouling layer. Thus, for a boundary layer subdivided into k sections, the flux given by
APt f(t)= (3.10) k ^ ^d, x x = 1
where x = increments k = boundary layer division Rd. x = resistance of each layer Rh = resistance of clean membrane
The model predicts the densest layer forms adjacent to the membrane, which is also observed experimentally. According to the model, increase in feed concentration and gel polarisation are the two factors which result in increase in the flux decline.
3.7 Retention Properties of Membranes
Separation property of MF and UF membranes is expressed in terms of their rejection coefficient which is defined as
R = l - ^ ...... (3 .1 1 ) where R = membrane rejection coefficient for a given component Cp = concentration of the rejected component in permeate Q = concentration of the rejected component in retentate
Other terms which give some idea of the membrane separation characteristics are the
42 membrane recovery rate and fractional product loss through the membrane. The recovery rate is the ratio of the permeate volume ( the product) to the feed volume. Thus,
' A = ^ E ...... (3.12) V 0 where A = recovery rate Vp = volume of filtrate V q = volume of feed
The fractional product loss is a measure of solute lost in the permeate. The fractional product loss is expressed in terms of R and A, i.e.,
8 = 1-(1 - A )Cl'R)...... (3 .1 3 )
where 8 = fractional product loss
In examining the separation properties of membranes when radioactive feeds are processed, it is convenient to express the separation capability of the membrane in terms of the ratio of concentration of radionuclides in the feed to that of the permeate. This ratio is known as decontamination factor. Therefore, decontamination factor or DF is defined as
D F = activity in the feed (3.14 ) activity in the permeate
In radioactive runs carried out in this work, DF values are evaluated to indicate the effectiveness of various membranes in decontaminating the radioactive feeds.
43 CHAPTER FOUR
Experimental
4.1 Summary
The first group of experiments are intended to examine the controlling mechanism that prevails during microfiltration of a highly fouling feed of magnesia suspension in water. The steady state behaviour of membranes are investigated with respect to variables such as crossflow velocity, transmembrane pressure, feed concentration and the membrane type. In these experiments PALL sintered stainless steel, DOULTON ceramic and FAIREY stainless steel membranes are used.
In the second series of experiments radioactive retention properties of a number membranes are examined using radioisotopes such as Am-241, Eu-154, Co-60, Sr-85 and Th-230. The membranes used are MILLIPORE filters, PALL, DOULTON and M4 and M6 CARBOSEP tubular membranes.
4.2 Apparatus 4.2.1 Flow Rig
The flow rig is illustrated in Figure 4.1. Slurry is pumped round the loop using a 1" Jabsco positive displacement pump ( No. 1 ) with a flexible self lubricating rubber impeller. The discharge from the pump passes through a three way valve where a portion of it is discharged via a bypass line ( No. 2 ) to the feed tank ( No. 3 ). The slurry passing the loop first goes through a Danfoss electromagnetic flow meter ( No. 4)
44 and then through the microfiltration module ( No. 5 ). There are two electronic pressure transducers ( No.6 ) mounted in this section of the apparatus to measure the pressure drop across the microfiltration module. Just downstream of the latter is a valve to maintain a positive pressure in the apparatus. The permeate flows through an electronic turbine meter ( No. 7) which measures the permeate production rate. As there is a finite pressure drop across the turbine meter there is a transducer upstream so that the transmembrane pressure can be measured accurately ( No.8 ). The fluid discharged from the flowmeter is passed into a sump ( No. 9) and is pumped back to the feed tank. Water or acid to back flush the microfiltration module is put into the sump ( No. 10) and is passed through a three way valve into the microfiltration module. The feed temperature is maintained at 20 °C by means of a commercial cooler ( No. 11 ). The voltage outputs (No. 12 ) from all three pressure transducers, magnetic flow meter and permeate turbine meter are fed via an ADC into a microcomputer ( No. 13 ). The computer converts all the signals into pressures and flowrates and stores them as disk files. Plate 1 shows the photograph of the microfiltration rig.
4.2.2 Filtration Module
Filter module consists of a tubular jacket which is designed to hold a single length of tubular membrane. The permeate collected from the microfilter flows out of the jacket through an opening welded on the wall of the jacket. Due to the variation of the membranes' geometry two different modules are designed . Figures 4.2 and 4.3 show their detailed drawings . The former is used for the PALL and DOULTON membranes while the latter is designed for the FAIREY microfilter.
45 10
5
1 Feed Pump 7 Turbine Meter
2 Bypass 8 Permeate Collection Sump 3 Feed Tank 9 Back Wash Sump 4 Magnetic Flowmeter 10 Signal Outputs 5 Filtration Module 11 Interface, Computer & Printer
6 Pressure Transducers 12 Temperature Controller
Figure 4.1 Flow Diagram of Microfiltration Rig.
46 Plate 1 Photograph of the Microfiltration Rig Used in the Fouling Experiments Adjustment Fitting
Figure 4.2 Detailed Drawing of Filtration Module Used for PALL and DOULTON Membranes. 0 1 = d i 0
Figure 4.3 Detailed Drawing of Filtration Module Used for FAIREY Membrane. ( All Dimensions in m m ) 4.2.3 Membranes 4.2.3.1 PALL Sintered Stainless Steel
This is an experimental product made of sintered stainless steel and has an equivalent pore size of 2.5 pm. The membrane has a uniform composition with an i.d. of 10 mm and a thickness of about 1.5 mm. Due to the uniform nature of the matrix considerable abrasion of the inside can take place without loss of microfiltration properties. The tubes are made by PALL and like other sintered metal products are tough and impact resistant.
4.2.3.2 DOULTON Ceramic
This type of membrane is made of graded alumina-silicate particles which are bound by a glass matrix forming a rigid and highly porous ceramic structure. This type of membrane is of composite type with two differing pore size zones. The fine surface zone typically has a pore size of 1.0 pm with a base pore size of 7.0 pm. The assembly is very strong and has the excellent abrasion resistance displayed by alumina based ceramics.
4.2.3.3 FAIREY Stainless Steel
The FAIREY stainless steel membrane comprises a surface layer of very fine powder ■ fiic which is sintered onto a porous substrate, therefore like^ DOULTON t^be, the membrane is composite in structure. The membrane is about 0.4 mm thick with typical pore size of 1.0 pm. This type of tubular membrane comes as a complete unit with fittings on either end.
All these membranes are capable of operating under reverse flow or back washing conditions up to 3-4 bars, whilst internal operating pressure can be as high as 10-15
50 bars. They are stable in all pH ranges and may also be steam sterilised. Table 4.1
»s the characteristics of the membranes used in the experiments. Plate2 shows photograph of the membranes (^CARBOSEP membrane wasused in the active experiments and will be discussed later).
Diameter Membrane Pore size Length Internal External (Hm) (m ) (mm) (mm)
PALL 2.5 0.27 1 0 . 0 1 2 . 0
DOULTON 1 .0 0.30 1 0 . 0 2 2 . 0
FAIREY 1 . 0 0.27 1 0 . 0 13.0
Table 4.1 Characteristics of the Membranes Used in Fouling Experiments. Plate 2 Photograph of the Tubular Membranes Used in the Experiments. (1 - PALL; 2 - DOULTON; 3 - FAIREY; 4 - CARBOSEP ) 4.3 Feed and its Characteristics
The slurry used in the experiments is a commercial brand of magnesia. Its particle size distribution was determined using a Malvern laser size analyser. In Figure 4.4 the distribution is given. The rheolgical properties of the slurry such as the viscosity and yield stress are determined for different solid concentrations using a capillary tube viscometer. Figure 4.5 shows the shear stress vs. shear rate behaviour of some of these runs. It is found that up to a concentration of 15% W/W the slurry behaves as a Newtonian fluid. At higher concentrations the slurry is found to be Bingham plastic in nature. Variation of the viscosity with the slurry concentration is shown in Figure 4.6. The solid content of different slurry concentrations are also found and plotted against density ( Figure 4.7 ).
4.4 Experiments 4.4.1 Determination of the Hydraulic Permeabilities of the Membranes
Hydraulic or clear water permeability of the membranes are required in order to estimate their resistance. Such parameters are usually quoted for membranes at a certain pressure and temperature, although they were not available for the membranes used here. The microfiltration rig, discussed earlier, is used with the appropriate filter in position to carry out runs using deionised water at 20 °C. The water is circulated through the filtration loop and permeate flux is measured at a transmembrane pressure of 30.0 KPa. The flux-pressure data gathered this way are used to estimate the hydraulic resistance of each membrane.
4.4.2 Preliminary Fouling Experiments
It is a well known fact that permeate flux of a membrane declines as the service time increases. These experiments are carried out to determine to what extent the flux changes
53 after the initial decline. The microfiltration rig is operated at constant transmembrane pressure and crossflow velocity. Different concentrations of magnesia slurry are circulated through the loop in each run for a period of two days. All three membranes mentioned earlier are used in these tests. Flux measurements for all feed concentrations show that after the initial decline, only an average of about 5 to 10 % flux decline occurs from 60 minutes to 48 hours of operation in all cases. Therefore, it is decided that an experimental run time of approximately 90 minutes is long enough to produce an indicative and reliable steady state flux.
Particle Size, pm
Figure 4.4 Particle Size Distribution of Magnesia Suspension.
54 40
Shear Rate, s-1
Figure 4.5 Shear Stress Against Shear Rate for Different Concentration of Magnesia Suspension.
Vi 03 CU
Vi V© Vi >
0 10 20 30
Concentration^ WAV
Figure 4.6 Variation of Viscosity of Magnesia Suspension with its Solid Concentration.
55 1300
Figure 4.7 Variation of Density of Magnesia Suspension with its Solid Concentration.
4.4.3 Experiments to Investigate Variation of Flux with Time and the Operating Conditions 4.4.3.1 Variation of Flux with Time
With the appropriate membrane sealed in the module the MF rig is used to carry out the following experiments. The experimental procedure for all three membranes is similar. The rig is operated in a closed-loop batch mode where the feed and resulting permeate are continuously recycled back to the feed tank to maintain a constant concentration. Each run lasted some 90 minutes during which the variation of the permeate flux with was were time A measured. Other parameters of interest monitored during the experiments A the feed flowrate, pressure at the inlet and the outlet of the membrane module and pressure on the permeate side. Between every run the membranes are back washed with 2M nitric
56 acid and then with distilled water for a few times to remove any fouling layer and restore the original membranes' permeabilities. Temperature of the feed is maintained at 20 °C o£- in all the runs. For a numberAfeed concentrations separate runs are made at various crossflow velocities and transmembrane pressures.
1. Effect of Crossflow Velocity
In order to find out the influence of crossflow velocity on the flux, fouling experiments are carried out where the transmembrane pressure is kept constant while the crossflow velocity is varied. In Tables A1 and A2 ( Appendix A ) details of the operating conditions are presented.
2. Effect of Transmembrane Pressure
Effect of pressure on the flux is examined in another set of experiments where the transmembrane pressure is varied while the crossflow velocity is kept constant. Tables
Bl, B2 and B3 ( Appendix B ) show a summary of the runs, their corresponding operating conditions and feed concentrations of these experiments.
4.5 Radioactive Retention Experiments 4.5.1 MILLIPORE Tests
The aim of this series of experiments is to determine decontamination factors ( DF) for a number of elements of interest using a laboratory " dead-end " microfilter and compare the results with a typical crossflow filter. The variation in DF is examined with respect to changes in the membrane pore size and changes in conditions of the feed solution ( pH and presence of fouling agent). MILLIPORE tests are carried out to yield order of
57 magnitude estimates of DFs.
4.5.1.1 Feed Preparation and Experimental Technique
The feed is labelled using a small sample of corroded Magnox sludge which was originally stored in concrete silos at Sellafield. Such a sample contains a variety of radionuclides of a and y emitting species. Three series of experiments are performed. In each experiment, an aliquot of active Magnox sludge is taken and diluted with freshly distilled water. The solution is then spiked with Sr-85 of known activity and is air-sparged for 24 hours. Spiking with Sr-85 obviates the need to determine Sr-90 DFs which involves time consuming methods of B-counting. The MILLIPORE filter is a circular flat sheet membrane which is placed in a plastic filtration module fed by a hypodermic pressure baiTel. 15 ml solutions are analysed for radionuclide contents prior and after filtration using a Gamma-X detector. The energies of chosen to represent the five radionuclide of interest are :
Am-241 at 58.850 Kev Eu-154 at 122.550 Kev Sb-125 at 428.808 Kev Sr-85 at 514.650 Kev Cs-137 at 662.420 Kev
The following experiments are earned out:
1. The first series of experiments are performed by diluting 1 ml of Sellafield Magnox sludge with 250 ml of distilled water. After the air sparging period of 24 hours, this solution has attained a pH of 8.70. Aliquots of this solution are then treated with MILLIPORE membranes of various pore diameters (0.1,0.45 and 0.8 pm ). A portion of this solution is then pH-adjusted to 11.50 with the aid of
58 1M NaOH. Aliquots of this high pH solution are then subjected to MILLIPORE filtration.
2. The second series of experiments are carried out by diluting 1 ml of Sellafield Magnox sludge with 250 ml of distilled water. After the air sparging for a period of 24 hours, 25 g of magnesium hydroxide is added to this solution ( 10% W/ V). This mixture is then air - sparged for a further period of 4 hours. A pH of 10.10 is recorded after sparging. Aliquots of this mixture is then treated with MILLIPORE membranes of similar pore diameters. A portion of this mixture is then pH adjusted to 11.50 with the aid of 1 M NaOH. Aliquots of this mixture are also MILLIPORE filtered.
3. The third series of experiments performed are similar to the first series with the dilution of 500 pi of Sellafield Magnox sludge with 200 ml of distilled water. A pH of 8.0 is recorded after the air-sparging operation of 24 hours. A portion of this solution is pH adjusted to 11.50. In this series of experiments, DF's for Am-241 and Pu-239 are found. Am analyses are performed by Gamma-
spectroscopy, while Pu analyses are carried out by a-detection subsequent to TTA extraction.
4.5.2 Retention Characteristics of M4 CARBOSEP Membrane
A number of experiments are performed on a single ultrafiltration CARBOSEP tube to investigate the nuclide retention properties of a crossflow membrane. Performance of M4 membrane is examined as a function of pH of the feed being processed for a number of radionuclides of interest, namely, Th-230, Co-60, Eu-152 and U-235. In another set of experiments, the variation of permeate flux with time and build-up of activity on the membrane surface are examined at the pH of interest
59 4.5.2.1 Apparatus and Membrane Used
A laboratory filtration module is designed and set up in a fume cupboard for the use in active experiments. Figure 4.8 shows a flow diagram of the rig. Feed is pumped through a variable flow peristaltic pump to the membrane module. Pressure across the membrane is controlled by means of a needle valve at the outlet of the module. The outlet from the membrane tube is either be recycled direcdy to the feed tank or to the drains. A single 14 cm long M4 CARBOSEP membrane portion is used in all the experiments. M4 CARBOSEP is a commercially available inorganic membrane with a sub-micron pore size. The membrane is a thin layer of zirconium oxide supported on a matrix of porous carbon. This type of membrane has found application in treating non-abrasive slurries such as dairy products. Like other membranes mentioned earlier CARBOSEP membrane is stable in all pH ranges and can stand pressures and temperatures of up to 25 bars and several hundred degrees respectively. CARBOSEP membrane is shown in Plate 2.
4.5.2.2 Experiments 1. Effect of pH on Radionuclide Decontamination
In these runs the removal of radioactive species by CARBOSEP tube as a function of the solution pH being processed is examined. The feed solution in each run is one litre of 1000 ppm nitrate solution of the chemical element of interest. Those studied are thorium, cobalt, europium and uranium. With the exception of the thorium solution all the solutions are spiked with a radioisotope of the element; Co-60, Eu-152 and U-235. The composition of the feed solution are as follows:
Th-230 4.08 Bq/ml Co-60 11.44 Bq/ml Eu-152 18.81 Bq/ml U-235 17.38 Bq/ml
60 The feed pH is increased stepwise during each run using 5M NaOH. Total reflux is employed where the outlet is recycled back to the feed tank throughout the experiments. Pressure across the membrane is kept around 1.25 bar and the feed flow rate is maintained at 0.1 lit/min. At each pH samples of feed and permeate are collected. Radionuclide contents of the samples ( except for Th-230 ) are measured by y-spectroscopy. Samples from the thorium run arc treated with Arscnazoin before being analysed spectrophotometrically at 660 nm. As active solution is being filtered and the volume of this solution is changing the feed activities in some of the runs is altered due to the retention of the active floe in the circuit. The uranium and cobalt experiments took a long time so feed activity changed quite a lot
2. Runs at Constant pH
In these experiments nuclide retention of the membrane is examined at constant pH conditions. The pH of the solution is held at 11.5 and the solution is continuously recycled through the apparatus. Throughout the experiments samples of feed and permeate are collected fory - analyses as before and the permeate flux is also measured. Particular attention in these run are devoted to studying changes in the permeate flowrate and activity. Two runs are earned out One with europium only and one with cobalt and europium present.
61 Concentrate Needle Valve
Figure 4.8 Flow Diagram of Microfiltration Rig Used in Active Experiments.
62 4.5.3 Radioactive Retention Properties of PALL, DOULTON and CARBOSEP Membranes
In these experiments radioactive retention of a number of tubular membranes including
PALL sintered stainless steel, DOULTON ceramic, M4 and 6 M CARBOSEP membranes are examined in the presence of a fouling agent. Magnesia suspension in water is used to investigate the effect of fouling layer on the decontamination of the membranes.
4.5.3.1 Apparatus and Feed
A filtration rig similar to that described in section 4.5.2.2 is used in all the runs. Due to varying diameters of the membrane tubes separate modules had to be made for each membrane. Table 4.3 shows characteristics of the membranes. In all the experiments one litre feed suspension is used. The following procedure is used in preparation of the suspension. An aliquot of corroded Magnox metal is dissolved in 2M nitric acid and sufficient volume of this solution is added to 1% WAV suspension of magnesium hydroxide in distilled water. This suspension is spiked with Sr-85 of known activity and stirred for 2 hours to form a uniform feed suspension. The presence of magnesium hydroxide in the feed keeps the feed pH around 10.0 and ensures the precipitation of most of the active species.
4.5.3.2 Experiments
With each membrane in position the feed suspension is circulated in the rig for a period of 5 hours during which samples of feed and permeates are collected and analysed for
Am-241, Eu-154, Sr-85 and Cs-137. In addition to the y - emitting radionuclei, analyses are made for Pu-239 in both feed and permeates. These samples are analysed
63 a-spectroscopically.
Membrane Length (cm) Hydraulic Resistance
( KPa m 2 hr lit-1)
PALL 16.0 0.14 DOULTON 11.3 0.217 M4 18.5 1.65
M6 18.5 0.82
Table 4.2 Characteristics of the Membranes used in the Active Experiments.
64 CHAPTER FIVE
Results and Discussion
5.1 Summary
In the first part of this Chapter results from the fouling experiments are presented and discussed. Data from these experiments are used to deduce the controlling mechanism occurring in microfiltration of highly fouling suspension. A model is derived based on the resistance in series approach which relates the steady state flux to cross flow velocity, transmembrane pressure and feed concentration. An Equation is also derived c\n to estimate the transient filtration period which givesAestimate of time before the steady state is reached. Application of other fouling models to the experimental results is considered and discussed. Results from the radioactive retention experiments are also used to evaluate and compare the performance of the membranes. Decontamination factors show that the membranes are quite effective in removing colloidal radioisotopes regardless of the type of the membrane used. Formation of the fouling layer does; not seem to increase the DFs.
It should be pointed out that a number of parameters with dimensions other than S.I. are introduced here. This is unavoidable since, for instance, the flux which is expressed in litres of permeate per square meter of the membrane area per hour of the operational time ( lit nr 2 hr _1 ) gives rise to a number of strange units once it is introduced in a relationship. Moreover, expressing flux in lit m■ 2 hr _l is more recognisable unit than m 3 m ■ 2 s -1 since the latter unit usually results in very small values . ‘
65 5.2 Results 5.2.2 Hydraulic Resistance of the Membranes
Hydraulic resistance of the membranes, ^m» may be found from the experimental data and Equation (3.3). Rm and clean flux values of the membranes under investigation are given in Table 5.1 below.
Membrane Rm ( KPa hr 1 -1 m 2 ) fo* ( lit m- 2hr _1 )
PALL 0.140 714.3 DOULTON 0.217 460.8 FAIREY 0 . 1 2 0 833.3
Table 5.1 Hydraulic resistance and clean water flux of the membranes. * at APt = 100 KPa
5.2.3 Fouling Experiments 5.2.3.1 Variation of Flux with Time and Crossflow Velocity at Constant Pressure
Table 5.2 shows the initial flux, flux after one hour of the start up, and the final flux after two days of the operation in the preliminary fouling runs. These results show that an operational run time of approximately 90 minutes produces a reasonable flux where it may be assumed that the plateau corresponding to the steady state flux value is reached. Figures 5.1 - 5.8 show the flux-time profiles of the main fouling experiments, with the variables in each run being the crossflow velocity. In general, the flux declines sharply within the first few minute or so of the runs. This period gets shorter as the slurry concentration increases. These Figures also show that the steady state flux increases
66 with the crossflow velocity and decreases with the feed concentration.
Membrane Feed Cone. V APt fQ fi hr 2 f days
% W/W ms" 1 KPa (lit m- 2 hr "1 ) (lit nr2 hr_l ) (lit m" 2 hr -1 )
PALL 2.25 1.34 76.17 544.1 285.1 276.3
it 18.5 1 . 6 6 94.22 673.0 130.2 1 2 1 . 2 DOULTON 6.25 0.94 112.52 518.5 157.5 133.5 M 17.5 0.91 50.0 230.4 69.6 60.0
FAIREY 2 . 0 1.7 1 0 0 . 0 833.3 500.0 489.1 II 13.0 1.5 62.0 516.7 175.5 161.6
Table 5.2 Initial, 1 Hour and 2 Days Flux of Runs with Magnesia Suspension at Various Feed Concentrations, Crossflow Velocities and Transmembrane Pressures.
67 500
u XS V=2.17m/s V=1.88 m/s B V=1.66 m/s V=1.34 m/s
Time, minutes
Figure 5.1 Variation of Flux with Time and Crossflow Velocity at Constant Transmembrane Pressure of 64.0 KPa and Feed Concentration of 6.0% W/W for PALL Membrane.
Figure 5.2 Variation of Flux with Time and Crossflow Velocity at Constant Transmembrane Pressure of 76.0 KPa and Feed Concentration of 6.0% W/W for PALL Membrane.
68 Figure 5.3 Variation of Flux with Time and Crossflow Velocity at Constant Transmembrane Pressure of 94.0 KPa and Feed Concentration of 18.5% WAV for PALL Membrane.
Figure 5.4 Variation of Flux with Time and Crossflow Velocity at Constant Transmembrane Pressure of 60.0 KPa and Feed Concentration of 28.5 % WAV forPALL Membrane.
69 250 1 V=1.02 m/s ♦ V=.943 m/s 200 - % ■ V=.480 m/s u • V=.307 m/s JS 150 . . n ■ i °• B 9 B %D ■ «♦ ■ ♦ Q B ♦ .ti 100
50 - ♦ ♦ ♦ • ♦ •
I | I I | I 1 ' I !■ 20 40 60 80 100 Time, minutes
Figure 5.5 Variation of Flux with Time and Crossflow Velocity at Constant Transmembrane Pressure of 60.0 KPa and Feed Concentration of 6.25% W/W for DOULTON Membrane.
Figure 5.6 Variation of Flux with Time and Crossflow Velocity at Constant Transmembrane Pressure of 100.0 KPa and Feed Concentration of 12.0% W/W for DOULTON Membrane.
70 Figure 5.7 Variation of Flux with Time and Crossflow Velocity at Constant Transmembrane Pressure of 50.0 KPa and Feed Concentration of 17.5 % W/W for DOULTON Membrane.
Figure 5.8 Variation of Flux with Time and Crossflow Velocity at Constant Transmembrane Pressure of « 60.0 KPa and Feed Concentration of 13.0% W/W for FAIREY Membrane.
71 5.2.3.2 Variation of Flux Time and Transmembrane Pressure at Constant Velocity
Apart from time, the main operating variable of interest here is transmembrane pressure,
APt. The results from these experiments show that the flux drops to a steady state value after a few minutes. Increase in AP(, however, increases the rate of decline of flux. In other words, the transient filtration period becomes shorter at higher driving pressures.
Influence of the rise in AP^ on the steady state flux is only noticeable at low transmembrane pressures and in few cases the steady state flux changes with APt. The results suggest that, in general, above a certain pressure the flux is no longer affected by pressure and the flux-time profiles of the membranes fall on the same value as shown in Figures 5.9 and 5.10 for two of such cases. Complete results of these experiments with their corresponding operating and feed conditions are given in Tables Bl, B2 and B3 (Appendix B ).
5.2.3.3 Further Investigation of the Variation of Flux with the Operating Variables
The steady flux values from the previous sections are plotted against their corresponding V and APt for various membranes. They are shown in Figures 5.13 - 5.19. The flux varies very little with the pressure especially at higher feed concentrations. That is, once a certain pressure is reached the flux is no longer affected by any further increase in the driving pressure across the membrane. These results indicate that the microfiltration is predominantly carried out beyond the pressure controlled region and is governed by the hydrodynamic properties of the flow. Therefore, the main factor affecting the flux is the crossflow velocity. As shown in the same Figures, the higher the crossflow velocity the higher the steady flux will be. It should be noted that higher feed concentrations results in lower fluxes.
72 Figure 5.9 Variation of Flux with Time and Transmembrane Pressure at Constant Crossflow Velocity of 1.343 m s"l and Feed Concentration of 6.0 % W/W for PALL Membrane.
Figure 5.10 Variation of Flux with Time and Transmembrane Pressure at Constant Crossflow Velocity of 1.02 m s’* and Feed Concentration of 12.0% W/W for Doulton Membrane.
73 5.3 Model Development
This model is based on the resistance in series approach discussed in Chapter Three. In general the flux based on strict S.I. units is
( 3 .7 ) Rm + R f
where fs = flux of a fouled membrane, m3 m"2 s _1
Rm = membrane resistance, Kg m"2 s
R^ = fouling layer resistance, Kg m' 2 s _1 APt = transmembrane pressure, Pa assuming that the boundary layer resistance, , is negligible.
For a fouled membrane it may be assumed that the rate of build up of the fouling layer reaches a steady value where the accumulation rate of the particles on the membrane surface becomes equal to their removal rate. Under such conditions the flux reaches a limiting value which is known as the steady state flux. In most practical applications of the membrane separation the operational period where this phenomenon occurs is long enough and such value may be used for design purposes. Referring to Equation (3.7 ),
Rm is found from water run with the appropriate membrane. Rf, however, is expected to be dependent on the hydrodynamic properties of the flow such as, V and APt and to some extent concentration of the feed. Such dependency may further be investigated by dimensional analysis which is presented in the next section.
74 5.3.1 Derivation of a Relationship Between Rf and Hydrodynamic Properties of Flow
Properties of the fouling layer and ultimately its resistance, Rf, is expected to be
dependent on the following variables:
Rf = f ( V, APt ,p ,p )...... (5.1)
where V = crossflow velocity, m s "1 |x = feed viscosity, Pa s
p = feed density, Kg m■ 3
\
By dimensional analysis
Rf = constant Va, APt^ , pc, p ^...... ( 5.2 )
and expressing Equation (5. 2) in the three fundamental units of mass, length and time
ML-2T-1 = constantLaT"a . Mb L'b T-2b .
•M d L-3d or
ML"2 T ' 1 = constant La' b-C‘3d • T'a‘2 b-C • Mb+c + d .
• • • ( 5.3 )
equating the two sides of Equation (5.3 )
M: b+ c+ d= 1
L i -a + b + c + 3d —2 T : a + 2b + c = l
75 and evaluating a, b and c in terms of d
a = 2 d -1 ; b = l-d ; c = 0 therefore, Equation (5.1) becomes
Rf = constant ( V2 d- 1 APt 1- d pd ) # . (5.4) .
or
Rf V . (5.5)
The groups ( Rf V / APt) and ( P / APt) are dimensionless. In order to keep dimensional consistency, " constant " and "d" in Equation ( 5.5 ) should also remain dimensionless.
The data for PALL membrane is used to correlate the two dimensionless groups found above. Figure 5.11 shows that plots of ( Rf V/APt ) against (V^p/APt ) for different feed concentrations are straight lines. Therefore, (RfV/APt ) and (V^ p /APt) are related by an equation of the form
Rf V = A - B (5.6) A P^
where, A and B are the intercept and gradient of the line representing by Equation(5.6) respectively. The gradients of the best fit lines through the appropriate points are found
76 to be roughly the same. Thus, from Figure 5.12
B= 5.361 X 10 “ 5
If the intercepts of the lines corresponding to different feed concentrations are plotted against the percentage solid content in the feed, Figure 5.11 is resulted which produces a relationship between the constant A and the feed concentration. Thus from Figure 5.11
A = 2.987 X 1 0 * 3 x 10 3.235 C . . . . ( 5 . 7 ) where C = feed concentration,Mass Fraction
A and B may be used in conjunction with Equation(5.6) to evaluate R f. It should be noted that both A and B are dimensionless quantities. The values found for A and B are for the magnesia suspension and may well be different for other type of feeds.
If Equation (5.6) is substituted back into Equation ( 3.7 ) for Rf, then
(5.8)
and after rearranging Equation ( 5.8 )
B V p . . .(5 .9 )
= accumulation rate - removal rate where Equation (5.9) has dimensions of mass / area x time.
77 3.0e-2
2.5e-2~
28.5% 2 .0©-2 -
RfV 1.5e-2- APt
18.5% 1.0©-2
5.00-3 -
0 20 40 60 80
V P APt
Figure 5.11 Plot of Equation (5.6) Using Experimental Data for PALL Membrane with the Gradients of Lines Being Equal to the Constant B in Equation ( 5.6).
78 Figure 5.11 Relationship Between the Constant A in Equation(5.6) and Magnesia Slurry Concentration.
If the crossflow velocity is set at a value Vc where no fouling layer is allowed to form then fs=f0 and Equation (5.9 ) becomes
A APt B V c p = 0 Vc or A A M V c = ( 5 .1 0 ) B p J
where Vc = critical crossflow velocity, i.e., the velocity at which no fouling occurs. 5.3.2 Application of the Model to the Experimental Data
Having estimated the constants A and B, it is possible to calculate the resistance of the fouling layer, Rf, at different crossflow velocities, transmembrane pressures and feed concentrations using Equation ( 5.6 ). Hydraulic resistance of the membrane, Rm, is also found from the pure water runs with the clean membrane as discussed in Section
5.2.2. Equation ( 5.8 ) is used to evaluate theoretical values of the flux, fs. Figures
5.13-5.19 show Equation (5.8) graphically using the experimental operating data (e.g.,
V, A Pt and feed concentration) of sections 5.2.3.1 and 5.2.3.2 for PALL, DOULTON and FAIREY membranes. Experimental flux values are also given in these Figures.
The standard deviation, Sn, calculated for individual membranes and concentrations show an error in the region of ± 1.0 to ± 3.8% for PALL, ± 6.0 % for DOULTON and ± 4.5 % for FAIREY membrane. The reason for a low error margin in PALL membrane case is obvious, since, the PALL experimental data is used for estimations of A and B in Equation ( 5.6 ). The effectiveness of the model is further proved when a plot of
^(experimental ) against f(predicted ) *s made. This is shown in Figure 5.20 which indicates a good agreement between the experimental and predicted flux values using data for PALL, DOULTON and FAIREY membranes and feed concentration range of 2.25 - 28.5 % WAV.
80 500
V=2.17 m/s 400 V=2.01 m/s h V=1.88 m/s JZ V=1.66 m/s NI V=1.34 m/s &
200
100 “ T“ 20 40 60 80 100 120
APt, KPa
Figure 5.13 Comparison of the Experimental Flux vs. Transmembrane Pressure Data at Different Crossflow Velocities with the Results Predicted by Equation (5.8 ) ( Continuous Lines ). Membrane : PALL, Feed Cone.: 6.0% WAV.
□ V=2.015 m/s • V=1.660 m/s B V=1.410 m/s • V=1.176 m/s
Figure 5.14 Comparison of the Experimental Flux vs. Transmembrane Pressure Data at Different Crossflow Velocities with the Results Predicted by Equation (5.8 ) (Continuous Lines ). Membrane : PALL, Feed Cone.: 18.5 % WAV .
81 D V=1.175 m/s ♦ V=.896 m/s B V=.763 m/s
*
APt, KPa
Figure 5.15 Comparison of the Experimental Flux vs. Transmembrane Pressure Data at Different Crossflow Velocities with the Results Predicted by Equation(5.8) ( Continuous Lines ). Membrane : PALL, Feed Cone.: 28.5% WAV .
o V=1.019 m/s • V=.943 m/s ■ V=.480 m/s o V=.307 m/s
APt, KPa
Figure 5.16 Comparison of the Experimental Flux vs. Transmembrane Pressure Data at Different Crossflow Velocities with the Results Predicted by Equation(5.8) ( Continuous Lines ). Membrane: DOULTON, Feed Cone.: 6.25 % WAV.
82 160
□ V=1.273 m/s • V=1.019 m/s ■ V=.727 m/s ♦ V=.661 m/s
Figure 5.17 Comparison of the Experimental Flux vs. Transmembrane Pressure Data at Different Crossflow Velocities with the Results Predicted by Equation(5.8) ( Continuous Lines). Membrane: DOULTON, Feed Cone.: 12.0% W/W.
o V=. 910 m/s • V=.796 m/s ■ V=.579 m/s
Figure 5.18 Comparison of the Experimental Flux vs. Transmembrane Pressure Data at Different Crossflow Velocities with the Results Predicted by Equation(5.8) ( Continuous Lines). Membrane: DOULTON, Feed Cone.: 17.5% W/W.
83 3 0 0
□ V=1.9m/s ♦ V=1.7m/s ■ V=1.5m/s c* • V=1.3m/s E ■ V=1.1m/s
"T" —r~ 1 o 30 50 70 APt, KPa
Figure 5.19 Comparison of the Experimental Flux vs. Transmembrane Pressure Data at Different Crossflow Velocities with the Results Predicted by Equation(5.8) ( Continuous Lines). Membrane: FAIREY, Feed Cone.: 13.0% WAV.
Figure 5.20 Experimental Values of Flux Against Those Calculated Using the Present Model.
84 5.3.3 Further Points About the Model
In development of the model it is assumed that the fouling film thickness is small compared to the membrane diameter and, therefore, its influence on the crossflow velocity is negligible. The fouling was not investigated with regard to the pipe diameter as the membranes available were generally of similar diameter. There is little microfiltration data available in the literature to apply the present model and it is difficult to determine to what extent the model is applicable to other feeds. However, as a means of comparison, the model due to Schock (*1986) is used to estimate the steady state flux for the present experimental results. The experimental flux values are plotted against the RHS of Equation ( 3.9 ) in Figure 5.21. According to the model, discussed in Section 3.6.2.1, such a plot should be a straight line with the gradient of the line being the constant of this model. As it can be seen the model is not strictly applicable to the experimental data especially at higher feed concentrations. The reason for this could be due to the low feed concentrations used in development of this model (Brambachet a l, 1989 ). However, if this model is further modified to take into account the effect of concentration ( in terms of the ratio of the viscosity of water to that of the feed) a better correlation is resulted. Thus, the modified Schock Equation becomes
(5.11) where Ci = constant m, = viscosity of water, Pa s |is = viscosity of slurry, Pa s
Figure 5.22 shows the application of the modified Schock model ( Eq. 5.11 ) to the same data used previously for plotting Figure 5.21. Clearly the results show a better straight line fit than in the previous case.
85 The Schock mcxiel assumes the effect of the transmembrane pressure is negligible. As shown in the experiments and in derivation of Equation ( 5.8 ), some degree of dependency on the pressure is envisaged. For instance, the pressure might be of importance if fouling layer is known to be affected by it. In conventional filtration it is known that rise in pressure causes compaction of the cake layer. The model expressed in the present work takes into account the effect of transmembrane pressure. Furthermore, the experimental results indicate some degree of dependency of flux on the pressure in the mass transfer controlled region. The pressure effect on the fouling layer is not investigated in these experiments and any further postulations could be premature.
86 500
Figure 5.21 Application of the Schock Model to the Experimental Results.
Figure 5.22 Application of the Modified Schock Model to the Experimental Results. ( Model Due to Brambachet a l, 1989)
87 5.4 Transient Filtration Period 5.4.1 Model Development
The semi - empirical model represented by Equation(5.8) predicts the steady stateflux of microfilters. The period before reaching the steady state, or transient filtration period, is also of interest since prolonging it results in more permeate production. The transient filtration period is also important when anti-fouling measures such as pulsating flow and electrically charged membranes arc used to restore the flux. Therefore, by predicting this period the optimum time may be chosen to apply the pulse or current. In this section a model for predicting the transient period is developed which may be used to estimate the cake permeability.
It has been pointed out by previous workers ( Faneet a l, 1987 ) that there are two mechanisms that can occur in the early stages of membrane fouling. In the first case a particle of solid or gel can partially block a pore in the membrane. In the second a film of insoluble matter can build up on the feed side of the microfilter. The controlling mechanism will depend on the surface permeability of the membrane, the concentration of the insoluble material in the feed and the hydrodynamics on the feed side of the filter as discussed earlier. The rate at which clogging occurs is assumed to be directly proportional to the product of the permeate flux ( f ) and the solid concentration ( C ) and both clogging mechanisms would predict a limiting flux. Thus, for the case where a solid fouling layer of thickness ( h ) is formed, the rate of its formation with time ( t ) is given by
(5.11)
where f = permeate flux, m 3 m * 2 s _1 f s = steady state flux when fouling removal
becomes equal to fouling accumulation, m3 m ■ 2 s _l
88 h = thickness of fouling layer, m C = concentration of solid in slurry, % W/W Cp = concentration of solid in fouling layer,% W/W t = time, s
On the other hand, the following relationship exists between the applied transmembrane pressure, flux, film thickness and permeability
= —+ c o n sta n t ...... (5.12) f k It is assumed thatthe permeability of the fouling layer, k, remains constant once the steady state flux isreached * where k = cake permeability, m 2 P a s ' 1 APt = transmembrane pressure, Pa
and "constant" refers to the resistance due to the membrane, assuming that no pore plugging occurs. Upon differentiation of Equation ( 5.12 ) with the boundary
conditions of t = 0, f = and t = t , f = fs and substitution of Equation ( 5.11 ) in the former Equation
APt df _ 1 dh _ C (f-fs) f2 dt k dt k(CF-C)
_ _ d f _ C dt . . . . ( 5.14 ) kAPt(CF-C) f2( f - 0
Now making the substitution X = f / fs in Equation (5.14)
dX C f, dt. . . .(5.15) x2(x-i) k APt( CF- C)
89 also substitution of 0 = k APt (Cp-C)/Cf2s in Equation (5.15) and its integration
. (5.16) results (5.17)
where at t = 0 , A = X q for flux of clear membrane and at t = t , A=A.
Cumulative permeate throughput is defined as
J q = f dt = f s f l X dt . . . . (5.18) A=0 t=0 where q = cumulative permeate, m ■ 3 of permeate
per m ' 2 of filtration area
Now substitute for / dt from Equation ( 5.18 ) into Equation (5.16) to get
q = - f se
•'AO therefore, after integration of Equation (5.19)
q = fse (5.20)
90 and substitution forX, Equation (5.20) becomes
q = fst + f ' e | i - l j . . . . .(5.21) or
q-fst = f ^ e /l- A . . . .(5.22)
where the LHS of Equation (5.22) represents the flux during the initial unsteady period up to time t, when the steady state period is reached. Therefore, at the end of the unsteady state period f = fs and Equation(5.22) becomes
cumulative flux before _ reaching steady state " M . . . .(5.23)
where qj = cumulative flux before s.s., m 3 m ■ 2 and substituting for 0 in above Equation
kAPt( C F-C) (5.24) C f s
Equations (5.21) and ( 5.24 ) may be represented graphically by Figure 5.23. The intercepts of the two lines qw = fo t and q= qj + fs t shown in Figure 5.23 also has a relationship to the time during which the permeate flux will vary. If the two lines cross at t = t c , then
91 ^0"" Qi • • • • ( 5.25 ) where t c = transient filtration period or qi *c” (5.26) fo-f.
Substituting the value obtained for qj fiom Equation (5.24)
kAP,( CF- C) C Cfsf0
Equation (5.27 ) may be used for estimation of t c.
Figure 5.23 Cumulative Flux Against Filtration Time Representing the Transient Filtration Model.
92 5.4.2 Application of the Model
In order to use the model derived in Section 5.4.1, represented by Equation ( 5.27 ), fg, fs, cake permeability, k, and fouling layer solid concentration, Cp, should be known, fg is found from clear water run with the membrane ( Table5.1 ). fs may be estimated from Equation (5.8) at given crossflow velocity, transmembrane pressure and feed concentration or directly from experiments. The clean membrane permeability is used in Equation 5.27( ) in the first instance, as there is no direct measurement of the fouling layer thickness. Finally, Cp is found by measuring the density of the back-flushed cake from the membrane.
Therefore, for a PALL membrane with a feed concentration of 2.25% WAV, crossflow velocity of 0.7 m s -1 and transmembrane pressure of 98.6 KPa, fs is found from Equation (5.8) to be
fs = 164.50, lit m • 2 h _1 or 4.57 x 10 ‘ 5 m 3 m " 2 s
and from Table 5.1 fg = 704.3, lit m ■ 2 h or 19.56 x 1 0 ‘ 5 m 3 m - 2 S -1
As a first approximation, permeability value corresponding to the case where the membrane itself is the main resistance will be used. The membrane permeability for a wall thickness of 2 x 10 ■ 3 m and hydraulic membrane resistance of 5.041 x 810
Pa s m ‘ 1 is
k = 1.90 x 10 _12 m 2 P a s -1 ( see Appendix D for calculations)
Density of the fouling layer was found to be 1350 Kg m _3, corresponding to a solid
93 concentration of around 50% ( Figure 4.7 ).
Therefore, from Equation (5.27)
t c = 928.8 seconds
The experimental run carried out at the above conditions is shown in Figure 5.24. From the graph it is evident that the predicted t c is in error since, the fouling layer resistance is significant.
Figure 5.24 Cumulative Flux As a Function of Filtration Time for a Run Carried Out with PALL Membrane at V = 0.7 m s _1, APt = 98.6 KPa and C = 2.25 % WAV.
In order to get a better estimation of k, Equation ( 5.24 ) is used for the conditions described by Figure 5.24. Thus from the plot
94 qj = 0.016 m 3 m ’ 2 and from Equation ( 5.24 )
k = 4.15 x 10 " 13 m 2 Pa-1 s - 1 As it can be seen, the new value of k, corresponding to the cake permeability, is markedly different from that of the membrane. In order to examine the new value of k let
V = 1.34 m s -1 APt = 76170 Pa C = 2.25 %W/W
From Equation (5.24) qj = 0.004 m ^ m ' 2 and from Equation (5.27 ) t c = 56.1 seconds ( see Appendix D for calculations)
Figure 5.25 shows the experimental cumulative flux as a function of time using the above conditions. Experimental t c is in close agreement with that found from the transient model approach. Thus, in addition to estimation of t c, the model may be used to calculate the cake permeability with a reasonable degree of accuracy.
95 Figure 5.25 Cumulative Flux as a Function of Filtration Time for a Run Carried Out with PALL Membrane at V = 1.34 m s _1, APt = 76.17 KPa and C=2.25 % WAV.
5.4.2.1 Estimation of the Film Thickness
Having calculated the cake permeability it is possible to find the film thickness. The film thickness is given by
h = Rf k ...... ( 5.28 )
Where Rf value used earlier in Equation ( 5.8 ) of the above example and the permeability value calculated from Equation ( 5.24 ) are used in Equation ( 5.28 ).
Thus, for R f = .128 KPa m 2 hr 1 "1 or 4.6x10s Pa s m _1
h= 1.910x10'4 m or « 0.2 mm
Brambach et al ( 1989 ) made approximate measurements of the film thickness of magnesia slurry by weighing the fouling deposit after back flushing of their sintered
96 stainless steel filters. Their results are of the same order of magnitude to that found above. A typical film thickness for a magnesia concentration of 2.25% W/W extrapolated from their plot of film thickness against concentration is around0 . 1 2 mm.
97 5.5 Results from the Active Experiments 5.5.1 MILLIPORE Filters 5.5.1.1 Effect of pH and Fouling Agent
In Chapter Three decontamination factor or DF is defined as
D F = activity in the feed (3.14 ) activity in the permeate
Table 5.3 summarises the DF ranges found for different pH and feed conditions of the radionuclides examined. DF's of Am and Eu are quite high indicating a good degree of retention by the filters. Raising the pH from 9.0 to 11.5 results in higher DF's. Increase in the pH may have caused the formation of more colloidal species of Am and Eu causing more of them to be retained by the filters. The presence of magnesia particles brings about a rise in the pH from 9.0 to 10.0 in addition to providing a sorption medium. However, the increase in DF's of Am and Eu in the presence of magnesia is not significant.
In the case of Sb, Sr and Cs the DFs are relatively low and these species are not retained effectively, moreover, the increase in pH hardly shows any change in DF's. In alkaline solution Sr, Cs and Sb are in aqueous phase. The results of MILLIPORE tests confirm this by producing low DF's. In the presence of magnesia suspension DF's of Cs, Sb and Sr increase although the rise in the DF of Sr is more than the others. Such increase is most probably due to the physical adsorption of these nuclides to the secondary cake forming on the filter. As the pH is raised under these conditions DF's of Cs remain more or less unchanged while those of Sb and Sr increase even more than before suggesting that the sorption of Sr and Sb by the magnesia particles is not purely physical and may be of a chemical nature.
98 5.5.1.2 Effect of Pore Size
Variations of DF with pore size of the MILLIPORE filters as a function of feed condition are shown in Figures 5.26 - 5.29. For colloidal Eu, DF's show some dependence on the pore size in all pHs and with or without the presence of the fouling agent. Dependency of Am DF's on the pore size is apparent only at pH of 11.5 and less apparent with magnesia present. DFs of Cs hardly show any variations with the pore size and feed conditions due to its presence in solution. Sr retention is also independent of the pore size in both pH of 9.0 and 11.5. However, in the presence of magnesia its DF's increase with decrease in the pore size. As the pH increases from 10.10 to 11.5 this variation no longer is apparent. Sb at pH 9.0 is soluble and is not retained by any of the filters. Although at pH 11.5 it is retained more by smaller pore size filters. Such trend is also observed when Mg (OH) 2 is present. As the pH is raised to 11.5, change of antimony's
DFs with the pore size becomes insignificant
PH 9.0 11.5 1 0 . 0 11.5
Feed (without magnesia) (with magnesia)
Am-241 74-176 603-large* 311 - large* large*
Eu-154 37-386 151-large* 58 - large* —
Sb-125 2.0-5.0 1 .1 -2 . 0 4.0- 6.4 7.2 - large*
Sr-85 1 .0 - 1 . 2 1 .2 -1.4 12.3-59 30-111 Cs-137 2.0-2.7 1.3-1.9 3.0 2.7-4.5
Table 5.3 Summary of the DF Ranges for Different pH and Feed Conditions Using MILLIPORE Filters. (* indicates that activity in the permeate is below detection)
99 Am-241 Eu-154 Sb-125 Sr-85 a Cs-137
Figure 5.26 Variations of Decontamination Factors ( DF ) with Pore Diameters of MILLIPORE Filters for Various Radionuclides at Feed pH of 8.7.
Am-241 Eu-154 Sb-125 Sr-85 Cs-137 Q
Figure 5.27 Variations of Decontamination Factors ( DF ) with Pore Diameters of MILLIPORE Filters for Various Radionuclides at Feed pH of 11.5.
100 Am-241 Eu-154 Sb-125 Sr-85 ft* Cs-137 a
Figure 5.28 Variations of Decontamination Factors ( DF ) with Pore Diameters of MILLIPORE Filters for Various Radionuclides at Feed pH of 10.10 in the Presence of Magnesia Suspension .
Figure 5.29 Variations of Decontamination Factors ( DF ) with Pore Diameters of MILLIPORE Filters for Various Radionuclides at Feed pH of 11.5 in the Presence of Magnesia Suspension .
101 Summary of the Main Results
The results showedthat:
1. At the pH of 8.7 - 9.0, Am and Eu are present as colloids and are retained satisfactorily by all the filters. Sr, Cs and Sb are in aqueous phase and hardly retained at all. At pH of 11.5, Sr and Cs are still in solution and are not retained while Sb, Eu and Am are in colloidal state. Sb is separated well at pore size of 0.1 pm while Am and Eu DFs remained as high as before.
2. In the presence of magnesia, at the pH of 10.10, Sr is sorbed by the magnesia particles and its average DF values increased. The sorption is believed to be of physical nature. DFs of Cs are hardly affected in the presence of magnesia. DFs of Sb increases to 4-6.4. Am and Eu DFs generally remain unchanged. As the pH is raised to 11.5, Sr removal increases but Cs remains in solution. Am and Eu DF values show only a slight change with the pore size in the presence of magnesia while Sb is retained more at smallest pore size.
3. Filters with pore sizes between 0.1 - 0.8 pm, which are within the microfilter pore size region, are capable of removing colloidal and particulate radioactivity quite effectively. In the case of soluble radionuclides, the magnesia cake layer has little effect on the retention of the soluble species. In general, alkaline pH is beneficial for retention of colloidal species.
5.5.2 M4 CARBOSEP Membrane 5.5.2.1 Retention of Nuclides at Different pH
Variations of DFs as a function of the feed pH are shown in Table 5.4 and in Figure 5.30. DF values in Table 5.5 indicate that there is very little retention of activity by the
102 membrane below pH 4. Th-230 and Co-60 in colloidal form are retained from pH of around 5.0 and 7.0 respectively. At around pH 10 no trace of thorium is detected in the permeate while retention of cobalt has reached a maximum value of 21.5. Increase in the pH of the cobalt feed in excess of 10.0 results in little change in its DFs. In the europium run resulting DF's show that europium species remain in the solution up to pH 6of . 8 - 7.0. Around pH 8.5 colloidal europium forms and further increase in the solution pH give rise to higher DF's.
The uranium results show a more complex behaviour. Uranium DFs indicate that it remains in solution up to pH 4.0. Around pH 6.0 colloidal uranium is formed which is retained by the CARBOSEP membrane. At pH 9.0 the uranium concentration in the permeate is a minimum yielding a maximum DF of 23.71 in this pH range. Above pH 9.0 uranium colloids start to go into solution possibly as uranium carbonate complexes, therefore, reducing the DFs at this particular pH. Further increase in the pH results in more retention of uranium colloids.
103 Th-230 Co-60 Eu-152 U-235
pH DF pH DF pH DF pH DF
3.3 1.16 2 . 0 1 . 1 0 3.0 1.17 4.0 1.05
4.5 1 . 2 0 6 .1 1.14 5.58 1.05 6 .1 7.38
6 . 0 9.10 7.79 1.05 6.84 1.05 7.0 18.16 7.0 50.0 8.62 10.76 8.5 4.24 9.0 23.71
1 0 . 0 large* 11.76 21.5 1 1 . 0 14.91 1 1 . 0 8.3
1 2 . 0 large* 13.11 2 0 . 0 1 2 . 0 15.03 1 2 . 0 16.83
Table 5.4 Variation of DFs with Feed pH for Different Elements in the Experiments with M4 CARBOSEP Membrane. (* indicates that activity in the permeate is below detection)
104 pH
Figure 5.30 Variation of Decontamination Factors ( DF ) with Feed pH for Various Radionuclides When M4 CARBOSEP Membrane is Used.
105 5.5.2.2 Retention of Nuclides at Constant pH 5.5.2.2.1 Europium Run
Table 5.5 shows the feed and permeate activities and decontamination factors for Eu-152 at feed pH of 11.5. Measurement of the activities of permeate indicate that europium retention by the membrane remain reasonably constant throughout the seven hours of the experimental run. They are also of similar order of magnitude to those found from the previous experiment with Eu-152 at high pH region ( Table 5.4 ). As the closed cycle UF experiment proceeds the activity in the feed stream increases from 16 to 30 Bq/ml. There is no evidence to suggest that the europium retention is affected by the increase in the feed activity. The permeate flux is constant and there is no evidence of the membrane loss of permeability due to fouling.
5.5.2.2.2 Europium and Cobalt Run
The result for the continuous run which both Eu-152 and Co-60 is given in Table 5.6. The overall activity of cobalt in the permeate is fairly constant even though there is quite a marked rise in the activity of the feed. During this period the cobalt DFs change from 44.0 to 66.0. The results for europium are similar to those in the previous experiment and it seems that the presence of cobalt in the feed has no effect on the retention of Eu-152. Permeate flowrates, as before, are fairly constant with very low activity.
106 Activity ( Bq/ml)
Time(h) Feed Permeate DF
0 16.83 ——
1 15.37 1.13 13.6
2 15.67 0.69 22.7
3 19.97 0 . 6 8 29.4
4 2 2 . 0 0 0 . 0 large*
5 26.20 0 . 0 large*
6 24.75 0.91 27.2 7 29.27 0.81 36.1
Table 5.5 Variations of Activities and DF's with Time for Europium Feed at pH 11.5 Using M4 CARBOSEP Membrane, (^indicates that activity in the permeate is below detection)
107 Activity ( Bq/ml) Activity ( Bq/ml) Eu-152 Co-60
Time (h) Feed Permeate DF Feed Permeate DF
0 1 2 .2 1 — — 62.30 — —
2 14.42 0 . 0 large* 73.17 0 . 0 large*
3 14.61 0 . 0 large* 78.92 0 . 0 large*
4 14.83 0 . 0 large* 83.90 1 .8 8 44.6 5 15.71 0.19 82.68 85.15 1.78 47.8
6 18.48 0.92 2 0 . 1 0 97.70 1.75 55.8
7 19.16 1.64 11.70 102.25 1.55 6 6 . 0
Table 5.6 Variations of Activities and DFs with Time for Europium and Cobalt Feed at pH 11.5 Using M4 CARBOSEP Membrane. ( * indicates that activity in the permeate is below detection)
108 Summary of the Main Results
1. M4 CARBOSEP, an ultrafiltration membrane, behaves in a similar way to MILLIPORE filters of microfiltration pore size range. It retains colloidal and particulate species effectively while letting through those in solution.
2. The solution pH is of great importance and with the aid of microfilters may be used to separate different radiocolloids in different pH regions. For thorium and other tetravalent elements like plutonium a pH of 7.0 is high enough for effective decontamination. Europium, a trivalent ion comes down at the higher pH of 9.0. Divalent cobalt precipitates at the same pH of 9.0. In a pH range of 11-13, DF values of around 20.0 is typical. Hexavalent uranium is relatively in soluble at a pH of 11.0. The highest DF found for U-235 from these experiments is 23.0 at the pH of 9.0.
5. The continuous tests at constant pH of 11.5 indicate that the permeate activity increasing remains constant despite the^total activity in the feed. Co-60 and Eu-152 DFs are generally high, the latter giving similar order of magnitude DFs to MILLIPORE tests. The permeate flux also remains more or less constant for during the runs.
5.5.3 M4,M6 CARBOSEP and PALL and DOULTON Membranes
Table 5.7 shows the DF ranges found from the experiments carried out over 4 hours using single tube MF/UF membranes. Pseudocolloids such as Am, Eu and Pu are retained and their corresponding DF values are of similar order of magnitude regardless of the type of the membrane. DFs of Sr-85 are in the range of 1.0 -1.9 indicating that it is not retained by any of the membranes.
109 Membrane Type
Feed M4 M6 PALL DOULTON
Am-241 26.0-large* 15.0-large* 87.0-large* 31.0-large*
Eu-152 5.4-large* 8 .6 -large* 61.0-large* 10.7 -large*
Sr-85 1.0-1.47 1.65-1.93 1 .1 1 .0 -1 .1
Cs-137 1.4-1.45 2 . 0 1 .2 -1.5 2.1-2.65 Pu-239 1054.5 893.4 588.2 322.2
Table 5.7 Summary of the DF Ranges of the Radionuclides of Interest for Different membranes. Vindicate that activity in the permeate is below detection)
As the filtration rig is a closed circuit where the outlet from the membrane module and the permeate are recycled back to the feed tank, the only source of loss of activity is through the sorption on the membrane and/or its accumulation due to the build-up of magnesia film on the membrane surface ( assuming that the pipes in the rig have negligible activity hold-up). Thus, the ratios of activity of the feed at the beginning and at the end of each run should give some indication of the adsorption properties of the membranes. Table 5.8 shows these ratios. Sorption of Am, Eu and Pu by the DOULTON membrane is more than^other membrane. Am, Eu and Pu hold-up of the
PALL membrane is also noticeable although to a lesser extent M6 tube also shows some degree of fouling . However, M4 shows no activity hold-up. This is most probably due to the very low permeability of this type of ultrafilter. Thus, adsorptive properties of the membranes decrease according to the series
110 DOULTON > PALL > M6 > M4
Observations during the experiments show that particulate magnesium hydroxide is washed-out of the feed tank gradually and deposited in the membrane. This is particularly marked in the experiments with DOULTON ceramic membrane where after only two hours most of the magnesia was removed from the feed tank.
Ratios of the activity of Sr in the feed at the beginning and the end of the runs remain virtually constant around the value of 1.1, indicating a negligible loss of Sr by sorption to the fouling layer and the membranes. In contrast Cs removal from the feed and its sorption by the membrane is more than Sr. Cs is retained in the membranes according to
DOULTON > M6 > M4 > PALL
It is thought that Cs is removed and trapped in the fouling layer formed on the membrane surface. However, since Cs is soluble in the experimental conditions its behaviour is different from Am, Eu and Pu. Cs hold-up could be due to the entrapment in the support structure of some of the membranes as a result of depth fouling. In DOULTON, 6M, M4 and PALL membranes the retention probably occurs due to the membrane fouling. However, this layer has smaller effect on the retention of Cs by M4 and PALL than other membranes.
Summary of the Main Results
1. The DF values show that all the membranes perform well in retention of Am, Eu and Pu. However, soluble Sr and Cs are not retained.
2. Comparison of the DF's from the present experiments with those of the MILLIPORE tests suggest that membrane fouling layer dose not enhance
111 retention of pseudocolloids and soluble species.
3. Sorption of Am, Eu and Pu is maximum in DOULTON ceramic membrane. Fouling follows the following pattern
M4 < M6 < PALL < DOULTON
4. Sorption of Cs by the membranes changes as follows
PALL < M4 < M6 < DOULTON
and sorption of Sr by all the membranes is negligible.
Membrane Type
Feed M4 M6 PALL DOULTON
Activity Ratios
Am-241 1.194 1.404 2.664 23.400 Eu-152 1.155 1.490 5.404 9.055
Sr-85 1.085 1.123 1.146 1 . 1 0 1 Cs-137 1.257 1.750 1.170 1.885 Pu-239 1.175 1.510 1.522 5.800
Table 5.8 Ratio of Activity of Feed at the Beginning and the End of the Runs Using Different Membranes.
112 CHAPTER SIX
Conclusions and Recommendations
6.1 Conclusions
The present work represents a pioneering study into microfiltration in the intermediate range between pressure driven and mass transfer controlled filtrationprocesses.Extensive fouling resulted during microfiltartion of magnesia suspension which affected the permeate flux considerably. The flux reached a near steady state value after the initial fast flux decline. In the mass transfer controlled region, where most of the experiments were carried out, a certain degree of pressure dependency was found which could be attributed to the changes in the fouling layer due to pressure. The steady state flux was found to vary with the crossflow velocity significantly, following a relationship of the type typical of this kind of operation. The steady state flux of the fouled membrane was modelled by a semi-empirical equation which relates the flux to the hydrodynamic properties of the flow. This model was applied successfully to the tubular inorganic microfilters used in the experiments and a good degree of accuracy was found between the experimental and predicted results. Therefore,
To use this model for design purposes the hydrodynamic properties of the flow are required which are basically the operating conditions and are set according to the operational requirements. There are two other parameters ( A and B ) which are related
113 to the feed and should be found when the model is used for feeds other than magnesia water suspension.
An expression was derived for the estimation of the period during which flux changes continuously before reaching a near steady state value. This period was referred to as the transient filtration period and experimental data was used to verify the model. In general, the model worked well for conditions where the transient filtration period was relatively long i.e., at low feed concentrations and slow crossflow velocity. At high feed concentrations the fouling rate was too fast and the transient filtration period was too short to be of any significance. This model is particularly useful for estimation of the cake permeability and ultimately the fouling film thickness. Cake permeability from this model is given by
k =
Permeability data is of great value since, there is very little published information in the literature.
Both models described are basically two dimensional in nature. It is likely that in some microfiltration processes the effect of a cylindrical geometry may ultimately be of significance.
In the active experiments filters with pore sizes between 0.1 - 0.8 pm, which are within the microfilter pore size region, were capable of removing colloidal and particulate radioactivity quite effectively particularly in alkaline pH region. They were effective in retention of radionulides such as Am-241, Eu-152, Eu-154, Co-60, Th-230, U-235 and Pu-239. In the case of soluble radionuclides including Sr-85 and Cs-137, the
114 membranes were not effective at all even in high pH conditions or in presence of the magnesia particles. Fouling experiments showed that, with the exception of M4 ( a UF membrane), the membranes' magnesia sorption ( or fouling) were significant. Fouling follows the following pattern
M4 < M6 < PALL < DOULTON
6.2 Recommendations for Further Work
In development of the model for estimation of the steady state flux the fouling was not modelled with regard to the channel geometry. Membrane diameter and its length could have some affect on the extent of fouling. It is worthwhile that some work is directed towards these factors. However, as most of the available tubular filters are of a similar diameter, one can use tubular inserts for changing their geometry.
There is little microfiltration data available in the literature for comparison purposes and it is difficult to determine to what extent the model is applicable to other feeds. Future experiments with other suspensions will be of value to examine the constants in the model and the function generalisation for other feeds.
No independent measurements of the film thickness were made in this study. The future rig should have facilities to measure directly the film thickness. Carrying out such measurement enables a better assessment of the transient filtration model. It will also help in elucidating greater details of feed hydrodynamics.
Measurement of transmembrane pressure along the membrane by means of a pitot tube could be of importance in order to examine the possible end effects as the fouling layer builds up in the first few millimetres of the filter membrane.
115 The present experimental results indicated some degree of dependency of flux on the pressure. Further work may be advisable on the effect of pressure on the properties of cake layer such as its compaction and permeability.
Though some interesting and innovative advances have been made in this work towards the derivation of a comprehensive theory of microfiltration processes, the modelling of such processes is quite complex. Relating microfiltration properties to rheological parameters of both the feed and fouling layer remains an important area of future work.
116 Nomenclature
A = constant of Equation (5.6) a = constant of Equation (3.8) B = constant of Equation (5.6) b = constant of Equation (3.8) C = concentration of solid in slurry,% W/W C = constant of Equation (3.9) Cp = concentration of solid in fouling layer,% W/W Ci = constant of Equation ( 5.11) Cg = bulk solute concentration,% W/W
Cq = gel or cake layer concentration,% W/W Cp = concentration of the rejected component in permeate, % W/W Q- = concentration of the rejected component in retentate, % W/W C\y = solute concentration at the membrane interface, % W/W
D = diffusion coefficient, m ^'s 1 d = channel diameter, m dp = particle diameter in Equations (3.9) and (5.11), m dfo = equivalent diameter, m dp = pore diameter, m f = permeate flux, lit m"2 hr-1 or m 3 m ■ 2 s f s = steady state flux when fouling removal becomes equal to fouling accumulation, 3m m ' 2 s _1 fO = initial flux, lit m"2 hr h = membrane wall thickness, m h = thickness of fouling layer, m k = boundary layer division k = membrane permeability, m 2 Pa_1 s _1 or KPa m 2 hr lit_1
117 n = exponent which varies with crossflow velocity ( also known as fouling constant) in Equation (3.6) Pl= inlet pressure, Pa
P2 = outlet pressure, Pa
P3 = permeate pressure, Pa q = cumulative permeate, m ■ 3 of permeate per m • 2 of filtration area qj = cumulative flux before s.s., m 3 m ■ 2 R = tube radius, m R = membrane rejection coefficient for a given component in Equation (3.11) r = radial position of the particle in the tube, m
Rb = boundary layer resistance, Kg m* 2 s _1 Re = p V d / p, Reynolds Number
Rf = resistance of fouling layer, Kg m■ 2 s -1
Rh = hydrodynamic resistance of membrane,'1 m
Rd, x = resistance of each layer, m_1
Rm = hydraulic resistance of medium ( or membrane), Kg 'm 2 s _1 rp = radius of diffusing particle, m r * = equilibrium radial position of the particle, m Sc = \i / p D, Schmidt Number Sh = d V p / |i, Sherwood Number Sd = standard deviation t c = transient filtration period, sec. t = time, seconds or minutes
U = mean crossflow velocity, m s-1
V = crossflow velocity, m s -1
Vc = critical crossflow velocity, m s *1
Vp = volume of filtrate, m 3
V q = volume of feed, m 3 x = boundary layer thickness, m
118 Symbols e = kAPt(CF-C)/Cf2s in Equation (5.15) e = porosity p = density, Kg m“3 x y = yield stress, Pa v = kinematic viscosity, m 2 s -1 5 = fractional product loss A = recovery rate APt = applied transmembrane pressure, Pa or KPa AX = thickness of membrane or porous medium, m tj = plastic viscosity, Pa s X = f / fs in Equation ( 5.14)
11 = viscosity of permeating fluid, Kg nr1 s*1 |i e = effective viscosity, Pa s |XS = viscosity of slurry, Pa s pw = viscosity of water, Pa s
119 References
Asaadi M. and D.A. White, Fouling Behaviour in Inorganic Tubular Membranes, International Technical Conference on Membrane Separation Processes, Brighton, UK, pp 143 - 156, 1989.
Banks J. and J.M. Parkes, Porous Ceramics : A Unique Medium in The Filtration Industry, DOULTON Ceramic Company Publication. Bedwell W. B., S. F. Yates, I. M. Brubaker and S. Uban, Crossflow Microfiltration-Fouling -> Mechanisms Studies, Separation Science and Technology, Vol.( 6 23 and 7), pp531-548,1988. Berger C. and F.C. Arrance, Presintered Metal Oxide Membranes for Fuel Cells, U.S. Patents 3,490,953; January 20,1970 and 3,462,314; August 19,1969.
Bergez P., J.M. Martinet, Elaboration of Mineral Membranes, Summer School on Membrane Science and Technology, Cadarache ( France), 1984.
Blatt W.F., Principles and Practice of Ultrafiltration, Membrane Separation Processes, Ch. 3, Edited by P. Meares, Elsevier Scientific Publishing Company, pp 119 -120, 1976.
Blatt W.F. A. David, A.S. Michaels and L. Nelson, Membrane Science and Technology, Edited by J.E. Flinn, Plenum Press, NY, 1970.
Booman G.L. and J.R. Delmastro, Radio Frequency Sputtering Technique Using Platinum and Silicon Nitrate, U.S. Patent 3,794,174; February 26,1974.
Brambach R., M. Asaadi and D. A. White, Studies in Fouling During Crossflow Microfiltration in Sintered Stainless Steel Membranes, Final Year Research Project,
120 Chemical Engineering Department, Imperial College, London, August 1989.
Charpin J., P. Bergez, F. Valin, H. Bamier, A. Maurel and J.M. Martinet, Inorganic Membranes : Preparation, Characterisation, Specific Applications, High Tech. Ceramics, Edited by P. Vincenzini, Published by Elsevier Science, Amsterdam, pp 2211-2225, 1987.
Cheryan M., Ultrafiltration Handbook, Technomic Publishing Company, 1986
Cheryan M. and B.H. Chiang, Engineering and Food, Edited by B.M. Me Kenna, Applied Science Publication, Vol. 1, London, UK, 1984.
Fane, A.G., C.J.D. Fell and M.T. Nor, Ultrafiltration / Activated Sludge System - Development of a Predictive Model, Ultrafiltration Membranes and Applications, Edited by A.R. Cooper, Planum Press, New York, pp 631 - 658, 1980.
Fane, A.G., C.J.D. Fell and A.G. Waters, The Relationship Between Membrane Surface Pore Characteristics and Flux for Ultrafiltration Membranes, Journal of Membrane Science, Vol. 9, pp 245 - 262,1981.
Fane, A.G. and C.J.D. Fell, A Review of Fouling and Fouling Control in Ultrafiltration, Desalination, Vol. 62, pp 117 - 138,1987.
Fleischer R.L., P.B. Price and R.M. Walker, Nuclear Tracks in Solids, Sci. Amer., pp 220 Vol. 30, 1969.
Gekas V, n. Resistance - in - Series Model, Characterisation of Ultrafiltration Membranes Proceedings, Orenas, Sweden, pp 179 - 211, 1987.
121 Grebenshchikova V.M. and Yu. P. Davydov, Radiokhmiya, Vol. 32, pp 155, 1961.
Gutman R.G., The Design of Membrane Separation Plant, The Chemical Engineer, pp 510-523, July 1977.
Jafftin M.Y., R. Ben Amar and B. B Gupta, Membrane Fouling Control in Crossflow Filtration of Wine with Mineral Membranes, International Technical Conference on Membrane Separation Processes, Brighton, UK, pp 157 - 165,1989.
Jonsson G. and P. M. Christensen, Separation Characteristics of Ultrafiltration Membranes, Membranes and Membrane Processes, Edited by E. Drioli and M. Nakagaki, pp 179 - 190,1986.
Kepak F., Adsorption and Colloidal Properties of Radioactive Elements in Trace Concentrations, Chemical Reviews, Vol. 71, No. 4, 1971.
Kroner K.H., W. Hummel, J. Volkel and M. -R. Kula, Effect of Antifoams on Crossflow Filtration of Microbial Suspension, Membranes and Membrane Processes, Edited by E. Drioli and M. Nakagaki, pp 223 - 232,1986.
Kuo K. P. and M. Cheryan, Journal of Food Science, 48:1113 ( 1983 )
Leenaars A.F.M., A.J. Burggraaf, The Preparation and Characterisation of Alumina Membranes with Ultrafine Pores, Part 2 The Formation of Supported Membranes, Journal of Colloid and Interface Science, Vol. 105, No. 1,1985.
Leenaars A.F.M., K. Keith and A.J. Burggraaf, The Preparation and Characterisation of Alumina Membranes with Ultrafine Pores, Part 1 Microstructural Investigation on Non - supported Membranes, Journal of Material Science, Vol. 19, pp 1077 -1088,
122 1984.
Lonsdale H.K., The Growth of Membrane Technology, Journal of Membrane Science, Vol. 10, pp 81 - 181, 1982.
Lunqvist R. and R. von Schalien, Mathematical Modelling of Colloidal Ultrafiltration : A Method for Estimation of Ultrafiltration Flux Using Experimental Data and Computer simulations, Membranes and Membrane Processes, Edited by E. Drioli and M. Nakagaki, pp 497-505,1986.
Madsen R.F., Hyperfiltration and Ultrafiltration in Plate - and - Frame Systems, Published by Elsevier Scientific Co., Amsterdam, 1977.
Maubois J.L., Journal of Society of Dairy Technology, 33:55,1980.
Murkes J and C.G. Carlsson, Crossflow Filtration, Published by John Wiley and Sons Ltd., 1988.
Olofsson V., B. Allard, K. Andersson and B. Torstanfelt, Formation and Properties of Americium Colloids in Aqueous Systems, Department of Nuclear Chemistry, Chalmers University of Technology, S-412 96, Gotemberg, Sweden, pp 191 - 199,1982.
Olofsson V., B. Allard, M. Bengtsson, B. Torstanfelt and K. Anderson, Formation and Properties of Actinide Colloids, Department of Nuclear Chemistry, Chalmers University of Technology, S-412 96, Gotemberg, Sweden, 1983. Orchard A.C.J., Recent Developments in Membranes for Critical Filtration Applications, International Technical Conference on Membrane Separation Processes, Brighton, UK, May 1989 Paper C2
123 Patel P.N., M.A. Mehaia and M. Cheryan, Crossflow Filtration of Yeast Suspensions, Journal of Biotechnology, pp 1 - 16,5,1987.
Porter M.C., in Synthetic Membranes : Vol. 1, Desalination, Edited by A.F. Turbak, Amer. Chem. Soc., Washington, D.C., pp 406, 1981.
Porter M.C., Ultrafiltration of Colloidal Suspensions, AIChE Symposium Series, Vol.
6 8 , No. 120, 1972.
Porter M.C., Concentration Polarisation with Membrane Ultrafiltration, Ind. Eng. Prod. Res. Develop., Vol. 11, No. 3, pp 234 - 248, 1972.
Probstein R.E., K.K Chan, R. Cohen and I. Rubenstein, Model and Preliminary Experiments on Membrane Fouling in Reverse Osmosis, Synthetic Membranes, Vol. 1 : Desalination, Edited by Turbak, pp 131 - 145,1981.
Rajan K.S. and A.J. Casolo, Thorium Hydrous Oxide, U.S. Patent 3,479,266 ; November 18, 1969.
Rautenbach R. and G. Schock, Mikrofiltration Kolloidaler Suspensionen und Ultrafiltration Makromolekul arer Losungen - Versuche zur Deckschichtbildung und zur Vorausberechnung des Permeatflusses, Chem.-Ing.-Tech., Vol. 59, No. 3, pp 242 - 243, 1987
Riesmeier B., Untersuchungen zur Crossflow Filtration Microbieller Suspension, Dissertation, TU Berlin, FB 13, 1987.
Schock G., Mikrofiltration an uberstromten Membranen, Dissertation, RWTH Aachen, IVT I, 1986.
124 Schweitzer P. A., Handbook of Separation Techniques for Chemical Engineers, Chapter 2.1, Me Graw - Hill Publication, 1979.
Sheppard J.D., D.G. Thomas and K.C. Channabasppa, Membrane Fouling Part IV. Parallel Operation of Four Tubular Hyperfiltration Modules at Different Velocities with Feeds of High Fouling Potential, Desalination, Vol. 11, pp 385 - 398,1972.
Skelland, A. H. P., Non-Newtonian Flow and Heat Transfer, Publ. John Wiley and Sons, 1967.
Strathmann H., Preparation of Microporous Membranes by Phase Inversion Process, Membranes and Membrane Processes, Edited by E. Drioli and M. Nakagaki, 1986.
Strathmann H., Membrane Separation Processes, Journal of Membrane Science, Vol. 9, pp 121 - 189, 1981.
Suki, A.B., A.G. Fane and C.J.D. Fell, Modelling of Fouling Mechanism in Protein Ultrafiltration, Journal of Membrane Science, Vol. 27, pp 181 - 193,1986.
Suki, A.B., A.G. Fane and C.J.D. Fell, Journal of Membrane Science, Vol. 21, pp 269-283, 1984.
Thomas D.G. and W.R. Mixon, Ind. Eng. Chem. Process Des. Dev., 11 ( 3 ), pp 339 , 1972.
Tsvetaeva N.E., V.M. Filin, T.K. Ragimov, B. Ya. Shcherbakov, L. Ya. Rudaya, and K. Yu. Shapiro, Comparative Behaviour of Americium and Plutonium in Wastewater, Radiokhimiya, Vol. 1, pp 129 - 133, 1986.
125 UKAEA Harwell, Electrolytic Filter Cleaning, Article in The Chemical Engineer, No. 463, August 1989.
Veyre R, Ultrafiltration by Inorganic Membranes for Liquid Sterilisation, Societe De Fabrication D’Elements Catalytiques ( S.F.E.C.), B.P. 201, 84500 Bollene, 1984.
Wakeman R.J. and E.S. Tarleton, Experiments Using Electricity to Prevent Fouling in Membrane Filtration, Proceedings of A Meeting on Membranes in Filtration Systems, Manchester, 1986.
126 Appendix A
Operating Conditions and Steady State Flux Values from the Fouling Experiments at Different Crossflow Velocities and Feed Concentrations When Transmembrane Pressure is Kept Constant.
127 Membrane PALL V(ms-l)
Feed Cone. 6.0% W/W 2.17 1 .8 8 1 . 6 6 1.34 APt= 64.0 KPa
fs(LMH) 420.1 328.5 290.0 2 2 0 . 6 Membrane PALL V(ms-l)
Feed Cone. 6.0% WAV 2.17 2 . 0 1 1 . 6 6 APt = 76.0 KPa fs(LMH) 450.0 375.4 300.2 Membrane PALL V(ms-l)
Feed Cone. 18.5% WAV 2 . 0 1 1 . 6 6 1.41 1.18 APt = 94.0 KPa f s ( LMH ) 166.0 130.2 112.7 94.2 Membrane PALL V(ms-l) Feed Cone. 28.5 % W/W 1.170 .896 .763 APt = 60.0 KPa fs(LMH) 43.0 33.2 30.4
Table A.l Variation of Steady State Flux with Crossflow Velocity and Feed Concentration for PALL Membrane at Constant Transmembrane Pressure.
128 Membrane DOULTON V(ms-l) Feed Cone. 6.25 %W/W 1.02 .943 .480 .307 APt = 60.0 KPa f s (LMH) 131.8 125.2 70.0 54.5 Membrane DOULTON V(ms-l) Feed Cone. 12.0 %W/W 1.27 1.02 .727 APt - 100 KPa fs(LMH) 120.3 109.0 87.8
Membrane DOULTON V(ms-l) Feed Cone. 17.5 %W/W .910 .796 .579 APt = 50.0 KPa fs(LMH) 69.6 60.2 45.3 Membrane FAIREY V(ms-l) Feed Cone. 13.0 %W/W 1.90 1.70 1.50 1.10 APt = 60.0 KPa f s (LMH) 225.1 193.8 175.5 115.0
Table A.2 Variation of Steady State Flux with Crossflow Velocity and Feed Concentration for DOULTON and FAIREY Membranes at Constant Transmembrane Pressure.
129 A p p en d ix B
Operating Conditions and Steady State Flux Values from the Fouling Experiments at Different Transmembrane Pressures and Feed Concentrations When Crossflow Velocity is Kept Constant.
130 Membrane PALL Feed Concentration 6.0% WAV V= 2.17 ms-1 V= 2.01 ms-1 V= 1.88 ms-1 APt, KPa fs (LMH) APt, KPa fs (LMH) APt, KPa fs (LMH) 58.26 400.0 46.20 348.8 52.75 311.7 64.47 420.7 52.75 364.5 58.26 335.0 76.20 412.1 76.00 375.0 64.47 341.1 104.52 431.8 104.82 388.7 104.82 350.8 Membrane PALL Feed Concentration 6.0% WAV V= 1.66 ms-1 V= 1.34 ms-1 APt, KPa fs (LMH) APt, KPa fs (LMH) 46.20 231.1 35.20 169.8 64.47 289.5 58.26 230.1 76.21 299.2 64.47 298.5 104.82 311.2 98.00 235.4 Membrane PALL Feed Concentration 18.5% WAV ( APt in KPa, fs in LMH) V= 2.01 ms -1 V= 1.66 ms -1 V= 1.41 ms -1 V= 1.17 ms -1 APt fs APt fs APt fs KPa fs 40.06 40.70 160.1 100.0 42.3 52.40 157.7 54.88 124.8 54.88 101.1 45.87 61.23 128.2 69.00 49.66 43.0 61.23 161.2 106.8 43.1 94.22 166.0 94.22 130.1 94.22 112.3 60.70
Table B.l Variation of Steady State Flux with Transmembrane Pressure and Feed Concentration for PALL Membrane at Constant Crossflow Velocity. ( Cone. 6.0 and 18.5% WAV)
131 Membrane PALL Feed Concentration 28.5% W/W V= 1.175 ms-1 V= .896 ms -1 V= .763 ms -1 APt, KPa fs (LMH) APt, KPa fs (LMH) APt, KPa fs (LMH) 45.87 42.3 38.94 32.6 42.21 30.1 49.66 43.0 49.66 32.8 49.66 28.0 60.70 43.1 60.70 33.0 60.70 30.1
Membrane DOULTON Feed Concentration 6.25% W/W ( APt in KPa, fs in LMH) V= 1.02 ms-1 V= .943 ms -1 V= .480 ms -1 V= .307 ms -1 APt fs APt fs APt fs KPa fs 43.91 59.31 69.7 59.31 54.5 127.5 51.51 122.3 72.18 90.1 77.24 68.3 54.83 132.6 59.31 125.0 61.25 89.50 86.8 91.25 65.0 135.2 112.52 157.5 132.24 87.3 120.05 65.1 Membrane DOULTON Feed Concentration 12.0% W/W ( APt in KPa, fs in LMH) V= 1.273 ms -1 V= 1.02 ms -1 V= .727 ms -1 V= .661 ms -1 APt fs APt fs APt fs KPa fs 38.29 102.8 35.30 75.5 59.43 86.0 42.42 64.1 42.42 110.3 53.11 105.4 55.00 119.5 75.52 80.6 62.50 73.5 65.52 105.4 100.05 87.8 110.15 75.7 100.00 116.8 100.07 106.0
Table B.2 Variation of Steady State Flux with Transmembrane Pressure and Feed Concentration for PALL ( Cone. 28.5% W/W ) and DOULTON ( Cone. 6.25 and 28.5 % W/W ) Membranes at Constant Crossflow Velocity.
132 Membrane DOULTON Feed Concentration 17.5 % W/W V= .910 ms-1 V= .796 ms -1 V= .579 ms -1 APt, KPa fs (LMH) APt, KPa fs (LMH) APt, KPa fs (LMH) 54.31 69.5 35.25 51.3 31.25 40.3 62.40 69.6 54.28 58.5 52.50 45.3 78.28 72.2 60.00 60.6 64.48 51.6 106.12 70.1 85.74 62.3 101.26 50.5
. Membrane FAIREY Feed Concentration 13.0 % W/W ( APt in KPa, fs in LMH) V= 1.90 ms-1 V= 1.70 ms-1 V= 1.50 ms -1 APt fs APt fs APt fs 22.50 33.21 212.4 163.1 24.05 144.7 55.84 240.1 36.84 185.8 53.00 36.54 159.1 59.05 224.8 195.2 61.81 175.0 61.80 193.5 Membrane FAIREY Feed Concentration 13.0% W/W V= 1.30 ms-1 V= 1.10 ms -1 APt, KPa fs, LMH APt, KPa fs, LMH
19.50 115.2 16.07 90.3 37.07 137.8 39.84 109.7 67.50 150.8 64.32 115.4
Table B.3 Variation of Steady State Flux with Transmembrane Pressure and Feed Concentration for DOULTON ( Cone. 17.5 % W/W ) and FAIREY (Cone. 13.0% W/W) Membranes at Constant Crossflow Velocity.
133 A p p en d ix C
Application of the Schock Relationship to the Experimental Data
134 C.l Notes on Calculation of Reynolds Numbers of Bingham Plastic Slurries in Tables C.2, C.3 and C.6
As discussed in Chapter Four, the magnesia slurry behaves as a Bingham plastic slurry above 15% W/W concentration. The Reynolds number for Bingham plastic fluids is given by
Re = iVe- He where d = channel diameter, m V = velocity, m s '1 p = density, Kg m " 3 fi e = effective viscosity, Pa s
fie is approximated for Bingham plastic fluids by ( Skelland, 1967 )
^ = - ^ - + 11...... (C .l) 6 V where x y = shear yield, Pa
ri = plastic viscosity, Pa s
There are other relationships that may be used for calculation of Re numbers of Bingham plastic fluids, however, Equation(Cl) is found to be just as effective as other methods.
135 Crossflow Reynolds Velocity ms-1 Number
.428 1.340 8824.0 .523 .560 1.660 10931.0 .685 1.880 12379.7 .802 .655 2.015 13268.7 .875 .715 2.170 14289.3 .961 .785
Table C.l Various parameters of the Runs with PALL Membrane for 6.0% WAV Magnesia Suspension and Parameters Evaluated from the Schock
Equation, p = 1042.0 Kg m-3, p =1.582e-3 Pa s
Crossflow Reynolds Velocity m s -1 Number
1.176 1852.4 .306 .129 1.410 2347.6 .390 .345 1.660 2888.3 .485 .213 2.015 3670.1 .626 .281
Table C.2 Various Parameters of the Runs with PALL Membrane for 18.5 % WAV Magnesia Suspension and Parameters Evaluated from the Schock
Equation, p = 1128.0 Kg m-3, p = 4.837 e-3 Pa s, ty = 1.64 Pa
136 Crossflow Reynolds Velocity ms-1 Number
.763 268.5 .120 .025 .896 352.5 .151 .034 1.175 551.2 .223 .054
Table C.3 Various Parameters of the Runs with PALL Membrane for 28.5 % W/W Magnesia Suspension and Parameters Evaluated from the Schock
Equation, p =1200.0 Kg m-3, 11 = 9.692 e-3 Pa s, ty = 11.2 Pa
Crossflow Reynolds Velocity m s -1 Number
.307 1976.8 .081 .066 .480 3094.0 .143 .116 .943 6072.1 .334 .270 1.019 6555.0 .368 .297
Table C.4 Various Parameters of the Runs with DOULTON Membrane for 6.25 % W/W Magnesia Suspension and Parameters Evaluated from the
Schock Equation, p = 1043.7 Kg m-3 , p = 1.621 e - 3 Pa s
137 Crossflow Reynolds Velocity ms-1 Number
.661 2537.4 .187 .118
.727 2791.0 . 2 1 1 .133 1.019 3908.0 .322 .204 1.273 4883.7 .426 .270
Table C.5 Various Parameters of the Runs with DOULTON Membrane for 12.0
% WAV Magnesia Suspension and Parameters Evaluated from the
Schock Equation, p = 1084.0 Kg m-3 , p = 2.825 e - 3 Pa s
Crossflow Reynolds Velocity ms-1 Number " “ © (" ■f t )
.579 871.8 .124 .051 .796 1341.7 .191 .083
.910 1597.8 .230 .1 0 1
Table C. 6 Various Parameters of the Runs with DOULTON Membrane for 17.5 % WAV Magnesia Suspension and Parameters Evaluated from the
Schock Equation, p = 1122.8 Kg m-3, p = 4.561 e-3 Pa s, ty =1.02Pa
138 Crossflow Reynolds Velocity ms-1 Number “(5)(dd^r) 1.070 3767.2 .335 .204 1.281 4507.8 .420 .255 1.494 5258.8 .510 .310 1.756 6180.0 .625 .380 1.921 6761.3 .700 .425
Table C.7 Various Parameters of the Runs with FAIREY Membrane for 13.0 % W/W Magnesia Suspension and Parameters Evaluated from the Schock
Equation, p = 1091.1 Kg m-3, |i = 3.10e-3 Pa s
139 Appendix D
Calculation of Transient Filtration Period. Cake Permeability and the Film Thickness
140 D. 1 Calculation of Transient Filtration Period
For the following conditions with a PALL membrane C = 2.25 % WAV p = 1016 Kg m-3
V = 0.7 m s -1
APt = 98.6 KPa
1) Calculate fs from Equation (5.8 )
A = 3.532 x 10' 3 from Equation (5.7 ) B = - 5.361x10-5
and Rm = 0.14 K Pahrl’ l m2 or 5.041 x 108Pa s m ( from Table 5.1) Therefore,
98.6 fs = ------72x 1016 98.6 0.14 + 3.532x 10 3- 5.361 x 10 98.6 .7
fs = 164.50 LMH or 4.57x 10' m3m-2s-1 5 and
„ _Apt _98.6 0 Rm -14
or f(> = 704.3 LMH or 19.56 x10-5 m 3nr2 s -1
141 2) Calculation of the Permeability
As a first approximation of the permeability, let the main resistance to the flow be due to the membrane only. Therefore the permeability is given by
where h = wall thickness of the membrane ~2 x 1 0 " 3 m Therefore,
k = 3.97 x 10 _12 m 2 Pa _1 s _1
3) Calculate qj from Equation (5.24)
3.97 x 10"12. 98.6 x 103. ( 0.5 - 0.0225 ) 4.57 x 10 -5 qi = ------^ ' 1- 0.0225 . 4.57 x 10 -5 19.56x10
qj = 0.1392 m 3 nr 2
where Cp = fouling layer solid concentration = %50 W/W
4) Calculate t c from Equation (5.26)
t .1392
(19.56 - 4.57 ) 10 '5
or t £ = 928.8 s
D.2 Estimation of t c from the Experimental Results
From Figure 5.24
qj = 0.016 m 3 nr2
142 This time calculate k from the experimental data of a run at the above conditions. From the experiment
fs = 153.0 LMH or 4.25x 10’ 5 m 3 n r2 s - 1 and from Equation (5.24 )
, 0.016 x 0.225 x 4.25 xlO' 5 k ------j ------—
98.6 x 103 ( 0.5 - 0.0225)< 1 - 4-25X 10 V 19.56x10 or k = 4.152 x 10 _13 m 2 Pa_1 s -1 and from Equation (5.26 )
t 0.016
(19.56-4.25 )10' 5 or t £= 104,5 s
D.3 Use of Experimental Permeability in Estimation of t c
For the following new conditions with a PALL membrane C = 2.25 % W/W p = 1016 Kg m"3
V = 1.34 m s' 1
APt = 76.17 KPa
1) Calculate fs from Equation (5.8 ) and fo as before A = 3.532 x 10' 3 from Equation (5.7 )
B = -5.361 x 10-5
143 and Rm = 0.14 from Table 5.1
Therefore,
u- 76.17 .34 x 1016 76.17 0.14 + 3.532x 10'3-5.361 x 10’ 5 76.17 1.34
fs = 285.1 LMH or 7.92 x 10 -5 m 3 m "2s -1 and
fo = 544.07 LMH or 15.11 x 10 - 5 m 3 m - 2 s - 1
2 ) Calculate qj from Equation (5.24)
4.152 x 10‘13. 76.17 x103. f 0.5 - 0.0225 ) 7.92 x 10' 5 Qi = ------1 - 0.0225 . 7.92 x 10 1.511 xlO' 5
qj = 0.0040 m 3 m" 2
where Cp = fouling layer solid concentration= 50% W/W
3) Calculate t c from Equation (5.26)
t 0.0040
(15.11-7.92 )10 ' 5 or t £=56.1 s
144 D.4 Calculation of the Film Thickness
Film thickness is given by h = R f k
Where R f is found from Equation ( 5.6 ) of the above example and the permeability value calculated from Equation (5.24) is used. Thus, for R f = .128 KPa m2 hr 1_1 or 4.6x10s Pa s m
h = 4.6 x 10 8. 4.152x 10 - 13 or h= 1.910x10-4 m or ~ 0.2 mm
145